Welcome to the electronic edition of Small-signal

stability, control and dynamic performance of power

systems.
The book opens with the bookmark panel and you will
see the contents page. Click on this anytime to return to
the contents. You can also add your own bookmarks.
Each chapter heading in the contents table is clickable
and will take you direct to the chapter. Return using the
contents link in the bookmarks.
The whole document is fully searchable.
Enjoy.

Small-signal stability,
control and
dynamic performance of
power systems

If a reader wishes to cite this publication the authors would be grateful

if the following format were used:
M.J. Gibbard, P. Pourbeik and D.J. Vowles, Small-signal stability, control
and dynamic performance of power systems, University of Adelaide Press,
Adelaide, 2015.

Small-signal stability,
control and
dynamic performance of
power systems
by

M.J. Gibbard
The University of Adelaide

P. Pourbeik
Electric Power Research Institute, USA

D.J. Vowles
The University of Adelaide

Published in Adelaide by
University of Adelaide Press
The University of Adelaide
Level 14, 115 Grenfell Street
South Australia 5005
[email protected]
www.adelaide.edu.au/press
The University of Adelaide Press publishes externally refereed scholarly books by staff of
the University of Adelaide. It aims to maximise access to the Universitys best research by
publishing works through the internet as free downloads and for sale as high quality printed
volumes.
2015 The authors
This work is licenced under the Creative Commons Attribution-NonCommercialNoDerivatives 4.0 International (CC BY-NC-ND 4.0) License. To view a copy of this
licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0 or send a letter to Creative
Commons, 444 Castro Street, Suite 900, Mountain View, California, 94041, USA. This
licence allows for the copying, distribution, display and performance of this work for
non-commercial purposes providing the work is clearly attributed to the copyright holders.
Address all inquiries to the Director at the above address.

For the full Cataloguing-in-Publication data please contact the National Library of Australia:
[email protected]

ISBN (paperback) 978-1-925261-02-8

ISBN (ebook) 978-1-925261-03-5

Book design: M.J. Gibbard and D.J.Vowles

Cover design: Emma Spoehr
Cover image: M.J. Gibbard
Paperback printed by Griffin Press, South Australia

Contents

Preface ...................................................................................... xvii

List of Symbols, Acronyms and Abbreviations......................... xxi
1

Introduction

1.1

Why analyse the small-signal dynamic performance of power systems?

1.2

The purpose and features of the book

1.3

Synchronizing and damping torques

............................................................................5
..........................................................................7

................1

........................................................................2

1.4

Definitions of power system stability

1.5

Types of modes.

1.6

Synchronous generator and transmission system controls

1.7

Power system and controls performance criteria and measures. ...........................13

1.7.1
Power system damping performance criteria ...........................................13
1.7.2
Control system performance measures .....................................................14

.............................................................................................................10
.....................................12

1.8

Validation of power system models

1.9

Robust controllers

...........................................................................15

1.10

How small is small in small-signal analysis?

1.11

Units of Modal Frequency

...........................................................................................18

1.12

Advanced control methods

.........................................................................................18

1.13

References

.........................................................................................................15
............................................................16

......................................................................................................................19

vi

Contents

Control systems techniques for

small-signal dynamic performance analysis

2.1

Introduction ...................................................................................................................23
2.1.1
Purpose and aims of the chapter ................................................................23

2.2

Mathematical model of a dynamic plant or system

2.3

The Laplace Transform

2.4

The poles and zeros of a transfer function.

2.5

The Partial Fraction Expansion and Residues ..........................................................31

2.5.1
Calculation of Residues ................................................................................31
2.5.2
A simple check on values of the residues ..................................................32

2.6

Modes of Response

2.7

The block diagram representation of transfer functions

2.8

Characteristics of first- and second-order systems ..................................................39

2.8.1
First-order system .........................................................................................39
2.8.2
The second-order system ............................................................................40

23

.................................................24

................................................................................................27
..............................................................30

......................................................................................................32
........................................37

2.9

The stability of linear systems

2.10

Steady-state alignment and following errors .............................................................45

2.10.1 Steady-state alignment error. ......................................................................46
2.10.2 The steady-state following error .................................................................48

.....................................................................................44

2.11

Frequency response methods

2.12

The frequency response diagram and the Bode Plot ..............................................51

2.12.1 Plotting the frequency response of the open-loop transfer function ...51
2.12.2 Stability Analysis of the closed-loop system from the Bode Plot .........59

......................................................................................49

2.13

The Q-filter, a passband filter

2.14

References

.....................................................................................61

State equations, eigen-analysis and applications

3.1

Introduction ...................................................................................................................63
3.1.1
Example 3.1. ..................................................................................................63
3.1.2
Example 3.2 ...................................................................................................65

3.2

The concept of state and the state equations

3.3

The linearized model of the non-linear dynamic system ........................................68

3.3.1
Linearization procedure ...............................................................................69

3.4

Solution of the State Equations ..................................................................................71

3.4.1
The Natural Response ..................................................................................71
3.4.2
Example 3.3 ...................................................................................................72
3.4.3
Example 3.4: Natural response ...................................................................73
3.4.4
The Forced Response ...................................................................................74
3.4.5
Example 3.4 (continued). .............................................................................74

......................................................................................................................62

63

...........................................................67

Contents

vii

3.5

Eigen-analysis ................................................................................................................74
3.5.1
The eigenvalues of the state matrix, A .......................................................74
3.5.2
A note on eigenvalues, modes and stability ..............................................76

3.6

Decoupling the state equations

3.7

Determination of residues from the state equations

3.8

Determination of zeros of a SISO sub-system

3.9

Mode shapes ..................................................................................................................80

3.9.1
Example 3.5: Mode shapes and modal responses ...................................82

3.10

Participation Factors .....................................................................................................83

3.10.1 The relative participation of a mode in a selected state ..........................83
3.10.2 The relative participation of a state in a selected mode ..........................84
3.10.3 Example 3.6: Participation factors .............................................................85

3.11

Eigenvalue sensitivities

3.12

References

Small-signal models of synchronous generators,

FACTS devices and the power system

4.1

Introduction

4.2

Small-signal models of synchronous generators ......................................................90

4.2.1
Structure of the per-unit linearized synchronous generator models .....90
4.2.2
Generator modelling assumptions .............................................................93
4.2.3
Electromagnetic model in terms of the per-unit coupled-circuit
parameters ......................................................................................................94
4.2.4
Alternative d- and q-axis rotor structures ...............................................111
4.2.5
Per-unit electromagnetic torque and electrical power output ..............113
4.2.6
Per-unit rotor equations of motion ..........................................................114
4.2.7
Non-reciprocal definition of the per-unit field voltage and current ...115
4.2.8
Modelling generator saturation .................................................................118
4.2.9
Balanced steady-state operating conditions of the coupled-circuit
model ............................................................................................................125
4.2.10 Interface between the generator Park/Blondel reference frame and
the synchronous network reference frame .............................................130
4.2.11 Linearized coupled-circuit formulation of the generator model
equations ......................................................................................................133
4.2.12 Transfer-function representation of the electromagnetic equations ...138
4.2.13 Electromagnetic model in terms of classically-defined standard
parameters ....................................................................................................145
4.2.14 Generator parameter conversions ............................................................156

4.3

Small-signal models of FACTS Devices ..................................................................157

4.3.1
Linearized equations of voltage, current and power at the AC
terminals of FACTS Devices: general results .........................................159
4.3.2
Model of a Static VAR Compensator (SVC) ..........................................163
4.3.3
Model of a Voltage Sourced Converter (VSC) .......................................165

...................................................................................76
...............................................77

........................................................78

.................................................................................................85

......................................................................................................................87

89

...................................................................................................................89

viii

Contents
4.3.4
4.3.5
4.3.6
4.3.7
4.3.8
4.3.9

Simplified STATCOM model ...................................................................172

Modelling of HVDC Transmission Systems ..........................................177
Model of a distributed-parameter HVDC transmission line or cable 178
Model of HVDC transmission with Voltage Sourced Converters
(VSCX) .........................................................................................................180
Model of HVDC transmission with Voltage Commutated
Converters ....................................................................................................181
Thyristor Controlled Series Capacitor (TCSC) ......................................192

4.4

Linearized power system model ...............................................................................195

4.4.1
General form of the linearized DAEs for a device and its controls ...196
4.4.2
General form of the network nodal current equations .........................197
4.4.3
General form of the linearized DAEs of the interconnected
power system ...............................................................................................198
4.4.4
Example demonstrating the structure of the linearized DAEs ...........199

4.5

Load models ................................................................................................................201

4.5.1
Types of load models .................................................................................201
4.5.2
Linearized load models ..............................................................................202

4.6

References

App. 4I

Linearization of the classical parameter model of the generator.

App. 4II

Forms of the equations of motion of the rotors of a generating unit

4II.1

Introduction

4II.2

Shaft equations expressed in terms of per-unit angular speed and torques

4II.3

Per-unit shaft acceleration equation in terms of rotor-speed and power

4II.4

Shaft acceleration equation neglecting speed perturbations in the

torque/power relationship .........................................................................................220

4II.5

A common misunderstanding in calculating the accelerating torque and

power ............................................................................................................................220

Concepts in the tuning of power system

stabilizers for a single machine system

5.1

Introduction

5.2

Heffron and Phillips Model of single machine - infinite bus system

5.3

Synchronizing and damping torques acting on the rotor of a synchronous

generator .......................................................................................................................227

5.4

The role of the Power System Stabilizer - some simple concepts

5.5

The inherent synchronizing and damping torques in a SMIB system ................232

5.5.1
Example 5.1 .................................................................................................233

5.6

Effect of the excitation system gain on stability

5.7

Effect of an idealized PSS on stability

5.8

Tuning concepts for a speed-PSS for a SMIB system

....................................................................................................................205
...........209
...... 216

.................................................................................................................216
.......218
...........219

223

.................................................................................................................223
.................226

.......................230

....................................................235

.....................................................................236
...........................................237

Contents

ix
5.8.1
5.8.2
5.8.3
5.8.4
5.8.5
5.8.6

Determination of compensating transfer function ................................238

The nature of the P-Vr characteristic ......................................................242
Example 5.2: Evaluate the P-Vr characteristics of the generator
and determine the PSS compensating transfer function. .....................243
Determination of the damping gain k of the PSS .................................245
Example 5.3. Calculation of the damping gain setting for the PSS ....245
Washout and low-pass filters .....................................................................246

5.9

Implementation of the PSS in a SMIB System .......................................................248

5.9.1
The transfer function of the PSS ..............................................................248
5.9.2
Example 5.4. The dynamic performance of the speed-PSS .................249
5.9.3
Analysis of the variation in the mode shifts over the range of
operating conditions ...................................................................................251

5.10

Tuning of a PSS for a higher-order generator model in a SMIB system ............257

5.10.1 The power system model ...........................................................................258
5.10.2 Calculation of the synchronizing and damping torque coefficients ....259
5.10.3 Calculation of the P-Vr characteristics for a SMIB system with
high-order generator models .....................................................................260
5.10.4 Example 5.5: Tuning and analysis of the performance of the PSS
for the higher-order generator model ......................................................261
5.10.5 The P-Vr characteristics for a SMIB system with a 6th order
generator model ..........................................................................................264
5.10.6 Tuning a speed-PSS for a SMIB system with a 6th order generator
model ............................................................................................................265

5.11

Performance of the PSS for a higher-order generator model

5.12

Alternative form of PSS compensation transfer function

5.13

Tuning an electric power-PSS based on the P-Vr approach .................................279

5.13.1 Example 5.6: Tuning of a power-based PSS ...........................................281

..............................273

....................................278

5.14

Summary: P-Vr approach to the tuning of a fixed-parameter PSS

5.15

References

.....................284

App. 5I

......................................................................................................................288

5I.1

K-coefficients, Heffron and Phillips Model of SMIB System

5I.2

Transfer function of the SMIB system with closed-loop control of

terminal voltage ...........................................................................................................289

5I.3

Model of the 6th-order generator and excitation system ......................................289

5I.3.1 State-space model .......................................................................................289
5I.3.2 Calculation of the inherent torque coefficients. .....................................291

Tuning of PSSs using methods based on

Residues and the GEP transfer function

6.1

Introduction

6.2

Method of Residues

....................................................................................................................286
.............................288

293

.................................................................................................................293
....................................................................................................294

Contents
6.2.1

Theoretical basis for the Method

.............................................................294

6.3

Tuning a speed-PSS using the Method of Residues ..............................................297

6.3.1
Calculation of the compensation transfer function of the PSS ...........297
6.3.2
Design Case C. Performance of the PSS with increasing PSS gain ....299
6.3.3
Significance of the PSS gains ....................................................................300

6.4

Conclusions, Method of Residues

6.5

The GEP Method

6.6

Tuning a speed-PSS using the GEP Method ..........................................................303

6.6.1
Example 2. Performance of the PSS based on Design Case C ...........303

6.7

Conclusions, GEP method

6.8

References

............................................................................300

.......................................................................................................302

........................................................................................306

....................................................................................................................307

App. 6I

......................................................................................................................309

6I.1

Algorithm for the calculation of stabilizer parameters

6I.2

Calculation of the nominal upper limit of the range of stabilizer gains

Introduction to the Tuning of

Automatic Voltage Regulators

7.1

Introduction .................................................................................................................313
7.1.1
Purposes .......................................................................................................313
7.1.2
Coverage of the topic .................................................................................314

7.2

The excitation control system of a synchronous generator

7.3

Types of compensation and methods of analysis

7.4

Steady-state and dynamic performance requirements on the generator and

excitation system .........................................................................................................316

7.5

A single-machine infinite-bus test system

7.6

Transient Gain Reduction (TGR) Compensation ..................................................320

7.6.1
Introduction .................................................................................................320
7.6.2
The performance of the generator and compensated excitation
system on-line ..............................................................................................321
7.6.3
The performance of the generator and compensated excitation
system off-line .............................................................................................324
7.6.4
Comparison of performance of the excitation control system
on- and off-line ...........................................................................................326

7.7

PID compensation ......................................................................................................327

7.7.1
PID Compensation: Theoretical Background ........................................329
7.7.2
Tuning methodology for PID Compensation Types 1 and 2A ...........335

7.8

Type 2B PID Compensation: Theory and Application to AVR tuning .............347

7.8.1
Tuning of Type 2B PID compensation ...................................................348
7.8.2
Example: Evaluation of Type 2B PID parameters. ...............................349

7.9

Proportional plus Integral Compensation

.........................................309
.............311

313

.................................314

...................................................316

...............................................................319

..............................................................351

Contents

xi
7.9.1
7.9.2

Simple PI Compensation ...........................................................................351

Conversion to a PID Compensator with an additional lead-lag
block ..............................................................................................................352

7.10

Rate feedback compensation .....................................................................................354

7.10.1 Method of analysis ......................................................................................354
7.10.2 Tuning of the Excitation System (ES) .....................................................354
7.10.3 Rate feedback compensation using Frequency Response Methods. ...356
7.10.4 Rate feedback compensation using the Root Locus Method ..............368

7.11

Tuning of AVRs with Type 2B PID compensation in a three-generator

system ...........................................................................................................................371
7.11.1 The three-generator, 132 kV power system ............................................371
7.11.2 The frequency response characteristics of the brushless exciter and
generator ......................................................................................................373

7.12

Summary, Chapter 7

7.13

References

App. 7I

...................................................................................................383

....................................................................................................................384

.....................................................................................................................386

7I.1

Generator and exciter parameters ............................................................................386

7I.1.1 Parameters for the 6th order generator and a simple exciter ...............386
7I.1.2 Parameters for the 5th order salient-pole generator and a
brushless AC exciter ...................................................................................386

7I.2

Models of the brushless AC exciter

7I.3

PI Compensation using positive feedback

7I.4

Integrator Wind-up Limiting

7I.5

A phase-matching method for constant phase margin over an

appropriate frequency range ......................................................................................392

Types of Power System Stabilizers

8.1

Introduction

8.2

Dynamic characteristics of washout filters ..............................................................399

8.2.1
Time-domain responses .............................................................................399
8.2.2
Frequency-domain responses ....................................................................401
8.2.3
Comparison of dynamic performance between a single and two
washout filters. .............................................................................................403

8.3

Performance of a PSS with electric power as the stabilizing signal. ...................404

8.3.1
Transfer function and parameters of the electric power pre-filter. .....404
8.3.2
Dynamic performance of a speed-PSS with an electric power
pre-filter. .......................................................................................................406

8.4

Performance of a PSS with bus-frequency as the stabilizing signal. ...................407

8.4.1
Dynamic performance of a speed-PSS with a bus-frequency
pre-filter ........................................................................................................408
8.4.2
Degradation in damping with the bus-frequency pre-filter ..................410

.........................................................................386
..............................................................388

.....................................................................................391

397

.................................................................................................................397

xii

Contents

8.5

Performance of the Integral-of-accelerating-power PSS ..................................413

8.5.1
Introduction .................................................................................................413
8.5.2
Torsional modes introduced by the speed stabilizing signal ................414
8.5.3
The electric power signal supplied to the pre-filter ...............................414
8.5.4
The Ramp Tacking Filter (RTF) ...............................................................415

8.6

Conceptual explanation of the action of the pre-filter in the IAP PSS ..............416
8.6.1
Action of the pre-filter, no washout filters .............................................416
8.6.2
Effect of the washout filters and integrators on the performance
of the pre-filter ............................................................................................419
8.6.3
Dynamic performance of the complete pre-filter .................................424
8.6.4
Potential causes of degradation in performance of the pre-filter
of the IAP PSS ............................................................................................429

8.7

The Multi-Band Power System Stabilizer

8.8

Concluding remarks

8.9

References

................................................................433

....................................................................................................436

....................................................................................................................438

App. 8I

...................................................................................................................... 441

8I.1

Action of the Ramp Tracking Filter (RTF)

8I.2

Steady-state conditions at the input and output of the RTF and associated
tracking errors for mechanical power input ...........................................................442
8I.2.1 With and without an Ideal Integrator ......................................................442
8I.2.2 With a Pseudo Integrator ..........................................................................444

8I.3

Multi-Band PSS transfer function

Basic Concepts in the Tuning of

PSSs in Multi-Machine Applications

9.1

Introduction .................................................................................................................447
9.1.1
Eigenvalues and Modes of the system .....................................................448

9.2

Mode Shape Analysis ..................................................................................................449

9.2.1
Example 1: Two-mass spring system .......................................................450
9.2.2
Example 2: Four-mass spring system ......................................................456

9.3

Participation Factors ...................................................................................................459

9.3.1
Example 4.3 .................................................................................................460

9.4

Determination of the PSS parameters based on the P-Vr approach with

speed perturbations as the stabilizing signal ...........................................................462
9.4.1
The P-Vr transfer function in the multi-machine environment ..........462
9.4.2
Transfer function of the PSS of generator i in a multi-machine
system ...........................................................................................................466

9.5

Synchronising and damping torque coefficients induced by PSS i on

generator i ....................................................................................................................469

9.6

References

.............................................................441

............................................................................444

447

....................................................................................................................471

Contents

xiii

10

Application of the PSS Tuning Concepts

to a Multi-Machine Power System

10.1

Introduction

10.2

A fourteen-generator model of a longitudinal power system ..............................477

10.2.1 Power flow analysis .....................................................................................479
10.2.2 Dynamic performance criterion ...............................................................479

10.3

Eigen-analysis, mode shapes and participation factors of the 14-generator

system, no PSSs in service .........................................................................................481
10.3.1 Eigenvalues of the system with no PSSs in service ...............................481
10.3.2 Application of Participation Factor and Mode Shape Analyses to
Case 1 ............................................................................................................482

10.4

The P-Vr characteristics of the generators and the associated synthesized

characteristics ...............................................................................................................487

10.5

The synthesized P-Vr and PSS transfer functions

10.6

Synchronising and damping torque coefficients induced by PSS i on

generator i ....................................................................................................................496

10.7

Dynamic performance of the system with PSSs in service ..................................500

10.7.1 Assessment of dynamic performance based on eigen-analysis ...........500
10.7.2 Assessment of dynamic performance based on participation and
mode-shape analysis ...................................................................................503
10.7.3 Assessment of dynamic performance based on time responses .........505

10.8

Intra-station modes of rotor oscillation

10.9

Correlation between small-signal dynamic performance and that following

a major disturbance .....................................................................................................509
10.9.1 A transient stability study based on the fourteen-generator system ...509
10.9.2 The analysis of modal interactions ...........................................................513

10.10

Summary: Tuning of PSSs based on the P-Vr approach

10.11

References

App. 10I

......................................................................................................................520

10I.1

Modes of rotor oscillation for Cases 2, 3, 5 and 6

10I.2

Data for steady-state power flow analysis

10I.3

Data for dynamic performance analysis ..................................................................527

10I.3.1 Excitation System Parameters ................................................................528

11

Tuning of FACTS Device Stabilizers

11.1

Introduction

11.2

A simplistic tuning procedure for a SVC

11.3

Theoretical basis for the tuning of FACTS Device Stabilizers

11.4

Tuning SVC stabilizers using bus frequency as a stabilizing signal

475

.................................................................................................................475

.................................................493

..................................................................507

......................................516

....................................................................................................................518
.................................................520

...............................................................522

531

.................................................................................................................531
..............................................................533
............................536
.....................538

xiv

Contents
11.4.1

Use of bus frequency as a stabilizing signal for the SVC, BSVC_4

....539

11.5

Use of line real-power flow as a stabilizing signal for a SVC

...............................545

11.6

Use of bus frequency as a stabilizing signal for the SVC, PSVC_5

11.7

Tuning a FDS for a TCSC using a power flow stabilizing signal ........................551
11.7.1 Gain range for the stability of TCSC with the FDS in service ............555
11.7.2 Inter-area mode trajectories with increasing stabilizer gain .................556

11.8

Concluding comments ...............................................................................................557

11.8.1 Improving the damping of inter-area modes using FACTS devices ...557
11.8.2 Robustness of FDSs ...................................................................................558
11.8.3 Estimated versus calculated mode shifts .................................................559
11.8.4 The notion of a nominal upper gain for FDSs ....................................559

11.9

References

12

The Concept, Theory, and Calculation of

Modal Induced Torque Coefficients

12.1

Introduction

12.2

The Concept of Modal Induced Torque Coefficients (MITCs) ..........................565

12.2.1 Conventional frequency response techniques versus modal analysis .565
12.2.2 Modal torque coefficients induced by the action of a power
system stabilizer ..........................................................................................565
12.2.3 Modal torque coefficients induced by the action of a FACTS
device stabilizer ...........................................................................................567
12.2.4 Modal torque coefficients induced by centralized stabilizers ...............569

12.3

Transfer function matrix representation of a linearized multi-machine

power system and its controllers ..............................................................................569

12.4

Modal torque coefficients induced by a centralized speed PSS

12.5

Modal torque coefficients induced by a centralized FDS

12.6

General expressions for the torque coefficients induced by conventional,

decentralized PSSs & FDSs .......................................................................................580
12.6.1 The total modal induced torque coefficients for systems with
both PSSs and FDSs ..................................................................................582
12.6.2 A relationship between modal induced torque coefficients and
incremental stabilizer gains .......................................................................582

12.7

References

App. 12I

...................................................................................................................... 585

.....................548

....................................................................................................................559

563

.................................................................................................................563

...........................574

....................................577

....................................................................................................................583

12I.1

Appendix: System response at a single modal frequency

12I.2

Reducing the TFM model in Figure 12.2 to those in 12.3 and 12.4

.....................................585

12I.3

Elements of the output matrix, C

...................586

............................................................................587

Contents

xv

13

Interactions between, and effectiveness of, PSSs

and FDSs in a multi-machine power system

13.1

Introduction

13.2

Relationship between rotor mode shifts and stabilizer gain increments ............590
13.2.1 Relationship between residues and MITCs in calculation of mode
shifts ..............................................................................................................593
13.2.2 Concept of interactions ...........................................................................594
13.2.3 Relationships between mode shifts, MITCs, participation factors
and stabilizer gains ......................................................................................595

13.3

Case Study: Contributions to MITCs/Mode Shifts by PSSs and generators ....596

13.3.1 Contributions to the MITC of each generator, local mode B .............599
13.3.2 Contributions of the mode shifts of each generator to mode B
damping ........................................................................................................602
13.3.3 Contributions to the MITCs of each generator, inter-area mode M ..604
13.3.4 Contributions of the mode shifts of each generator to the Mode M
damping ........................................................................................................606

13.4

Stabilizer damping contribution diagrams

13.5

Comparison of the estimated and actual mode shifts for increments in

stabilizer gain settings .................................................................................................614

13.6

Summary .......................................................................................................................616
13.6.1 Interactions ..................................................................................................616
13.6.2 Relative Effectiveness of Stabilizers .........................................................617

13.7

References

14

Coordination of PSSs and FDSs using Heuristic

and Linear Programming Approaches

14.1

Introduction

14.2

The 14-generator power system ...............................................................................623

14.2.1 Evaluation of the transfer function for the SVC at bus 212. ...............624

14.3

A Heuristic Coordination Approach ........................................................................626

14.3.1 Coordination of stabilizers for damping the inter-area modes ............626
14.3.2 Coordination of local-area modes. ...........................................................633

14.4

Simultaneous Coordination of PSSs and FDSs using Linear Programming .....634

14.4.1 Introduction .................................................................................................634
14.4.2 Comment on the LP solution: optimality versus uniqueness ..............635
14.4.3 Coordination of PSSs and FDSs ..............................................................636

14.5

Case study: Simultaneous coordination in a multi-machine power system

of PSSs and FDSs using linear programming .........................................................639
14.5.1 Scenario 1: Inter-area modes. Maximum PSS & FDS gain 40 pu. ......642
14.5.2 Scenario 2: Inter-area modes. Limits PSS gains 40 pu, FDSs 20 pu ...644
14.5.3 Scenario 3: Inter-area modes. Limits: all stabilizer gains 20 pu ...........645

589

.................................................................................................................589

..............................................................609

....................................................................................................................618

621

.................................................................................................................621

xvi

Contents
14.5.4
14.5.5

Scenario 4: Inter-area modes. Various limits ..........................................645

Scenario 5: Inter- and local-area modes .................................................645

14.6

Concluding remarks

14.7

References

....................................................................................................647

....................................................................................................................650

Preface

We have written this book in the hope that the following engineers, or potential engineers,
will benefit from it:

Recent graduates in electrical engineering who need to understand the tools and techniques currently available in the analysis of small-signal dynamic performance and
design.

Practicing electrical engineers who need to understand the significance of more recent
developments and techniques in the field of small-signal dynamic performance.

Postgraduate students in electrical engineering who need to understand current developments in the field and the need to orient their research to achieve practical, useful
outcomes.

Undergraduate electrical engineering students in courses oriented towards electric

power engineering in which there is an introductory subject in power system dynamics
(for access to basic material).

Managerial staff with responsibilities in power system planning, and system stability
and control.

An aim of the book is to provide a bridge between the mathematical/theoretical and physical/practical significance to the topic. Some of the fundamental background relevant to the
main topics of the book is presented in the early chapters so that the necessary material is
readily available to the reader in the one book.

Because the emphasis is on controllers for generators, for FACTS and other devices,
the pertinent topics in classical control and eigenanalysis techniques are provided in
Chapters 2 and 3.

The authors have covered in Chapter 4 a wide range of small-signal generator models,
equations, and associated material. Third- to eighth-order generator models in their
coupled-circuit and operational parameter versions are described. The following features are also included in the generator models: (i) the 'classical' and 'exact' definitions

xvii

xviii

Preface
of the operational parameters; (ii) the various approaches to the modelling of saturation; (iii) the formulation of the differential-algebraic generator equations to exploit
sparsity. These models and features are employed in the Mudpack software package.
Small-signal equations and models of FACTS devices employed in the software are
also described. Devices covered include SVCs, STATCOMS, Thyristor Controlled
Series Compensators, HVDC links with Voltage Source Converters or with line-commutated converters.

In Chapters 5, 9 and 10 there is an emphasis on practical robust techniques, based on

the P-Vr method, for the design of robust stabilizers for generators in multi-machine
systems.

Two other techniques for the tuning of stabilizers, the GEP and the Method of Residues, are examined in Chapter 6. The benefits and limitations of the two as well as
those of the P-Vr method are reviewed.

An introduction to the tuning of Automatic Voltage Regulators is provided in Chapter

7. The authors have attempted to outline in some detail potential design approaches
for (i) TGR (transient gain reduction), (ii) PI and various types of PID control; (iii)
brushless, static and conventional excitation systems.

A detailed analysis of the 'Integral of accelerating power' stabilizer for generators, not
previously published, is provided in Chapter 8, together with other practical PSS
structures.

In Chapter 11 the tuning of stabilizers for various FACTS devices in a large power system for operation over a range of conditions is described and illustrated, together with
the merits and limitations of the design.

The concept, theory, and calculation of Modal Induced Torque Coefficients are outlined in Chapter 12,. This is a new method of analysis, developed by one of the
authors and forms the basis for Chapter 13. The synchronizing and damping torques
induced on generator shafts at the modal frequencies by both PSSs and FACTS device
stabilizers are derived.

The interactions between, and effectiveness of, PSSs and FACTS device stabilizers in a
multi-machine power system are analysed in Chapter 13. A new and potentially valuable tool, which is based on the Stabilizer Damping Contribution Diagrams (SDCD)
and developed by the authors, enables the engineer to assess the effectiveness of stabilizers installed on generators or FACTS devices in enhancing the damping and stability
of the power system.

In Chapter 14 the coordination of PSSs with PSSs, or PSSs with FACTS device stabilizers is achieved by either heuristic or an optimization techniques. In either case it is
based on the newly developed tool, the SDCD, of Chapter 13.

Preface

xix

A number of chapters in the book are based on a PhD Thesis by Pouyan Pourbeik, Design
and Coordination of Stabilisers for Generators and FACTS devices in Multi-machine Power
Systems, The University of Adelaide, Australia, 1997. The comprehensive, small-signal
modelling of devices is a major contribution by David Vowles to the development of the
Mudpack software, as is the architecture of - and graphics in - the package. Many of these
developments were based on the earliest versions of the software written by Rainer Korte.
The authors are indebted to the following organizations for their support through R & D
grants for the development of Mudpack software and research in the field of power system
dynamics: Australian Energy Market Operator, Powerlink Queensland, TransGrid (New
South Wales), Transend Networks (Tasmania) and ElectraNet (South Australia).
Finally, we wish to thank our families for their support and patience over the long period of
the gestation of the book.
Michael Gibbard

Pouyan Pourbeik

David Vowles

List of Symbols, Acronyms

and Abbreviations

Mathematical symbols
a ij , a i * , a i , A
*

(i,j)th element, the ith row and the ith column vector of the matrix A,
respectively. All elements may have real of complex values.

1
T
A , A , A

Inverse, the transpose complex conjugate transpose of the matrix A.

Real and imaginary parts of a complex number.

I, 0, diag a i

Identity, null and diagonal matrices.

j; s
x , x 0 , x

1 ; the Laplace operator.

Time derivative, initial condition, and perturbation of variable x

X 3 = D x1 x2 x3 Diagonal matrix X 3 , elements x1 x2 x3

Variables and parameters

A, B
C, D

System state and input matrices.

System output and direct-transmission matrices.

B , C

Mode controllability and observability matrices.

D, Kd

Damping coefficients

Susceptance

E fd , E q

Field voltage, and voltage proportional to d-axis flux linkages.

xxi

xxii

List of Symbols, Acronyms and Abbreviations

Frequency of a sinusoidal signal

f0

System frequency in Hertz ( 50 Hz unless otherwise stated)

Gs

A transfer function

Inertia constant of the rotating generating or motoring unit

Transfer Function Matrix (see Section 12.3 for definitions of various
TFMs)

Hi s
H ij s

Transfer function from input j to output i

I dc

DC current through HVDC link.

PSS damping gain.

Kd , D

Damping coefficients

Twice the inertia constant, M = 2H

differential operator d dt
Torque of electromagnetic origin; prime-mover torque (power, if in
per unit)
Reactive power output or flow
General symbol for various resistances

rh

Residue of the mode h

Laplace operator
Time (s)
General symbol for various time constants (s)

T hij

Complex modal induced torque coefficient

V, W

Right and left modal matrices of the state matrix A.

Reference signals at the controller summing junction of an AVR of a
FACTS device; of a generator

Pe , Pm
Q

Ur , Vr
Vt

Terminal voltage

General symbol for various reactances

Real part of a complex quantity, e.g. of an eigenvalue in Np/s
Imaginary part of a complex quantity, e.g. of an eigenvalue in rad/s

Total modal induced torque coefficient ( T j )

Generator rotor angle (degrees)

x implies perturbation in variable x ; also
in parameter k
Damping ratio

implies perturbation

List of Symbols, Acronyms and Abbreviations

Angle (normally in degrees)

Real or complex eigenvalue or mode
Matrix of diagonal eigenvalues
Damping ratio
Participation factor
Real part of an eigenvalue or mode
Speed-torque coefficient transfer function; also a time constant
Flux linkages
Stabilizing signal for FACTS device

Base value of speed

Rotational speed in rad/s, or imaginary part of a complex quantity,

e.g. of an eigenvalue

Frequency of an exciting sinusoidal signal in rad/s

Undamped natural frequency in rad/s

Synchronous speed (= 2f0 ) in rad/s

ohm
Symbol for real part of a complex number
Symbol for imaginary part of a complex number

List of Acronyms and Abbreviations

AVR
cc
dB
DAE
DC
ECS
em
ES
FACTS
FDS
FVT
GEP(s)
GEPSDD

Automatic Voltage Regulator

Couples-circuit
decibel
Differential-algebraic equation
Direct current (or may imply zero frequency)
Excitation control system
Electromechanical
Excitation system
Flexible AC transmission systems
FACTS device stabilizer
Final Value Theorem
Generator, excitation system and power system transfer function
GEP with shaft dynamics disabled

xxiii

xxiv

List of Symbols, Acronyms and Abbreviations

HVDC
IAP

High voltage direct current

Integral of accelerating power (PSS)

IPE

1Integral of electric power signal, -----P dt

2H e

IPM

1P dt
Integral of mechanical power signal, -----2H m

LPF
MB-PSS
MIMO
MITC
MMPS
MS
o.c
O.C.C.
*op
PF
PI
PID

Low pass (filter)

Multi-band PSS
Multi-input multi-output
Modal induced torque coefficient
Multi-machine power system
Mode shape
open circuit
Open-circuit characteristic

PSS
PSSE
PTI
pu
P-Vr TF
PWM
RMS
RI
RTF
SDCD
SI
SISO
SMIB
STATCOM
SVC

nth order operation impedance (model of a generator)

Participation factor
Proportional plus integral (compensation)
Proportional plus integral plus derivative (compensation)
Power Oscillation Damper (referred to as a FACTS device stabilizer,
FDS, in the following chapters)
Power system stabilizer
Power System Simulator (Software package from PTI)
Power Technologies International.
per-unit
P-Vr transfer function
Pulse Width Modulation
root-mean-square
A synchronously rotating network reference frame
Ramp tracking filter
Stabilizer damping contribution diagram
International System of Units
Single-input single-output
Single machine infinite bus (system)
Static compensator
Static var compensator

TCCX

Voltage-commutated thyristor-controlled converters with a HVDC link

TCR

Thyristor controlled reactor

POD

List of Symbols, Acronyms and Abbreviations

TCSC
TF
TFM
TGR
TSP
VSC
VSCX
XPVR
WO

Thyristor controlled series compensator

Transfer function
Transfer function matrix
Transient gain reduction
Transmission Service Provider
Voltage sourced converter
Voltage source converters with HVDC transmission
Extended P-Vr
Washout (filter)

xxv

Chapter 1
Introduction

1.1 Why analyse the small-signal dynamic performance of power systems?

We shall be concerned mainly with the analysis of the dynamic performance and control of
large, interconnected electric power systems in the following chapters. The differential-algebraic equations which describe the behaviour of a power system are inherently non-linear.
Among the non-linearities are functional types (e.g. sin ), product types (e.g. voltage
current), limits on controller action, saturation in magnetic circuits, etc. The general method
of assessing the performance of the system, with all its non-linearities, is through a time-domain simulation which reveals the response of the system to a specific disturbance, e.g. a
fault, the loss of a generating unit, line switching. Typically, it may be necessary to conduct
many such studies with disturbances applied in various locations in the system to ascertain
its stability and dynamic characteristics. Even with many such studies, many of the characteristics of the dynamic behaviour may be missed and insights into system performance lost.
In small-signal analysis of dynamic performance of multi-machine systems the stability and
characteristics of the system are readily derived from eigenanalysis and other tools. Furthermore, in such linear analysis the design of controllers and their integration into the dynamics
of plant are facilitated.
Modern linear control system theory contains many powerful techniques, not only for determining the stability and dynamic characteristics of large linear systems, but also for tuning
controllers that satisfy steady-state and dynamic performance specifications. Fortunately

Introduction

Ch. 1

and importantly, Henri Poincar [1] showed that if the linearized form of the non-linear system is stable, so is the non-linear system stable at the steady-state operating condition at
which the system is linearized. Moreover, the dynamic characteristics of the system at the
selected operating condition can be established from linear control system theory and, as
long as the perturbations are small, the time-domain responses can be calculated. With such
information the design of linear controllers may be undertaken and the resulting controls
embedded in the non-linear system. In practice, if the modelling of the devices is adequate,
small-signal tests involving generator controls, for example, have revealed close agreement
between simulation and test results. Continuously-acting controllers of interest for synchronous generating units are Automatic Voltage Regulators (AVRs), Power System Stabilizers
(PSSs) and speed governors. In a later chapter an analysis of the controls and stabilizers for
FACTS-based devices is conducted; such stabilizers are commonly called a Power Oscillations Dampers (PODs). Many of the techniques and control concepts are also applicable to
the small-signal analysis of dynamic performance of wind turbine-generators and other technologies.

1.2

The purpose and features of the book

The main purpose of this book is to introduce the graduate engineer to the concepts and
applications of small-signal analysis and controller design for the enhancement of the dynamic performance of multi-machine power systems. To this end, the analyses of the control and dynamic performance are illustrated by examples based on an interconnected highvoltage system comprising fourteen generating stations and various types of FACTS devices. An emphasis in the book is on more recent theoretical developments and application to
practical issues which are amenable to small-signal analysis using a comprehensive software
package. In addition, the tools in - and features of - such a software package for analysis and
controller design are illustrated.
The aim and features of the book are illustrated in the following summary.
1.

In the following chapters it is assumed that the reader has already been introduced to
the basics of: (i) the steady-state and dynamic performance of power systems [2], [3],
[4], [5], [6], [7], [8] and (ii) control system theory [9], [10]. However, because controller design and tuning are described in later chapters of this book, Chapters 2 and 3
are devoted particularly to those aspects of basic control theory and the associated
analysis and design techniques which are employed in later. Practical insights and limitations in control theory and analysis are emphasized in order to isolate that material
which is important for application in later chapters.

2.

For the practical design of robust power system stabilizers (PSSs), a tuning approach
based on the generator P-Vr characteristics, which are a development of the GEP
Method for single generator applications, is applied to multi-machine systems [11],
[12]. The uses of the P-Vr characteristics are (a) analysed in detail in Chapter 5 for a
single generator power system in order to explain the features of the P-Vr Method

Sec. 1.2

The purpose and features of the book

and its limitations; (b) applied in Chapters 9 and 10 to the multi-machine system
over a range of operating conditions.
3.

It is shown in Chapters 5 and 10 that the rationale in the tuning of PSSs based on
the P-Vr method is that:
(i)

there are two important components in the PSS transfer function kG s which are
essentially decoupled for practical purposes 1:

(a)

(b)
(c)

(ii)
(iii)

(iv)

the rotor modes are more-or-less directly left-shifted 2 by the PSS

compensating transfer function G s with increase in the PSS damping gain, k 3;
the extent of the left-shift of the rotor modes is determined by the
damping gain, k;
the incremental left-shifts of the rotor modes are linearly related to
increments in damping gain for changes about the nominal values (typically 5 to 10 pu on generator MVA rating) [11];

the tuning of the PSS is based on a more extreme set of encompassing operating conditions;
the PSS damping gain has special significance: it is also the damping torque
coefficient induced by the PSS on the shaft of the generator. It forms the basis for
the theoretical developments in Chapters 5, 10, 12 to 14.
the PSS damping gain can be adjusted to swamp out any inherent negative
damping torques;

(v)

as a result of (i), (ii) and (iv) above, the PSS transfer function kG s is robust
over the encompassing range of normal and outage operating conditions [12];

(vi)

the PSS damping gain, when expressed in per unit on generator MVA rating,
is a meaningful quantity, unlike the term PSS gain currently used. PSS
damping gains less than 10 pu are low, are normal between 10 and 30 pu, and
greater than 30 pu tend to be high.
As opposed to the application of advanced control techniques, the significance of the above rationale is that the natural characteristics of the generator
and the system are employed and thus meaningful insight and explanations
for the dynamic behaviour of the system can be established.

1. The features described in items (i)(a) and (i)(b) are illustrated for six operating conditions
in Figure 10.26
2. By direct left-shift is implied that the mode shift is j0 , 0 . As explained in
Chapter 13, deviations from the direct left-shift of modes are mainly due to interactions
between multi-machine PSSs and non-real generator participation factors.
3. The damping gain of the PSS is defined in Section 5.4.

Introduction

Ch. 1

4.

Methods other than the P-Vr method for the design of PSSs, namely the commonlyused Method of Residues [13] and the GEP Method [14], are described in Chapter 6.
By means of an example the merits, deficiencies and limitations of the latter Methods and the P-Vr approach are examined [15]. (See item 4 in Section 6.7.)

5.

Various concepts and methods for the tuning of automatic voltage regulators
(AVRs) are introduced and examined in Chapter 7. Some simplifications in the
approaches to the commonly-used techniques are suggested.

6.

A more fundamental and detailed examination is undertaken - than previously conducted - to explain, and understand more fully, not only the performance of certain
devices but also the theory behind certain tools. Examples are: (a) the performance
of the Integral of accelerating power PSS in Chapter 8; (b) the characteristics of two
tools, Mode Shapes and Participation Factors in Chapter 9 (these are used in the
analysis of the performance of multi-machine systems).

7.

The tuning of power oscillation dampers for FACTS devices (PODs, also referred to
as FACTS Device Stabilizers (FDSs)) is described in Chapter 11. Some of the problems encountered in the design are revealed in the case of a multi-machine system
for a wide range of operating conditions [16], [17].
Due to the short-comings of existing techniques for the tuning of FACTS Device Stabilizers, their robustness is more difficult to achieve compared the tuning of PSSs for
robustness (see item 3 above).

8.

The concept, theory, and calculation of Modal Induced Torque Coefficients

(MITCs) are described in Chapter 12. In this chapter the synchronizing and damping
torques induced on a generator by both PSSs and FDSs at the modal frequencies are
explained; this chapter provides the theoretical basis for Chapter 13 [16], [17].

9.

The interactions between, and effectiveness of, PSSs and FDSs in a multi-machine
power system are analysed in Chapter 13. A valuable aid in establishing the relative
effectiveness of stabilizers are the Stabilizer Damping Contribution Diagrams
(SDCDs) [18]. Extending the concepts introduced in Chapter 13 the SDCDs form a
basis for the heuristic coordination of power system stabilizers and FACTs device
stabilizers. Both the latter approach as well as an optimization approach based on
linear programming are illustrated in Chapter 14 and [20].

10.

A comprehensive set of the various small-signal models of synchronous generators

and FACTS devices are provided in Chapter 4. These models are intended to be
embedded in a set of differential and algebraic equations (DAEs) which are
employed to take advantage of sparsity in the system equations [21], [22].

11.

The practical theme throughout this book is based on consulting projects for industry and queries raised by industry on practical problems that they have encountered.
Many of the queries relate to some lack of understanding of the theoretical or practical backgrounds to the issues raised.

Sec. 1.3

Synchronizing and damping torques

An aim of the design of controllers is to enhance the damping of the rotor modes of oscillation, either to stabilize unstable oscillations or to ensure that the damping criteria for the
power system are satisfied. Therefore the concepts of synchronizing and damping torques
[23] - which operate on the shafts of generating units - are introduced in Section 1.3, followed by the concepts and definitions of stability in Section 1.4.

1.3 Synchronizing and damping torques

It is assumed that the reader is generally familiar with the basic concepts of power system
stability, such as the equal area criterion; these concepts are covered in many texts [2] - [7].
However, let us consider a somewhat simplistic scenario which may reveal not only the nature of dynamic interactions between a generator rotor and a power system following a disturbance but also the actions that occur within the rotating system of the generator itself.
Assume a synchronous generator is connected through its transformer by two parallel transmission lines to a receiving-end transformer and a large system. This scenario is summarized
in Figure 1.1(a).

End 1
x1

V1

End 2

V2

xL

x2
V

E
Generator

(a)

Infinite
bus

I
Two lines in

Electrical Power

0.6
0.5

Pe

0.4

e0

0.3
0.2

Pt

0.1

10

20

40
(deg)

Figure 1.1

One line in
0.6
Pp

0.4

Pm0=Pe0
0.2

Pt

0
0

0.8

60

(c)

0
0

90
Angle (deg)

(b)
180

(b)

(a) Single-machine infinite-bus power system. (b) Power-angle characteristics.

(c) Expanded view about 10 , Pe0.

Introduction

Ch. 1

The per unit voltage behind transient reactance of the generator is E, and the large system
is represented by an infinite bus, voltage V per unit. The reactance x1 comprises the transient
and transformer reactances of the generating unit; x2 represents the reactances of the receiving-end transformer plus the Thvenin equivalent of the large system. The effective reactance of the parallel lines is xL; their series resistance and shunt capacitance are negligible.
All reactances are in per-unit (pu) on the generator MVA rating. The direction of current
flow I is consistent with the generator producing power. The system is assumed to be lossless.
Because the power flow is from the generator to the large system, the rotor of the generator
leads that of the infinite bus by an angle (rad). The power output of the generator is
EV
P e = ---------------------------- sin
x1 + xL + x2

(per unit).

(1.1)

The associated power-angle curve is shown in Figure 1.1(b). The power output of the generator increases from zero at zero rotor angle and reaches a maximum value of
EV
P em = ---------------------------- at = 2 rad., or 90 (pu).
x1 + xL + x2

Let us assume that (i) when two lines are in service an equilibrium or steady-state condition
exists in which the power output of the prime-mover Pm0 is equal to the electrical power
output of the generator, Pe0 , at synchronous speed and the rotor angle is 0 ; (ii) the power
output of the prime-mover remains constant during a disturbance on the electrical system;
(iii) at time zero, one of the two lines is opened. Because the effective reactance of the lines
is now 2xL , it follows from (1.1) that immediately after the disturbance the electric power
output of the generator falls to Pt at 0 on the one-line-in power-angle characteristic shown
in Figure 1.1(c). The net torque acting on the shaft of the generator will cause it to accelerate
with respect to the system. The rotor angle of the generator, , immediately starts to increase from 0 on the latter characteristic thereby increasing the electrical power flow from
the generator. Once the electrical power output exceeds the prime-mover power output Pm0
at 10 the generator decelerates but, due to the inertia of the rotor, the rotor angle continues
to increase until the speed falls to synchronous. At this time the electric power output and
the rotor angle are at their peak values, Pp and p . However, the net decelerating torque continues acting on the shaft to reduce both the electrical power flow and the rotor angle along
the lower characteristic until zero net accelerating torque once more arises at 10 in
Figure 1.1(c). Due to inertia, the electric power output and rotor angle continue to decrease
and reach their minimum values at Pt and 0 at synchronous speed. Thereafter the process
repeats itself with the electric power output and rotor angle oscillating about Pm0, 10 between peak and trough values Pp, p and Pt, 0 , respectively.

Sec. 1.4

Definitions of power system stability

In the absence of damping, these oscillations will continue indefinitely. Synchronism in this
scenario is maintained by the electrical power flow, given by (1.1), between the generator
and the system, resulting in a synchronizing torque P e t being produced on the shaft of the
generator. For the oscillations to decay away a damping torque must also be established on
the rotor of the generator, typically by means of a PSS. The inherent damping torques acting
on the shaft of the generator are typically associated with eddy currents flowing in the rotor
iron and/or the damper (amortisseur) windings installed on the rotor, together with windage, friction and other losses.
Throughout the analysis which follows in the later chapters it will be shown that the production of sufficient positive synchronizing and damping torques on the shaft of a generator
is a continuing requirement for stable dynamic performance. Following a disturbance, the
perturbation in the electromagnetic torque of a synchronous machine, either a generator or
a motor, can be resolved into the two components defined as follows:
1.

synchronizing torque component: a torque in phase with rotor angle perturbations,

and

2.

damping torque component: a torque in phase with rotor speed perturbations.

Those familiar with the equal area criterion will recognize that for the scenario shown in
Figure 1.1(b) the power system is stable if positive damping torques are present. What are
implied by power system stability and some associated concepts are reviewed in the following section.

1.4 Definitions of power system stability

Power system stability is the ability of an electric power system, for
a given initial operating condition, to regain a state of operating equilibrium after being subjected to a physical disturbance, with most system
variables bounded so that practically the entire system remains intact.
[24].
This definition is intended to apply to an interconnected system in its entirety, however, it
also includes the instability and timely disconnection of an element such as a generator without the system itself becoming unstable.
There are three main categories of power system stability: rotor-angle stability, voltage stability and frequency stability. Ensuring stability in all its categories is the primary focus of
the design of power system controllers. Such controllers are designed to ensure, through the
specifications on the dynamic performance of the system, that adequate margins of stable
operation are attained over a range of normal operating conditions and contingencies.

Introduction

Ch. 1

In the following chapters we are concerned with rotor-angle stability which is defined as follows [24]:
Rotor angle stability refers to the ability of synchronous machines of
an interconnected power system to remain in synchronism after being
subjected to a disturbance. It depends on the ability to maintain/restore
equilibrium between electromagnetic torque and mechanical torque of
each synchronous machine in the system. Instability that may result occurs in the form of increasing angular swings of some generators leading
to their loss of synchronism with other generators.
Rotor-angle stability, and oscillations of the rotors of synchronous generators, are essentially
governed by the equations of motion of the unit; the relevant versions of the equations are
derived in Chapter 4. In terms of the per unit rotor speed , synchronous speed 0 and
the per unit prime-mover torque and the torque of electromagnetic origin, T m and T g , the
equations of motion are given by (4.58) and (4.59) which are repeated below:
p = b 0 and
1
p = ------- T m T g D 0 .
2H

.H is the inertia constant (MWs/MVA) of the generating unit and D (pu torque/pu speed)
is the damping torque coefficient. From the above equation it is apparent that a steady-state
condition exists when the torques are in balance and thus there is no change in rotor angle
or in speed about synchronous speed. However, a disturbance on an element of the electrical
system will result in an imbalance in the torques and cause the rotor to accelerate or decelerate, in turn causing the rotor angle to increase or decrease. The shaft equation is linear so
it is applicable to large and small disturbances. For large disturbances the term transient stability is defined as follows [24]:
Large-disturbance rotor angle stability or transient stability, as it
is commonly referred to, is concerned with the ability of the power system to maintain synchronism when subjected to a severe disturbance,
such as a short circuit on a transmission line.
On the other hand, rotor-angle stability for small disturbances is defined as [24]:
Small-disturbance (or small-signal) rotor-angle stability is concerned with the ability of the power system to maintain synchronism
under small disturbances. The disturbances are considered to be sufficiently small that linearization of system equations is permissible for
purposes of analysis.

Sec. 1.4

Definitions of power system stability

Associated with transient stability are severe events or disturbances such as system faults,
the opening of a faulted line - or a heavily loaded circuit, the tripping of a large generator,
the loss of a large load. As indicated in Section 1.1, for transient stability analysis the dynamic
behaviour of certain devices are modelled by their non-linear differential and algebraic equations. The presence of the various types of non-linearities in the equations results in transient stability analysis, in practice, being conducted by simulation studies in the time domain.
The basis of such analyses is a power flow study, an equilibrium or steady-state operating
condition to which the relevant disturbance is applied.
Stable, large-disturbance performance of a multi-machine power system depends on adequate synchronizing power flows being established between synchronous generators to prevent loss of synchronism of any generator on the system. High gain excitation systems are
employed to increase synchronizing power flows and torques. The decay of oscillations, not
only following the initial transient (usually the first swing) but also following cessation of
limiting action by controllers, is dependent on the development of damping torques of an
electro-magnetic origin acting on the generator rotors. Damping torques may be degraded
significantly by high gain excitation systems such that, if the net damping torque is negative,
instability occurs. To counter this type of instability, positive damping torques can be induced on generators by installing continuously-acting controllers known as stabilizers.
Small-disturbance or small-signal rotor-angle stability is associated with disturbances
such as the more-or-less continuous switching on and off of relatively small loads. The analysis of small-signal rotor-angle stability is conducted for a selected steady-state operating
condition about which the non-linear differential and algebraic equations and other non-linearities are linearized. This process produces a set of equations in a new set of variables, the
perturbed variables. Important features of small-signal analysis are: (i) as shown by Poincar,
information on the stability of the non-linear model at the selected operating condition,
based on the stability of the linearized system, is exact; and (ii) all the powerful tools and
techniques in linear control system analysis are available for the design and analysis of dynamic performance. The design of power system stabilizers for inducing damping torques
under normal and post-contingency conditions is conducted using such facilities.
Two forms of spontaneous small-signal instability may be: (i) a steady increase in rotor angle
due to inadequate synchronizing torque, or (ii) rotor oscillations of increasing amplitude due
to insufficient damping torque. Most generally in practice, however, the latter is of concern
in small-signal rotor-angle stability analysis.
While we are mainly interested in small-signal rotor-angle stability in the following chapters,
the definitions of voltage and frequency stability are quoted from [24] for information and
for the sake of completeness.
Voltage stability refers to the ability of a power system to maintain
steady voltages at all buses in the system after being subjected to a dis-

10

Introduction

Ch. 1

turbance from a given initial operating condition. It depends on the ability to maintain/restore equilibrium between load demand and load
supply from the power system. Instability that may result occurs in the
form of a progressive fall or rise of voltages of some buses.
(In system planning studies it is occasionally found that analyses which suggested rotor angle
instability are, in fact, associated with voltage instability; at times, it may be difficult to discriminate between them.)
Frequency stability refers to the ability of a power system to maintain
steady frequency following a severe system upset resulting in a significant imbalance between generation and load. It depends on the ability
to maintain/restore equilibrium between system generation and load,
with minimum unintentional loss of load. Instability that may result occurs in the form of sustained frequency swings leading to tripping of
generating units and/or loads.

1.5

Types of modes.

The term mode is used to refer to the natural or characteristic response to a disturbance of
the small-signal dynamics of the power system. Such modes may be oscillatory or monotonic and in the time domain are of the forms:
y1 t = A1 e

1 t

sin 1 t + 1

or y 2 t = A 2 e

2 t

, respectively.

A mode, and its typical frequency range, is usually identified with a phenomenon of one of
the following types.

Global Mode. This is a low-frequency mode of 0.05 - 0.2 Hz (approx. 0.3 - 1.2 rad/s) in
which all generating units move in unison. Such a phenomenon is observed, for example, in isolated systems connected to an AC system through a HVDC link [25].

Low-frequency mode. This is a localized oscillatory mode of frequency 0.01 - 0.05 Hz

Local-area mode. This is an oscillatory electro-mechanical mode and is usually associated with the rotors of synchronous generating units 1 in a station swinging against the
rest of the power system, or against electrically-close generating station(s). Its frequency range is normally in the range 6 - 12 rad/s (1 - 2 Hz).

(approx. 0.05 - 0.3 rad/s). For example, they have been associated with interactions
between the water column and the governors on hydro-generators [26], [27].

1. By a synchronous generating unit is implied the synchronous generator, its prime mover
and their controls.

Sec. 1.5

Types of modes

11

Intra-station or intra-plant mode. This oscillatory electro-mechanical mode is associated

with units within a generating station swinging against each other. The range of modal
frequencies is normally 10 - 15 rad/s (1.5 - 2.5 Hz).

Inter-area mode. This is an oscillatory electro-mechanical mode and is associated with

a group of generating stations in one area of the system swinging against a group of
stations in one or more other areas of the system. Inter-area modes are usually associated with (possibly) weak interconnecting ties between geographically separated areas
of the system. The range of modal frequencies is typically 1.5 - 6 rad/s (0.25 - 1.0 Hz).

Torsional modes. These modes are normally associated with oscillations between the

Control modes. These modes may be oscillatory or monotonic and may be identified
with the controls of generating units or FACTS devices.

rotating masses on the prime-mover - generator shaft. The frequencies of these

modes are normally greater than 50 rad/s (8 Hz) for nuclear units, and greater than 95
rad/s (15 Hz) for other small and large generating units.

A number of the above modes are referred to as the electro-mechanical modes. As stated
earlier the associated oscillations are the characteristic or natural modes of the system, the
frequency and damping of which generally change with the operating conditions, i.e. changes in the system configuration and the loading conditions. Such changes in operating condition may cause the system to drift - or be forced 1- towards a small-signal stability limit.
Instability of torsional modes caused by interactions with other controllers or other devices
has been observed [2].
In recent times it has been necessary to operate power systems closer to stability limits because of environmental and economic considerations. Furthermore, lightly damped interarea modes are becoming more common since interconnections between power systems are
also increasing. This is because interconnections allow adjacent systems (i) to share spinning
reserve, (ii) to reduce costs by better utilisation of the more efficient generating stations, and
(iii) to reduce the environmental impact by using the most efficient units, thereby facilitating
the postponement of investment in new generation. Methods are continually being sought
for increasing the power transfer over existing (possibly weak) interconnections thereby reducing the damping of the already lightly-damped modes. Consequently, ensuring that the
damping of modes of rotor oscillation in power systems provides adequate margins of stability has been - and still is - of concern to system planners and operators.
A most important feature of small-signal analysis is that it provides an understanding of the
underlying modal structure of a power system and gives insights into a system's dynamic
characteristics that cannot easily be derived from time-domain simulations. For example, in
the time-domain response following a major fault shown in Figure 10.32 only three of the
thirteen modes appear to be excited; the nature and location of the fault does not significantly excite the local-area modes outside the faulted area at all. Understanding the nature
1. In the aftermath of a major disturbance on the system.

12

Introduction

Ch. 1

of the modal behaviour as revealed from small-signal analysis yields a synoptic view of the
system characteristics which would require many large-signal studies of faults and other disturbances in different locations to gain similar, but not exact, information.

1.6

Synchronous generator and transmission system controls

In Figure 1.2 are shown the basic controllers and control signals for a prime mover, synchronous generator and static VAR compensator (SVC). These are the basic devices with which
we will be concerned in later chapters in the context of the design of controllers, however,
other Flexible Alternating Current Transmission System (FACTS) devices such as Thyristor
Controlled Series Capacitors (TCSCs), HVDC transmission links, and their controls will also
be considered [28], [29].
The control objectives are to ensure that system voltages and frequency lie within specified
ranges during normal and abnormal operating conditions. For system voltage control this is
achieved by adjusting the voltage references (or set points) to appropriate levels on generators and SVCs, by on-load tap-changing transformers, or by injecting or absorbing reactive
power by switching of capacitor banks or reactors. System frequency control may be implemented in several ways (e.g. through generator dispatch by adjustment of the speed or load
set-points on governors). For the purpose of small-signal stability analysis our focus is on
the dynamic behaviour of continuously-acting control systems. Discontinuous controls
such as transformer tap-changers, or switching operations of capacitor banks or reactors,
typically incorporate dead-bands, hysteresis, and time delays. For small-signal stability analysis at a given operating condition, the outputs of discontinuous controllers are usually assumed to be fixed at their initial steady-state values. This is legitimate because for smalldisturbances the changes in the discontinuous controller inputs are assumed to be negligibly
small and insufficient to trigger changes in the controller output.

Sec. 1.7

System and controls performance criteria

Voltage
reference
Speed
Speed
reference
Load
reference

Terminal
voltage

PSS

Governor
controls

Excitation
system
controls

Prime
mover

Generator

Generator i of n
synchronous
generators

13

Terminal voltage

P
O
W
E
R
S
Y
S
T
E
M

Stabilizing
signal
POD
Automatic
Voltage
Regulator

Voltage
reference

SVC

SVC j of c SVCs

Figure 1.2 Basic controls for a synchronous generating unit and a SVC. Other reactive
and voltage controls include other types of FACTs devices, reactors and capacitor banks.
When investigating the small-signal performance of a system, not only is its behaviour under
normal conditions of interest but also is its performance in the immediate post-fault condition before tap-changer and reactive switching operations have had time to occur and also
when all such operations have been completed following the disturbance. Any investigations
would include establishing if the margin of stability for such conditions is adequate.

1.7 Power system and controls performance criteria and measures.

1.7.1 Power system damping performance criteria
The criteria for small-signal damping performance of the power system are the more relevant in the analyses in the following chapters. These measures take different forms 1, namely:

The damping ratios of the dominant local or inter-area modes of rotor oscillation
should exceed a specified value, e.g. 0.03 or 3%.

The time constant of the dominant mode should less than a nominated value. (The
time constant of the mode is time taken for the mode to decay to 37% of its initial
value).

1. Values for these measures used by a number of organizations are given in a 1996 report
[30].

14

Introduction

Ch. 1

The settling time of the dominant mode should be shorter than a specified value. For
example, depending on the definition, the mode will decay to within 10%, 5% or 2%
of its initial value within 2.3, 3 or 4 times the time constant of the mode.

As an example, a slightly different form of the criterion used in South-Eastern Australia is,
as stated in detail in the National Electricity Rules [31]:
Damping of power system oscillations must be assessed for planning
purposes according to the design criteria which states that power system
damping is considered adequate if after the most critical credible contingency event, simulations calibrated against past performance indicate
that the halving time of the least damped electromechanical mode of oscillation is not more than five seconds.
To assess the damping of power system oscillations during operation,
or when analysing results of tests ..... , the Network Service Provider
must take into account statistical effects. Therefore, the power system
damping operational performance criterion is that at a given operating
point, real-time monitoring or available test results show that there is
less than a 10 percent probability that the halving time of the least
damped mode of oscillation will exceed ten seconds, and that the average halving time of the least damped mode of oscillation is not more
than five seconds.
The above criteria imply that (i) the damping constant of the mode should be less than
0.139 Np/s for a halving time of 5 s, and (ii) for an inter-area mode of 2 rad/s (0.32 Hz),
say, the damping ratio should be greater than 0.07.
1.7.2 Control system performance measures
The measures commonly quoted to characterize the performance of a control system are
(i) in the time domain: rise time, percentage overshoot, settling time, steady-state error; (ii) in
the s-domain: damping ratio, damping constant; (iii) in the frequency domain: phase and
gain margins, gain-crossover frequency, bandwidth. However, these measures are determined not only by the controller(s) but also by the device under control; such measures are
considered in more detail in Chapter 2.
For specific control systems such as synchronous generator excitation systems, governing
systems, etc., detailed performance criteria are the subject of various technical standards. For
example, in the case of excitation systems, two sources of relevant information are (i) the
IEEE Std. 421.2-1990 [32] which is a guide that presents dynamic performance criteria, definitions, and test objectives for excitation control systems as applied in power systems; (ii)
Clause S5.2.5.13 Voltage and reactive power control of the Australian National Electricity

Sec. 1.8

Validation of power system models

15

Rules [31] which specify dynamic performance criteria for excitation systems of generation
connected to the Australian grid.

1.8 Validation of power system models

Power system simulations for the purpose of detailed dynamic performance analysis are required for a variety of purposes including real-time operational control (e.g. dynamic security assessment), operational and long-term planning, controller design and tuning. If the
results of such dynamic simulations are to be trusted then the models of the generating
plants and their controls, the system loads and the interconnecting network, on which on
which the simulations are based must be accurate. In a number of jurisdictions, to achieve
the required level of modelling accuracy it is required that field tests be performed to establish the parameters of generating plants and their controls, and to verify that the model accurately represents the dynamic performance of the plant at its point of connection to the
network. For example, in Australia the Generating System Model Guidelines [33] - which
are enforceable under the National Electricity Rules [31] - stipulate requirements for steadystate, short-circuit and dynamic models of generating plant that are connected to the grid.
In 2007 the Western Electricity Coordinating Council (WECC) in the United States similarly
developed and maintained guidelines for generating system technical data, testing and model
validation [34]. An IEEE Task Force published guidelines for validation testing of generator
models for rotor-angle stability analysis [35]. Models for rotor-angle stability analysis are
typically required to be accurate in the frequency range from DC to at least 5 Hz (31.4 rad/
s). In some jurisdictions the minimum accuracy requirements are expressed quantitatively;
in others engineering judgment is applied on a case-by-case basis to assess the accuracy of
models.
In addition to validating the models of individual generating plant and control systems, system wide tests are typically undertaken from time to time to establish the validity of the integrated system models. Furthermore, the results of such tests may be used to calibrate the
system models. Such system tests may be staged with the intent of exciting key system dynamic responses, for example, by switching transmission circuits, energizing braking resistors. System disturbances may be unstaged with important system dynamic responses being
excited, say, by a system fault. Recently, on-line modal estimation schemes have also been
employed to assist in validating and/or calibrating small-signal models of power systems
[36].

1.9 Robust controllers

In what follows it is assumed - based on some design procedure or technique - that the configuration and parameter values of a robust controller are fixed, i.e. gain scheduling, conditional path switching or some such form of variation is not employed. For the purposes of
the analyses in later chapters a fixed-parameter controller is said to be robust if over the defined range of modal frequencies, and over a prudently-chosen, encompassing range of N
and N-1 operating conditions, the stabilizers induce adequate positive levels of damping and

16

Introduction

Ch. 1

synchronizing torques on the shafts of the generators such that the power system is stable
subject to the relevant stability margins. (See Section 1.2, item 3) (Note: as is shown in
Chapter 13, for some inter-area modes the stabilizer may degrade the damping torques on
other generators.)

1.10

How small is small in small-signal analysis?

For the purpose of small-signal analysis it is often convenient or necessary to express the
shaft equation for the synchronous machine in terms of the perturbations in the per-unit
mechanical and airgap torques T m and T g , acting on the shaft. The equation for the
small-signal motion of the shaft, evaluated at a steady-steady operating condition, is derived
in (4.59), and is expressed in a commonly-used per-unit form as:
2H

d
+ D =T m T g ,
dt

where is the per-unit perturbation in the angular speed of the shaft, H is the inertia constant of the prime mover, shaft and generator (MWs/MVA), and D is a damping coefficient
(pu torque/pu speed perturbation). It should be noted that for small perturbations in speed:
(i) P m = T m and P g = T g (per-unit); (ii) the perturbation in the electrical power output of the generator ( P e ) is related to the air-gap power ( P g ) by P e = P g 2r a I 0 I
where r a is the stator resistance (per-unit), I 0 and I are the steady-state and the perturbations in stator current (per-unit), respectively. Typically, r a is very small and consequently
P e = P g .
In using the linearized equations for the generator and other elements of the power system,
or when conducting field tests on devices, it is necessary to decide for what size of disturbance the system response can deemed to be linear. For example, for what peak-peak swing
in electrical power output is the generator response essentially small signal? For the particular application an analysis may be required to determine for what size perturbations the performance of the system can be considered small signal.
In the context of the small-signal analysis of the dynamic performance of power systems
the question often arises in practice how small is small. As mentioned earlier, we know
from Poincar that the non-linear model of the power system at a selected operating condition is stable if its linearized system at that operating condition is stable. However, information on the perturbations in the variables in the linearized system becomes exact only as their
magnitudes tend to zero. In practical applications such as staged tests, this implies that the
magnitude of any perturbations must be kept small, e.g. for step-changes in the reference
input signal for the testing of a closed-loop control system whose design is based on linear
control theory.
For illustrative purposes consider the classical power-angle characteristics, treated earlier, of
a simple transmission line, reactance X in per-unit. The non-linear power-angle function is

Sec. 1.10

How small is small in small-signal analysis?

17

P = V s V r X sin = P max sin ,

(1.2)

in which P is the power (per-unit) transmitted from the sending to receiving end, is the
difference in voltage angle (rad) between the sending- and receiving-end buses; V s and V r
are the respective voltage magnitudes (per-unit). For convenience we define
P max = V s V r X .
For small-disturbances the non-linear power-angle function is linearized about an initial
steady-state operating point P 0 , 0 employing, for example, the method used in
Section 3.3. The resulting small-signal characteristic is described by:
P = P max cos 0

per unit at P 0 , 0 .

(1.3)

Let P max = 1000 MW 1 pu . In Figure 1.3 two regions of linear operation on the non-linear characteristic are shown. For these regions the steady-state operating conditions are respectively P0 = 500 MW, 0 = 30 and P0 = 940 MW, 0 = 70 . The criterion employed
for a linear range is, say, that the maximum power deviation between the linear and nonlinear characteristics is less than 8.5 MW.

1100
Electrical Power (MW)

1000
90 MW

900
800

15 deg

700
600
500

300 MW

400
300
200

20 deg

100

0=30 deg

0
0

20

0=70 deg

40
60
Angle (deg)

80

100

Figure 1.3 Regions of the power-angle characteristic about 30 and 70 for which
peak to peak oscillatory swings in power and rotor angle can be considered linear.
Over these ranges the maximum deviations between the linear and non-linear power
characteristics are less than 8.5 MW (i.e. 0.0085 pu)

18

Introduction

Ch. 1

Based on our criterion, for a continuous oscillatory angular swing of 20 peak-peak about
the steady-state angle of 30 ; the power swing of 300 MW peak-peak is linearly related to
the angular swing. Similarly, at the steady-state angle of 70 the maximum power swing is
restricted to 90 MW peak-peak and is linearly related to a smaller angular swing of 15
peak-peak. Clearly, under more stressed conditions the range of perturbations over which
the system performance can be considered more-or-less linear is much smaller. Therefore,
depending on the application and the type of disturbance, engineering judgement - and analysis - is required to establish how small is small.

1.11

Units of Modal Frequency

Throughout this book the preferred unit of frequency of the mode j is in radian/
second (rad/s) and the damping constant is in Neper/second (Np/s). The damping ratio
is defined in Section 2.8.2.1 as:
= 2 + 2 if 0 0.3 , say.

Knowledge of the value of is important, in particular the system is stable if 0 and

unstable if 0 . Moreover, the value of not only provides a measure of the margin of
stability but yields the settling time of the modes, e.g. the time for the envelope of an oscillatory mode to decay to a value of 5% of the final value is three time constants, 3 .
For lightly damped modes it is useful to remember that if 0 0.3 , and if is in rad/
s, then .
In reports and papers of some organizations, it is common to characterize a mode by its values of and (in Hz) rather than (in Np/s) and (in rad/s). Therefore, in order to
estimate (mentally) the value of for a lightly damped stable mode it is necessary multiply
(in Hz) by 2 . For example, if = 0.03 and = 0.2 Hz, then = 1.26 rad/s and
= 0.038 Np/s; it may then be considered that the associated margin of stability of
0.038 Np/s is too low.
If the frequency is specified in rad/s an approximate value is sometimes given in Hz, and
vice-versa; for example = 6.3 rad/s or 1 Hz.

1.12

Advanced control methods

It should be emphasized that, in this book, techniques for the analysis and design of controllers employ the inherent characteristics of the components of the power system, for example the P-Vr characteristic of the generator for the tuning of its power system stabilizer.
These techniques are mainly based on the so-called classical control theory. Other techniques, sometimes called Advanced Control Methods [37], tend not to utilize the inherent
system characteristics and in the case of large, multi-machine systems advanced methods of

Sec. 1.13

References

19

control have had limited application in practice to date. Nevertheless, combining the ideas
in advanced control methods, and utilizing the inherent characteristics of the system and
its devices, may not only be a fruitful line of research but may also lead to practical outcomes.

1.13 References
[1]

Jules Henri Poincar, b. 1853, d. 1912. See web references to Poincar maps and stability.

[2]

P. Kundur, Power system stability and control. New York: McGraw-Hill, 1994.

[3]

J. J. Grainger, and W. Stevenson, Power system analysis, McGraw-Hill Education Europe, 1994.

[4]

A. R. Bergen and V. Vittal, Power system analysis. Second Edition, Prentice Hall,
1999.

[5]

H. Saadat, Power system analysis. Third Edition. PSA Publishing, 2010.

[6]

J. D. Glover, M. S. Sarmi, and T. Overbye, Power system analysis and design. PrenticeHall, 1998.

[7]

P. W. Sauer and M. A. Pai, Power System Dynamics and Stability, Prentice Hall, New
Jersey, 1998.

[8]

L. L. Grigsby (Ed.), Power System Stability and Control, Electric Power Engineering
Handbook, 3rd Edition, CRC Press, Taylor & Francis Group, Boca Raton, 2012.

[9]

G. F. Franklin, J. D. Powell and Abbas Emami-Naeini, Feedback Control of Dynamic

Systems, Prentice Hall, 6th Edition, Oct. 2009.

[10]

R. C. Dorf and R. H. Bishop, Modern Control Systems, Prentice Hall, 12th Edition,
July 2010.

[11]

M. J. Gibbard, Coordinated design of multimachine power system stabilisers based

on damping torque concepts, IEE Proceedings C Generation, Transmission and Distribution, vol. 135, pp. 276-284, 1988.

[12]

M. J. Gibbard, Robust design of fixed-parameter power system stabilizers over a

wide range of operating conditions, Power Systems, IEEE Transactions on, vol. 6, pp.
794-800, 1991.

[13]

F. L. Pagola, I. J. Perez-Arriaga, and G. C. Verghese, On sensitivities, residues and

participations: applications to oscillatory stability analysis and control, Power Systems, IEEE Transactions on, vol. 4, pp. 278-285, 1989.

[14]

E. V. Larsen and D. A. Swann, Applying power system stabilizers: Part I III, Power Apparatus and Systems, IEEE Transactions on, vol. PAS-100, pp. 30173046, June
1981.

20

Introduction

Ch. 1

[15]

M. J. Gibbard and D. J. Vowles, Reconciliation of methods of compensation for

PSSs in multimachine systems, Power Systems, IEEE Transactions on, vol. 19, pp.
463-472, 2004.

[16]

P. Pourbeik and M. J. Gibbard, Damping and synchronizing torques induced on

generators by FACTS stabilizers in multimachine power systems, Power Systems,
IEEE Transactions on, vol. 11, pp. 1920-1925, 1996.

[17]

P. Pourbeik, M. J. Gibbard, and D. J. Vowles, Proof of the Equivalence of Residues

and Induced Torque Coefficients for Use in the Calculation of Eigenvalue Shifts,
Power Systems, IEEE Transactions on, IEEE, vol. 22, pp. 58-60, 2002.

[18]

M. J. Gibbard, D. J. Vowles, and P. Pourbeik, Interactions between, and effectiveness of, power system stabilizers and FACTS device stabilizers in multimachine systems, Power Systems, IEEE Transactions on, vol. 15, pp. 748-755, 2000.

[19]

M. J. Gibbard and D. J. Vowles, Reconciliation of methods of compensation for

PSSs in multimachine systems, Power Systems, IEEE Transactions on, vol. 19, pp.
463-472, 2004.

[20]

P. Pourbeik and M. J. Gibbard, Simultaneous coordination of power system stabilizers and FACTS device stabilizers in a multimachine power system for enhancing
dynamic performance, Power Systems, IEEE Transactions on, vol. 13, pp. 473-479,
1998.

[21]

K. E. Brennan, S. L. Campbell and L. R. Petzold, Numerical Solution of Initial-Value

Problems in Differential-Algebraic Equations, Society for Industrial and Applied Mathematics, Philadelphia, 1996.

[22]

I. S. Duff, A. M. Erisman and J. K. Reid, Direct Methods for Sparse Matrices, Oxford
University Press, Oxford, 2003.

[23]

F. P. de Mello and C. Concordia, Concepts of Synchronous Machine Stability as Affected by Excitation Control, Power Apparatus and Systems, IEEE Transactions on,
vol. PAS-88, pp. 316-329, 1969.

[24]

IEEE/CIGRE Joint Task Force on Stability Terms and Definitions, Definition and
Classification of Power System Stability, Power Systems, IEEE Transactions on, vol.
19, no. 2, May 2004, pp. 1387 - 1401.

[25]

R. Grondin, I. Kamwa, G. Trudel, L. Gerin-Lajoie, and J. Taborda, Modeling and

closed-loop validation of a new PSS concept, the multi-band PSS, in Power Engineering Society General Meeting, 2003, IEEE, 2003, p. 1809 Vol. 3.

[26]

F. R. Schleif, G. E. Martin, and R. R. Angell, Damping of System Oscillations with

a Hydrogenerating Unit, Power Apparatus and Systems, IEEE Transactions on, vol.
PAS-86, pp. 438-442, 1967.

Sec. 1.13

References

21

[27]

H. V. Pico, J. D. McCalley, A. Angel, R. Leon, and N. J. Castrillon, Analysis of Very

Low Frequency Oscillations in Hydro-Dominant Power Systems Using Multi-Unit
Modeling, Power Systems, IEEE Transactions on, vol. 27, pp. 1906-1915, 2012.

[28]

Narian G. Hingorani and Laszlo Gyugyi, Understanding FACTS: concepts and technology of flexible AC transmission systems, IEEE Press, Piscataway, NJ, 2000.

[29]

Xiao-Ping Zhang, C. Rehtanz and B. Pal, Flexible AC Transmission Systems: Modelling

and Control, Springer-Verlag Berlin Heidelberg, 2012.

[30]

CIGRE Task Force 07.01.38, Analysis and Control of Power System Oscillations,
CIGRE Technical Brochure No. 111, Dec. 1996.

[31]

Australian Energy Market Commission, National Electricity Rules, Vers. 61, March
2014.

[32]

IEEE Guide for Identification, Testing, and Evaluation of the Dynamic Performance of
Excitation Control Systems, IEEE Std. 421.2-1990, 1990.

[33]

Australian Energy Market Operator, Generating System Model Guidelines, Planning

Department, Doc. No. 118-0009,Vers. No. 1.0, Feb. 2008.

[34]

WECC Guideline: Generating Facility Data, Testing and Model Validation Requirements, Western Electricity Coordinating Council, 13 July 2013. Available: http://
www.wecc.biz/library/WECC%20Documents/Documents%20for%20Generators/Generator%20Testing%20Program/WECC%20Gen%20Fac%20Testing%20and%20Model%20Validation%20Rqmts%20v%207-13-2012.pdf.

[35]

L. Hajagos, J. Barton, R. Berube, M. Coultes, et al., Guidelines for Generator Stability Model Validation Testing, in Power Engineering Society General Meeting, 2007.
IEEE, pp. 1-16.

[36]

Juan Sanchez-Gasca (Task Force Chairman), Identification of Electromechanical

Modes in Power Systems, Special Publication, TP462, Task Force on Identification of
Electromechanical Modes. Power System Stability Subcommittee of the Power System Dynamic Performance Committee of the IEEE PES, July 2012, 282 pages.

[37]

B. Pal and B. Chaudhuri, Robust Control in Power Systems, Springer Science+Business

Media, Inc., New York, 2005.

Chapter 2
Control systems techniques for
small-signal dynamic performance analysis

2.1 Introduction
2.1.1 Purpose and aims of the chapter
As emphasized in the Section 1.1 the equations describing an electric power system and its
components are inherently non-linear. The equations contain non-linearities such as the
product of voltage and current, functional non-linearities such as sine and cosine, and nonlinear characteristics such as magnetic saturation in machines. The analysis of dynamic systems with non-linearities is complex, particularly for power systems which are large and have
a variety of non-linear elements. On the other hand, in the case of linear control systems,
there is a comprehensive body of theory and a wide range of techniques and tools for assessing both the performance and stability of dynamic systems.
For small-signal analysis of power systems, the non-linear differential and algebraic equations are linearized about a selected steady-state operating condition. A set of linear equations in a new set of variables, the perturbed variables, result. For example, on linearization,
the non-linear equation y = f x 1 x 2 x n = f x becomes a linear equation in the perturbed variables, y = k 1 x 1 + k 2 x 2 + + k n x n , at the initial steady-state operating
condition Y 0 X 10 X 20 X n0 . The constant coefficients k i depend on the initial condition.
The question now is: how does the assessment of stability and dynamic performance based

23

24

Control systems techniques

Ch. 2

on the analysis of the linearized system relate to those aspects of the non-linear system? As
also mentioned earlier, a theorem by Poincar states that information on the stability of the
non-linear system, based on a stability analysis of the linearized equations, is exact at the
steady-state operating condition selected. However, information on the variable
x i = x i + X i0 becomes exact only as x i 0 . That is, for practical purposes, the perturbations must be small - typically a few percent of the steady-state value.
Small-signal analysis of power systems, based on the linearized dynamic equations, provides
a means not only of assessing the stability and the damping performance of the system
(through eigenanalysis and other techniques), but also for designing controllers and determining their effectiveness. The various applications of small-signal analysis in the field of
power systems dynamics and control are the subjects of later chapters. The purpose of this
chapter is to introduce and extend some of the concepts in linear control theory, analysis
and design which are particularly relevant to understanding of later material.

2.2

Mathematical model of a dynamic plant or system

Why model? Maybe a reason is that we wish to describe the behaviour of the plant when
subjected to some disturbance or to the action of a control signal. One means of characterizing its behaviour is by determining its time response to a test signal such as a step input.
However, in order to calculate the response we require a mathematical model of the plant.
Such a model can be derived from tests but often is most simply described by a set of differential equations which are derived from first principles. Let us consider two simple examples the results of which will be of interest in later cases. In these examples let p represent
the differential operator d dt , i.e. p = d dt . Note that we can manipulate expressions
in p as we would any algebraic variable.
Example 1.
A simple resistive-inductive circuit is shown in Figure 2.1. Write down the equations which
describe the behaviour of the current i t when an arbitrary voltage v s t is applied to the
circuit. The circuit resistance is r (ohm) and its inductance is L (Henry); the inductor is aircored.
i(t)

vs(t)

Figure 2.1 Simple resistive-inductive circuit

Because there is no iron in the magnetic circuit the voltage drop across the inductance is linearly related to the current through it. Thus, from Kirchoffs voltage law, the voltage-current

Sec. 2.2

Mathematical model of a dynamic plant or system

25

relationship at time t is
di t
v s t = i t r + L ----------- ,
dt

or

vs t
di t
r
----------- + i t --- = ----------.
dt
L
L

The latter equation can be expressed in a more convenient form involving the operator
p = d dt , i.e.
p + r L i = v s L .

(2.1)

Note that in (2.1) we recognise that the variables are instantaneous quantities in the time domain and therefore we have dropped the dependency on time, (t).
------------------Example 2
A load is driven by a d.c. motor at an angular speed of t (rad/s) as illustrated in
Figure 2.2. The speed of the motor and load is controlled by varying the DC supply voltage
v s t (volts), the field current being held constant. The back-emf developed by the motor is
v b t ; the torque of electromagnetic origin developed by the motor is T e t (Nm) and the
opposing load torque, T L t (Nm), is proportional to shaft speed. The combined polar moment of inertia of the rotors of the load and motor is J (kg-m2). The resistance and inductance of the armature winding are r (ohm) and L (Henry), respectively. The effect of armature
reaction on the field flux is negligible.
i(t)

r
TL

vs(t)

vb(t)

J
Te

load

Figure 2.2 DC motor and load

(i) Write down the equation that describes the behaviour of the load current i(t) when an
arbitrary voltage v s t is applied to the circuit.
Based on Kirchoffs voltage law, the voltage-current relationship at time t is:
di
v s = v b + ir + L ----- ,
dt
1L
i = ------------------- v s v b .
p+rL

or
(2.2)

26

Control systems techniques

Ch. 2

(ii) Derive an expression which describes the behaviour of the shaft speed when the motor
torque is varied.
Based on Newtons Second Law of Motion, the accelerating torque on the shaft is
d
T e T m = J ------- .
dt

(2.3)

Let the load torque-speed characteristic be defined by T m = k L . Following substitution of

the latter expression for T m in (2.3), we find:
1J
= --------------------- T e .
p + kL J

(2.4)

(iii) Derive the differential equation which describes the variation of motor speed with
changes in supply voltage.
Because the field flux is independent of the load current the electrical torque is proportional
to load current, T e = k i i . The back e.m.f. is proportional to speed, v b = k .
Substitute for T e in (2.4), with the result that:
1J
= --------------------- k i i .
p + kL J

(2.5)

1L
1J
= k i ------------------- --------------------- v s k .
p + r L p + k L J

(2.6)

Substitution of (2.2) in (2.5) yields

Speed , the dependent variable, can then be expressed in terms of the supply voltage, the
independent variable v s , in the following form:
kL r + k ki
ki
2
r kL
p + p --- + ----- + ------------------------- t = ------ v s t .
L J
LJ
LJ

(2.7)

------------------The plant equations such as (2.1) and (2.7) are simple; more complex cases will be considered later. The significance of these equations is that they represent simple forms of the following general form of the differential equation:
n

p + an 1 p

n1

+ + a 1 p + a 0 y t = b m p + b m 1 p

m1

+ + b 0 u t , m n . (2.8)

Notice all the terms in the output or dependent variable y t are collected on the left-hand
side of the equation, those in the input variable u t on the right-hand side. Importantly, the
resulting equation is a single, nth order differential equation in the dependent variable y t ,
i.e. the nature of the time-domain response depends on the form of the input variable u t .

Sec. 2.3

The Laplace Transform

27

For example, the input signal may be a step function or a sinusoidal function of time, both
of which are commonly-used test signals.
In the general case m n . The system of order n is said to be proper when m = n , or
strictly proper when m n .
Example 3
For the plants described in (2.1) and (2.7), n = 1, m = 0 and n = 2 and m = 0, respectively;
both systems are therefore strictly proper.
-------------------

2.3 The Laplace Transform

The theory and application of Laplace Transforms are covered in detail in the literature [1]
[2], hence only features of significance to the understanding of the material in this and the
following chapters are reviewed.
A valuable application of Laplace Transforms is the solution of linear differential equations
of the form of (2.8).
Example 4
Form the Laplace Transform of the following second-order differential equation which describes the dynamics of a plant:
p 2 y + a 1 py + a 0 y = b 1 pu + b 0 u .

(2.9)

Let F(s) be the Laplace Transform of a function f(t). The following results are derived from
a table of Laplace Transforms:
L af t = aF s ,
L pf t = sF s f 0 ,

a is a constant coefficient;
f(0) is the value of f(t) at time zero;

L p 2 f t = s 2 F s sf 0 pf 0 ,

pf(0) is the value of the derivative at

time zero.
A convention is adopted that the input u t commences at time t 0 + , hence at time zero
u t and all its derivatives are zero, i.e. u 0 = pu 0 = = 0 . Initial conditions on the
dependent variable are specified at t = 0 .
Substituting the expressions for the Laplace Transforms in (2.9), and accounting for initial
conditions on u t , we find

28

Control systems techniques

Ch. 2

s 2 + a 1 s + a 0 Y s s + a 1 y 0 py 0
= b 1 s + b 0 U s b 1 u 0

Thus the output response Y(s) can be expressed in terms of both the forcing function U(s)
and the initial conditions on the output as:
b1 s + b0
s + a 1 y 0 + py 0
Y s = ------------------------------- U s + --------------------------------------------------2
s + a1 s + a0
s 2 + a1 s + a0
forced response

and u 0 = 0.

(2.10)

natural response

Note that we can manipulate the Laplace operator s as any other algebraic variable.
There are two terms in the output response Y(s) in (2.10). The first term, called the forced
response, is determined by the nature of the forcing function, the input U(s). The second
term, called the natural response, is determined only by initial conditions on the dependent
variable; in this case the output signal and its derivatives at time zero. If all initial conditions
are zero, only the forced response is present in the output. Similarly, only the natural response exists in the output in the absence of an input signal (U(s) = 0).
If all initial conditions are zero, equations of the form of (2.10) can be written as:
b1 s + b0
Ys
----------= ------------------------------- = G s .
Us
s2 + a1 s + a0
G s is called the Transfer Function (TF) between the input U(s) and output Y(s).

------------------In summary, therefore, if it is assumed that:

the initial conditions on the dependent variable and all its derivatives are zero, i.e.
2

y t 0 = py t 0 = p y t 0 = = 0 , and

the input signal is applied at time t 0 + (so that u(t0) and all its derivatives are zero),

then the Laplace Transform of (2.8) and (2.9) can be simply formed from the differential equation
by replacing the differential operator p by the complex Laplace operator, s. The time-domain
variables y(t) and u(t) become the Laplace variables Y(s) and U(s), respectively. Applying
these results directly to (2.8), we find the general form of the differential equation describing
the plant becomes
n

s + an 1 s

n1

+ + a 1 s + a 0 Y s = b m s + b m 1 s

m1

+ + b 0 U s , m n . .(2.11)

Sec. 2.3

The Laplace Transform

29

The transfer function, which can be expressed a ratio of two polynomials in s, is thus
m

m1

bm s + bm 1 s
+ + b0
Ys
G s = ----------- = --------------------------------------------------------------------------- , for m n .
n
n1
Us
s + an 1 s
+ + a1 s + a0

(2.12)

Notice that a significant short-cut for writing the transfer function directly from the differential
equation is demonstrated.

Note that if m = n then the transfer function can be written in the form:
n1

b n 1 b n a n 1 s
+ + b 1 b n a 1 s + b 0 b n a 0
b n + ------------------------------------------------------------------------------------------------------------------------------------------- .
n
n1
+ + a1 s + a0
s + an 1 s

The above transfer function consists of two paths, one being the direct path between input
and output through the gain b n ; this path is simple to accommodate in any analysis - for
example in that of lead and lag transfer functions. Note that in the following analysis and
chapters it will be assumed that m n , i.e. all transfer functions are proper or strictly proper.
It will be found in the analyses that follow in this chapter that the transfer function (2.12) is
a more useful and practical form of the plant or system model than the form described by
the nth order differential equation, (2.8).
An example illustrating an application of the above results is discussed below.
Example 5
Find the time response y(t) of a plant described by (2.12) given an input signal u(t).
If the input function U(s) is known, the solution for the response Y(s) is
m

Y s = bm s + bm 1 s

m1

+ + b0 s + an 1 s

n1

+ + a 1 s + a 0 U s .

The Laplace Transforms for a range of input signals are given in tables of Laplace Transforms; for example:
step input:

u(t) = R0 for all t > 0, u(t) = 0 for t < 0:

sinusoidal input: u t = A sin t :

U s = R0 s ;
2

U s = A s + .

As an illustration, let us determine the response of the current through the inductor in
Example 1 to a step increase in the input voltage of V0 volts. Assuming zero initial conditions, replacing p by s in (2.1) and setting U(s)=V0/s, we find
Vo 1
1 L Vo
1
I s = ------------------ ------ = ------ --- ------------------ .
s+rL s
r s s + r L

30

Control systems techniques

Ch. 2

From a table of inverse Laplace Transforms [1] [2], the time responses for the terms 1 s
and 1 s + r L are found to be 1 and exp r L t , respectively. Hence the time response of the inductor current is:
i t = Vo r 1 e

r L t

.
------------------The concepts and results discussed above are of particular value in analyzing and designing
power system controllers. The so-called classical methods for the analysis and design of
linear control systems are based on the single-input, single-output transfer-function model
of a plant or system. The examination of the properties of the transfer function, and of the
information on system dynamic performance derived from those properties, is the subject
of the following sections.

2.4

The poles and zeros of a transfer function.

Considerable information on the stability and dynamic performance of a plant is derived

from the knowledge of the location of the poles and zeros of its transfer function. Importantly, such knowledge is obtained without having to solve the differential equation of the
form of (2.8) for the plant. Not only may a solution of the differential equation be time-consuming to conduct, but the information derived from the associated time response is not as
extensive as that extracted from an analysis of its associated transfer function (2.12).
The transfer function G s of (2.12) can be expressed in the following pole-zero form,
namely,
A s z 1 s z 2 s z m
P s
G s = ----------- = ------------------------------------------------------------------ , m n ,
Qs
s p 1 s p 2 s p n

when P s and Q s are factorized into factors with real roots or complex-conjugate pairs
of roots. The denominator polynomial Q s , when set equal to zero, is known as the characteristic equation:
Q s = an s n + an 1 s n 1 + + a0 = 0 .

(2.13)

The roots of the characteristic equation p 1 , ,p n are known as the zeros of Q s or the
poles of G s ; for example the poles lie at s = p 1 s = p n in the complex s-plane. Likewise, z 1 , ,z m are the zeros of both P s and G s and lie at s = z1 s = z m in the complex s-plane.

Sec. 2.5

The Partial Fraction Expansion and Residues

31

Note that the poles and zeros of G s may be real or complex. For example the poles of the
transfer function,
13 s + 5
G s = ----------------------------- ,
2
s + 4s + 13

are a complex-conjugate pair and lie at s 1 = 2 + j3, s 2 = 2 j3 in the complex s-plane;

the zero lies at s = 5 .
The significance of the poles and zeros of G s will be examined in more detail in the following sections.

2.5 The Partial Fraction Expansion and Residues

2.5.1 Calculation of Residues
If all the poles are distinct, i.e. there is not more than one pole at any location, the transfer
function given by (2.12) can be written as
K1
K2
Ki
Kn
P s
G s = ----------- = ------------- + ------------- + + ------------ + + ------------- , m n .
Q s s p1 s p2
s pi
s pn

(2.14)

If p 1 and p 2 , say, are a complex-conjugate pair, K 1 and K 2 are also a complex-conjugate

pair.
The coefficient K i is calculated by multiplying both sides of (2.14) by s p i (this isolates
th

K i in (2.14)), and then setting s equal to p i (this sets all other terms, except the i , to zero).
Thus, for example, to isolate and calculate K 1 we evaluate the expression,
A p 1 z 1 p 1 z 2 p 1 z m
P s
K 1 = s p 1 ----------= ---------------------------------------------------------------------------- , m n ,
Q s s = p1
p 1 p 2 p 1 p 3 p 1 p n

using the procedure illustrated in the following example.

Example 6
Evaluate the partial fraction expansion for the following transfer function.
K1
K2
A s + 1
G s = --------------------------------- = ---------------- + ---------------- , hence p 1 = 2 , p 2 = 3 .
s + 2s + 3
s + 2 s + 3

Multiply terms two and three by (s+2) to isolate K1,

K2 s + 2
As + 1
-------------------- = K 1 + ----------------------.
s + 3
s + 3

Then set s = 2 , thus

A
2 + 1 -
2 + 2
-----------------------= K 1 + K 2 --------------------- = K 1 ,
2 + 3
2 + 3

hence

K1 = A .

(2.15)

32

Control systems techniques

Ch. 2

Following the same procedure for the pole at s = 3 , we find K 2 = 2A.

------------------When the transfer function G s is expressed as a summation of first-order transfer functions, as in (2.14), the constant K i is also known as the residue, r i , of the pole at s = p i .
For a pair of complex poles, the residue K i and its complex conjugate exists.
The case when there is a zero in close proximity to a pole is of particular interest. Say, there
is a zero at s = z 1 , close to the pole at p 1 . The factor in p 1 z 1 in (2.15), will be small or
negligible and so too will be the residue r 1 . This useful result will be employed in different
contexts later. (To confirm this observation, try evaluating the residues when the zero at
s = 1 in the transfer function G s of Example 6 is replaced by one at s = 2.01.)
The special case of multiple poles at any location is not considered here; see [1] or [2].
2.5.2 A simple check on values of the residues
The following result assists in the calculation - or in checking the calculation - of residues.
Let us multiply out the right-hand side of (2.14), i.e.
r 1 s p 2 s p 3 s p n + r 2 s p 1 s p 3 s p n +
G s = ------------------------------------------------------------------------------------------------------------------------------------------------------------ ,
s p 1 s p 2 s p n

(2.16)

where ri = Ki. The numerator term of (2.16) is thus

r1 + r2 + + rn s

n1

+ lower order terms in s .

(2.17)

The order of the denominator is n. Hence, if the order of the numerator is:
* one less than the denominator, i.e. m = n 1 , the coefficient of s
n

n1

must be

ri = A ,

i=1

* two or more less than the denominator, the coefficient of s

n

n1

vanishes, i.e

ri = 0.

i=1

What can you say when m = n ?

2.6

Modes of Response

Let us examine the response of a plant to a step input. If U s is a unit step input, i.e.
U s = 1 s , then the response Y s is
1
Y s = G s --- .
s

(2.18)

Sec. 2.6

Modes of Response

33

This can be expressed in partial fraction form as:

K0
K1
K2
Kn
Y s = ------ + ------------- + ------------- + + ------------- .
s s p1 s p2
s pn

(2.19)

Using a table of inverse Laplace transforms yields the time response to the step input,
y t = K0 + K1 e

p1 t

+ K2 e

p2 t

+ + K n e

pn t

(2.20)

Note that

the response y t comprises a steady-state component K 0 and transient terms; K 0 is

the response after all the transients components have decayed away;

K i , i 0 , is the amplitude at time zero of the transient terms, K i e

if the input step-size were increased by a factor A, all the terms in (2.19) and (2.20) are
multiplied by the same factor;

the form of the transient response y t is determined by the n roots of the characteristic polynomial, Q s , or by the n poles of G s .

if p 1 and p 2 are a complex-conjugate pole pair, i.e. p 2 = p 1 , then the associated res-

pi t

idues are also a complex-conjugate pair, K 2 = K 1 .

For the present time, assume all roots of Q s have negative real parts. In general, there will
be real and complex roots of Q s of the form,
th

p i = i for the i real root,

and p k = k j k and p k + 1 = k + j k for the complex pair k, k + 1 .

Associated with each real root p i = i , there is a term in the partial fraction expansion
Ki s i .
Taking the inverse transform of the latter term, as in (2.19) - (2.20), results in a term in the
time-domain response K i e i t . If i is negative, this response is a monotonically, exponentially-decaying mode 1.

Likewise, for the complex pair, p k = k j k and, p k + 1 = k + j k there are terms in the
partial fraction expansion of the form
1. See a note on the term mode in Section 3.5.2.

34

Control systems techniques

Ch. 2

K k*
Kk
--------------------------------+ ---------------------------------- , where K k and K k* are complex conjugates.
s k j k s k + j k

Correspondingly, there will be a term in the time-domain response of the form

t

Ae k sin k t + k . If k is negative, this response is an oscillatory, exponentially-decaying,

sinusoidal mode. In power systems analysis the term mode usually refers to a broader set

of properties that characterize the physical behaviour of the natural system responses. Other
modal characterisations include whether it is an electro-mechanical mode, for example, or a
controller mode, etc.
The significance of the above analysis is that it reveals, by a simple examination of the poles
of the transfer function, the nature of the transient response in the time domain. The real
part of the pole, k , measured in Neper/s (Np/s), indicates how rapidly the modes in the
response decay away. The imaginary part, k , of a complex pole pair is the frequency in rad/
s of the damped sinusoidal oscillation. These results provide another valuable short-cut in
linear analysis: there is no need to solve the differential equations to determine the nature of the
transients in the response.
Let us illustrate some of the important concepts outlined above by means of examples. They
are intended to provide some useful insights into the dynamic behaviour of systems.
Example 7
In the following cases, find the time-response of the plant to a step input of magnitude A
units.
Case 1. The plant is described by the second-order differential equation
2

p y t + 5py t + 6y t = 6pu t + 6u t ,

where u(t) and y(t) are the input and output signals, respectively.
Assuming initial conditions are all zero, the plant transfer function is found by replacing the
differential operator p by the Laplace operator s, i.e.
Ys
6s + 1
6s + 1
----------- = G s = -------------------------- = --------------------------------- .
2
Us

s
+
2s + 3
s + 5s + 6

For a step input, U s = A s . The response can be found using a partial fraction expansion,
i.e.
6A s + 1
A
3A
4A
Y s = G s U s = ------------------------------------ = --- + ---------------- ---------------- .
ss + 2 s + 3
s s + 2 s + 3

Note that the sum of the residues in this case is zero since n m 2 . Using the inverse
Laplace transform tables, the response in the time domain is found to be
y t = A + 3Ae

2t

4Ae

3t

Sec. 2.6

Modes of Response

35
2t

3t

Note that the coefficients of the modes e and e , 3A and 4A respectively, are the initial
amplitudes of the transient response and are of comparable magnitude.
Case 2. Assume that the zero in the previous plant transfer function lies at s = 2 Np/s
2t

instead of at s = 1 Np/s. The coefficient of the mode e is then zero. (Note that the
pole at s = 2 Np/s still exists in the plant; this aspect is considered later.) This case shows
that pole-zero cancellation causes the amplitude of mode to become small or negligible.
Pole-zero cancellation, or close cancellation, is sometimes used in control system design.
However, it should be used with caution because the mode e
servable - or even unobservable - in the output.

2t

is then only partially ob-

Case 3. Let the plant transfer function be:

Ys
20
20
----------= G s = -------------------------------- = ------------------------------------ .
2
Us

s
+
2

s + 10
s + 12s + 20

The time response to the step input U s = A s is

y t = A 1.25 Ae

2t

+ 0.25Ae

10t

In this case the pole at s = 2 Np/s is much closer to the origin of the complex s-plane
10t

at a time equal to its time

than that at s = 10 Np/s. The response of the fast mode e
constant, i.e. t = T = 1 10 s, is about 37% of its initial amplitude (see Section 2.8.1). The
2t

response of the slow mode e at t = 1 10 s is about 82% of its initial amplitude. In this
case the contribution to the overall response of the fast mode rapidly diminishes with time.
Thus the response of the slow mode with a time constant of 1/2 s, dominates until it itself
decays after a time equal to four time constants (2 s).
Case 4. Furthermore, it can be shown that the initial amplitude of the response of the fast
mode becomes smaller as the pole at s = 10 is moved further into the left-half of the splane relative to the location of the pole of the slow mode. For example, for the transfer
function:
Ys
200
----------- = --------------------------------------- ,
Us
s + 2 s + 100

the step response is y t = A 1.0204 Ae

2t

+ 0.0204Ae

100t

Note that the initial amplitude 0.0204A and therefore the time response of the fast
mode is almost negligible in comparison to the slower mode.

36

Control systems techniques

Ch. 2

Case 5. Let the plant transfer function be:

Ys
20 s + 1
20 s + 1
----------- = G s = -------------------------------- = ------------------------------------ .
2
Us

s
+
2 s + 10
s + 12s + 20

The time response to the step input is y t = A + 1.25Ae

2t

2.25Ae

10t

Although the poles in this case are the same as those in Case 3, the initial amplitude of the
10t

fast mode e
is relatively much larger. Due to this mode the overall response of the plant
is much faster, although it settles in a time determined by the slow mode.
The significance of the results illustrated in Cases 3 and 5 is that the slower modes tend to
dominate the response. Case 5 reveals that the placement of a zero at an appropriate position
can speed up the response of a sluggish system. In this case the location of the zero at
s = 1 Np/s relatively close to the pole s = 2 Np/s diminishes the effect of the slow
mode. This concept is commonly used in classical control system design for speeding up the
response of a sluggish system.
Case 6. The plant transfer function of Case 1 is modified so that the damping ratio of its
second-order poles is less than one, say,
Ys
6s + 1
s + 1
6s + 1
----------= -------------------------- = 6 ---------------------------- = ------------------------------------------------------------- .
2
2
Us

s
+
1
+
j
5 s + 1 j 5
s + 2s + 6
s + 1 + 5

There are a pair of complex poles at s 1 2 = 1 j 5 . The response to a step of magnitude

A results in the following partial fraction expansion
1
1 + j 5
1 j 5
Y s = A --- ---------------------------------------- ---------------------------------------- .
s 2s + 1 + j 5 2s + 1 j 5

Replacing each term by its inverse Laplace transform, we find the closed-form expression
for the oscillatory response is:
1 + j 5 t

1 j 5 t

1 + j 5 e
1 j 5 e
y t = A 1 ------------------------------------------------- ------------------------------------------------2
2
= A + Ae

e j 5t e j 5t e j 5t + e j 5t
5 -------------------------------- ---------------------------------
j2
2

= A + Ae 5 sin 5t cos 5t
t

= A 1 + 6e sin 5t 24.1
2

The denominator of the transfer function, s + 2s + 6 , is of the form s + 2 n s + n . (The

following results are described in Section 2.8.2.1). The undamped natural frequency is

Sec. 2.7

Block diagram representations

37

n = 6 rad/s. The oscillatory mode thus has a damping ratio = 2 2 6 = 0.41; the
damping constant is 1 Np/s. If this were a rotor mode having a frequency of oscillation

of 5 rad/s, it would be considered well damped.

t

As has been pointed out earlier, the form of the mode e sin 5t can be derived by
inspection of the poles of the transfer function without having to solve for the time response
to a step input of the system described by the second-order differential equation,
2

p + 2p + 6 y t = 6 p + 1 u t .

----------------Let us emphasize some important results:

The response to an input of a linear system consists of a steady-state response and a

transient response.

The steady-state term bears a direct relationship to the input function (e.g. doubling
the amplitude of the input signal doubles the amplitude of the response).

The transient terms are determined by the initial magnitude of the input function (at time
t(0+). However the transient response has a form which is characteristic of the system,
and may by identified with the position of the poles of the transfer function; these poles
are. the zeros of the characteristic equation.

The concept not only of modes, poles and zeros, together with information provided
through the partial fraction expansion, provide important engineering short-cuts for
predicting the characteristic response of the plant in the time domain. By inspection of
the factorised denominator of the plant transfer function - thus revealing the pole
positions - we can ascertain: whether the plant responses will contain monotonic or
oscillatory (i.e. sinusoidal) components, how rapidly transients decay away, and the frequency of any oscillation.

2.7 The block diagram representation of transfer functions

Block diagrams are a very convenient way of communicating knowledge about the structure
of the plant and its control system. Typically, a block diagram contains a number of blocks
that represent the transfer functions of elements or components in the system. Moreover,
such blocks can be combined or eliminated in a series of operations that reduce or modify
the system to a form that is amenable to analysis. Both the basic transfer function blocks
and the operations on the blocks are outlined below.

38

Control systems techniques

Rs

(a)

Gs

R1 s

Cs

C s = G s R s

G1 s
Rs

G2 s

(c)

Figure 2.3

C2 s

Vs

G2 s

C2 s

C2 = G2 V = G2 G1 R1

(b)
C1 s

G1 s

Ch. 2

Cs

C 1 = G 1 R,
C2 = G2 R
C = C 1 C 2 = G 1 G 2 R

+-

(a) The basic transfer function block. (b) Combining blocks in series.
(c) Combining blocks in parallel.

Assume the input to a block such as that in Figure 2.3(a) represents the transfer function of
system elements that have high input impedances and negligible output impedances (i.e.
there is no loading by the elements). Figure 2.3(b) shows that the cascading of two or more
blocks can be represented by a single block whose transfer function is the product of the
individual transfer functions, G 2 s G 1 s . Likewise, in Figure 2.3(c) it is demonstrated that
blocks in parallel can be represented by the sum or difference of the individual transfer functions, G 1 s G 2 s . These building blocks form the basis for the analyzing, manipulating
or reducing block diagrams representing more complex systems.
A transfer function of particular interest is that representing a closed-loop control system
with negative feedback as shown in Figure 2.4. A purpose of the automatic control of stable
closed-loop systems is to minimise the error E s , i.e. the difference between the input and
feedback signals, so that the output signal C s aligns closely with the reference input R s .
In the following the transfer functions G s , H s , G s H s and W s are known as the
forward-loop, feedback-loop, open-loop and closed-loop transfer functions, respectively.
Rs

+
-

E s

Cs
Gs

Hs

Figure 2.4

The elements of the basic closed-loop control system, W s .

Sec. 2.8

Characteristics of first- and second-order systems

39

The relation between the error signal and the output and reference signals are:
C s = G s E s ,
E s = R s H s C s .
Eliminating the error signal and rearranging the terms, the closed-loop transfer function is
found, i.e.
W s = C s R s = G s 1 + G s H s .

(2.21)

This result is basic to the analysis of closed-loop control systems and is referred to frequently
in the following sections and chapters.

2.8 Characteristics of first- and second-order systems

In an earlier section we noticed that the poles of the system determined its modes - and thus
the form of the transient response. We will now examine how the pole locations determine
the characteristics of that response. It is important to note that, by understanding the nature
of the transient responses for simple first- and second-order systems, it is possible to predict
the characteristics of the dynamic behaviour of higher-order systems.
2.8.1 First-order system
The transfer function of the first-order system has the forms:
Ys
1
a
----------- = --------------- = ----------Us
1 + sT
s+a

where

1
a = --- ;
T

(2.22)

T is called the time constant of the system, U(s) and Y(s) are its input and output signals,
respectively.
a
A
1
1
The response to a step input, U s = A s , is Y s = ----------- --- = --- ----------- A,
s+a s
s s+a

or

y t = A 1 e at .

(2.23)

The time-domain response y t is shown if Figure 2.5. Note the following important properties of the first-order system:

At a time t equal to the time constant, t = T = 1 a , the term Ae at in the above

response is Ae at = Ae a a = Ae 1 = 0.368A , i.e. this term has decayed to 36.8% of
its initial value, Ae at t = 0 = A . The value of the response, however, is
y t = T = 0.632A , i.e. 63.2% of its final value.

After a time equal to four time constants, the response y t lies within 2% 1 of the
final value, A units, i.e. in effect, the transient response Ae at has completely decayed
away. This time is known as the 2% settling time, t s = 4T . Similarly, a 5% settling
time is often quoted for which t s = 3T .

1. Actual value at four time constants is 1.83%; for three time constants it is 4.98%.

40

Control systems techniques

Ch. 2

From the denominator of (2.22) we note there is a real pole at s = a Np/s. The associated
term in the transient response is K e at . Thus a real pole is associated with an exponentially
decaying mode in the response if a is negative.
A
y(t)

Us

0.5A

Figure 2.5
2.8.2

T=1/a

2T
time (s)

1 -------------1 + sT

Y(s)

3T

Time response of a first-order system to a step input of magnitude A units.

The second-order system

2.8.2.1

The characteristics of the second-order system

The typical form of the differential equation of a second-order system is:

p 2 + a 1 p + a 0 y t = b 0 u t .
Having taken the Laplace Transform and assuming zero initial conditions, we can express
the transfer function in the following two forms:
b0
n2
Ys
----------- = ------------------------------ = ---------------------------------------Us
s2 + a1 s + a0
s 2 + 2 n s + n2

if a 0 = b 0 .

(2.24)

The second form is the ideal or classical form of the second-order transfer function which
has complex poles. The parameters and n in (2.24), and associated quantities marked as
important (*), are defined below:

n is the undamped natural frequency (rad/s), n =

is the damping ratio, 0 1 ; = a 1 2 n .

a0 ;

The poles of the above transfer function are complex when 0 1 and are of the form
s 1 2 = j d where

Sec. 2.8

Characteristics of first- and second-order systems

41

d (rad/s) is the frequency of the damped oscillations in the transient response,

is the damping constant (Neper/s), and

the relation between radian frequency and frequency f in Hertz (Hz) is = 2f .

The characteristic equation for the system of (2.24) is

s 2 + 2 n s + n2 = 0 , with roots s 1 2 = n n 2 n2 .

Since the damping ratio lies in the range 0 1 the complex poles of the second-order
transfer function are:
s 1 2 = j d = n j n 1 2 .

(2.25)

Based on the above definitions the frequency of the damped oscillations and the damping constant are given by d = n 1 2 rad/s and = n Np/s, respectively. Solving for

from the latter two relations it is found that the damping ratio is:
= 2 + 2d d

if 0.3 d .

The time-domain response of the second-order transfer function (2.24) to a step input of
magnitude A units is:
n2
A
Y s = ---------------------------------------- --- .
2
2
s + 2 n s + n s

The associated time response consists of two terms, the steady-state term of value A and a
transient component (an exponentially decaying sinusoid); it is:
A
n t
y t = A ------------------ e
sin d t + , 0 1 ,
1 2

where cos = and sin =

(2.26)

1 2 .

It is often useful to refer to the characteristics of the step response of the ideal second-order
system with a complex pole pair. The damped oscillatory response is shown in Figure 2.6 in
which some meaningful measures that characterize the response are defined.

42

Control systems techniques

Ch. 2

2
1.8

peak overshoot

1.6

period = 2/

Response y(t)

1.4
envelope et
1.2

2% settling time

1
t

0.8

0.6
0.4
0.2
0
0

10

15

20

25

Time (s)

Figure 2.6

Characteristics of the response to a unit step input of the ideal second-order

system with a complex pole-pair.

The frequency in Hertz of the damped oscillation is:

The period of the oscillation is:

f d = d 2 Hz.

t f = 1 f d = 2 d s.

The time to the first peak is 1/2 of the period:

t p = d = n 1 2 s.

For a unit step input the peak overshoot occurs at t = t p and is:
M0 = y tp 1 = e

t p

2
= e 1 .

As already stated, the settling time t s is the time for envelope to decay to a value of 2% (actually 1.8%) of the final value of y t and is equal to four time constants:
i.e. t s = 4 = 4 n s.1
The useful reference family of normalised-time responses for a step input is shown in
Figure 2.7 for values of between 0.1 and 1. The figure can be interpreted as follows. If,
say, n = 1 rad/s and = 0.1 then the first peak in the transient occurs at 3.2 s, however,
if n = 2 the first peak is reached at 1.6 s, etc.

1. A 5% settling time is equal to three time constants

Sec. 2.8

Characteristics of first- and second-order systems

43

2
1.8

Response y(t)

1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0

10

15

Normalised Time
= 0.10
= 0.25
= 0.50

0.15
0.30
0.71

0.20
0.40
1.0

Figure 2.7 Response of the second-order transfer function to a unit step input as a function of normalised time, n t , for damping ratios between 0.1 and 1.
2.8.2.2

Implications for the dynamic performance of power systems

Certain electro-mechanical modes of oscillation, associated with the rotors of generators, are
typically complex and lightly damped and of the form j d . The important features of
these modes to which frequent reference will be made are listed in Section 2.8.2.1.
In the analysis of power system dynamic performance the damping ratio, is used in several contexts, e.g. a criterion for the dynamic performance of the system is that the damping
ratio for all rotor modes should be better than, say, 0.05 (or 5%). For the second-order transfer function given in (2.24), the complex poles j d vary as shown in Figure 2.8 for 0 .
A line of constant damping ratio makes an angle with the negative real axis such that
= cos . Note that the loci of the poles in the complex s-plane is along a semi-circle of
radius n . The nature of the time-domain response for a mode located on the semi-circle
can be ascertained directly by inspection.

44

Control systems techniques

Ch. 2

line of constant
damping ratio

= 0
n
s1

d = n 1

= 1

cos =

1
= n

s2
= 0

Figure 2.8

2.9

Trajectory of the complex poles of the second-order transfer function, (2.24),

for the damping ratio 0 .

The stability of linear systems

A plant or a system is described by a mathematical model such as a nth order differential

equation or transfer function. If the system is passive there is no internal source of energy.
A bounded input signal is the only available source of the energy to provide the energy necessary in the output signal; moreover, some of the input energy may be dissipated internally
within the system. Such a system is inherently stable since the output must also be bounded.
A system is defined as stable if a bounded input always produces a bounded output.
In an active system there is an internal energy source from which the energy contained in
the output signal is derived. In this case the possibility of a bounded input producing an unbounded output exists. Such a system may therefore be unstable.
We saw earlier, for linear systems, that the response consists of a steady-state and a transient
component. If the system is to be stable all transient terms must decay to zero. For this to
occur, we noted in the examination of the modes for first and second order system that the
real part of the mode must be negative. Thus for a system to be stable, the poles of the
transfer function - or the roots of the characteristic equation (2.13) - must all have negative
real parts. An alternative way of stating the same result is: for stability, all the poles of the
transfer function must lie in the left-half of the complex s-plane.
If, in the characteristic equation, the real part of a complex pair of poles, s k = k j k and
s k + 1 = k + j k is positive ( k 0 ), the pair lies in the right-half of the s-plane. The associated instability is manifested in a term in the time response which is an exponentially in-

Sec. 2.10

Steady-state alignment and following errors

45

creasing sinusoid. A similar result applies to a first order pole, i.e. an exponentially increasing
response.

If any poles lie on the imaginary axis, all others being in the left-half of the s-plane, the system is said to be marginally stable. However, in practical linear control systems it is not possible to locate and maintain poles exactly on the imaginary and therefore marginal stability
is of academic interest only.

2.10

Steady-state alignment and following errors

One reason for the use of closed-loop control systems is to automatically control the output
of a system to align with a reference input or a set-point as closely as possible. If the reference is fixed, i.e. set to a constant value, the difference between the set-point and the controlled output in the steady-state is called the alignment error. However, the reference input
may be time varying; in this case it is necessary for the controlled output to track or follow
the reference as closely as possible. In order to assess how well a closed-loop control system
aligns with - or follows - a reference input a set of test reference signals is devised that provides a measure of the quality of system performance. These tests signals are analysed in the
following sections.
The steady-state value of a time-varying signal x t , i.e. its value after all oscillations associated with any transients have died away, x ss , can be derived from the final-value theorem
(FVT), i.e.
x ss = lim

x t = lim

s0

sX s .

(2.27)

From this result the steady-state value of the output of a system, the alignment and following
errors for a given test reference-input can be determined.
The closed-loop control system under study is shown in Figure 2.9, where C s is the controlled output signal, R s is the reference input signal and the error signal is
E s = R s H s C s . With forward-loop and feedback-path transfer functions G s
and H s , respectively, the transfer function of the closed-loop system of Figure 2.9 has
been shown in (2.21) to be
Cs
Gs
W s = ----------- = --------------------------------- .
Rs
1 + G s H s

46

Control systems techniques

Rs

E s

Ch. 2

Cs
Gs

Hs

Figure 2.9 Block diagram of the classical closed-loop system

2.10.1 Steady-state alignment error.
The response of the closed-loop system to a step change in the reference input is shown in
Figure 2.10. Under steady-state conditions following the change in the reference input r t ,
the steady-state output c ss may not be equal to the constant value of the reference, r ss .
Based on the final value theorem the steady-state alignment error e ass is defined as:
R0
R0
e ass = r ss c ss = lim s ------ W s ------ , where C s = W s R s .
s
s
s0
ct

rt

(2.28)

r ss

R0
e ass
c ss = lim c t
t

Figure 2.10

The steady-state alignment error e ass following a step input.

Thus from (2.28) a general result for the alignment error follows, i.e.
e ass = lim

s0

R 0 1 W s .

(2.29)

In the following we will derive the alignment errors for a unity-feedback closed-loop system,
i.e. when H s = 1. In the case of a closed-loop system in which the feedback back is not
unity gain, the alignment errors can be derived by using the general result (2.29) or by direct
application of the final value theorem. Let us consider two special cases for the unity-feedback system, namely without and with integration in the forward path.

Sec. 2.10

Steady-state alignment and following errors

47

2.10.1.1

(a) No integration in the forward path, H(s)=1.

K 1 + sT b1 1 + sT b2
Let us assume G s takes the form G s = ------------------------------------------------------------------------------.
2
1 + sT a1 1 + C a1 s + C a2 s

In the steady-state following a transient, G s K as s 0 . Note the form of the factors

1 + sT - not s + a - in the transfer function is important; with no integration in the forward path the gain K is often called the static gain. According to (2.29) the associated alignment error is
Gs
e ass = lim R 0 1 W s = lim R 0 1 ---------------------------- ,
1
+
Gs 1
s0
s0
1
i.e. e ass = lim R 0 --------------------- ,

(2.30)

1
e ass = R 0 ------------- ,
1+K

(2.31)

1 + Gs

s0

since G s K as s 0 .
The result in (2.31) provides a very useful insight, namely, the higher the static gain K the
smaller is the alignment error. However, there is a downside to high gain settings without
suitable compensation, i.e. the dynamic performance of the closed-loop system may become
more oscillatory and even unstable. The latter effect will revealed through the analysis of the
stability and performance of the closed-loop system using the Bode plot in Section 2.12.2.
2.10.1.2

(b) Single integration in the forward path, H(s)=1.

1 + sT 1 + sT

K
b1
b2
In this case G s takes the form G s = ---- ------------------------------------------------------------------------------ . Let s 0 ,
2
s 1 + sT 1 + sC + C s
a1
a1
a2

then G s K s . Substitute this limit for G s in (2.30) above; the steady-state error is thus
1
s
e ass = lim R 0 ------------- = lim R 0 ------------ = 0 .
K
s+K
s0
s

0
1 + ---s

This is a very useful result; it shows that a single integration in the forward path integrates
out to zero any error e t that develops between the reference input and the controlled
output. If there is no integration in the forward path the introduction of proportional plus
integral (PI) compensation [1], [2] into that path ensures zero alignment error in the steady
state. This is highly desirable in some types of closed-loop control systems, however, like the
case above, there is a disadvantage. For example, introduction of pure integration 1 s also
introduces a phase-lag of 90 in the open-loop transfer function. In turn, the Phase Margin
is reduced, and consequently the closed-loop system may become unstable. This is discussed
in Section 2.12.2.

48

Control systems techniques

Ch. 2

2.10.2 The steady-state following error

A test reference input of a suitable form for defining the following error is the ramp signal,
r t = R 0 t . Based on the final-value theorem the steady-state following error e fss is given
by:
R0
R0
e fss = lim r t c t = lim s ------ W s ------ , or
t
s0
s2
s2
R0
e fss = lim ------ 1 W s .
s0 s

(2.32)

This is a general result for the following error. However, let us again consider a unity-feedback system and two special cases, without and with integration in the forward path.
2.10.2.1

(a) No integration in the forward path, H(s)=1.

K 1 + sT 1 + sT

b1
b2
Assume G s again takes the form G s = ------------------------------------------------------------------------------. In the steady2

1 + sT a1 1 + sC a1 + C a2 s

state following the decay of the transient we find G s K as s 0 . By substitution of

this limit in (2.32), the following error becomes
R0
R0
Gs
1
e fss = lim ------ 1 ---------------------------- = lim ------ --------------------- ,
1
+
G

1
s
s
1
+
Gs
s0
s0

(2.33)

R0
e fss = lim --------------- , i.e. e fss .
s 0 s + sK

(2.34)

The result reveals that the following error becomes increasingly large with time; the latter is
illustrated in Fig. 2.11. The controlled output thus cannot follow a reference input that
changes linearly with time. Such a closed-loop system cannot track, for example, a satellite
passing overhead.
2.10.2.2

(b) Single integration in the forward path, H(s)=1.

K
As shown in Section 2.10.1.2 above, G s ---- as s 0 . Substitution for G s in (2.33)
s

above yields a following error given by:

R0 1
R0
R
------ ------------- = lim ----------- = -----0- .
K
K
s0 s
s0 s+K
1 + ---s

e fss = lim

(2.35)

Thus after any oscillations have died away following the application of the ramp, there is a
constant difference, or error, between the ramp input and the controlled output given by
(2.35), as illustrated in Figure 2.11.

Sec. 2.11

Frequency response methods

49

rt
ct

e fss
ct

rt

(one integration)

ct
(no integration)

time (s)

Figure 2.11

The following error e fss resulting from a ramp input.

Exercise. Show that if the forward-loop transfer function G s in a unity feedback system
contains two integrations, the following error for a ramp reference input is zero, but is finite
2

for a parabolic input r t = R 0 t 2 .

Example 8
A unity feedback system has a forward-loop transfer function:
G s = K s + 2 s s + 5 s + 10 .

Find the value of the gain K such that the steady-state following error e fss 0.01 units for
a ramp input of 2 unit/second.
Change the factors in G s into the form 1 + sT , i.e
1 + 0.5s
G s = K --------------------------------------------------- ,
s 1 + 0.2s 1 + 0.1s

2K
where the static gain is K = -------- .
50

R
2
In order to satisfy the specification, the error e fss = -----0- = ---- < 0.01 , i.e. K 200 , or
K

5000 .

-----------------

2.11

Frequency response methods

The purpose of this section is to outline briefly frequency response methods of analysis and
to provide insight and understanding of those features of the analysis that are relevant to
power system dynamics and control.
In frequency response analysis the injected frequency will be represented by f (rad/s) rather than by - which will refer to rotor speed in later chapters.

50

Control systems techniques

Ch. 2

Frequency response methods assume that a sinusoidal signal of constant amplitude is applied to system. That is, when a signal r t = A sin f t of frequency f (rad/s) is injected
into the system G s , a steady-state sinusoidal signal y t of the same frequency appears at
the output after the transient terms have died away. In the Laplace domain the response of
the system is
Y s = G s U s ,
2

where, for a sinusoidal signal, U s = A f s + f . The response is thus

f
Y s = G s A ------------------- .
2
2
s + f

Forming the partial fraction expansion as described in Section 2.5, we find

AG j f
AG j f
terms of the form
,
Y s = -------------------------- + ---------------------------------- +
2j s j f 2 j s + j f
Ki s pi

where p i are the poles of G s and K i are their residues.

Taking the inverse Laplace transforms, the time-domain response is
AG j f j f t AG j f j f t
+ ------------------------ e
+ transient terms which .
y t = --------------------- e
2j
2 j
tend to zero as t

Let G j f = G j f e

j f

and G j f = G j e

j f

(2.36)

, where the phase angle of

G j f is a function of frequency. Following substitution for these terms in (2.36), the time-

domain response tends to

j f t + f

j f t + f

e
e
y t = A G j f ---------------------------------------------------------------------- = A G j f sin f t + f ,
2j

after all transient terms have decayed away.

Hence y t = A G j f sin f t + j f is the steady-state response. The steady-state output y t is a sinusoid of amplitude A G j f with phase shift j f with respect to the
input signal. The frequency response of G s is a plot of G j f and j f as f (rad/s)
is varied over a range of the frequencies. Typically the frequency response is plotted in two
forms, the polar plot and the Bode plot; we shall employ the Bode-type responses in later
chapters.

Sec. 2.12

Frequency response diagrams and Bode Plots

51

It is interesting to note that the analysis so far has allowed us to take several short-cuts. Firstly, G s was formed by replacing, for zero initial conditions, the differential operator p by s
in the differential equations for the system. Secondly, the frequency response is obtained
simply by replacing s by j f in G s .

2.12

The frequency response diagram and the Bode Plot

The following treatment of the Bode Plot contains two main features, firstly, the graphical
plotting of the frequency response and, secondly, its application to the determination of the
stability of closed-loop control systems. Although the frequency response plot of a given
transfer function is readily obtained using the appropriate software, the ability to visualise
the frequency response plot is a very useful skill, or conversely, to deduce the form of a
transfer function when its frequency response plot is presented. By understanding the basis
of the frequency response plot it becomes easier to carry out the required visualisation or
interpretation expeditiously.
For the analysis of stability of closed-loop control systems the frequency response plot of
interest is that of the open-loop transfer function, G j f H j f , where G s and H s are
the forward-loop and feedback-path transfer functions, respectively, shown in Figure 2.9.
The associated log-magnitude and phase responses of G j f H j f are known as the Bode
Plot from which information on stability can be deduced - subject to certain conditions.
Consequently, the following is a brief description of the basis of techniques not only for
drawing the Bode Plot, but also for interpreting the Plot to assess both the margin of stability of the closed-loop control system and its dynamic performance.
The system is excited by a sinusoidal signal r t of unity amplitude and frequency f (rad/
s). The two plots of the Bode diagram are both plotted as a logarithmic function of frequency, log f . The first plot is of the log (on base 10) of the magnitude (LM) of the open-loop
transfer function, i.e. 20 log G j f H j f in dB; the second plot is of the argument (Arg)
- or phase - j f = G j f H j f of the open-loop transfer function in degrees.
It should be noted that for any fairly simple transfer function the frequency response can be
drawn manually using the graphical technique described in the following section.
2.12.1 Plotting the frequency response of the open-loop transfer function
Any transfer function can be divided into a number of basic first- or second-order factors
which form the numerators or denominators of the element. The magnitude and phase response of the basic factors are simple to derive and recall. By combining the responses of
the factors, the overall frequency response of the transfer function is generated.

52

Control systems techniques

Ch. 2

Any transfer function (with s = j f ) contains factors, F j f , of the following types in its
numerator or denominator.
F j f =

Fs =

(i)

from

K , a scalar gain

(ii)

j f n

from

s n

(iii)

1 + j f T m

from

1 + sT m

(iv)

2
p

------f + j 2 ------f
1

n
n

from

s
s 2-
---------1
2
+
+

n 2

Note the forms of the type (iii) and (iv) factors, i.e. the polynomial form 1 + sa 1 + , rather
than pole-zero form s + a .

Example 9
Find log-magnitude and phase of the open-loop transfer function:
K
G j H j = --------------------------------- .
j f 1 + j f T

The log-magnitude and phase responses are:

LM = 20 log GH = 20 log K log j f log 1 + j f T ;
Arg GH = Arg K Arg j f Arg 1 + j f T .

------------------This example demonstrates that multiplication or division of the four types of factors become addition or subtraction of their log-magnitudes and of their phase contributions. This
simplifies analysis because it involves simple addition or subtraction of the component
terms; such operations are the basis for plotting manually the frequency response of transfer
functions.
Let us consider the log-magnitude and phase plots of each of the factors F j f as a function of log f .
2.12.1.1

(i) A factor in the transfer function is a constant gain K.

The transfer function is F j f = K , and the associated log-magnitude and phase are:
LM = 20 log K and j f = 0 if K 0 , (or j f = 180 if K 0 ) respectively. The

magnitude response is shown in Figure 2.12.

Sec. 2.12

Frequency response diagrams and Bode Plots

53

20 log K

LM (dB)

0.1

Figure 2.12

(Horizontal line)

10

100

Frequency rad/s

Transfer function of a constant gain K

Increasing or decreasing the gain K results in the plot of 20 log K moving vertically up or
down.
2.12.1.2 (ii) A factor in the transfer function contains pure differentiation or integration of
multiplicity n
The transfer function is of the form: F j f = j f n , n = 1 2 . Differentiation is as-

sociated with +n and integration with -n. The log magnitude of the transfer function is:
LM = 20 log j f n = 20n log f dB.

This asymptote, when plotted against

(2.37)

log f , is a straight-line having a slope

20n db/decade . According to (2.37) it intersects the frequency axis when LM = 0 dB at

log f = 0 , i.e. at f = 1 rad/s. The phase response is Arg j f n = n90 , i.e. it is a

constant for a given n over the entire frequency range. The log-magnitude and phase responses are plotted in Figure 2.13.
+20

integration

+n = +1

LM

+90
Arg
(deg)

(dB)
0.1

10

-90
-20

differentiation

0.1

10 f

-n = -1

integration

-n = -1

(b)

(a)

Figure 2.13 (a) Magnitude and (b) phase responses of j

2.12.1.3

+n = +1

differentiation

when n=1

(iii) A factor in the transfer function contains a real pole / zero of multiplicity n:

The transfer function is of the form: F j f = 1 + j f T n , n = 1 2 . Its log magnitude is:

54

Control systems techniques

LM = 20 log F = 20n log 1 + 2 T 2 1 / 2 dB.

Ch. 2
(2.38)

Consider the asymptotes for (a) a low frequency case when f T 1 , (b) a high-frequency
case when f T 1 .
(a)

f T 1 ,

LM = 20n log 1 = 0 dB. This low-frequency asymptote is a horizontal

line when plotted against log f .

(b)

f T 1 ,

LM = 20n log f T = 20n log f + log T dB.

This log-magnitude asymptote, when plotted against log f , is a straight-line. The slope of
the line is 20n dB for each decade of frequency (i.e. for each log 10 = 1 unit). The lowfrequency and high-frequency straight-line asymptotes of the log-magnitude plots are shown
in Figure 2.14 for zeros (+n) or poles (-n). The actual plot is also drawn for a transfer funcn
tion 1 1 + j f T over a frequency range about the so-called corner frequency c where
the high- and low-frequency asymptotes intersect; c = 1 T rad/s. The differences between the actual plot and the straight-line asymptotes are easily remembered. In the case of
multiple poles the actual log-magnitude plot is 3n dB down at the corner frequency and n
dB down at an octave above and below the corner.
Zeros:
slope 20n dB/dec
0

LM
(dB)

c 1 T

c 2

log

2 c

-n dB

Poles:
slope -20n dB/dec

-3n dB
actual plot

-n dB

Figure 2.14 Magnitude response for a real pole or zero of order n.

n

Consideration of (2.38) reveals that for multiple zeros 1 + j f T the actual log-magnitude
plot and the straight-line asymptotes are the mirror image about the frequency axis of those
for multiple poles as shown in Figure 2.14.
Consider now the plot of the phase shift j f for the transfer function 1 + j f T n .

Sec. 2.12
(a) for

Frequency response diagrams and Bode Plots

f T 1 :

55

= n atan f T 0 ,

hence 0 ;

= n atan f T ,

hence n90 ;

(b) for

f T 1 :

(c) for

f T = 1 (corner):

= n atan 1,

hence = n45 .

The phase response for real poles of order n is shown in Figure 2.15. A straight-line approximation of the phase response is employed from a decade below the corner, phase 0 , to a
decade above the corner at n 90 . The straight-line approximation to the response, which
passes through the corner frequency at n 45 , differs at most from the actual by
5 to 6 (for n = 1 over a decade in frequency on either side of the corner. The phase response for multiple zeros is the mirror image about the frequency axis of those for the multiple poles shown in Figure 2.15.
Phase
n 5.7 0 0.1 c

c 2

c = 1 T
2 c

10 c log

n 5

-n45
n 5

n 5.7

-n90
Straight line
approximation

Figure 2.15

actual plot

Phase response for real poles of order n.

2.12.1.4 (iv) A factor in the transfer function contains a complex pair of poles or zeros of multiplicity n:

The transfer function is of the form: F j = 1 ------f + j 2 ------f .

n
n

The frequency responses of this factor for a single pair of complex poles ( n = 1 ) are
shown in Figure 2.16 for damping ratios 0.1 1 . The responses are given for a normalised frequency f n , where n is the undamped natural frequency. The straight-line approximations which can be employed are crude and therefore the more accurate plots
shown in the figure are used as templates when sketching the frequency responses of complex poles or zeros. Note that for a single pair of complex zeros the magnitude and phase
plots are those shown in Figure 2.16 rotated a half-turn about their respective frequency axes; note that for n = 1 the associated phase varies between zero and 180 .

56

Control systems techniques

Ch. 2

20
Magnitude Response (dB)

15
10
5
0
5
10
15
20
25
30
1
10

10

10

Phase Response (deg)

20
40
60
80
100
120
140
160
180
200
1
10

= 0.10
= 0.25
= 0.50

Figure 2.16

10
Frequency (rad/s)
0.15
0.30
0.71

10

0.20
0.40
1.0

Magnitude and phase responses as a function of the normalized frequency

f n for a pair of complex poles.

Example 10. Lead compensation

Lead compensation is often employed as a more practical form of derivative compensation
over a range of frequencies. In its application in power system dynamics and control it is
used to provide phase lead over a desired range of frequencies; the design of cascade phaselead compensation for conventional closed-loop control systems is covered in [1].
1 + sT
1 + s T
ble gain. The range of values for is typically 0.1 1 when the transfer function is implemented using analog devices. For = 0.1 the phase lead is 56 . For values of 0.1

The form of the lead transfer function is A G LD s = A ------------------------ , where A is an adjusta-

Sec. 2.12

Frequency response diagrams and Bode Plots

57

the additional phase lead provided is small. For example, cascading two identical lead networks with = 0.25 produces the same maximum phase shift as a single lead network
with = 0.025 .

LM (dB)
0

1
--T

1
------T

+ 90
Phase
(deg)

20 log 1

1
10 log ---

m
m

Figure 2.17 Frequency response for a lead transfer function.

The form of the frequency response of the lead compensator is shown in Figure 2.17; the
important feature of this response is the phase lead introduced by the compensator. The
maximum phase lead, m , occurs at the geometric mean of its corners, m , where
1 1 12
1
m = ------- ---
= ----------- . The maximum phase lead can be shown
T T
T
1
1

m = asin ------------- . Note that at m the log-magnitude is LM = 10 log --- .

1 +

to be

--------------------2.12.1.5

Exercise.

Show that the magnitude and phase plots of the lag block,
1 + sT
A G LG s = A ------------------------ ,
1 + s T

1 ,

are those shown in Figure 2.17 reflected in their respective frequency axes. It is a practical
form of integral compensation over a range of frequencies; identify that range.
--------------------Example 11. Plotting the frequency response
It is often useful to visualize or sketch the frequency response of a given transfer function
based on the straight-line approximations to the frequency responses of the component factors.
Draw the straight-line approximations to the frequency response of the following open-loop
transfer function of a unity-feedback control system. Show both the straight-line asymptotes
and the actual plot.

58

Control systems techniques

10 4
B s = --------------------------------------------- .
s s + 10 s + 100

Ch. 2

(2.39)

Note the linear factors are in the form s + a rather than 1 + sT form required for plotting the asymptotes. By dividing the denominator factors by 10 and 100 the form of (2.39)
10
is changed to B s = ------------------------------------------------------ ; note the gain is 10, or 20 dB. Set s = j f .
s 1 + s0.1 1 + s0.01

The corner frequencies of the transfer function are c = 1 T , i.e 1 0.1 = 10 and
1 0.01 = 100 rad/s for the two factors 1 + s0.1

and 1 + s0.01 , respectively. The

straight-line representations for the four factors are shown in Figure 2.18. These are combined, making allowance for the deviations of the actual responses from the straight-line approximations as shown in Figure 2.14 and Figure 2.15, to form the response of the transfer
function.

Magnitude Response (dB)

50

Gain x10 (20dB)

1/s
1/(1+0.1s)
1/(1+0.01s)
Transfer function plot
Combined straightline approximation

50

100
1
10

10

10

10

10

Phase Response (deg)

0
50
100
150
180 deg
200
250
1

10

10

10
Frequency (rad/s)

10

10

Figure 2.18 Frequency response plots of the transfer function B s

using straight-line approximations.

Sec. 2.12

Frequency response diagrams and Bode Plots

59

2.12.2 Stability Analysis of the closed-loop system from the Bode Plot
The transfer function W s of the closed-loop system shown Figure 2.9 was derived in
Gs
Section 2.7; the transfer function is W s = --------------------------------- .
1 + G s H s

As mentioned earlier, the important feature of the Bode Plot is that the stability of the
closed-loop system can be derived from the plot of the open-loop transfer function
G s H s . The theoretical basis for this result requires that no poles or zeros of G s H s
lie in the right-half of the complex s-plane, i.e. it is open-loop stable and minimum phase
[1]. (If zeros of G s H s lie in the right-half of the s-plane the latter transfer function is
called non-minimum phase.) The criterion for the stability of closed-loop systems based on
the Bode Plot of open-loop stable transfer functions follows from the more generally applicable Nyquist criterion that covers both open-loop unstable and non-minimum phase systems [1].
Assume the Bode Plot in Figure 2.19 is drawn in the vicinity of the gain cross-over frequency, co , for the open-loop transfer function G j f H j f . (This analysis may be carried
out for the transfer function B s in Figure 2.18; B s satisfies the condition that it is openloop stable and minimum phase.)
Log Magnitude Plot (dB)

|GH|

co
f

Gain Margin

Phase Plot (deg)

Frequency f

Arg GH

-180

Phase Margin

-180

co is the gain cross-over frequency (rad/s)

Figure 2.19

Gain and Phase Margins defined on the Bode Plot of the G s H s .

It can then be shown that when the phase shift f = 180 , the corresponding value of
the LM must be negative for stability.The amount by which the gain can be increased before
instability results, is called the Gain Margin.

60

Control systems techniques

Ch. 2

The term, Phase Margin is defined as the difference between the 180 line and the phase
plot when the Log Magnitude Plot crosses the zero dB axis, i.e. when LM = 0 . The Phase
Margin for a minimum-phase system must be positive for stability. The Phase Margin can
also be interpreted as the amount of phase lag that can be introduced at unity loop-gain before instability of the closed-loop system results.
Example 12. Derive the information on stability and dynamic performance of the
closed-loop system from the Bode Plot
Assume that the transfer function B s in the Example 11 represents the open-loop transfer
function of a unity gain feedback system, i.e. B s = G s H s = G s . The Gain and
Phase Margins for this system are illustrated in Figure 2.20 for the Bode Plot for G s H s .
The following information can be extracted from the Plot.

Because the Gain and Phase Margins are positive the closed-loop system is stable.

If the gain in the forward-loop is increased by 21 dB the closed-loop system becomes

marginally stable and, ideally, it would oscillate with a constant amplitude at a frequency of 32 rad/s. Note that the gain-crossover frequency shifts to the right increasing in frequency from 8 to 32 rad/s as the magnitude plot shifts vertically
upwards.

Further indicative data on the dynamic performance of the closed-loop system is

revealed by the Phase Margin (PM ). A rule of thumb is, if the closed-loop system
has a pair of dominant complex poles, the damping ratio of the closed-loop poles is
approximately PM /100 [1].

For a good servo-system transient response, the Phase Margin should be about 70 for a
closed-loop system that has a dominant pair of complex poles. For such a Phase Margin it
would be necessary to reduce the gain by 9.3 dB in the case of Figure 2.20. Note the gain
crossover frequency is reduced to 3.3 rad/s; this implies that the frequency of the damped
sinusoidal response to a step change at the input of the closed-loop system would also be
reduced, possibly to 4 - 5 rad/s.
The significance of this example is that it illustrates how a variety of useful information for
the design of the performance of the closed-loop system can be derived from the Bode Plot of the
open-loop system.

-------------------

Sec. 2.13

The Q-filter, a passband filter

61

Magnitude Response (dB)

50

Gain Margin = 21 dB
Gain cross-over
frequency = 8 rad/s

50

100
1
10

10

10

10

10

Phase Response (deg)

0
50
100
Phase Margin = 47 deg

150

-180 deg

200
250
1

10

Figure 2.20

10

10
Frequency (rad/s)

10

10

Bode Plot of the open-loop transfer function showing the gains

and phase margins.

As stated, in applications to power system analysis one should be aware of open-loop systems that are non-minimum phase when using the Bode Plot for stability analysis. An example of such a case is the model of a Francis turbine in a hydro-electric plant. This model
contains a right-half plane zero (which causes the turbine power output to rise initially as the
wicket gates are closed).

2.13

The Q-filter, a passband filter

The Q-filter is a bandpass filter which passes a selected band of frequencies and attenuates
those which lie outside the bandwidth of the filter. The transfer function of the filter is
2 m s
Q s = --------------------------------------------------------------- ,
1 + 2 m s + s m 2

(2.40)

62

Control systems techniques

Ch. 2

where m (rad/s) is the centre or resonant frequency, i.e. the frequency at which the magnitude of Q(s) is a maximum; is the damping ratio. The 3 dB bandwidth is
2 1 = 2 m where 1 2 = 2m ; the quality factor of the filter is defined as
Q = 1 2 . The frequency response characteristics of the filter are shown in Figure 2.21

as a function of the damping ratio; note that the phase responses pass through zero degrees
at resonance. These characteristics are relevant to a type of stabilizer in section Appendix 8
I.3.

Magnitude (dB)

20
0
20
40
60
80
2

10

10

10

90
Phase (deg)

60
30
0
30
60
90
2

10
=0.1
=0.5
=2.0

Figure 2.21

10
Frequency (rad/s)
=0.2
=0.6
=4.0

=0.3
=0.8
=7.0

10

=0.4
=1.0
=10

Frequency response of the Q-Filter for damping ratios from 0.1 to 10.

2.14 References
[1]

G. Franklin, J.D. Powell and A. Emami-Naeini, Feedback Control of Dynamic Systems,

6th edition, Prentice Hall, 2009.

[2]

R.C. Dorf and R.H. Bishop, Modern Control Systems, 12th edition, Prentice Hall,
2010.

[3]

P. Kundur, Power System Stability and Control, McGraw-Hill, Inc., 1994.

Chapter 3
State equations, eigen-analysis and applications

3.1 Introduction
The description of the dynamics of large systems, such as power systems, by their transfer
functions is unsatisfactory for a number of reasons. For example, for a system of order n,
say 100, the characteristic polynomial has degree 100 and 101 coefficients of s. Moreover,
such systems typically have more than one output variable and more than one input signal.
modelling based on the multi-input multi-output state equations of the system is simpler and
problems of loss of accuracy are reduced. Moreover, such modelling has a number of advantages and features some of which are described in the following sections. To illustrate
the formation of the state equations of a plant or an electro-mechanical system, let us consider two examples.
Much of the material on linear systems analysis in the later sections is covered by [1].
3.1.1 Example 3.1.
Find a set of state and output equations for the simple RLC circuit shown in Figure 3.1. The
voltage supplied by an ideal source is v s t , and the required outputs are the capacitor voltage are v C t and inductor current i L t .

63

64

State equations; eigen-analysis

Ch. 3

iL(t)

vs(t)

vC(t)

Figure 3.1 RLC circuit

v

Loop voltages: v s = v L + v C = Lpi L + v C or pi L = -----C- + ----s ,

(3.1)

where p is the differential operator d dt .

vC
i L = ------ + Cpv C
R

vC iL
pv C = -------- + ---- .
(3.2)
RC C
Note that each of the right-hand equations is a first-order differential equation with the derivative specifically sited on the left-hand side of the equation.

Current flow:

or

There are two independent energy storage elements, C and L. Because the instantaneous en2
1
1 2
ergy stored in C and L is --- Cv C and --- Li L , respectively, the variables x 1 = v C and

x 2 = i L are natural selections for states. Hence, the state equations are formed as follows:

From (3.1),

vC vs
pi L = ------ + ---- ,
L L

which becomes

1
1
x 2 = --- x 1 + --- v s ,
L
L

and, from (3.2),

vC iL
pv C = -------- + ---- ,
RC C

we have

1
1
x 1 = -------- x 1 + ---- x 2 .
C
RC

The two output equations required are

y1 = vC = x1

and

y2 = iL = x2 .

The state and output equations can thus be written in matrix form as follows:
1 1

-------- ---0
x1
RC C x 1 +
=
vs

--11
x2
x2
--- 0
L
L

or

x
= 1 0 1 + 0 vs
0 1
0
y2
x2

or

y1

x1
x
= A 1 + b v s , and

x2
x2
y1
y2

= C

x1
x2

+ d vs ,

Sec. 3.1

Introduction

1
-------RC
A =
1
--L

where

65

1
---C ,

b =

0
1 ,
--L

C = 1 0 , and
0 1

d = 0.
0

--------------------------------3.1.2 Example 3.2

A drive system shown in Figure 3.2 consists of a DC motor driving an inertial load through
a speed-reducing gearbox. The controlled DC supply voltage to the armature is supplied by
a power amplifier. The motor field current is maintained constant (i.e. the flux/pole is constant). Write down the equations of motion for this system.
ra

La

ia(t)

vs(t)

TLa

vb(t)
Ta

TL
J

Gearbox
n:1

Figure 3.2 Load driven by a DC motor through a gearbox

The equations describing the dynamic behaviour of the system are developed below. The
variables, parameters and their units are:
v s t and v b t are respectively the supply voltage and the back emf of the motor (V);
i a t is the armature current (A);
T a t is the motor (armature) torque (Nm);
T L t and T La t are load torques referred to the load and motor shafts, respectively (Nm);
a t and t are the motor and load speeds (rad/s), n:1 is the gearbox ratio;
r a t and L a are the resistance (ohm) and self-inductance (Henry) of the armature circuit;
J is the inertia constant of the rotating system (kg-m2);
B is the coefficient of viscous friction (Nm/rad/s);
k k t are constants of proportionality.

Armature circuit equation:

v s = r a i a + L a pi a + v b

Back emf of motor:

vb = k a

v b flux/pole a

Torque developed by motor:

Ta = kt ia

T a flux/pole i a

66

State equations; eigen-analysis

Opposing load torque:

T L = Jp + B

Speed of motor shaft:

a = n

Power transfer across gearbox:

T La a = T L ;

Ch. 3

T La = T L n

Eliminating T a t , T L t , T La t , a t and v b t from the above equations, and inserting

the derivative terms on the left-hand side of the relevant equations, we find
ra
nk
vs
pi a = ----- i a --------- + ----- ,
La
La
La

(3.3)

nk t
B
p = --- + ------- i a .
J
J

(3.4)

Assume the outputs of interest are: i a , , a , T a .

Let the states be x 1 = i a and x 2 = , then substituting these quantities in (3.3) and (3.4)
we find:
r a nk

----- --------1
x1
x
----La La
=
1 + La vs t

x2
nk t
x2
B
0
------- --J
J

or

x1
x
= A 1 + b vs t .

x2
x2

(3.5)

The outputs of interest can be expressed as:

y1 = ia ,
y1

1
y2
0
=
0
y3
k
t
y4

y 2 = ,

y 3 = a = n,

0
0
x1
1

+ 0 vs
n
0
x2
0
0

y 4 = T a = k t i a , then
y1

or

y2
y3

= C

x1
x2

+ d vs ,

d = 0.

(3.6)

y4

Note there are 2 state equations, 1 input, 4 outputs. Thus A is dimension 2 2 , b is 2 1 ,

C is 4 2 , d is 4 1 and is a zero vector.
Equations (3.5) are called the state equations for this second-order system, (3.6) are its output
equations. There are two independent energy storage elements, the field-circuit inductance
and the inertia of the rotating system, L and J. As stated, the energy stored in these elements
2

is Li a 2 and J 2 , respectively at any time t, and are uniquely determined by the instantaneous values of i a t and t . Thus x 1 = i a and x 2 = are thus natural selections for
the state variables. In general there are as many states in the system as there are independent
energy-storage elements - with the addition of those states and state equations representing

Sec. 3.2

Concept of state and the state equations

67

pure integration (e.g. the relation pd = v between speed v and distance d).
---------------------------------

3.2 The concept of state and the state equations [1]

Definition: The state of a system at any time t = t i is a minimum set of numbers which, together with the input function u t for t t i and the equations describing the dynamics, are

sufficient to determine the future behaviour of the states and the output of the system.
The state of the system at time t is described by a vector of n state variables
x1 t x2 t xn t

. A knowledge of the initial values of the state variables

x1 t0 x2 t0 xn t0

at time t0 and the m input signals u 1 t u 2 t u m t

for

t t 0 , is sufficient to determine the future values of the state variables and the output.

These concepts lead to the state-space model of a system in terms of a set of n first-order differential equations (in contrast to a single nth - order differential equation from which the
conventional transfer function is derived). Recall that in each first-order equation the derivative of the state variable is placed on the left-hand side of the equation and all terms in the
state variables and inputs on the right-hand side. A general form of the n state equations,
with m input signals, is thus
x 1 = a 11 x 1 + a 12 x 2 + + a 1n x n + b 11 u 1 + b 12 u 2 + + b 1m u m
x 2 = a 21 x 1 + a 22 x 2 + + a 2n x n + b 21 u 1 + b 22 u 2 + + b 2m u m

x n = a n1 x 1 + a n2 x 2 + + a nn x n + b n1 u 1 + b n2 u 2 + + b nm u m

(3.7)

The p output equations are:

y 1 = c 11 x 1 + c 12 x 2 + + c 1n x n + d 11 u 1 + d 12 u 2 + + d 1m u m

y p = c p1 x 1 + c p2 x 2 + + c pn x n + d p1 u 1 + d p2 u 2 + + d pm u m

(3.8)

A more compact arrangement of the above equations is the matrix form:

x t = Ax t + Bu t
y t = Cx t + Du t

(3.9)

68

State equations; eigen-analysis

where

A =

a 11 a 12 a 1n
a 21 a 22 a 2n
.

. .
a n1 a n2 a nn

B =

b 11 b 12 b 1m
b 21 b 22 b 2m
.

. .
b n1 b n2 b nm

Ch. 3

(3.10)

(3.11)

of dimension n n and n m , respectively, and

where

C =

c 11 c 12 c 1n
c 21 c 22 c 2n
.

. .
c p1 c p2 c pn

D =

d 11 d 12 d 1m
d 21 d 22 d 2m
.

. .
d p1 d p2 d nm

of dimension p n and p m .
T

The n 1 state vector is

x t = x1 t x2 t xn t

the m 1 input vector is

u t = u1 t u2 t um t

the p 1 output vector is

y t = y1 t y2 t yp t

,
T

, and

Examples of state and output equations have been given in Examples 1 and 2. (Also see [1],
[2] and [3].)

3.3

The linearized model of the non-linear dynamic system

The equations describing the power system and its dynamics are non-linear. For analyzing
the dynamic performance of the non-linear plant and the system, typically following a large
disturbance such as a fault, a step-by-step integration of the non-linear equations is carried
out to calculate the time-domain responses of the system variables. Such variables are generator speeds and rotor angles, bus voltages, controller outputs, etc. Because the dynamic
behaviour of the system depends very much on the location and the severity of the disturbance, as well as the operating conditions, it is necessary to conduct a large number of socalled transient stability studies to characterise the dynamics of the system [4].
Linearizing the set of non-linear equations for a selected operating condition results in a new
set of equations in a new set of variables. These variables are the perturbations x about the
steady-state quantities x 0 . The variable x in the non-linear equations is related to the former
pair by x = x + x 0 . The advantages of forming the linearized equations of a system are:

All the powerful analytical methods developed in linear control theory are available for
the analysis of the linearized dynamic system.

Sec. 3.3

Linearized models

69

If the linearized system is stable at the selected steady-state operating point then,
according to a theorem by Poincar [5], the non-linear system is also stable at that
operating point.

The dynamic performance of the linearized system can be characterized by the location of its poles in the complex s-plane. Based on results in the theory covered in Sections 2.8 and 2.9, the real parts of these poles (or damping constants) provide the
information on stability, how well damped the modes 1 are, the nature of the transient
response, etc. Such information cannot be gleaned directly from the results of timedomain analysis of the non-linear system.

If the transient responses of the linearized system to a disturbance are calculated, a question
is how accurate are the responses. Poincar proved that the response of the linearized system to a disturbance is exactly the same as that of the original non-linear system if the disturbance is vanishingly small. As explained in Section 1.1, the responses predicted by the
linearized model are often sufficiently accurate for practical analysis and design purposes.
However, care must be exercised to take into account the nature of the non-linearities, the
operating point and the size of the perturbation when deciding if the linearized model is
practically applicable to the analysis being performed.
3.3.1 Linearization procedure
In [6] the set of nonlinear differential-algebraic equations (DAEs) describing the dynamic
behaviour of the integrated power system are derived and are shown to be of the form:
x = f x u , 0 = g x u , y = h x ,
where the vector x represents the n states of the system, the r algebraic variables, u the
m system input variables, and y the p output variables.
At the steady-state operating condition, which is the equilibrium point about which the system is to be linearized, implies by definition that all rates of change are zero, x = 0 , thus
f x 0 , 0 ,u 0 = 0 ,

0 = g x 0 , 0 ,u 0 ,

y 0 = h x 0 , 0 .

(3.12)

Assume the system is subjected to a small perturbation from the steady state such that
x = x 0 + x ,

= 0 + ,

u = u 0 + u ,

y = y 0 + y .

(3.13)

The perturbed the variables must satisfy (3.12). For example, in the case of the output y in
(3.13)
y = y 0 + y = h x 0 + x , 0 + .
Because the perturbations are small, the nonlinear function y = h x , can be expressed as
a first-order Taylors series expansion. Consider the ith output, y i , i = 1 , ,p :
1. See a note on eigenvalues, modes and stability in Section 3.5.2.

70

State equations; eigen-analysis

Ch. 3

h i
h i
h i
h i
y i = y i0 + y i = h i x 0 , 0 + -------- x 1 + + -------- x n + -------- 1 + + ------- r
x
x

x a

where the partial derivatives --------i , a = 1 , ,n ; and --------i , b = 1 , ,r are evaluated at the initial steady-state operating point x 0 , 0 ,u 0 .
Because y i0 = h i x 0 , 0 in the above equation, it reduces to an equation in the perturbed
variables:
h i
h i
h i
h i
y i = -------- x 1 + + -------- x n + -------- 1 + + ------- r .
x 1
x n
1
r

(3.14)

Similar expressions can be derived for the two remaining functions in (3.12), for j=1, ... , n:
x j =

m f
r f
f j
j
j
------------- u c ,
+
+
x

x a b ------
u
c=1 c
a=1 a
b=1 b
n

0=

(3.15)

k
- x a + -------k- b + -------k- u c .
------

u
x
c=1 c
a=1 a
b=1 b

(3.16)

The sets of linearized equations, (3.14) to (3.16) are more conveniently expressed in matrix
form,
x = J fx x + J f + J fu u
0 = J gx x + J g + J gu u

(3.17)

y = J hx x + J h + J hu u

or more compactly:
J fx
x
0 = J gx
y
J hx

where the i a

th

J f J fu x
J g J gu
J h J hu u

(3.18)
f

element of the J fx sub-matrix of the system Jacobian matrix is -------j- evalx a

uated at the initial steady-state operating point. The elements of the other sub-matrices are
similarly defined.
The formulation of the linearized equations of the system as a set of DAEs possesses a number of significant advantages, including:.

The natural formulation of the equations for devices and their controllers is
exploited when building the set of system equations;

Sec. 3.4

Solution of the State Equations

71

The equations for large systems are inherently highly modular and extremely sparse
(i.e. the Jacobian matrix contains very large numbers of zeros). Highly efficient computational algorithms for processing modular and sparse matrices are exploited when
computing frequency responses, transfer-function residues - as well as computing a
subset of the system eigenvalues within a selected region of the complex s-plane.

The modularity and sparsity of the system equations can be exploited when computing eigenvalue sensitivities.

Elimination of the algebraic variables from the DAEs yields the conventional ABCD form
of the state equations, i.e.

x t = Ax t + Bu t ,

y t = Cx t + Du t ,

,where

B = J fu J f J g J gu

A = J fx J f J g J gx
1

C = J hx J h J g J gx

(3.19)

(3.20)

D = J hu J h J g J gu

Equations (3.19) are the state and output equations and are a form commonly used in the
literature on linear control theory to describe a system.

3.4 Solution of the State Equations

The solution of the state and output equations (3.19) in matrix form can be obtained by taking the Laplace Transform of each of its first-order differential equations, i.e.
sX s x 0 = AX s + BU s , or
sI A X s = x 0 + BU s ,

(3.21)
1

where I is the nth order identity matrix. Pre-multiplying both sides by sI A , we obtain:
1

X s = sI A x 0 + sI A BU s .

(3.22)

The first term on the right of (3.22) is the natural response and the second is the forced response. The time-domain solution is of the form:
At

xt = e x0 + e
0

At

Bu d
At

y t = Cx t + Du t .

The solution involves the matrix-exponential e , whose numerical computation for high
order systems is very challenging [7]. The following analysis reveals important theoretical aspects of the time response. However, it is emphasized that these methods are not employed
in practice to numerically calculate time-responses.
3.4.1 The Natural Response
The Natural Response is derived from (3.22):

72

State equations; eigen-analysis

1

X s = sI A x 0
Y s = CX s

or
or

Ch. 3

At

x t = e x 0 .
y t = Cx t

(3.23)

The matrix sI A is known as the resolvent matrix. The solution for X s in (3.23) can
be expressed in the form:
1
Adjoint sI A
X s = sI A x 0 = ------------------------------------------------------ x 0 .
Determinant sI A

Let B = Adjoint A . The i j

where M ji is the j i

th

th

element of B , b ij is defined as b ij = 1

(3.24)
i + j

det(M ji)

minor of A ; M is the n 1 n 1 matrix obtained by deleting

the jth row and ith column of A [1].

3.4.2
Let

Example 3.3
1 3 4
A = 2 1 3 ; compute B = Adjoint A .
4 2 3
b 11 = 1

1 + 1

b 12 = 1

1 + 2

det 1 3 = 1 1 3 3 2 = 3 ,
2 3

det 3 4 = 1 3 3 4 2 = 17 , and so on, to yield

2 3

3 17 13
B = 18 13 11 .
8 14 5

--------------------------------For a solution which is non-trivial it is necessary that:

Determinant sI A = 0 .

(3.25)

This is also the characteristic equation of the system; the zeros of (3.25) are the poles of the
system defined by the state matrix A . This is an important result because it reveals that the
characteristic dynamic behaviour of system is encapsulated in the state matrix. Unfortunately, the solution for X s based on (3.24) becomes unmanageable for fourth- and higher-order systems.
At

We can write the natural response, x t = e x 0 in the form x t = t t 0 x 0 where

t t 0 is known as the n n state transition matrix. It describes the transition of the n
At

states over the interval t 0 t and is equal to e . If t 0 = 0 , t t 0 is replaced by t ,

Sec. 3.4

Solution of the State Equations

73

x t = t x 0 .

(3.26)

We note, by examination of (3.24) and (3.26), that the Laplace Transform of t is

s = sI A
sI A

. Hence t can be determined by evaluating the resolvent matrix,

, and taking the inverse Laplace Transform.

3.4.3 Example 3.4: Natural response

The state, input, output and direct-transmission matrices in this example are:
A = 0 2 , B = 0 , C = 2
1 3
1
0

0 , D = 0 respectively.
1

(a) Calculate the state transition matrix t for the state-matrix.

1

Firstly, we evaluate the resolvent matrix sI A using the result for the inverse given in
(3.24). The Adjoint (Adj) matrix is the transpose of the matrix formed by replacing the elements aij by their cofactors (signed minors). Thus,
sI A =

Hence

2
s
2
, and det sI A = s + 3s + 2 = s + 1 s + 2 .
1 s + 3

s = sI A

s+3
--------------------------------
s
+
1 s + 2
=
1
------------------------------- s + 1 s + 2

2
-------------------------------- s + 1 s + 2 .
s
------------------------------- s + 1 s + 2

Note the poles of the system, s 1 2 = 1 2 , are associated with each element of the matrix.
Taking the inverse Laplace Transform, we find: t = e

(b) Find the responses of the outputs y =

At

2t

= 2e e
t
2t
e e

y1 t

= Cx = 2
0
y2 t

2e + 2e
t

e + 2e

0 x1 t
1
x2 t

x 1 0 = 1 and x 2 0 = 1 .

The responses of the two states are

x1 t
x2 t

11 12
x 0
1
,
21 22
x2 0

2t

x 1 0 + 2e + 2e

2t

x 1 0 + e + 2e

x 1 t = 2e e
x2 t = e e

2t

2t

i.e.
t

x 2 0 = 4e 3e
t

x 2 0 = 2e 3e

2t

2t

2t

2t

when

74

State equations; eigen-analysis

t

y 1 t = 2x 1 t = 8e 6e

The output responses are thus:

y 2 t = x 2 t = 2e 3e

Ch. 3

2t

2t

Note for this second-order system there are two eigenvalues in the responses, e
---------------------------------

and e

2t

3.4.4 The Forced Response

Assuming all initial conditions are zero in (3.22), the state and output responses are:
1

X s = sI A BU s ,

Y s = CX s + DU s . Eliminating X(s), we find:

Y s = G s U s

or Y s = C sI A B + D U s .

(3.27)

From (3.27) the multi-input, multi-output (MIMO) transfer function is defined as:
G 11 s G 1m s
G s = C sI A B + D =
.
G p1 s G pm s
1

Clearly, if there are no pure gain paths directly between input and output, then D = 0 . The
MIMO transfer function is then:
1

G s = C sI A B .

3.4.5

(3.28)

Example 3.4 (continued).

(c) Find G(s) given: sI A

1
= --------------------------------- s + 3
s + 1s + 2 1

2 ,
s

B = 0 .
1

The elements of C are given in part (b) of Example 3.4. Hence, by substitution in (3.28),
1
G s = --------------------------------- 2 0 s + 3
s + 1s + 2 0 1
1

4
--------------------------------2 0 = s + 1 s + 2 .
s
s
1
-------------------------------s + 1s + 2

----------------------------------

3.5

Eigen-analysis

3.5.1

The eigenvalues of the state matrix, A

th

The h eigenvalue of a real, n n matrix A is the real or complex scalar quantity, h , which
is the non-trivial solution of the equation
(3.29)
Av h = h v h .

Sec. 3.5

Eigen-analysis

75
T

The n -element column vector, v h = v 1h v 2h v nh , is called the right eigenvector of

the matrix A corresponding to the eigenvalue h .
To calculate the eigenvalue, let us rewrite (3.29) in the form
h I A v h = 0 .

For a solution which is non-trivial, Determinant h I A = 0 .

This is also the characteristic equation of the system and is the same as (3.25) with s replaced
by h . The eigenvalues of the state matrix A are thus the same as the poles of the transfer function, and are independent of which variables in the system are selected to be inputs or outth

puts. For the n -order system there are n eigenvalues which are real or exist in complexconjugate pairs.
Likewise, there exists a n-element row vector, w h = w h1 w h2 w hn , which satisfies the
equation
wh A = wh h ;

(3.30)

w h is called the left eigenvector of the matrix A corresponding to the eigenvalue h .

Let us form the n n matrices of right and left eigenvectors V and W , respectively, corresponding to the eigenvalues 1 2 h n , i.e.
T

V = v 1 v 2 v n , and W = w T w T w T .
1 2
n

(3.31)

Also let us form the diagonal matrix of eigenvalues, diag h h = 1n .

Equations (3.29) and (3.30) can then be expressed respectively as
AV = V , and

WA = W .

If V is non-singular, which is usually the case for realistic power system models, then:
1

V AV = and WA = V AVW ,

from which we may conclude that W = V

or

WV = I .

The latter equation reveals a useful orthogonality property of eigenvectors corresponding to

a selected eigenvalue, h , namely
wh vh = 1,

(3.32)

76

State equations; eigen-analysis

Ch. 3

whereas, for eigenvectors corresponding to different eigenvalues, g and h , w g v h = 0 .

V and W are respectively known as the right and left modal matrices of the state matrix A .

3.5.2 A note on eigenvalues, modes and stability

As described in Section 3.5.1, an eigenvalue i of the n n system matrix A is a real or
complex number. The system matrix has n eigenvalues, some real and some in complex-conjugate pairs. Each real eigenvalue ( 1 ) is associated with a monotonic mode which in the
t

time-domain has the form y 1 t = A 1 e 1 . Each complex-conjugate pair of eigenvalues

( 2 j ) is associated with an oscillatory mode which in the time-domain has the form
y2 t = A2 e

2 t

sin t + 2 . The unforced response of a linear time-invariant system is a

(weighted) superposition of the response of each of the system modes. In power systems
analysis the term mode usually refers to a broader set of properties than just the damping and
frequency of oscillation in order to characterise more completely the physical behaviour of
the natural modes of system response in the time domain. Other modal characterisations include, for example, whether it is an electro-mechanical mode, or a controller mode, etc. In
the case of electro-mechanical modes we refer to sub-classifications such as inter-area
modes, local modes, etc. (see Section 1.5). In conjunction with the engineers detailed
knowledge of the system structure, the eigen-decomposition of the system, including eigenvalues, eigenvectors and participation factors, is a tool employed to characterize the system
modes. In the following text the term mode is, at times, used instead of eigenvalue. Note that
h

there are occasions when the symbol refers to either the hth mode or the hth eigenvalue;
the application depends on the context. See Section 9.1.1 for further details.
It has been noted in Section 2.9 that the system is unstable if a pole or a pair of complex
poles lies in the right-half of the s-plane. Correspondingly, instability arises if the real part of
any eigenvalue is positive; the associated mode therefore increases exponentially with time.

3.6

Decoupling the state equations

Let us consider the response of the states in the state equations of (3.9) when the system

responds to a set of initial conditions with zero input. Substituting x = Vz and x = Vz in

the resulting state equation,

x t = Ax 0 , we find

(3.33)

1
z = V AVz 0 = z 0 .

(3.34)

Cross-coupling between states exists in the state equations (3.33) which represent the physical system. However, because the matrix is a diagonal matrix of eigenvalues, no crosscoupling terms exist in (3.34); the latter equation is a said to be a decoupled form of the state
equations, with pseudo-states z t .

Sec. 3.7

Residues from the state equations

77

In the above analysis we have invoked two transformations,

1

and z 0 = V x 0 = Wx 0 .

x = Vz

3.7 Determination of residues from the state equations

In the following sections much of the material relating to the applications to power system
dynamics is covered in [8] to [12].
The full set of state equations (3.9) can be expressed in a decoupled form by letting x = Vz .
The resulting state equations in the new state vector z are:
z = z + B U ,

y = C z + D U,

(3.35)

where = WAV is a diagonal matrix of the system eigenvalues,

B = WB is the mode-controllability matrix,

C = CV is the mode-observability matrix, and

D = D (usually a null matrix in power system applications).

A mode h is observable in an output signal y m if and only if c mh 0 . Consequently,

c mh = c m v h is a measure of the observability of the mode in the output y m . Similarly,

b = w b is a measure of the controllability of the mode from an input u .
hq

h q

Assume that there are n distinct eigenvalues. The transfer function between the q

th

input

th

and the m output is

ym s
1
------------- = c m sI b q

uq s
1
1
1

= c m diag -------------- -------------- -------------- b q

n
1
2

c mh b hq
--------------------
h = 1 s h
n

The form of the transfer function displayed in the last equation is the same as that in (2.14)
in which K h is the residue associated with the eigenvalue at s = h . Here, the residue for
the eigenvalue s = h is defined as
h

r mq = c mh b hq ,

(3.36)

78

State equations; eigen-analysis

Ch. 3

which, according to the above definitions, is a combined measure of observability and controllability.
Based on the concepts of controllability the mode of concern is highly controllable from
th
the q input of the state-space model of the system if the magnitude of b hq is large relative

to that evaluated at all other inputs - subject to a note of caution 1. Likewise, the signal at
m

th

output of the model is highly observable - and is suitable as a feedback or stabilizing

signal for the selected mode - if the magnitude of c mh is large relative to the values found
for other possible feedback or stabilizing signals. Hence, for a candidate controller or a stath

bilizer in the path from the m output to the q

residue

3.8

r hmq

th

input to be effective, the magnitude of the

must be relatively large at the modal frequency h .

Determination of zeros of a SISO sub-system

Although power systems are MIMO systems there is often interest in the analysis of a singleinput, single-output (SISO) subsystem. Consider the SISO from the ith input to the jth output of the MIMO system described by the general set of state-equations in (3.9). Let b i be
the ith column of the input matrix B , c j the jth row of the output matrix C and d ij be the
ijth element of the direct-transmission matrix. The resulting SISO sub-system is described

by:

x t = Ax t + b i u i t

(3.37)

y j t = c j x t + d ij u i t

in which u i t and y j t are respectively the ith and jth input and output.
The objective is to find the set of zeros of the SISO subsystem: that is, complex frequencies
h t

h , h = 1 , ,m n for which the forced response to the non-zero input u ih t = u ih0 e

results in the output y jh t 0 for all time. It is also assumed that none of the zeros are equal

to any of the system poles, i.e. there are no pole-zero cancellations. The forced response of
the state-variables is x h t = x h0 e

h t

and their rate of change is x h t = x h0 h e

h t

. Sub-

stituting for the driving input and associated forced responses in (3.37) results in:

1. A note of caution: because the various output or input signals will be different (e.g. output signals may be voltage, power, speed, etc.), care should be taken in assessing the
effects of scaling.

Sec. 3.8

Zeros of a SISO sub-system

A h I
cj

79

bi

x h0

= 0 .
0
d ij u ih0

(3.38)

The non-trivial solution of (3.38) requires:

A I
h
Det

cj

b i
= 0.
d ij

(3.39)

Expansion of the above determinant results in a polynomial in h and the associated roots
correspond to the zeros of the SISO sub-system.
Equation (3.39) does not provide a tractable means for computing the zeros of systems with
more than a few state variables. Thus, (3.38) is rewritten in the form of a generalized eigenvalue problem as follows:
Hv h = h Mv h ,

where H =

bi

cj

d ij

(3.40)

x
, M = I 0 and v h = h0 ;
0 0

u ih0

(3.41)

the objective is to compute all finite values of h for which there exist non-trivial solutions
of (3.40).
The well-known QZ algorithm developed by Moler and Stewart [13] is a numerically robust
procedure for computing h . The Fortran LAPACK library [14] provides a suite of subroutines in the public domain that implement the QZ algorithm. The Matlab function qz provides an interface to the appropriate LAPACK library routines.
At the heart of the QZ algorithm is the determination of unitary matrices Q and Z such
that both S = QHZ and T = QMZ are upper diagonal. As stated in [13] the two eigenvalue problems Syh = h Ty h and Hv h = h Mv h are unitarily equivalent: they both have the
same eigenvalues h and their eigenvectors are related by v h = Zy h . Suppose the hth diagonal entries of S and T are respectively s hh and t hh then h = s hh t hh . If t hh is zero, or
computationally very close to zero, then h = . If both s hh and t hh are zero, or computationally very close to zero, then the system is said to be degenerate.
Extensions of the above approach to the determination of the properties and computation
of the various kinds of zeros of MIMO systems have been devised by a number of investigators [15, 16, 17]. Methods for computing dominant zeros in large systems have been developed and implemented by Martins, et al. [18, 19, 20].

80

3.9

State equations; eigen-analysis

Ch. 3

Mode shapes

Let us assume that we able both to excite a particular mode and to evaluate the time responses of the states of the system. In (3.22) it was noted that the response can be separated into
its natural and forced components. Assuming (i) non-zero initial conditions on the states,
and (ii) no forcing signals applied at the inputs to the system, we can write the equation for
1

the natural response of the states in the form X s = sI A x 0 .

In the previous section a decoupled form of the state equations is derived (3.35),

z = z + B U ,

assuming the pseudo-states and the original states, z and x respectively, are related by
x = Vz . With no external excitation at the inputs and initial conditions z 0 , the natural
response is
1
1
1
1
1
1
z t = L sI z 0 = L diag -------------- -------------- -------------- z 0 ,
s

n
1
2

where L f t is the Laplace transform of f t ;

e

i.e.

zt=

1 t

0 0
2 t

t
t t
0 z 0 = diag e 1 e 2 e n z 0 .
0 e

0 e

(3.42)

n t

On expressing the latter equation in terms of the original state variables, but retaining the
decoupled modes, the response becomes
n t
1 t 2 t
x t = V diag e e e W x 0 ,

the right and left modal matrices ( V and W ) being defined in (3.31). An alternative form of
(3.42) is
n

xt=

vh e

h t

wh x 0 .

(3.43)

h=1

If we account for the fact that the inner product of two vectors is a scalar we can rewrite
(3.43) as
n

xt=

v h w h x 0 e

h t

h=1

Let us assume that initial conditions on the states are set equal to the right eigenvector of
the ith eigenvalue, i.e. x 0 = v i . The latter equation becomes

Sec. 3.9

Mode shapes

81
n

xt=

v h w h v i e

h t

(3.44)

h=1

From (3.32), w h v i = 1 when h = i and is zero when h i . Hence the above equation reduces to
x t = vi e

i t

(3.45)

Thus only mode i is excited. Moreover,

i t

x 1 t = v 1i e ,

x k t = v ki e

i t

, ,

x n t = v ni e

i t

(3.46)

Note that each of the modal responses, x 1 t to x n t , has an identical form but their shapes
are determined by the initial amplitude v ki in each response. Thus for a given mode, the
relative amplitudes or shapes of the responses are determined by the associated right eigenvector. Consequently, we can plot the mode shapes for selected modes, these shapes revealing
not only the relative amplitude of the states in the mode, but also the relative phase between
the responses of the states. From (3.46) the relative amplitudes of states at time t are
x 1 : x k , x 2 : x k , ... ,1, ... x n : x k , or v 1i : v ki , v 2i : v ki , ... ,1, ... v ni : v ki ,

(3.47)

where v ki is the element with the largest magnitude among the selected states whose mode
shapes are to be displayed; this result will be employed later in Chapters 9 and 10.
A note of caution. Prior to the development of participation factors (see Section 3.10), the
element v ki of the right eigenvector was employed to determine the involvement of the
state variable x k in mode i. A large relative value of v ki was assessed as representing a significant involvement of x k in the ith mode. However, this is misleading as the numerical values of the elements v ki depend on the units selected (e.g. speed in pu, angle in rad.) for the
associated state variables, i.e. they are not dimensionless, they are scaling dependent. It
should be noted, therefore, that the relative amplitude of the component v jh v mh revealed in (3.47) should not be interpreted as implying the relative participations of states j
and m in mode h. The concept of participation is considered in Section 3.10.
In a practical application, the elements in the right eigenvector corresponding to the speed
states of all generators are selected to reveal the speed mode-shape, say, for an inter-area
mode. For such a mode the relative phase between the speed states reveal, for example, that
machines in areas A and B swing against generators in area C. The plots of mode shapes and
their significance in the analysis of dynamic behaviour of power systems will be discussed in
more detail in Chapters 9 and 10.

82

State equations; eigen-analysis

Ch. 3

The significance of modal response and the mode shapes are most simply illustrated by
means of a numerical example.
3.9.1

Example 3.5: Mode shapes and modal responses

The state matrix of a system is given by A =

0 2 .
4 6

Determine its modal responses and mode shapes.

The eigenvalues of A are evaluated from the characteristic equation, det h I A = 0 ; there
are two eigenvalues, a slower at 1 = 2 and a faster at 2 = 4 .
1 1
. The right eigenvector associated with the eigen2 2

Let the right modal matrix be V =

value 1 = 2 is found from (3.29),

Av h = h v h ,

i.e.

0 2 1 = 2 1
4 6
2
a2

0 2 1 = 4 1
4 6
2
2

Likewise, for 2 = 4 ,

Let 1 = 1 = 1 , then

V =

1 1 =
v1 v2
1 2

and

or

or

2 = 1 .

2 = 2 1 .

W=V

2 1 = w1 .
1 1
w2

The time response, as given by (3.43), is

xt =
x1 t
x2 t

2t
0 2 1 x1 0 ,
1 1 e
4t
1 1
1 2
x2 0
0 e

or

1 2x 0 + x 0 e 2t + 1 x 0 x 0 e 4t .
1
2
1
2
2
1

The form of this response illustrates a number of significant points.

As stated earlier, if the initial value of the states is equal to either of the right eigenvectors,
the associated mode is the only mode present in the response, e.g. for eigenvalue 1,
1 = 2 ,
x1 0
x2 0

1
2

1 ,
1

x1 t
x2 t

1 2x 0 + x 0 e 2t = 1 e 2t .
1
2
1
1

Sec. 3.10

Participation Factors

83

Correspondingly, for eigenvalue 2, 2 = 4 :

x1 0
x2 0

1
2

1 ,
2

x1 t
x2 t

1 e 4t .
2

Note that in each of the two modal responses, the responses x 1 t and x 2 t are related by
a constant factor for all t , i.e. x 1 t = x 2 t for 1, and x 1 t = 1 2 x 2 t for 2. We observe that
(i) for both modes, the responses of the states x 1 t and x 2 t vary in anti-phase;
(ii) for the slow mode 1 = 2 the relative amplitudes of the two states are same, but for
the fast mode 2 = 4 they differ by a factor of two.
Thus, for a given mode, the mode shape reveals not only relative phase between the time responses of the states but also the relative amplitudes of the states in the modal responses; furthermore, the mode shape is determined by the right eigenvector of the associated
eigenvalue.
Further insight into the physical significance of mode shapes is provided in Section 9.2.
---------------------------------

3.10

Participation Factors

We will make fairly extensive use of participation factors later, mainly to determine the degree to which certain states of certain generators or other devices participate in a selected
mode. For example, by examining the speed states of generators, the generators which are
involved in a selected mode of rotor oscillation can be found.
3.10.1 The relative participation of a mode in a selected state
We will analyse participation factors in two stages. In the first it is assumed that the initial
conditions on the states are such that only the k
vector. Then (3.43) becomes:
x1 t

v 11

1 t
x k t = v k1 w 1k e + +

v n1
xn t
th

th

state is excited, i.e. x 0 = e k , the unit

v 1h

v 1n

h t
v kh w hk e + +

v nh

n t
v kn w nk e .

v nn

Note that, although only the k state is excited, all eigenvalues are excited by the unit vector.

84

State equations; eigen-analysis

Ch. 3

Let us consider the state x k t , i.e

x k t = v k1 w 1k e

1 t

= w 1k v k1 e

1 t

+ + v kh w hk e

h t

+ + w hk v kh e

h t

+ v kn w nk e

n t

+ w nk v kn e

n t

(3.48)

The participation of hth eigenvalue in state k is defined as

p hk = w hk v kh = v kh w hk .

Thus, (3.48) becomes x k t = p 1k e

1 t

+ + p hk e

h t

(3.49)

+ p nk e

n t

p jk e

h t

h=1

However, from (3.32) we know

n

p hk = 1 .

w hk v kh =

(3.50)

k=1

k=1

The inner product w h v h (= 1) is dimensionless, therefore the numbers w hk v kh are also dimensionless and are invariant under changes in the units of the state variables of the system.
Hence p hk in (3.49) provides a measure of the relative extent to which hth eigenvalue participates
in state k at time t = 0 ; p hk is therefore known as the participation factor of the hth eigenth

state. (Note that since each mode in the system decays at a different rate the
relative amplitude of each mode in the response (3.48) does change with time.)

value in the k

3.10.2 The relative participation of a state in a selected mode

For the second stage it is assumed that the initial conditions on the states are such that each
state is excited in turn by the unit vector, i.e. x 0 = e j , j = 1 n . Using (3.43) and the
principle of superposition, the form of (3.48) is modified to the following,
for j = 1 :

x 1 t = w 11 v 11 e

for j = k :

x k t = w 1k v k1 e

for j = n :

x n t = w 1n v n1 e

1 t

1 t

1 t

+ + w h1 v 1h e

+ + w hk v kh e

+ + w hn v nh e

h t

h t

h t

+ + w n1 v 1n e

+ + w nk v kn e

n t

n t

+ + w nn v nn e

n t

Sec. 3.11

Eigenvalue sensitivities

85

If we replace w hk v kh by the participation factors, p hk , the coefficients of exponential

terms in the above set of equations can be formed into the following column arrays,
p 11
e

1 t

p 1k ,

p h1
, e

h t

p 1n

p hk ,

p n1
, e

n t

p hn

p nk .

p nn

Consider the hth eigenvalue. According to (3.50), the sum of the participation factors in the
column array p hk is unity. Hence, when each state is excited in turn by the unit vector, the
participation factor p hk also provides a measure of the relative extent to which each of the n
states participates in the hth eigenvalue at time t = 0 .

3.10.3 Example 3.6: Participation factors

In Example 3.4 the right and left modal matrices are
V =

1 1 =
v1 v2
1 2

and

W=

2 1 = w1 ;
w2
1 1

these are associated with the eigenvalues 1 = 2 and 2 = 4 .

The participation factors, p hk = w hk v kh , for states x 1 and x 2 in each of the two eigenvalues are:
p 11 = 2, p 12 = 1 , for 1 = 2 , and p 21 = 1 , p 22 = 2, for 2 = 4 .
Likewise, the participations of the eigenvalues 1 and 2 in each of the two states are:
p 11 = 2, p 21 = 1 , for x 1 , and p 12 = 1 , p 22 = 2, for x 2 .

Note for both eigenvalues the participation factors sum to one.

Further insight into the significance of participation factors is described in Section 9.3.
---------------------------------

3.11

Eigenvalue sensitivities

In the analysis of power system dynamics, it is of interest to assess the effect of the change
of a system parameter, or some element of the state matrix. Later we will need to examine
the effect on certain modes of a change in an element a ij of the state matrix, A . Earlier we
defined the relationship between the state matrix, the eigenvalues and right eigenvectors, i.e.
A vh = h vh .

86

State equations; eigen-analysis

Ch. 3

Differentiation of this expression with respect to a ij yields,

v

v
A
--------- v h + A --------h- = --------h- v h + h --------h- .
a ij
a ij
a ij
a ij

Collecting terms and pre-multiplying by the left eigenvector, we find

v h
w h A h I --------- = w h
a ij

A
--------h- -------- v .
a ij a ij h

Since, by definition w h A = w h h , the left-hand side vanishes. Furthermore w h v h = 1,

and all elements except the ij

th

of A a ij are zero. The sensitivity of the eigenvalue h to

a change in the element a ij is thus

A
--------h- = w hi --------- v jh .
a ij
a ij

(3.51)

In its simplest form A a ij = 1, and the sensitivity is the product of the elements
w hi and v jh of the left and right eigenvectors,

--------h- = w hi v jh .
a ij

(3.52)

In the case i = j , and a ij = a ii , the eigenvalue sensitivity is --------h- = w hi v ih = p hk , i.e. the

a ij

participation factor of the hth eigenvalue in state i .

Note, however, that the element a ij in (3.52) may be a function of a device parameter, q ,
i.e. a ij = f q . Moreover, several elements of A may be functions of the same parameter.
The above analysis can then be extended to evaluate the sensitivity of h to changes in the
parameter q .

Sec. 3.12

3.12

References

87

References

[1]

T. Kailath. Linear Systems, Prentice Hall, 1980.

[2]

R.C. Dorf and R.H. Bishop, Control Systems, 12th edition, Prentice Hall, 2010.

[3]

J. D. Aplevich, Linear State Space Systems, John Wiley & Sons Inc. , 2000.

[4]

P. Kundur, Power system stability and control. New York: McGraw-Hill, 1994.

[5]

Jules Henri Poincar, b. 1853, d. 1912. See web references to Poincar maps and stability.

[6]

B. Stott, Power system dynamic response calculations, Proceedings of the IEEE, vol.
67, pp. 219-241, 1979.

[7]

C. Moler and C. V. Loan, Nineteen Dubious Ways to Compute the Exponential of a Matrix, Twenty-Five Years Later, SIAM Review, vol. 45, pp. 3-49, 2003.

[8]

I. J. Perez-Arriaga, F. L. Pagola, G. C. Verghese, and F. C. Schweppe, Selective Modal Analysis in Power Systems, in American Control Conference, 1983, 1983, pp. 650655.

[9]

I. J. Perez-Arriaga, L. Rouco, F. L. Pagola, and J. L. Sancha, The role of participation

factors in reduced order eigenanalysis of large power systems, in Circuits and Systems, 1988., IEEE International Symposium on, 1988, pp. 923-927 vol.1.

[10]

F. L. Pagola, I. J. Perez-Arriaga, and G. C. Verghese, On sensitivities, residues and

participations: applications to oscillatory stability analysis and control, Power Systems, IEEE Transactions on, vol. 4, pp. 278-285, 1989.

[11]

G. C. Verghese, I. J. Perez-Arriaga, and F. C. Schweppe, Selective Modal Analysis

With Applications to Electric Power Systems, Part II: The Dynamic Stability Problem, Power Apparatus and Systems, IEEE Transactions on, vol. PAS-101, pp. 31263134, 1982.

[12]

L. Rouco and F. L. Pagola, Eigenvalue sensitivities for design of power system

damping controllers, in Decision and Control, 2001. Proceedings of the 40th IEEE Conference on, 2001, pp. 3051-3055 vol.4.

[13]

C. B. Moler and G. W. Stewart, An Algorithm for Generalized Matrix Eigenvalue

Problems, SIAM Journal on Numerical Analysis, vol. 10, pp. 241-256, 1973.

[14]

E. Anderson, Z. Bai, C. Bischof, S. Blackford, et al., LAPACK Users' Guide, Third ed.:
Society for Industrial and Applied Mathematics, 1999.

[15]

H.H. Rosenbrock, State-Space and Multivariable Theory, Nelson, 1970.

[16]

E. J. Davison and S. H. Wang, Properties and calculation of transmission zeros of

linear multivariable systems, Automatica, vol. 10, pp. 643-658, 1974.

88

State equations; eigen-analysis

Ch. 3

[17]

A. J. Laub and B. C. Moore, Calculation of transmission zeros using QZ techniques, Automatica, vol. 14, pp. 557-566, 1978.

[18]

N. Martins, H. J. C. P. Pinto and L. T. G. Lima, Efficient methods for finding transfer function zeros of power systems, Power Systems, IEEE Transactions on, vol. 7, pp.
1350-1361, 1992.

[19]

N. Martins, P. C. Pellanda and J. Rommes, Computation of Transfer Function

Dominant Zeros With Applications to Oscillation Damping Control of Large Power
Systems, Power Systems, IEEE Transactions on, vol. 22, pp. 1657-1664, 2007.

[20]

J. Rommes, Methods for eigenvalue problems with applications in model order reduction,
PhD Thesis, University of Utrecht, 2007.

Chapter 4
Small-signal models of synchronous generators,
FACTS devices and the power system

4.1 Introduction
In this chapter various models of synchronous generators, FACTS devices and of the power
system are developed in forms which are employed in software for the analysis of the smallsignal dynamic performance of multi-machine systems. Small-signal models for the synchronous generator are formulated in Section 4.2. An essential feature of this analysis is that the
higher order coupled-circuit representation of the generator electromagnetic dynamic behaviour is formulated in Section 4.2.3. This is treated as the fundamental model from which
the following two alternative but equivalent formulations of the electromagnetic model are
derived. The first is the Operational Parameter formulation described in Section 4.2.12. The
second, described in Section 4.2.13, is referred to as the Classical Parameter formulation and
is expressed directly in terms of the classically-defined standard parameters of the generator.
The Classical Parameter formulation is presented because it is employed in widely used power system simulation software packages such as Siemens PTI PSSE [1] and GE PSLF [2].
The parameters for the fundamental coupled-circuit formulation are the resistances and inductances of the d- and q-axis circuits. The parameters for the Operational Parameter representation of the electromagnetic equations are the gains and time constants of the transferfunction representations of the respective axes and are collectively referred to as the exactlydefined standard parameters. The Classical Parameter formulation requires the classicallydefined standard parameters. The relationship and conversion between the three parameter
sets are outlined in Section 4.2.14.

89

90

Generators, FACTS devices & system models

Small-signal models of a range of FACTS devices are formulated in Section 4.3 and include
those of the Static VAR Compensator (SVC), Voltage Sourced Converter (VSC), Static Synchronous Compensator (STATCOM), and HVDC transmission links. The general purpose
VSC model formulated in Section 4.3.3 is used as a component in the simplified STATCOM
model in Section 4.3.4 as well as for the rectifier and inverter in the model of the VSC
HVDC transmission link in Section 4.3.7. A general model for a voltage-commutated thyristor-controlled AC/DC converter is formulated Section 4.3.8; this model is then used in a
modular fashion to represent the rectifier and inverter of a line-commutated HVDC transmission link. A methodology to formulate the small-signal equations of the power system is
described in Section 4.4. Finally, in Section 4.5 a general purpose small-signal representation
of a static load model is described.

4.2

Small-signal models of synchronous generators

4.2.1 Structure of the per-unit linearized synchronous generator models

Before developing the details of the per-unit linearized model of the synchronous generator
the overall mathematical structure of the model is described by means of the block diagram
in Figure 4.1. The linearized model is formulated in the rotating direct- and quadrature-axis
(dq) coordinate system in which the d-axis is aligned with the magnetic north pole of the field
winding and the q-axis leads the d-axis by 90 deg. (electrical). The model is linearized about
an initial steady-state operating point that is defined by the initial stator terminal quantities
which are typically obtained from a power flow solution. The calculation of the generator
initial conditions from the specified terminal quantities is described in Section 4.2.9. The
two principal components of the linearized generator model in Figure 4.1 are the electromagnetic (em) equations and the shaft equations of motion. The em equations comprise a
set of differential equations that describe the dynamic characteristics of the d- and q-axis rotor-winding flux linkages together with algebraic equations for the stator voltage components.
As described in Section 4.2.10 the connection of the generator stator to the network requires
the transformation of the perturbations in the stator voltage and current components in the
generator dq coordinate system to the corresponding components in the synchronously rotating RI-network coordinate system1.
There are two control inputs to the generator. The first is the perturbation in the field voltage ( E fd ) developed by the excitation system. As explained in Section 4.2.7 it is important
to note that E fd is expressed in the non-reciprocal per-unit system of the generator field.
The field voltage input in this per-unit system must be converted to the reciprocal per-unit
system which is employed in the formulation of the generator electromagnetic equations.
Similarly, the field current in the reciprocal per-unit system must be converted to the non1. RI refers to the Real and Imaginary components in the network coordinate system.

Sec. 4.2

Structure of per-unit linearized generator models

91

reciprocal per-unit system for use in the excitation system model. The second input is the
perturbation in the mechanical torque ( T m ) developed by the turbine / governor system
which is expressed in per-unit on the generator base value of mechanical torque. Models for
the excitation and turbine / governor systems are not included in this chapter.
The generator models are designated by a code of the form ndmq-c{0,1} in which n and m
are the number of d- and q-axis rotor-windings respectively; c1 and c0 are used to indicate,
respectively, that unequal mutual coupling between the d-axis rotor windings is represented
or neglected. The 1d0q-c0 model comprises three state-variables: the rotor-angle, rotorspeed and d-axis field flux-linkages. This is the basis for the Heffron-Phillips model [3, 4]
that is frequently used for developing concepts for generator controls. It is, however, not
recommended for use in power system analysis. The 3d3q-c1 model, the most complex model considered in this work, comprises a field winding, and two damper windings in the d-axis
and three q-axis damper windings; unequal mutual coupling between the d-axis rotor windings is represented. This model with eight state-variables comprising the six rotor-winding
flux-linkage variables and rotor angle and rotor speed is the most complex model encountered in small-signal analysis of large power systems. The most commonly employed models
in large scale small-signal stability studies are the fifth and sixth order models 2d1q-c0 and
2d2q-c0 in which unequal mutual coupling effects are neglected.
The formulation of the em equations for the 3d3q-c1 model described in Section 4.2.3 is
based on the ideal coupled-circuit representation of the synchronous machine for which the
model parameters are the resistances, mutual and leakage inductances of the windings. As
explained in Section 4.2.4 the em equations developed for this model are readily modified
to represent machine models with fewer damper windings in the respective axes. In particular, the structure of the em equations and their interface with other components in the overall model of the generator are unaffected by changes in the number of damper windings.
The linearized coupled-circuit formulation of the state- and algebraic equations of the complete generator model are given in matrix form in equation (4.117) on page 133 followed in
Table 4.9 by a step-by-step procedure for calculating the associated coefficient matrices.

92

Generators, FACTS devices & system models

Network

On generator per-unit
system
V t P Q I

i RI

v RI

KI

Compute
perturbations in
stator quantities
(Section 4.3.1)

i RI

RI
Transformation from
generator dq to network
RI coordinate system
(Section 4.2.10)

KV

Conversion from generator

(g) to network (n) per-unit
system. (Section 4.2.10)

v RI

R 0 v dq + U 0 v dq0

R 0 i dq + U 0 i dq0

dq
Input Signals
(Includes
Generator
Outputs)

Conversion
between reciprocal (R)
and non-reciprocal
(NR) field winding
per-unit systems
(Section 4.2.7)
E fd

Excitation
System
Model

r fd
--------L ad

e fd

I fd

NR

L ad

i fd
u

i dq

v dq

Electromagnetic Equations (see Note)

x r

A r B ri 0
px r
b
i dq + re e fd
=

C vr D vi I
0
0
v dq

Rotor Equations of Motion (Sections 4.2.5 & 4.2.6)

Turbine /
Governor
Model

Input Signals
(Includes
Generator
Outputs)

T m

0 b
0
0
0
0

p
1 T
D
1

T g + ------0
p = 0 ------- ------- 0
m
2H
2H
2H
i dq
0

0
0
0
1 C tidq C tvdq
v dq

Note: Coupled-circuit formulation of linearized em equations is developed in Section 4.2.11. The Classical Parameter formulation is developed
in Section 4.2.13.4. In the Classical Parameter formulation the conversion
between the reciprocal and non-reciprocal per-unit systems of the field
winding is embedded in the em equations.

Figure 4.1 Structure of the per-unit linearized model of the synchronous generator. (Refer
to Tables 4.3 and 4.4 for descriptions of the parameter and variable symbols in this figure).

Sec. 4.2

Generator modelling assumptions

93

Test procedures that are used to identify synchronous generator models for dynamic analysis commonly employ the Operational Parameter representation of the generator. As explained in Section 4.2.12 this representation comprises three d-axis transfer-functions and
one q-axis transfer-function to completely characterise the machine. The test procedures
identify the gains and time constants of these transfer-functions. These transfer-function
constants are referred to as the standard parameters such as T d0' , T d0'' , L d' , L d'' , etc. In
order to employ the coupled-circuit formulation of the em equations when only the standard
parameters are provided it is necessary to transform the standard parameters to the coupledcircuit parameters as outlined in Section 4.2.14. A troublesome aspect of using the standard
parameters is that over the years two alternative and inconsistent definitions of the parameters have evolved. The Exact definitions correspond to the exact roots of the above transfer-functions. In the Classical definitions the d-axis standard parameters are related to the
parameters of the equivalent circuit of the machine by the classical relationships which are
based on the assumptions that (i) during the transient period the damper winding resistances
are infinite; (ii) during the subtransient period the resistance of the field winding is zero and
the resistances of the second damper winding is infinite; (iii) finally, during the sub-subtransient period the resistances of the field and first damper winding are assumed to be zero. In
the q-axis, analogous assumptions are made to arrive at the classical definitions of the q-axis
standard parameters in terms of the coupled-circuit parameters. It is important to know if
the generator standard parameters that are provided conform to the Exact or Classical
definitions and if necessary to transform them appropriately to suit the requirements of the
simulation model in use. This is particularly important for the q-axis parameters.
The em equations in some widely-used simulation packages are formulated directly in terms
of the classically-defined standard parameters. This is referred to as the Classical Parameter
formulation in this book. It is emphasised that the Classical Parameter formulation is exactly
equivalent to the coupled-circuit formulation provided: (i) that the unequal coupling between the
d-axis rotor windings is neglected, and (ii) that the same method for representing magnetic saturation is employed in the two models.

4.2.2 Generator modelling assumptions

For the purpose of rotor-angle small-signal analysis the following generator modelling assumptions are made:

The d-axis is aligned with magnetic axis of the field winding and the q-axis leads the daxis by 90 degrees (electrical).

The following non-reciprocal Park/Blondel transform [5, 6] is used to transform variables in the stationary abc coordinate system to the rotating dq coordinate system:
cos cos cos +
2 sin sin sin +
2
, = ------ .
T = --3
3
1
1
1
------2
2
2

(4.1)

94

Generators, FACTS devices & system models

In the stator voltage equations the transformer voltage terms, p d and p q , are
omitted because, in the bandwidth of interest, they are negligible in comparison with
the speed voltage terms, d and q [7].

For the purpose of calculating the d- and q-axis stator voltages the perturbation of the
per-unit rotor-speed from the per-unit synchronous speed 0 is assumed to be
negligible and thus, for this purpose, = 0 . Rotor angle and speed perturbations
are, necessarily, represented in the rotor equations of motion.

A consequence of the above assumption is that the per-unit power transferred across
the airgap is independent of perturbations in the rotor speed. Thus, it is also necessary
to assume that the mechanical power developed by the turbine is independent of perturbations in rotor speed.

The generator equations are expressed in per-unit form in which the base quantities
are summarized in Section 4.2.3.2. In particular, the Lad-base reciprocal per-unit system is chosen for the rotor windings [8, 9, 10].

As recommended in IEEE Std. 421.5 [11] a non-reciprocal per-unit system is assumed

to be employed for the representation of generator excitation systems. As explained in
Section 4.2.7 it is therefore necessary to appropriately adjust the scaling of the field
current and voltage at the interface between the generator and excitation system models.

4.2.3 Electromagnetic model in terms of the per-unit coupled-circuit parameters

The development of the per-unit generator equations in the rotating dq coordinate system
from the ideal coupled-circuit representation of the synchronous machine in the stationary
abc coordinate system is provided in reference [12]. In this section the per-unit coupled-circuit formulation of the electromagnetic equations of the machine are listed. These equations
are then linearized about the initial steady-state operating point of the machine in
Section 4.2.11. The Operational Parameter and Classical Parameter formulations for the em
equations are summarized in Sections 4.2.12 and 4.2.13 respectively.
The third to eighth-order coupled-circuit machine models require data in the form of the dand q- axis equivalent circuit winding resistances and leakage and mutual inductances.
Though not commonly used the 3d3q-c1 model has been included because there is evidence
in the literature that this number of rotor circuits may be required to adequately represent
some machines. Test methods have already been developed to identify machine models with
this number of rotor windings, for example [13, 14, 15, .16] Figure 4.2 shows the stator and
rotor winding flux linkages in terms of the winding mutual and leakage inductances and the
winding currents. The unequal mutual coupling between the d-axis rotor windings is repre-

Sec. 4.2

Vector and matrix nomenclature

95

sented by the inductances L c1 and L c2 which are referred to as the Canay inductances. The
shielding of the field winding by the damper windings has been identified in the literature as
important in correctly predicting the field voltage and current [13, 17, 18]. Unequal mutual
coupling between the q-axis rotor windings is not represented since these windings are not
directly observable. The d- and q-axis equivalent circuits for the 3d3q-c1 model are shown
in Figure 4.3.
i fd

Flux linkage contributions:

fl = L fd i fd

e fd

fl

i 1d

f1

2l = L 2d i 2d
l = Ll id

i 2d
2l f12

1d
2d

id
l

Figure 4.2

f12 = L c2 i fd + i 1d + i 2d
ad = L ad i fd + i 1d + i 2d i d

d-axis rotor winding flux linkages:

1l

p d

f1 = L c1 i fd + i 1d

1l = L 1d i 1d

ad

fd = ad + fl + f1 + f12
= L ad + L fd + L c1 + L c2 i fd + L ad + L c1 + L c2 i 1d + L ad + L c2 i 2d L ad i d
1d = ad + 1l + f1 + f12
= L ad + L c1 + L c2 i fd + L ad + L 1d + L c1 + L c2 i 1d + L ad + L c2 i 2d L ad i d
2d = ad + 2l + f12
= L ad + L c2 i fd + L ad + L c2 i 1d + L ad + L 2d + L c2 i 2d L ad i d

d-axis stator winding flux linkages:

d:

d = ad + l
= L ad i fd + i 1d + i 2d L ad + L l i d

Per-unit d-axis flux linkage distribution showing the unequal mutual coupling
between the rotor windings as described by Canay [17, 18].

Importantly, in the following analysis, once the equations for the eighth-order model are defined, the lower-order coupled-circuit models are readily derived. All lower-order coupled-circuit models are formed by deleting the equations and variables associated with those damper
windings to be omitted in the formulation of the simpler model.
Two equivalent approaches to the representation of magnetic saturation are accommodated
in the formulation of the model. The specific details of the non-linear saturation functions
and their linearization are provided in Section 4.2.8.
4.2.3.1

Notes on vector and matrix nomenclature

The nomenclature in the following table applies to matrices and vectors that are extensively
employed in the formulation of the models.

96

Generators, FACTS devices & system models

Table 4.1

Vector and matrix nomenclature

Symbol

Meaning

X
T

x ,X

Denotes a column vector

Denotes a matrix or, depending on the context, a vector.
T

Denotes vector and matrix transposition respectively

Null matrix or vector. The null matrix has zero rows and/
or zero columns. The possibility of a non-zero number of
rows or columns ensures dimensional consistency of
matrix equations. The dimension is to be inferred from
the context.
The identity matrix. The dimension is to be inferred from
the context and in some situations may be the scalar identity (i.e. 1) or the null or empty matrix.
The zero matrix or vector. The dimension is to be
inferred from the context and in some situations may be
the scalar zero (i.e. 0) or the null matrix.
Denotes a column vector whose entries are all ones. The
dimension is to be inferred from the context.

u = 1 1 1

If u is a n x 1 vector then u u is a n x n matrix of

all ones.

D X Y = X 0
0 Y

4.2.3.2

u x is the sum of all the elements in x

Denotes block diagonalization which is the result of

appending the mx x nx matrix X and the my x ny matrix

Y to create the (mx + my) x (nx + ny) matrix as shown. If

X and Y are scalars or diagonal matrices the result is a
diagonal matrix.

Summary of the generator per-unit system

The principal base quantities for the machine are normally V usb (kV), S usb (MVA) and f usb
(Hz) which are respectively values for the stator RMS line-to-line voltage, the stator threephase apparent power and the stator frequency. Usually, but not necessarily, the generator
rated values of these quantities are chosen. Additionally, the base value of time t b is chosen
to be one-second. Finally, the relationship between mechanical and electrical angles requires
knowledge of the number of rotor pole pairs, n pp .

Sec. 4.2

Summary of generator per-unit system

r q

rs

L
-----lb

i fd + i 1d + i 2d
L c2
------ b

id
d
p ------
b

vd

ad
L
------ b

ad
p --------
b

97

i fd + i 1d
c1
L
------ b
1d
L
------ b

2d
L
------ b

i 2d

r d

q
p ------
b

r fd

r fd
e fd = ---------- E fd
L ad
u

r 1d

L
-----lb
iq

vq

fd
L
----- b

i 1d

r 2d

rs

i fd

aq
L
------ b

aq
p --------
b

3q
L
------ b

2q
L
------ b

i 3q
r 3q

1q
L
------ b

i 2q
r 2q

i 1q
r 1q

Figure 4.3 The d- (top) and q-axis (bottom) equivalent circuits for the 3d3q-c1 generator
model represented by three rotor windings in each axis and unequal mutual coupling between the d-axis rotor windings. (Note: inductance & flux-linkage values are scaled by 1 b
because the base value of time is one second).
The base values of the rotor winding currents and flux-linkages are determined such that (i)
the per-unit mutual inductances between all pairs of windings are reciprocal; (ii) the mutual
inductances between all d-axis rotor windings and the stator are equal to the per-unit d-axis
mutual inductance L ad ; (iii) the mutual inductances between all q-axis rotor windings and
the stator are equal to L aq . Furthermore, the base values of rotor-winding voltages are chosen such that the form of the rotor winding voltage equations in SI units and in per-unit are
identical. This choice of base values for the rotor quantities is equivalent to that recommended by Rankin in 1945 [8, 9] and is referred to as the Lad-base reciprocal per-unit system
[10]. It has gained very wide, if not universal, acceptance in the power system analysis field.
On the above basis for the generator per-unit system are derived the base values for the mechanical, stator winding and rotor winding quantities in Table 4.2.

98

Generators, FACTS devices & system models

Table 4.2 Base values for generator quantities (Note: a bar above a quantity (e.g. x )
means the SI value and the subscript b denotes the base value of a quantity.).
Base
Quantity
V usb
S usb
f usb

SI Units

Principal base quantities from which all other base quantities are derived
kV (rms, Arbitrary choice, but usually rated RMS phase-to-phase stator voltage
ph-ph)
(sometimes referred to as VBASE).
Arbitrary choice, but usually three-phase MVA rating of the machine
MVA
(sometimes referred to as MBASE).
Arbitrary choice, but usually rated generator frequency.
(This is not necessarily the same as nominal frequency of the system to
Hz
which the generator is connected. For example, when a generator rated
at 60 Hz is connected to a 50 Hz system or vice-versa).

n pp
tb

Description

Number of pole pairs.

s

Base value of time is chosen to be 1 second.

Derived base quantities
Base value of the time differential operator:
1
d
1d
d
p =
=
= ---- = p b p where p b = ---- = 1 s-1.
tb
d tb t
tb d t
d t

pb

s-1

(elec)
rad/s

v sb

V(peak,
ph-n)

Sb

VA

Machine three-phase VA (apparent power) base: S b = S usb 10 .

Ub

Joules

Base energy: U b = S b t b = S b 1 .

i sb

A (peak,
line)

Stator base current: peak value of line current i sb = 2 3 S b v sb .

Z sb

Stator base resistance / impedance: Z sb = v sb i sb .

L sb

Stator base inductance: L sb = Z sb b = sb i sb .

sb

Wb-turns Stator base flux-linkages: sb = L sb i sb = v sb b .

mb

(mech)
rad/s

Base mechanical rotor speed: mb = b n pp .

T mb

Nm

Base mechanical torque: T mb = S b mb .

Tb

Nm

Base electrical torque: T m = T mb T mb .

Base electrical frequency: b = 2f usb .

Stator base voltage: peak value of phase to neutral voltage
v sb =

2 3 V usb 10 .
6

Sec. 4.2

Parameter and variable definitions

Base
Quantity

SI Units

99

Description
d-axis rotor quantities

(per-unit unsaturated d-axis mutual inductance L ad = L ad L sb is used in the following.)

Base field current in the reciprocal per-unit system of units:

i fdb

e fdb

Base field voltage in the reciprocal per-unit system: e fdb = S b i fdb .

I fdb

Base field current in the non-reciprocal per-unit system:

I fdb = i fdb L ad .

E fdb

i kdb

v kdb

i fdb = L ad sb L afd = L ad L afd i sb .

Base field voltage in the non-reciprocal per-unit system:

E fdb = r fd I fdb , r fd is the field resistance in at the specified temperature.
Base current of the d-axis damper windings:
i kdb = L ad sb L akd = L ad L akd i sb , k = 1 2 .
Base voltage of the d-axis damper windings:
v kdb = S b i kdb , k = 1 2 .

fdb , 1db ,
Base flux-linkages of d-axis rotor windings:
Wb-turns
2db
fdb = v fdb b , 1db = v 1db b & 2db = v 2db b .
r fdb , r 1db ,
r 2db

Base resistance of the d-axis rotor windings:

r fdb = v fdb i fdb , r 1db = v 1db i 1db & r 2db = v 2db i 2db .
q-axis rotor quantities

(per-unit unsaturated q-axis mutual inductance L aq = L aq L sb is used in the following.)

Base current of the q-axis damper windings:

i kqb

v kqb

kqb

Base flux-linkages of the q-axis damper windings:

Wb-turns
kqb = v kqb b , k = 1 3 .

r kqb

4.2.3.3

i kqb = L aq sb L akq = L ad L akq i sb , k = 1 3 .

Base voltage of the q-axis damper windings:
v kqb = S b i kqb , k = 1 3 .

Base resistance of the q-axis damper windings:

r kqb = v kqb i kqb , k = 1 3 .

Parameter and variable definitions

The parameters and variables used in the formulation of the model are listed in Tables 4.3
and 4.4 respectively together with their base values as defined in Table 4.2.

100

Generators, FACTS devices & system models

Table 4.3 Summary of the parameters in the per-unit coupled-circuit
representation of the 3d3q-c1 synchronous machine model.

Per-unit Parameter

Base Value
(see Tab. 4.2)

n/a

2U b b

T mb mb ,

Tb b

rs

Z sb

Ll

L sb

L ad , L aq
u
u

L sb

L adq = L ad L aq

r fd , r 1d , r 2d

L sb
r fdb , r 1db ,
r 2db

L fd , L 1d , L 2d

L c1

See Note (1)

The base frequency (elec. rad/s) which appears explicitly in the per-unit equations due to the choice of one
second as the base value of time.
Aggregate inertia constant of the generating unit.
Refer to Appendix 4II.2 for derivation.
Aggregate incremental mechanical damping torque
coefficient of the generating unit. Refer to
Appendix 4II.2 for derivation.
Stator resistance, assumed identical in the d- and qaxes.
Stator leakage inductance, assumed identical in the dand q- axes.
Respectively the d- and q-axis unsaturated airgap
mutual inductance between the corresponding stator
and rotor windings.
The operating point dependent values of the d- and qaxis mutual inductances. (Note: L adq may be a varia
ble depending on the method used to represent magnetic saturation).
Resistances of the field winding and the first and second d-axis damper windings respectively.
Leakage inductances of the field winding and the first
and second d-axis damper windings respectively.
These inductances represent flux that links only their
respective windings.
Mutual inductance between the field and first damper
winding which represents flux linkages between these
windings but which do not link the stator or the second damper winding. To neglect unequal coupling
between the d-axis rotor windings L c1 = L c2 = 0 .
Mutual inductance between the three d-axis rotor
windings which represents flux that links all three daxis rotor windings but not the stator.

L c2
r 1q , r 2q , r 3q

Description

r 1qb , r 1qb ,
r 2qb

Resistances of the three q-axis damper windings.

Sec. 4.2

Parameter and variable definitions

Per-unit Parameter

L 1q , L 2q , L 3q

Base Value
(see Tab. 4.2)
See Note (2)

101

Description
Leakage inductances of the three q-axis damper windings. Note that unequal coupling between the q-axis
rotor windings is not represented since the q-axis is
observable only from the stator.

(1) The d-axis rotor-winding per-unit inductance matrix L rd = L ad uu T + Llrd = -----b- i rdb Lrd i rdb is
Sb
defined in terms of the corresponding matrix L rd in terms of SI units where the d-axis rotorL ffd L f1d L f2d
winding leakage inductance matrix Llrd is defined in (4.19), L rd = L f1d L 11d L 12d (H) and
L f2d L 12d L 22d
i rdb = D i fdb i 1db i 2db .

(2) The q-axis rotor-winding per-unit inductance matrix L rq = L aq uu T + Llrq = -----b- i rdq Lrq i rqb is
Sb
defined in terms of the corresponding matrix L rq in terms of SI units where the q-axis rotor-windL 11q L 12q L 13q
ing leakage inductance matrix L lrq is defined in (4.20), L rq = L 12q L 22q L 23q (H) and
L 13q L 23q L 33q
i rqb = D i 1qb i 2qb i 3qb .

In the literature on models of generators reference is often made to per-unit machine reactances (e.g. X d , X ad , etc.) rather than per-unit machine inductances (e.g. L d , L ad ). In this
book we adopt per-unit machine inductances. It should be noted that in the per-unit system
used the values of per-unit reactances and inductances can be used interchangeably.

102

Generators, FACTS devices & system models

Table 4.4

Summary of variables in the per-unit coupled-circuit representation of the

3d3q-c1 synchronous machine.

Variable (in per-unit)

T

v dq = v d v q

i = id iq
dq

adq = ad aq

dq = d q

rd = fd 1d 2d

rq = 1q 2q 3q

v rd = e fd 0 0

v rq = 0 0 0

d- and q-axis stator terminal voltage respectively.

i sb

d- and q-axis stator winding current respectively. Direction of positive stator current is
from the generator into the network.

sb

d- and q-axis airgap flux linkages respectively.

sb

d- and q-axis stator flux linkages respectively.

1qb 2qb 3qb

i fdb i 1db i 2db

fdb 1db 2db

v 1qb v 2qb v 3qb

T

Description

v sb

e fdb v 1db v 2db

i = i fd i 1d i 2d
rd

Base Value
(see sec. Tab. 4.2)

d-axis rotor flux linkages in which subscripts

fd refers to the field winding and 1d and 2d
refer respectively to the first and second d-axis
damper windings.
q-axis rotor flux linkages in which the subscripts 1q, 2q and 3q refer respectively to
the three q-axis damper windings.
d-axis rotor-winding voltages in which e fd is
the per-unit field winding voltage in the reciprocal base system. The damper windings are
short-circuit so their voltages are zero.
q-axis rotor winding voltages are identically
zero since the damper windings are short-circuit.
d-axis rotor winding currents. i fd is the perunit field winding current in the reciprocal
base system.

q-axis rotor winding currents.

i = i 1q i 2q i 3q
i 1qb i 2qb i 3qb
rq
Note: For compactness the d- and q-axis rotor winding variables are aggregated as follows:

I fd , E fd

v
i
rdq = rd , v rdq = rd , i rdq = rd .

rq
v rq
i

rq

Per-unit field current and voltage respectively

I fdb , E fdb
in the non-reciprocal base system.

Sec. 4.2

Summary of the generator coupled-circuit equations

Variable (in per-unit)

i
= i sd i sq
sdq

I sdq = I sd I sq

I sd = L ad i
u

sd

I sq = L aq i
u sq

Base Value
(see sec. Tab. 4.2)

i fdb i 1qb

where

i fdb L ad

i 1qb L aq

103

Description
The demagnetizing components of the d- and
q-axis excitation current that is required to
account for the effects of magnetic saturation
in the respective axes. This saturation excitation current is incorporated in the model only
if the second method of saturation modelling
in Section 4.2.8.2 is employed.
As above, but the non-reciprocal per-unit system is employed. This representation of the
demagnetizing effects of magnetic saturation is
employed in the Classical Parameter formulation of the em equations in Section 4.2.13.
Relative rotor angle being the angular position
of the d-axis with respect to the synchronously
rotating network reference (in elec. rad).
Stationary rotor angle being the angular position of the d-axis with respect to a stationary
reference (in elec. rad).

(elec. rad)

(elec. rad)

Rotor-speed.

Synchronous speed. Note, if the nominal system frequency is equal to the generator base
frequency then 0 = 1 .

Pe

Sb

Electrical power output.

Tg

T mb

Electromagnetic (or airgap) torque.

4.2.3.4
tions

Summary of the coupled-circuit formulation of the generator electromagnetic equa-

Summarized below are the per-unit coupled-circuit equations describing the electromagnetic behaviour of the generator in the rotating dq coordinate system. These equations are developed from first principles in [12].
The d- and q-axis rotor-winding voltage equations are respectively:
r fd 0 0 i fd
fd
1
1
------ p
= 0 e fd 0 r 1d 0 i 1d and
1d
b
0
2d
0 0 r 2d i 2d

(4.2)

104

Generators, FACTS devices & system models

1q
r 1q
1----p 2q = 0
b
3q
0

r 2q 0

i 1q
i 2q ,

(4.3)

0 r 3q i 3q

These equations are conveniently expressed in the following compact forms:

p rd = b red e fd b r rd i rd in which r rd = D r fd r 1d r 2d ; b red = b 0 0

and p rq = b r rq i rq in which r rq = D r 1q r 2q r 3q .

(4.4)
(4.5)

The above equations are combined as follows.

Rotor Voltage Equations
p rdq = b r rdq i rdq + b re e fd ,

(4.6)

in which r rdq = D r rd r rq and

(4.7)

b re =

b red
0

where b red = b 0 0 .

(4.8)

The per-unit flux linkage equations for the d-axis are presented in terms of the winding mutual and leakage inductances, and the winding currents based on Figure 4.2 on page 95. As
mentioned earlier, the flux linkages of the q-axis windings neglect unequal coupling between
the q-axis rotor windings. This is valid because the q-axis is observable only from the stator.
Consequently an equivalent circuit that assumes equal coupling between the q-axis rotor
windings can be identified that represents the observable q-axis behaviour.
Referring to Figure 4.2 the d-axis mutual (or airgap) flux linkages are:
T

ad = L ad i fd + i 1d + i 2d i d = L ad u i rd i d , where u = 1 1 1

and analogously the q-axis mutual flux linkages are:

(4.9)

(4.10)
aq = L aq i 1q + i 2q + i 3q i q = L aq u i rq i q .

The values of L ad and L aq depend on the method used to represent magnetic saturation.
Two mathematically equivalent methods for representing magnetic saturation are considered. In the following, the method is denoted by the parameter s m = 1 if the first method
is being used, or s m = 2 for the second method. If magnetic saturation is to be neglected
then s m = 0 .

Sec. 4.2

Summary of the generator coupled-circuit equations

105

The first method for representing magnetic saturation ( s m = 1 ) is described in detail in

Section 4.2.8.1. In this method it is assumed that the airgap mutual inductances are non-linear functions of the airgap flux linkages whereas the leakage inductances in both axes and
the unequal mutual inductances between the d-axis rotor windings are assumed to independent of the fluxes linking them and are therefore constant parameters. Thus, in this method:
L ad = L ad ad aq , L aq = L aq ad aq .

(4.11)

In the second method ( s m = 2 ), for which the details are provided in Section 4.2.8.2, the
unsaturated values of the airgap mutual inductances are retained; instead the components of
the excitation current, i sd and i sq , necessary to represent the demagnetizing effect of saturation in the respective axes, are deducted from the excitation current in equations (4.9) and
(4.10) respectively, i.e.
ad

aq

L ad 0 u T 0
0 L aq 0 u T

i
i
i
rd d s 2 sd
i
iq
i sq
rq

1 if s m = 2

where s 2 =

(4.12)

(4.13)

0 otherwise
If method 2 is being used to represent magnetic saturation then the unsaturated values of the airgap mutual inductances are used in the equations (i.e. L ad = L ad and L aq = L aq ).
u

As explained in Section 4.2.8.2 the demagnetizing currents which account for the effect of
saturation are non-linear functions of the airgap flux linkages:
i sd = i sd ad aq , i sq = i sq ad aq .

(4.14)

Defining u 2 = D u u results in the following compact matrix equation for the d- and qaxis airgap flux linkages:
Airgap Mutual Flux Linkage Equations
T

adq = L adq u 2 i rdq i dq s 2 i sdq , in which

(4.15)

D L L if s = 1
ad ad aq
aq ad aq
m

L adq =
D L ad L aq = L adq otherwise
u
u
u

(4.16)

1 if s m = 2

and s 2 =

0 otherwise

(4.17)

106

Generators, FACTS devices & system models

The d- and q-axis rotor winding flux-linkages are expressed in terms of the mutual and leakage flux linkages as follows:
rd = ad u + L lrd i rd and rq = aq u + L lrq i rq .

in which the d- and q-axis rotor leakage inductance matrices are respectively:

L lrd =

L fd + L c1 + L c2

L c1 + L c2

L c2

L c1 + L c2

L 1d + L c1 + L c2

L c2

L c2

L c2

L 2d + L c2

and

L lrq = D L 1q L 2q L 3q .

(4.18)

(4.19)

(4.20)

From (4.18) the rotor winding flux linkage equations are written in the following compact
form:
Rotor winding flux-linkage equations
rdq = u 2 adq + L lrdq i rdq in which L lrdq = D L lrd L lrq .

(4.21)

The stator winding flux linkage equations are:

d

ad
aq

Ll 0

id

0 Ll iq

(4.22)

which have the following compact form:

Stator winding flux-linkage
equations

(4.23)

dq = adq L l i dq

The generator d- and q-axis stator voltage equations in which, consistently with the modelling assumptions in Section 4.2.2, both the transformer voltages and the rotor-speed perturbations are neglected are:
vd
vq

d
rs 0 id
+ 0 0 1
,
1 0 q
0 rs iq

which are expressed in the following compact form:

(4.24)

Sec. 4.2

Linearized coupled-circuit equations

107

Stator winding voltage equations

v dq = r s i dq + 0 W dq dq where W dq = 0 1 .

1 0

(4.25)

The electromagnetic behaviour of the machine is characterised by the differential equations

for the rotor winding voltages (4.6) and the algebraic equations for the airgap flux linkages
(4.15), the rotor winding flux linkages (4.21), the stator winding flux linkages (4.23) and the
stator voltages (4.25).
4.2.3.5

Linearization of the coupled-circuit formulation of the electromagnetic equations

The coupled-circuit formulation of the generator electromagnetic equations in the previous

section are now linearized about the initial steady-state operating point of the machine. The
procedure for determining the initial steady-state values of the generator quantities from the
power flow solution of the generator stator terminal quantities is given in Section 4.2.9. All
of the electromagnetic equations are linear, apart from the airgap flux-linkage equations. The
linear equations are linearized trivially by replacing the variables with their perturbed values
(i.e. replace x with x ).
Equation (4.15) for the airgap mutual flux-linkages is firstly linearized for the case in which
saturation method 1 (i.e. s m = 1 ) is used. The initial steady-state values of the saturated airgap mutual inductances are L ad L aq and the corresponding flux linkages are
0

ad aq . The perturbations in the airgap flux linkages are:

0
0
T

adq = L adq u 2 i rdq i dq + s 1 C aldq L adq ,

0
0

in which L adq = D L ad L aq ,
0

ad aq
C aldq = D -----------0 -----------0 and
0
L ad 0 L aq 0

1 if s m = 1
.
s1 =
0 otherwise

(4.26)
(4.27)

(4.28)

If saturation method 2 (i.e. s m = 2 ) is used then the linearized airgap flux linkage equations
are:
T

adq = L adq u 2 i rdq i dq s 2 i sdq .

u

(4.29)

It is noted that the two methods of representing magnetic saturation are mutually exclusive.
Thus, it is convenient to define the saturation variable z s depending on the method used
to represent magnetic saturation:

108

Generators, FACTS devices & system models

L adq if s m = 1

z s = i sdq if s m = 2

if s m = 0

(4.30)

As mentioned earlier the definition of the airgap mutual inductance matrix also depends on
which saturation method is employed, as formalized by the following definition:
L
= D L ad L aq if s m = 1
adq 0
0
0
L adq =
L adq = D L ad L aq otherwise
u
u
u

(4.31)

From equations (4.26) and (4.29) the saturation coefficient matrix C asdq is defined as:
0

C aldq 0 if s m = 1

C asdq = L adq if s m = 2
0

if s m = 0

(4.32)

Thus the linearized matrix equation for the airgap mutual flux linkages, which is applicable
to either method of representing magnetic saturation, is:
T

adq = L adq u 2 i rdq i dq + C asdq z s .

0

(4.33)

It will be shown in Section 4.2.8 that the perturbation in the saturation variable is related to
perturbations in the airgap mutual flux linkages by the operating point dependent matrix
C sadq as follows:
0

z s = C sadq adq ,
0

(4.34)

C
if s m = 1
ladq0

in which C sadq = C madq0 if s m = 2

0

if s m = 0

(4.35)

where C ladq is defined in (4.73) on page 123 and C madq is defined in (4.80) on page 124.
0

Thus, from equations (4.6) on page 104, (4.21), (4.23), (4.25), (4.33) and (4.34) the differential and algebraic equations describing the electromagnetic behaviour of the machine are linearized about the initial steady-state operating point of the machine to yield:

Sec. 4.2

Linearized coupled-circuit equations

109

p rdq = b r rdq i rdq + b re e fd

0 = u 2 adq + L lrdq i rdq rdq

T
0 = L adq u 2 i rdq i dq + C asdq z s adq
0

0 = adq L l i dq dq

0 = C sadq adq z s
0

0 = r s i dq + 0 W dq dq v dq

(4.36)

The above equations are rewritten in the following matrix form:

rdq

p rdq
b re
i rdq
I L lrdq
u2
0
0
0
0

0
adq
0
T

0 L adq u 2
I
0
C asdq L adq 0
0
0
=
dq + 0 e fd
0

0
0
0
I
I
0
Ll I 0
z s
0
0

0
I
0
0
0
0
C sadq
0
i
0
0
dq
0
0
0
0 W dq
0
rs I I
v dq

0 b r rdq

(4.37)

In order to consolidate the structure of the above equations it is convenient to define the
following consolidated vector of n z algebraic variables:
i rdq

adq
z e =

dq

z s

and the associated consolidated matrix coefficients:

(4.38)

110

Generators, FACTS devices & system models

L lrdq
u2
I
T
L adq u 2 I
C re = b r 0 0 0 , C er = 0 , C ee =
rdq
0
0
I
0
0
C sadq
0

C ei =

0
L adq
Ll I

, C ve = 0 0 0 W dq 0

0 C asdq
0 ,
I
0
0
I

(4.39)

Substitution of the quantities in (4.38) and (4.39) into (4.37) results in the following compact
form of the linearized electromagnetic equations:
0 C re 0 0
p rdq

= C er C ee C ei 0
0
0 C ve r s I I
0

rdq

b re
z e
+ 0 e fd

i dq

0
v dq

(4.40)

These equations preserve the important structural characteristics of the underlying coupledcircuit model and the elements in the coefficient matrices have a simple form. For these and
other reasons it may be desirable to implement the generator model in the form of (4.40)
without eliminating the internal algebraic variables z e . However, it is also straight forward

to eliminate z e from (4.40) as follows:

A r B ri 0
p rdq
=

C vr D vi I
0

rdq

b
i dq + re e fd

0
v dq

(4.41)

in which A r = C re J er , B ri = C re J ei , C vr = C ve J er and D vi = C ve J ei r s I , (4.42)

1

where J er = C ee C er and J ei = C ee C ei .

(4.43)

The perturbations in the algebraic variables are given by:

z e = J er rdq + J ei i dq .

(4.44)

Suppose a subset of the generator algebraic-variables in (4.38) is required as output variables

y g from the generator. Let k i , i = 1 n y be the index in z e of the ith output variable.

Then,
y g = S gz z e ,

(4.45)

Sec. 4.2

Alternative d- and q-axis rotor structures

in which S gz

ny x nz

111

, S gz i k i = 1 , i = 1 n y and all other elements of S gz are ze-

ro. Substituting for z e from (4.44) into (4.45) yields the following output equation in terms

of the generator state-variables and stator current components:

y g = S gz J er rdq + S gz J ei i dq .

(4.46)

For example, in order to monitor (i) the generator field-current set k i = 1 , or (ii) the d-axis
airgap flux-linkages set k i = n d + n q + 1 .
The equations for perturbations in AC terminal quantities such as the voltage magnitude and
angle, real and reactive power output, current magnitude, are formulated for FACTS Devices in Section 4.3.1. These equations, which are calculated in terms of the perturbations in
the voltage and current components in the RI network frame of reference are also applicable
to the generator stator terminal.
4.2.4 Alternative d- and q-axis rotor structures
The model development so far has been based on a representation with three rotor windings
in each axis and in which the unequal mutual coupling between the d-axis rotor windings is
represented by means of the Canay inductances (i.e. model 3d3q-c1). A suite of simpler model structures that are commonly employed in practice is readily extracted from the 3d3q-c1
model. The modifications to the rotor winding variables and parameters required to represent a range of alternative rotor structures are summarized in the following. Once these
modifications are made the formulation of the linearized equations of the generator in matrix form proceeds independently of the number of rotor windings.
Models of d-axis rotor structures with a field winding and respectively nkd = 0, 1 and 2
damper windings for a total of nd = 1, 2 and 3 rotor windings are developed. Cases with two
or three d-axis rotor windings represented are provided which either neglect (c = 0) or include (c = 1) unequal coupling between the rotor windings. Definitions are given in
Table 4.5 for the d-axis rotor-winding variables rd and i rd , the associated resistance and

leakage inductance parameters r rd and L lrd , and the input matrix b red for the resulting five
alternative structures.

112

Generators, FACTS devices & system models

Table 4.5 Summary of d-axis rotor variables and parameters with one, two or three
rotor windings and with the inclusion (c = 1) or exclusion (c = 0) of unequal coupling
between the rotor windings.

nd

rd

i
rd

b red

r rd

L lrd

fd

i fd

r fd

L fd

D L fd L 1d

fd

i fd

1d

i 1d

fd

i fd

1d

i 1d

2d

i 2d

2
3

D r fd r 1d

4
5

L fd + L c1

L c1

L c1

L 1d + L c1

0
3

D r fd r 1d r 2d

0
0

D L fd L 1d L 2d
L fd + L c1 + L c2

L c1 + L c2

L c2

L c1 + L c2

L 1d + L c1 + L c2

L c2

L c2

L c2

L 2d + L c2

Note that the leakage inductance matrices for rotor structures #2 and #4 which neglect unequal coupling between the rotor windings (i.e. c = 0) can be obtained from the corresponding matrices for rotor structures #3 and #5 (i.e. c = 1) by setting, in the latter structures, the
rotor mutual inductances L c1 and L c2 to zero.
Table 4.6 presents definitions of the q-axis rotor-variables rq , i rq and the associated re

sistance and inductance parameters r rq and L lrq for four q-axis rotor structures with
nq = 0, 1, 2 or 3 damper windings.
Table 4.6

Summary of q-axis rotor variables and parameters with zero to three rotor
windings.

nq

rq

i
rq

r rq

L lrq

1q

i 1q

r 1q

L 1q

1q 2q

D r 1q r 2q

D L 1q L 2q

1q 2q 3q

D r 1q r 2q r 3q

D L 1q L 2q L 3q

i 1q i 2q
T

i 1q i 2q i 3q

Sec. 4.2

Per-unit electrical torque and power

113

4.2.5 Per-unit electromagnetic torque and electrical power output

The per-unit equations for the electromagnetic torque and electrical power output of the
machine presented below are developed from first principles in [12].
Under balanced conditions the non-linear per-unit equations for the electrical power output
and electromagnetic torque are respectively:
P e = i d v d + i q v q and T g = d i q q i d .

(4.47)

It is instructive to express the per-unit generator power output in terms of the stator currents
and flux linkages by eliminating from equation (4.47) the stator voltages by substitution
from equation (4.24) on page 106 to yield:
2

Pe = 0 d iq q id rs id + iq .

(4.48)

It is important to recall at this point that (4.24) neglects the transformer voltage terms and
rotor-speed perturbations in the speed voltage terms. Therefore the electrical power equation in (4.48) also neglects these effects.
Then, substituting for d i q q i d in the preceding equation from (4.47) gives the following relationship between the per-unit electrical power and electromagnetic torque:
2

Tg = Pe + rs id + iq 0 .

(4.49)

Substituting for P e from (4.47) into (4.49) gives the electromagnetic torque in terms of the
stator voltage and current components:
2

Tg = id vd + iq vq + rs id + iq 0 .

(4.50)

The airgap power is the electrical power output of the generator inclusive of the stator resistive losses:
2

P g = P e + r s i d + i q , and thus

(4.51)

Pg = 0 Tg .

(4.52)

This confirms that the relationship between the airgap power and torque neglects the perturbations in rotor speed which occurs as a consequence of neglecting perturbations in rotor
speed in the calculation of the stator voltage.
The electromagnetic torque equation (4.50) is linearized about the initial steady-state operating point v d v q i d i q to yield:
0

T g = C tvdq v dq + C tidq i dq ,

(4.53)

114

Generators, FACTS devices & system models

1
in which C tvdq = ------ i d i q
0

1
and C tidq = ------ v d + 2r s i d
0

v q + 2r s i q .
0

(4.54)

Note that if, as normally is the case, the per-unit synchronous speed 0 = 1 then from
(4.49) the per-unit electrical power output of the generator is equal to the electromagnetic
torque less the resistive losses in the stator winding.
An essential point, that is overlooked in some commercial software packages, is that when
calculating the mechanical torque developed by the turbine from the mechanical power it
necessary to neglect perturbations in the rotor speed. That is:
Pm = 0 Tm .

(4.55)

This is to be consistent with neglecting the rotor speed perturbations in the relationship between airgap power and torque revealed in equation (4.52). It is shown in Appendix 4II.5
that if the relationship between mechanical power and torque does include rotor speed perturbations then the effect is to erroneously increase the generator damping constant D by
2

P m 0 per-unit.
0

4.2.6 Per-unit rotor equations of motion

The per-unit rotor equations of motion in which the rotor-position, , is measured with respect to the synchronously rotating network reference frame (see Section 4.2.10 and
Appendix 4II) are:
p = b 0 and

(4.56)

1
p = ------- T m T g D 0 .
2H

(4.57)

in which T g is defined in (4.47) and equivalently in (4.50). The mechanical torque T m developed by the turbine is treated as a generator model input.
The equations of motion are linearized about the steady-state operating point, in which it is
assumed that the machine is rotating synchronously with the network reference frame at a
speed = 0 per-unit to yield:
p = b and

(4.58)

1
p = ------- T m T g D .
2H

(4.59)

Substituting the expression for the perturbation in the electromagnetic torque T g from
(4.53) into the preceding equations yields the following formulation of the shaft acceleration

Sec. 4.2

Definition of non-reciprocal per-unit field voltage and current

115

equation in terms of perturbations in the rotor-speed, the stator voltage and current components and mechanical torque:
D
1
1
1
(4.60)
p = ------- ------- C tvdq v dq ------- C tidq i dq + ------- T m .
2H
2H
2H
2H

Importantly, the above formulation is independent of the model employed to represent the rotorwinding structure.

To facilitate analysis in later chapters an alternative formulation of the linearized acceleration

equation is now developed in terms of the perturbation of acceleration power. Multiplying
the acceleration equation (4.57) by the per-unit rotor-speed yields:
1
p = ------- P m P g D 0 .
2H

(4.61)

Substituting for P g from (4.51) in the preceding equation gives:

2
2
1
p = ------- P m P e D 0 r s i d + i q .
2H

(4.62)

Linearizing the preceding equation yields:

1
p = -------------- P m P e D 0 2r s i d i d + i q i q .
0
0
2H 0

(4.63)

If, as is normally the case 0 = 1 and if r s 0 , it follows that:

1
p = ------- P m P e D .
2H

(4.64)

In the preceding equation 2H is sometimes replaced by M = 2H .

4.2.7 Non-reciprocal definition of the per-unit field voltage and current
Although the Lad-base reciprocal per-unit system has a number of advantages from the perspective of representing the generator it is usually the case that a different per-unit system,
referred to as the non-reciprocal or unity-slope per-unit system, is used when representing the excitation system of the generator. Thus, it is necessary to establish the relationship
between these two per-unit systems for the purpose of interfacing between the field winding
of the generator and excitation system models.
To begin, consider the generator represented by the per-unit equations in the reciprocal perunit system and neglecting magnetic saturation. Suppose that the generator is on open-circuit and rotating steadily at one per-unit speed. Under this steady-state condition
i d = i q = 0 , = 0 = 1 and the rates of change of all variables in the dq coordinate system are zero. From the d-axis rotor voltage equations (4.2) on page 103 it is deduced that
e fd = r fd i fd and the d-axis damper winding currents are zero (i.e. i 1d = i 2d = 0 ). From the
corresponding q-axis equations (4.3) the q-axis damper winding currents are also found to
be zero (i.e. i rq = 0 ). Given these initial values of the winding currents it is deduced from

116

Generators, FACTS devices & system models

the d- and q-axis airgap flux-linkage equations (4.15) on page 105 and the d- and q-axis fluxlinkage equations (4.23) on page 106 that d = L ad i fd and q = 0 . From the d- and q-axis
u

stator voltage equations (4.24) it follows that v d = q = 0 and v q = d = L ad i fd . If the

u

field current in the reciprocal per-unit system is one per-unit then the stator voltage is L ad

per-unit (neglecting saturation) and the field voltage is e fd = r fd per-unit.

Thus, an equivalent way of defining the base field current in the Lad-base reciprocal per-unit
system is:
The base field current i fdb in the reciprocal per-unit system is that field
current, in Amperes, which is required to generate L ad per-unit stator
u

voltage on the airgap line when the machine is open-circuit and rotating
steadily at one per-unit speed. The base field voltage e fdb is the corresponding field voltage in Volts divided by the per-unit field winding
resistance 1 at the specified field winding temperature.
The above reciprocal definition of the base field current and voltage is not consistent with
the non-reciprocal definition of the base values of the field quantities which is recommended in Annex B of IEEE Std. 421.5 [11] for the modelling of excitation systems. The following definition of the non-reciprocal per-unit system for the field current and voltage is
consistent with that given in IEEE Std. 421.5. Note that in the reciprocal per-unit system
quantities related to the field current and voltage are denoted by lower case i and e respectively whereas the corresponding quantities in the non-reciprocal per-unit system are
denoted by upper-case I and E.
The base field current I fdb in the non-reciprocal per-unit system is that
field current, in Amperes, which is required to generate 1.0 per-unit stator voltage on the airgap line when the machine is open-circuit and rotating steadily at one per-unit speed. The base field voltage E fdb in this
per-unit system is the field voltage in Volts, corrected to the specified
field winding temperature, required to generate the base field current
I fdb .

1. From Table 4.2, the per-unit field resistance is r fd = r fd r fdb where r fd is the field
2

resistance in ohms and r fdb = S b i fdb (ohm) is the base value of field resistance in the
reciprocal per-unit system.

Sec. 4.2

Definition of non-reciprocal per-unit field voltage and current

117

The above definitions lead to the following mathematical conversions between the per-unit
field current I fd (and voltage E fd ) in the non-reciprocal per-unit system and the corresponding value of i fd ( e fd ) in the reciprocal per-unit system:
L ad
I fd = L ad i fd and E fd = ----------u e fd .
u
r fd

(4.65)

Opencircuit stator voltage (pu)

and fluxlinkages (pu)

The conversion between the reciprocal and non-reciprocal definitions of the field current is
shown graphically in the generator open-circuit characteristic in Figure 4.4 in which three
field current scales are shown: (i) Amperes, (ii) per-unit on the reciprocal base system; and
(iii) per-unit on the non-reciprocal base system.

Airgap
line

adu

O.C.C.
1.0
Reciprocal
Nonreciprocal

Ifdb

Ifd (A)

fdb

= Ladu Ifdb
1/L

adu

1.0

1.0
Ladu

ifd (pu)
(reciprocal)
Ifd (pu)
(nonreciprocal)

Figure 4.4 Generator open-circuit characteristic with the field current scaled in Amperes,
and in per-unit according to the reciprocal and non-reciprocal per-unit systems.
The scaling required at the interface between the model of the exciter and the generator field
winding is depicted in Figure 4.5. It is assumed that the output from the exciter is the field
voltage in per-unit in the non-reciprocal per-unit system and the input to the generator is
the per-unit field voltage in the reciprocal system. It is assumed that the generator per-unit
field current in the reciprocal system is, from model signal flow perspective, an output signal
from the generator which is input to the model of the exciter in per-unit in the non-reciprocal per-unit system.

118

Generators, FACTS devices & system models

E fd

Exciter
Model

r fd
--------L ad

e fd

I fd

L ad

i fd

Generator
Model

Figure 4.5 Interface between the generator and exciter model taking account of the conversion between the reciprocal and non-reciprocal per-unit systems in the
respective models.
In the reciprocal per-unit system it follows from (4.2) on page 103 that under steady-state
condition e fd = r fd i fd . By applying the conversion in (4.65) to this relationship it follows
0

that in the non-reciprocal per-unit system the steady-state value of the field voltage and current are equal (i.e. E fd = I fd ).
0

Some of the reasons why the non-reciprocal per-unit system is preferred [10] are:

The measured generator open-circuit characteristic (O.C.C.) rarely extends beyond a

stator voltage of 1.1 per-unit and never to L ad . Thus, direct graphical determination
u

of the base field current I fdb from the measured O.C.C. is straight-forward in the
non-reciprocal system whereas supplementary calculation is required to determine the
base field current i fdb in the reciprocal system.

The numerical value of per-unit field-voltage e fd in the reciprocal per-unit system is

very small whereas, under steady-state conditions, E fd = I fd in the non-reciprocal system.

Although IEEE Std. 421.5 recommends the use of the non-reciprocal per-unit system for
modelling of the excitation system, it is sometimes the case that vendors or testing contractors provide excitation system model parameters on a different per-unit system. For example, sometimes the base value of field current is defined as that field current, in Amperes,
that is required to produce rated stator voltage when the generator is operating at rated output and frequency. Therefore, it is essential that those who are entering data into simulation
programs understand the basis on which model parameters are supplied and, if necessary,
adjust parameter values to comply with the per-unit system assumed by the simulation program being used.
4.2.8 Modelling generator saturation
Before proceeding further the two methods for representing the effects of magnetic saturation introduced in Section 4.2.3.4 are described in some detail. It is emphasised that these two

Sec. 4.2

Modelling generator saturation

119

methods are strictly equivalent and yield identical results. Both methods are employed in dif-

ferent simulation packages. In the first method the airgap mutual inductances in the respective axes are assumed to be subject to magnetic saturation. The objective in this case is to
show that the perturbations in these inductances, L ad and L aq , about their steady-state
saturated values L ad and L aq can be expressed in terms of the perturbations in the airgap
0

flux linkage components in the respective axes. In the second method, the component of
excitation current necessary to account for the demagnetizing effect of magnetic saturation
is deducted from the respective axes. The d-and q-axis components of the saturation demagnetizing current are referred to as i sd and i sq respectively. The objective in the following is show that the perturbations in the saturation demagnetizing currents i sd and i sq
about their steady-state values of i sd and i sq can also be expressed in terms of the pertur0

bations in the airgap mutual flux-linkages. Provision is made for these two representations
when formulating the generator equations in Section 4.2.3.4 by including the perturbations
in either the mutual inductances or saturation demagnetizing currents depending on the
method employed.
In both methods, the saturation level is determined from the user-supplied, open-circuit saturation characteristic(s). The user may choose to supply only the d-axis characteristic and
select one of several functions for determining the q-axis characteristic from the d-axis characteristic. Alternatively, the manufacturer or testing contractor may supply a separate characteristic for each axis.
Terminal
voltage (pu)

Airgap line
(unity slope)

Va
Vt

Open-circuit
characteristic

Excitation (pu)
Vt

Va

Figure 4.6 Open-circuit characteristic of the generator in the non-reciprocal per-unit system. (Note that the per-unit values of terminal voltage and airgap flux linkages are equal
when the machine is open-circuit and rotating steadily at base rotor speed.)

120

Generators, FACTS devices & system models

An open-circuit saturation characteristic is shown in Figure 4.6 in the non-reciprocal or unity-slope per-unit system. If V t is the terminal voltage and V a is the corresponding voltage
on the airgap line, then the saturation function S V t is defined as
S Vt = Va Vt Vt .

(4.66)

For the given open-circuit characteristic of the generator, saturation is characterized by the
values of S 1 and S 2 such that 2 1 0 together with the selection of a function
to interpolate between the latter two points on the saturation characteristic. Typically,
1 = 1.0 pu and 2 = 1.2 pu. Several commonly employed interpolation functions are
detailed in Table 4.7 although other functions may be used.
Table 4.7
Type
Exponential

Interpolation functions for saturation characteristics.

S = f A B

A, B

2 S 2
2
B = ln -------------------- ln ------
1 S 1 1

A = 1 S 1

Quadratic

Ontario
Hydro
[10]

Linear

if A
0 otherwise

B A

B I

B 2

A = 2 a 1 1 a

Ae

A B 1

B
1

if I
0 otherwise;

I 1
B A
---------------------

B = 1 S 1 1 A
a =

, where

A
B 1.0 ---

2 S 2 1 S 1

2 S 2
B = ln -------------------- 2 1
1 S 1
A = 1 S 1 e

B 1 I

A = 2 a 1 1 a ,
B = 1 S 1 1 A ,

where

a = 2 S 2 1 S 1

1
S B --

AB
------2

So far only the open-circuit characteristic has been considered. However, when the generator is loaded the open-circuit characteristics no longer apply. A common approximation is
that the resultant airgap flux ag is indicative of the level of saturation when the generator
is on-load. This is based on the fact that when the generator is on open-circuit and rotating
at one per-unit speed the terminal voltage and airgap flux are equal in the per-unit system
used. Other approximations for the level of saturation which are employed in widely-used
software packages are described in Section 4.2.13.2.
Note that in Table 4.7 the value of to be used depends on the context. When determining
the parameters A and B of the interpolation function the values of are the o.c. flux-link-

Sec. 4.2

Modelling generator saturation

121

ages (equivalently o.c. voltages) obtained from the open-circuit-characteristic. When evaluating the saturation function when the machine is loaded then is the value of the selected
saturation level indicator such as the resultant airgap flux ( ag ), the resultant kth-transient
k

flux-linkages ( ), etc.
The resultant airgap flux linkages are defined as:
ag =

ad + aq .

(4.67)

To determine the steady-state operating value of the airgap flux, it is noted that ag is also
equal to the voltage behind the stator resistance and leakage inductance in the per-unit system used, taking into account any difference between the synchronous speed and base frequency of the generator (i.e. to account for the situation when 0 1 ).
v ag
r s P 0 + 0 L l Q 0 2 0 L l P 0 r s Q 0 2

1
ag = ----------0 = ------ V t + ------------------------------------ + ------------------------------------ ,
0 0
0
0
Vt
Vt

0
0

(4.68)

in which V t , P 0 and Q 0 are respectively the initial steady-state values of the generator sta0

tor terminal voltage, and real and reactive power output.

The perturbations in the resultant airgap flux linkages about the operating point ag de0

fined in (4.68) and in which the corresponding steady-state values of the d- and q-axis flux
linkages are ad and aq respectively are obtained by linearizing equation (4.67) to give:
0

ad
aq
1
ag = -----------0 ad + -----------0 aq = ----------- ad
ag
0
ag0
ag 0
0

aq adq .
0

(4.69)

It should be noted that the saturation characteristic interpolation functions listed in

Table 4.7 are intended to be used when the machine is operating within its normal range of
steady-state operating conditions, i.e. ag is expected to range at most between about 0.8
0

and 1.3 pu. The interpolation functions may require modification at higher flux levels that
may occur under some large disturbance conditions.
The d- and q-axis saturation characteristics are denoted by S d and S q respectively. In the
situation where a q-axis saturation characteristic is not provided one of the rules in Table 4.8
can be used to derive the q-axis characteristic from the d-axis characteristic provided.

122

Generators, FACTS devices & system models

Table 4.8

q-Axis saturation characteristics as a function of those for the d-axis

Sq Sd

S q S d

Sd

Note A: L aq L ad = L aqu L adu

N/A
N/A
Note B: The points S q 1.0 and S q 1.2 on the q-axis open-circuit saturation characteristic are specified. The same interpolation function used for the d-axis characteristic is employed.
Sd
--------------------------------------------------------------------------------------S d L aqu L adu 1 + L aqu L adu

D
E

Note C: L ad L aq = L adu L aqu = L sal . This characteristic is based on empirical

results reported by Shackshaft [19]; the variation of L ad and L aq with rotor position
is neglected.
0
0
Note D: The q-axis is unsaturated. Useful in modelling salient pole machines.
L du L qu S d

L du L qu

Note E: This option is employed in the saturation model of some software.

Z

L du L qu S d ,

Sd
L aqu
----- 1 S d S q ----------1
L adu
Sq

where

Z = ln S d 1.2 S d 1.0 ln 1.2

L du L qu

Note F: This option is employed in the saturation model of some software.

4.2.8.1

Method 1: Non-linear airgap mutual inductances

As mentioned earlier the first method for representing generator magnetic saturation is to
treat the d- and q-axis airgap mutual inductances as non-linear parameters that depend on
the resultant airgap flux linkages. It is assumed that the leakage inductances are not subject
to magnetic saturation and are thus assumed to be constant parameters.
The values of the d- and q-axis airgap mutual inductances are expressed in terms of their respective saturation characteristics by:
L ad
L aq
u
u
- and L aq = ----------------------------.
L ad = ---------------------------1 + S d ag
1 + S q ag

(4.70)

The steady-state saturated values of the airgap mutual inductances L ad and L aq are ob0

tained by substituting ag = ag in (4.70).

0

Sec. 4.2

Saturation modelling method 2: Saturation demagnetization current

123

The perturbations in the airgap mutual inductances about the operating point ag defined
0

in (4.68) are obtained by linearization of the equations for the non-linear airgap mutual inductances in (4.70) to yield:
2

L ad 0 S d ag
---------- --------------------- L ad u ag 0
L ad
=
ag
2
L aq
L
aq 0 S q ag
---------- --------------------- L aq u ag 0

(4.71)

Substituting for the perturbations in the resultant airgap flux linkages from equation (4.69)
into the preceding equation results in the following expression for the perturbations in the
airgap mutual inductances in terms of the perturbations in the airgap flux-linkages:
Saturation Method 1:
Perturbations in airgap mutual inductances
L adq = C ladq adq ,
0

(4.72)

in which
2

L ad 0 S d ag
---------- --------------------- L ad u ag 0
1

C ladq = ---------- ad aq
ag
0
0
0
2
0
L aq 0 S q ag
---------- --------------------- L aq u ag 0

4.2.8.2

(4.73)

Method 2: Saturation demagnetization current

The second method for representing the effects of magnetic saturation involves deducting
non-linear components of d- and q-axis saturation demagnetization current i sd and i sq
from the excitation of the d- and q-axis windings respectively. In this formulation the model
utilizes the fixed unsaturated airgap mutual inductances. This is especially advantageous
when representing saturation in the Classical Parameter Formulation of the generator model
because it is unnecessary to adjust the classically-defined standard parameters to account for
the effects of saturation.
To determine the expression for i sd it can be deduced from (4.12) on page 105 that:
ad
T
i sd = u i rd i d ---------- .
L

ad u

(4.74)

124

Generators, FACTS devices & system models

Alternatively, if the effects of magnetic saturation are represented by non-linear mutual airgap inductances according to (4.70) (i.e. by Method 1) then from (4.12) with s 2 = 0 and
with L ad defined according to (4.70) it follows that:
1 + S d ag ad
T
u i rd i d = ------------------------------------------- .
L ad

(4.75)

Substituting from the preceding equation for u i rd i d into (4.74) yields:

ad
i sd = S d ag ---------- .
L ad

(4.76)

The q-axis saturation demagnetization current component is similarly derived:

aq
i sq = S q ag ---------- .
L

(4.77)

aq u

Equations (4.76) and (4.77) are combined to yield:

S d ag S q ag
i sdq = S dq ag adq in which S dq ag = D ------------------- ------------------- .
L aq

L ad u

(4.78)

Linearizing the preceding equation about the operating point ag ad aq yields the
0

following expression for the perturbations in the saturation demagnetizing current components in terms of the perturbations in the airgap flux-linkages.
Saturation Method 2:
Perturbations in saturation demagnetizing currents
i sdq = C madq adq ,
0

(4.79)

where

C madq = S dq +
0
0

ad S
d ag
-----------0 ---------------------L ad ag 0
u

aq S
q ag
-----------0 ---------------------L aq ag 0

ad 0
-----------
ag 0

aq0
-----------
ag0

(4.80)

S d ag S q ag
L aq
L ad u
u

0
0
and S dq = D -------------------- --------------------.
0

(4.81)

Sec. 4.2

Steady-state conditions, coupled-circuit model

125

4.2.9 Balanced steady-state operating conditions of the coupled-circuit model

The initial steady-state values of the coupled-circuit generator model variables when operating under balanced conditions are now calculated. In the following analysis the subscript
0 denotes the steady-state value of the variable.
It is assumed that the steady-state generator stator voltage magnitude, V t0 , and the real and
reactive power output P 0 = P e and Q 0 of the generator are given, in per-unit on the gen0

erator base quantities. These initial values are usually obtained from the power flow solution
on which the dynamic analysis is to be based.
Under steady-state conditions p = 0 so from (4.56) on page 114 it follows that = 0
per-unit; and p = 0 so from (4.57) on page 114 it follows that T m = T g . Note, that normally the generator rated frequency is the same as the system nominal frequency and so normally 0 = 1 . However, if, for example, a generator rated for 60 Hz is connected to a 50
Hz system and b is chosen to be 2 60 , then 0 = 5 6 per-unit.
Under balanced steady-state operating conditions the stator voltage is represented as a phasor V t in the complex plane in which the d-axis corresponds to the real axis of the complex
plane and the q-axis to that of the imaginary axis so that:
j 0
V t = v d0 + jv q0 = V t0 e = V t0 cos 0 + jV t0 sin 0 .

(4.82)

where 0 is the angle by which the voltage phasor leads the d-axis. (Note that 0 is defined
differently than the load angle which is the angle by which the voltage phasor lags the qaxis. The use of 0 is convenient analytically and the results are consistent.).
The phasor representing the generator current output is:
P 0 jQ 0
j +
I = i + ji = -------------------- = I0 e 0 0 ,
d0
q0
j 0
V t0 e

(4.83)

where
2

P0 + Q0
I 0 = ----------------------V t0

(4.84)

is the magnitude of the current; and

0 = atan 2 Q 0 P 0 1

is the angle by which the current phasor leads that of the voltage.

(4.85)

126

Generators, FACTS devices & system models

If the effects of magnetic saturation are being represented then, for the purpose of calculating the initial steady-state operating conditions, the following steady-state saturated values
of the d- and q-axis airgap mutual inductances obtained from (4.70) are used. This applies to
both of the methods of representing the effects of magnetic saturation in the dynamic model
of the machine.
L ad = L ad 1 + S d ag and L aq = L aq 1 + S q ag ,
0
u
0
0
u
0

(4.86)

in which the resultant airgap flux-linkages, ag , are obtained from (4.68) on page 121.
0

The saturated values of the d- and q-axis synchronous inductances at the steady-state operating point are:
L d = L ad + L l and L q = L aq + L l .
0
0
0
0

(4.87)

Since, under steady-state conditions, p rq = 0 and since v rq = 0 it is deduced from (4.5)

on page 104 that i rq = 0 . Consequently, from equation (4.12) on page 105,

aq = L aq i q
0
0 0

(4.88)

and then from (4.22) on page 106 it follows that:

q = aq L l i q = L aq + L l i q = L q i q ,
0

(4.89)

a result that is independent of the number of rotor windings.

From (4.24) and (4.89), under steady-state conditions,
vd = rs id 0 q = rs id + 0 Lq iq ;
0
0
0
0
0 0

(4.90)

which again is independent of the number of rotor windings.

The d-axis damper winding currents are zero in the steady-state so from (4.9) on page 104,
(4.22) on page 106 and (4.65) on page 117 it follows that:
ad = L ad i fd i d = L ad L ad I fd L ad i d and
0

1. Definition of = atan 2 y x :
If x = y = 0 then arbitrarily define = 0 ;
else if x = 0 then if y 0 , = 2 ; else = 2 ,
otherwise let z = y x and define = atan z then
First quadrant: x 0 and y 0 then = ;
Second quadrant: x 0 and y 0 then = ;
Third quadrant: x 0 and y 0 then = + ; and
Fourth quadrant: x 0 and y 0 then = .

(4.91)

Sec. 4.2

Steady-state conditions, coupled-circuit model

L ad
d = ad L l i d = L d i d + L ad i fd = L d i d + ----------0 I fd .
0
0
0
0 0
0
0
0 0
L adu 0

127

(4.92)

From (4.24) on page 106 and (4.92) the following expression for the steady-state q-axis voltage is obtained:
L ad
v q = r s i q + 0 d = r s i q 0 L d i d + 0 ----------0 I fd .
0
0
0
0
0 0
L adu 0

(4.93)

The stator voltage phasor is obtained by combining (4.90) and (4.93) to yield:
L ad

V t = v d + jv q = r s + j 0 L q I + j 0 ----------0 I fd L d L q i d
0
0
0
0
0
0
L
0
ad u

(4.94)

in which I = i d + ji q

(4.95)

L ad

E q = V t + r s + j 0 L q I = j 0 ----------0 I fd L d L q i d = jE q .
0
0
0
0
L
0
adu

(4.96)

Rearranging (4.94) yields:

The artificial voltage phasor E q = jE q is aligned with q-axis and corresponds to the voltage
behind the impedance Z = rs + j 0 L q .
0

Now, substituting for V t and I from equations (4.82) and (4.83) into the preceding equation
yields:
Eq e
E q e

j 2

j 2 0

j 0 j 0
= V t + r s + j 0 L q I 0 e e
0

. (4.97)

= V t + r s I 0 cos 0 0 L q I 0 sin 0 + j r s I 0 sin 0 + 0 L q I 0 cos 0

0
0
0

By equating the arguments of both sides of (4.97) yields 0 , the angle by which voltage phasor leads the d-axis:

128

Generators, FACTS devices & system models

0 = --- atan 2 I 0 r s sin 0 + 0 L q cos 0 V t + I 0 r s cos 0 0 L q sin 0

0
2
0
0

in which, from equations (4.84) & (4.85) respectively:

2

(4.98)

P0 + Q0
I 0 = ----------------------- and 0 = atan 2 Q 0 P 0 .
V t0

This result is independent of the number of d- or q-axis rotor-windings or of the representation of coupling between the d-axis rotor windings. Figure 4.7 is a phasor diagram showing
the computation of E q , 0 and the associated location of the d- and q-axes with respect to
the voltage phasor. The d- and q-axis components of the voltage and current phasors are
also shown in this diagram.
qaxis
q = Vt + (rs + j0 Lq )I
E
0

j0 Lq0 I

vq0
iq0
Vt
0
vd0

rs I
I

id0

daxis

Figure 4.7

Phasor diagram showing the computation of E q , 0 and the location of the d-

and q-axes in relation to the voltage and current phasors. The d- and q-axis components of
the voltage and current are also shown.

Sec. 4.2

Steady-state conditions, coupled-circuit model

129

Having calculated the steady-state saturated values of the airgap mutual inductances and synchronous inductances according to (4.86) and (4.87) and the values of 0 , I 0 and 0 in
(4.98) the initial steady-state values of those generator variables that are independent of the
rotor winding structure are readily found to be:
v d = V t cos 0
0
0
v q = V t sin 0
0
0

i fd = d + L d i d L ad
0
0
0 0
0

i d = I 0 cos 0 + 0
0

I fd = L ad i fd
0
u
0

i q = I 0 sin 0 + 0
0

e fd = r fd i fd
0
0

d = vq + rs iq 0
0
0
0
ad = d + L l i d
0
0
0
q = vd + rs id 0
0
0
0

E fd = I fd
0

(4.99)

Tg = d iq q id = P0 + rs I0 0
0

Tm = Tg = Pm 0

aq = q + L l i q
0
0
0

The calculation of 0 , the initial steady-state value of the angle by which the d-axis leads the
R-axis of the synchronously rotating network reference frame is deferred until
Section 4.2.10.
For the 3d3q-c1 model the steady-state values of the following d- and q-axis rotor winding
current and flux-linkage and variables are:
i 1d = i 2d = 0
0
0
fd = ad + L fd + L c1 + L c2 i fd
0
0
0
1d = ad + L c1 + L c2 i fd
0
0
0

i 1q = i 2q = i 3q = 0
0
0
0

(4.100)

1q = 2q = 3q = aq
0
0
0
0

2d = ad + L c2 i fd
0
0
0

For generator models which neglect unequal coupling between the d-axis rotor windings
L c1 = L c2 = 0 . For models with only one d-axis damper winding the variables i 2d and 2d
do not exist and the non-existence of L c2 is represented by setting its value to zero in the
above equations. Similar trivial modifications are made to (4.100) so they can be applied to
coupled-circuit models with other rotor structures.

130

Generators, FACTS devices & system models

4.2.10 Interface between the generator Park/Blondel reference frame and the synchronous network reference frame
The generator equations are developed in the Park/Blondel co-ordinate system in which, as
mentioned earlier, the d-axis is aligned with the magnetic axis of the rotor field winding and
the q-axis leads the d-axis by 90 electrical degrees. The dq reference frame rotates in an anticlockwise direction at the speed of the generator rotor per-unit. As explained in the development of the generator equations of motion in [12] the generator rotor angle t (elec.
rad) is measured relative to a synchronously rotating reference. In the analysis of multi-machine systems the R-axis of the synchronously rotating network RI reference frame is chosen
as the reference for the rotor angle of each generator.
To facilitate the analysis of multi-machine systems it is necessary to transform the stator current and voltage at the machine terminals between the generator dq reference frame and network RI reference frame.
g

In Figure 4.8 the stator current phasor, I , is shown at the instant t. The superscript (g)
denotes that the current phasor is in per-unit of the generator base quantities. At this instant
the d-axis leads the R-axis by t (rad) and the current phasor leads the d-axis by t (rad).
g
In the generator dq reference frame the current phasor I is expressed as:

I g = i + ji = I g e j t ,
d
q

and in the RI reference frame it is:

I g = i g + ji g = I g e j t + t = e j t I g e j t = e j t i + ji .
R
I
d
q

Expanding the preceding equation yields the following relationship between the current
phasor components in the dq and RI reference frames.
g

i R + ji I

= i d + ji q e = i d + ji q cos + j sin
= cos i d sin i q + j sin i d + cos i q

Equating respectively the real and imaginary components in the above equation yields the
following matrix relationship between the current components in the respective reference
frames:
g

iR

iI

cos
sin

sin i d ,
cos i q

(4.101)

In compact matrix form this equation is denoted as:

g

i
= R i dq ,
RI

(4.102)

Sec. 4.2

Interface between generator and network

in which i RI = i g i g , i dq = i d i q
R
I

131

and R = cos
sin

sin .
cos

(4.103)

The inverse relationship is:

1 g

i = R i RI .
dq

(4.104)

The unitary matrix R is referred to as the rotation matrix and its inverse is equal to its
transpose:
R

= R = R =

cos
sin

sin .
cos

(4.105)

When linearizing the model the partial derivative of the rotation matrix is required:
R
U = --------------- = sin

cos

T
R
and -------------------- = U .

cos
sin

(4.106)

The transformation of the stator terminal voltage between the respective reference frames
is similar to that for the stator current:
g

1 n

v RI = R v dq

and v dq = R v RI

(4.107)

(g)

vI

V (g)

vq
I(g)

(g)

iI

iq

(t)

(t)

(t)
id
vd

(t)

d
(t)

R
(g)
vR

(g)
iR

Figure 4.8 Relationship between the Park/Blondel (dq) and Network (RI) reference
frames.

132

Generators, FACTS devices & system models

Again referring to Figure 4.8 the angle by which the voltage phasor leads the R-axis is t .
The initial steady-state value of this angle, 0 , is typically obtained from the power flow
solution. Given this angle, it follows that the initial steady-state value of the rotor angle 0 is:
0 = 0 0 ,

(4.108)

where 0 is given by (4.98). From the above value of 0 and the initial steady-state values
g

and v dq from (4.99), the initial steady-state values i RI and v RI are deduced from
i
dq0
0
0
0

equations (4.102) and (4.107) respectively.

g

i
= R 0 i dq
RI0
0

and v RI = R 0 v dq

(4.109)

Linearizing the current transformation equations (4.102) and (4.104) about the steady-state
g

operating point 0 i dq i RI yields:

i RI = R 0 i dq + U 0 i dq

0
g

and i dq = R 0 i RI + U 0 i RI .

(4.110)
(4.111)

The voltage transformation equations (4.107) are similarly linearized about the steady-state
g

operating point 0 v dq v RI to give:

v RI = R 0 v dq + U 0 v dq

0
g

and v dq = R 0 v RI + U 0 v RI .

(4.112)
(4.113)

There is normally a change in the apparent power base between the generator model and
that of the network. For generality it will also be assumed that there is a change in base voltage between the generator and that of the network bus to which the generator is connected.
However, most simulation programs assume that the respective base voltages are identical.
g

Let S usb MVA and V usb kV be respectively the generator three-phase MVA base and linen

to-line RMS voltage base values of the generator (see Table 4.2 on page 98). Let S usb and
n

V usb be the corresponding quantities for the network bus to which the generator is connect-

ed. The base stator currents in the respective per-unit systems are then deduced as:

Sec. 4.2

Complete set of linearized generator coupled-circuit equations

Ib
g

133

S usb 10
S usb 10
n
= ------------------------- (A) and I b = ------------------------- (A).
g
n
3V usb
3V sub

(4.114)

Let v RI and v RI be the stator voltage components in the network RI reference frame in the

per-unit systems of the generator and network, respectively. It follows that the two are related by:
n

v RI = K V v RI in which K V = V usb V usb ,

(4.115)

and similarly for the current components in the respective per-unit systems:
n

i
= K I i RI in which K I = I b I b
RI

= K S K V and K S = S usb S usb .

(4.116)

4.2.11 Linearized coupled-circuit formulation of the generator model equations

The linearized state- and algebraic-equations of the coupled-circuit generator model incorporating generator saturation and the transformation between the dq generator and network
RI reference frames that were developed in the previous sections are now combined in matrix form. A systematic procedure for formulating the coefficient matrices is also provided.
The generator equations are linearized about the initial steady-state operating condition calculated in Sections 4.2.9 and 4.2.10.
The linearized generator equations comprising (i) the electromagnetic state and algebraic
equations (4.41) on page 110; (ii) the rotor equations of motion in (4.58) and (4.60) on
page 115; (iii) the transformation of the stator current and voltage components from the dq
to RI coordinate systems in equations (4.110) and (4.112) on page 132; and (iv) the conversion from the generator to network per-unit systems for the voltage and current components in equations (4.115) and (4.116) are combined to form the following matrix equation,
A m B mi B mv 0 0
x m
px m
Bm
0 0

C vgx D vgi I 0 0
i dq
0 0
0
0
n

i
D
0

I
0
C
inx
ing
v dq + 0 0
0
RI + 0 u m .
=

n
0
0
C vnx 0 D vng 0 I
0
0
g
v RI
i RI

I 0
0
0

0
0
0 KI I 0
g
0

I
0
0
0
0
0
0 K V I v RI

(4.117)

In the preceding equation the generator state- and input-variable vectors are respectively:
x m =

e fd
and u m =

T m
x r

(4.118)

134

Generators, FACTS devices & system models

and the vector of electromagnetic state-variables are:

rd
.
x r = rdq =

rq

(4.119)

The step-by-step procedure to formulate the coefficient matrices in (4.117) is given in the
following table.
Table 4.9

Step-by-step procedure to compute the coefficient matrices in the linearized

coupled-circuit equations of the generator.

Step
1

3
4

7
8

Operation

Source

Obtain the model parameters according to the number of d- and q-axis

rotor windings n d and n q respectively.
If magnetic saturation is neglected set s m = 0 . Otherwise obtain the
saturation model parameters and set (i) s m = 1 if the incremental
inductance; or (ii) set s m = 2 if the demagnetizing current representation of saturation is employed.
Compute the generator steady-state initial conditions in the dq reference
Section 4.2.9
frame.
Compute the initial values of the generator rotor-angle ( 0 ) and stator
voltage and current in the synchronously rotating RI reference frame.
If s m = 1 then in the following L ad = L ad and L aq = L aq other0
0
wise L ad = L ad and L aq = L aq
u
u

Section 4.2.10

(4.31) pg. 108

Define rd , i rd , r rd , L lrd and b red depending on n d the number

Tab. 4.5

of d-axis rotor windings and whether or not unequal mutual coupling

pg. 112
between the d-axis rotor windings is to be represented.
Define rq , i rq , r rq and L lrq depending on the number of q-axis Tab. 4.6

pg. 112
damper windings.
(4.7) pg. 104 &
Construct r rdq = D r rd r rq and L lrdq = D L lrd L lrq
(4.21) pg. 106
b red

in which b red = b 0 0

Construct b re =

10

Construct the airgap mutual inductance matrix L adq = D L ad L aq

(4.8) pg. 104

(4.31) pg. 108

Sec. 4.2

Complete set of linearized generator coupled-circuit equations

Step

Operation

135

Source

Construct the coefficient matrices in the stator voltage equations:

11

(4.25) pg. 107

W dq = 0 1
1 0
Calculate the coefficient of the saturation variable in the airgap mutual
flux linkage equation:

12

ad 0 aq0

- ----------- if s m = 1
C aldq0 = D --------- L ad 0 L aq 0

C asdq =
0
if s m = 2
L adq

if s m = 0

(4.32) pg. 108

13

Calculate the coefficient matrix C sadq of the saturation variable equa0

tion (4.34) on page 108 depending on the method used to represent saturation.
(4.35) pg. 108
C
if s m = 1
ladq0

C sadq = C madq 0 if s m = 2 , where

0

if s m = 0

13(a)

L ad0 S d ag
---------- ----------------------
ag 0
1 L adu

C ladq = ---------- ad aq

0
0
0
ag
2
0
L aq0 S q ag
---------- ----------------------
ag 0
L aq

and

(4.73) pg. 123

ad S
d ag
-----------0 ---------------------L ad ag 0 ad
u
0
C madq = S dq +
-----------
0
0

aq S
ag 0
q ag
-----------0 ---------------------13(b)

L aq
ag 0
u

S d ag 0 S q ag 0
S dq = D -------------------- ---------------------.
0
L aq
L ad u
u
... Continued on following page ...

aq 0
----------- in which
ag 0
(4.80) pg. 124 &
(4.81) pg. 124

136

Generators, FACTS devices & system models

Step

Operation

Source

The d- and q-axis saturation functions S d ag , S q ag and their

S d
S q
derivatives ----------- and ----------- depend on the selected interpolation func ag
ag
tion in Table 4.7 on page 120. If parameters for the q-axis saturation
characteristic are not specified then one of the relationships in Table 4.8
may be used to express S q as a function of S d .
Compute the coefficient sub-matrices in the electromagnetic state- and
algebraic-equations of the machine (4.41) on page 110.
0
I
L

adq
C re = b r rdq 0 0 0 , C er = 0 , C ei =
0
Ll I
0
0
L lrdq

u2

L adq u 2

0 C asdq

0
0

14

C ee =

I
C sadq 0
0

(4.39) pg. 110

0
I

, C ve = 0 0 0 W dq 0

(4.43) pg. 110

J er = C ee C er and J ei = C ee C ei

A r = C re J er , B ri = C re J ei , C vr = C ve J er & D vi = C ve J ei r s I (4.42) pg. 110

15

Compute the coefficient matrices C tidq and C tvdq in the shaft acceleration equation (4.60) on page 115
1
C tidq = ------ v d + 2r s i d v q + 2r s i q and
(4.54) pg. 114
0
0
0
0
0
1
C tvdq = ------ i d i q
0 0 0
Compute the transformation matrices between the generator dq and
network synchronously rotating RI coordinate systems.

16

R 0 =

cos 0

sin 0

sin 0

cos 0

sin 0
R
U 0 = --------------- =
0
cos 0

(4.103) pg. 131,

(4.106) pg. 131
cos 0
sin 0

Sec. 4.2

Complete set of linearized generator coupled-circuit equations

Step

Operation

137

Source

Construct the coefficient matrices in the generator state- and algebraicequations in (4.117):
b

1
D
A m = 0 ------ 0 , B mi = ------- C tidq ,
2H
2H
0

17

B ri

Ar
0

0
1
, Bm =
B mv = ------ C
2H tvdq

1
0 ------- , C vgx = 0 0 C vr ,
2H
b re

(4.41) pg. 110,

(4.58) pg. 114 &
(4.60) pg. 115

D vgi = D vi , C inx = U 0 i dq 0 0 , D ing = R 0 ,

0
C vnx = U 0 v dq 0 0 and D vng = R 0
0
18

Compute the generator to network voltage and current scaling factors:

KV =

g
V ub

n
V ub ,

KS =

g
S ub

n
S ub ,

KI =

g
Ib

n
Ib

= KS KV ,

(4.115) pg. 133,

(4.116) pg. 133

138

Generators, FACTS devices & system models

4.2.12 Transfer-function representation of the electromagnetic equations

In the previous sections fundamental electromagnetic circuit concepts were employed to
provide a sound physical basis for modelling the dynamic performance characteristics of
synchronous generators in terms of coupled circuits. However, tests to directly identify the
coupled-circuit parameters (resistances and inductances) of these models are not routinely
performed in practice. Rather, parameter identification is commonly based on the so-called
operational or, in modern terminology, transfer-function representation of the machine
depicted in Figure 4.9. In the transfer-function representation the d-axis of the machine is
represented by a two-port network comprising the stator and field terminals and the q-axis
by a single port network.
r s q s
rs

id s

i fd s

s d s
---------------b

vd s

s fd s
-----------------b

Zd s

r fd
e fd s

Two-port
d-axis network
r s d s
rs
vq s

iq s
s q s
--------------b

Zq s

Single-port
q-axis net-

Figure 4.9

General transfer-function representations of the d- and q-axis of the synchronous generator.

The transfer impedance representation of the incremental form of the d-axis two port network can be shown to be [22]
b Z s
= ------ 11
s Z s
fd s
12
d s

Z 12 s i d s
Z 22 s i fd s

(4.120)

Note that the impedance matrix is symmetrical since the network is passive and linear. The
scaling of the impedance matrix by the base electrical frequency occurs because one second
is the base value of time. The following hybrid formulation of the transfer characteristic is

Sec. 4.2

Transfer-function representation of generator em model

139

useful because the input quantities are the d-axis stator current and field voltage and is a formulation frequently encountered in the literature (e.g. [6, 15, 20, 21, 22, 23]):
d s
i fd s

Ld s

Gs

id s
.
s
------ G s Y fd s e fd s
b

(4.121)

The three d-axis operational parameters are:

b
L d s = ------ Z 11 s Z 12 s Y fd s Z 12 s
s
b
G s = ------ Z 12 s Y fd s
s
Y fd s = r fd + Z 22 s

(4.122)

in which L d s is referred to as the d-axis operational inductance of the stator with constant field voltage; G s is the stator-to-field transfer-function and Y fd s is the operational admittance of the field with the stator open-circuit.

The single port representation of the q-axis is:

b
q s = ------ Z q s i q s = L q s i q s .
s

(4.123)

L q s is referred to as the q-axis operational inductance.

It is important to note that no assumptions have been made at this stage about the internal
electromagnetic structure of the machine.
4.2.12.1 Exact relationship between coupled-circuit and transfer-function representations of
the electromagnetic equations

In [12] the exact relationships between the coupled-circuit formulation of the electromagnetic equations in Section 4.2.3.4 and the transfer-function representations are established.
It is found that for a machine with nd d-axis rotor windings and nq q-axis rotor-windings the
transfer-functions have the following forms:
N Ld s
NG s
NY s
N Lq s
L d s = L d ----------------- , G s = G 0 --------------- , Y fd s = Y 0 -------------- and L q s = L q ----------------- (4.124)
D s
D s
D s
D s
d

in which the d-axis numerator polynomials are:

nd

N Ld s = 1 +

nd 1

nd 1

k
a kd s , N G s = 1 + b kd s , N Y s = 1 + c kd s . (4.125)
k=1

The d-axis denominator polynomial is:

k=1

k=1

140

Generators, FACTS devices & system models

nd

k
d kd s .

Dd s = 1 +

(4.126)

k=1

The q-axis numerator and denominator polynomials are:

nq

nq

k
a kq s and D q s = 1 + d kq s .

N Lq s = 1 +

k=1

(4.127)

k=1

The coefficients in the above polynomials (i.e. a kd , b kd , etc.) are functions of the coupledcircuit parameters.
It is normal practice to factorize the d-axis numerator and denominator polynomials such
that:
nd 1

nd

N Ld s =

nd 1

1 + sT dk , N G s = 1 + sT Gk , N Y s = 1 + sT Yk , (4.128)
k=1

k=1

k=1

nd

Dd s =

1 + sT d0k

(4.129)

k=1

where T dk T d k + 1 , T d0k T d0 k + 1 , T Gk T G k + 1 and T Yk T Y k + 1 .

Similarly, for the q-axis:
nq

N Lq s =

nq

1 + sT qk and D q s = 1 + sT q0k
k=1

(4.130)

k=1

such that T qk T q k + 1 , T q0k T q0 k + 1

The time constants T dk , k = 1 n d are referred to as the d-axis principal short-circuit
time constants and are denoted as the d-axis transient ( T d' ), subtransient ( T d'' ) and subsubtransient ( T d''' ) short-circuit time constants. The denominator time constants T d0k ,
k = 1 n d are referred to as the d-axis principal open-circuit time constants and are

denoted as the d-axis transient ( T d0' ), subtransient ( T d0'' ) and sub-subtransient ( T d0''' )
open-circuit time constants. Analogous definitions apply to the q-axis principal short- and
open-circuit time constants.
As shown in [12] the d- and q-axis operational-inductances can also be expressed as sums of
partial fractions as follows:

Sec. 4.2

Relationship between classical and coupled-circuit parameters

141

d
sT dk
1
1
1
1
------------- = ----- + -------- ------------------- ------------------- , where L d0 = L d and
L dk L d k 1 1 + sT dk
Ld s
Ld
k=1

(4.131)

q
sT qk
1
1
1
1
------------- = ----- + -------- ------------------- ------------------- , where L q0 = L q .
L qk L q k 1 1 + sT qk
Lq s
Lq
k=1

(4.132)

The inductances L dk , k = 1 n d in (4.131) are referred to as the d-axis principal dynamic inductances and are denoted as the d-axis transient ( L d' ), subtransient ( L d'' ) and sub-subtransient ( L d''' ) inductances. Analogous definitions apply to the q-axis principal dynamic
inductances.
The exact mathematical relationships between (i) the coupled-circuit parameters and (ii) the
time constants and principal dynamic inductances in the above factorizations of the operational parameters are very complex when there is more than one rotor-winding in an axis.
Exact bi-directional transformations have been derived between the coupled-circuit parameters of the machines and the transfer-function parameters in [12]. Software for performing
these exact transformations are available on the website of this eBook [24].
4.2.12.2 Classical definitions of the standard parameters in terms of the coupled-circuit
parameters

The opportunity for considerable simplification in the relationships between the coupledcircuit and standard parameters of the d-axis has long been recognized (e.g. Concordia [20]).
If the d-axis is assumed to be represented by a field winding and one damper winding the
simplification is based on the observation that the d-axis damper winding resistance r 1d is
usually much larger than the field winding resistance r fd . Consequently, following a disturbance the d-axis damper winding flux-linkages tend to decay much more rapidly than those
of the field. Thus, the relatively rapid rate of decay of the damper winding flux-linkages in
the so-called subtransient period immediately following a disturbance can be approximated by
assuming that r fd = 0 and that the field flux linkages remain constant during this period.
The behaviour of the generator in the transient period which follows the subtransient period
is approximated by neglecting the damper winding (i.e. r 1d = L 1d = ) on the assumption
that the damper winding current has decayed to zero. Finally, the behaviour in the steadystate period, following the decay of the transient response, is approximated by neglecting
both the damper and field windings. By applying these approximations, the classical definitions of the standard-parameters in terms of the coupled-circuit parameters result. By applying analogous approximations to the q-axis and extending the approximations to axes
with three rotor-windings, classical definitions of the standard parameters in terms of the
coupled-circuit parameters are obtained in [12]. In the case of a d-axis represented with two
rotor-windings it is usually found that there is little difference between the standard parameters which are specified in accordance with the exact definitions and those which are spec-

142

Generators, FACTS devices & system models

ified in accordance with the classical definitions. However, if the q-axis is represented with
two or more windings, or if the d-axis is represented by three rotor windings, the accuracy
of the classical approximations of the standard parameters may be considerably diminished.
This is because there is greater overlap between the transient and subtransient periods of the
q-axis than occurs in the d-axis. Similarly, when three rotor windings are represented there
is a tendency, particularly in the q-axis, for there to be overlap between the transient, subtransient and sub-subtransient periods with consequential inaccuracy of the classical approximations of the standard parameters.
Table 4.10 gives the transformation from given coupled-circuit parameters to the classicallydefined principal open- and short-circuit time constants and the principal dynamic inductances for the d- or q-axis represented by one, two or three rotor-windings respectively.
Table 4.11 gives the inverse transformation (i.e. classically-defined standard parameters to
coupled-circuit parameters). If the transformations are being performed for the d-axis the
subscript a in these tables is substituted with d; 1 with f; 2 with 1 and 3 with 2. Thus,
L a is replaced by L d ; L 1a by L fd ; r 2a by r 1d ; T a' by T d' , etc. If the q-axis parameters are
being converted then the subscript a is substituted with q. An exception to the above rules
is that the airgap mutual inductance L aa is substituted with L ad or L aq depending on which
axis is being converted. These transformations assume that unequal coupling between the
d-axis rotor windings is neglected (i.e. L c1 = L c2 = 0 ).

When the q-axis is represented with a single rotor winding it is conventional practice to label
the standard parameters as being the subtransient parameters (i.e. (i.e. T q0'' , T q'' and L q'' )
instead of the transient parameters T q0' , T q' and L q' . This is because the principal q-axis
time constants for a machine whose q-axis is adequately represented by a single rotor winding tend to be short. Thus, in this scenario, in order to apply the conversions in Tables 4.10
and 4.11 it is necessary to substitute the transient parameters in the tables with the given subtransient q-axis parameters.
The electromagnetic equations in some widely-used simulation packages are formulated directly in terms of the classically-defined standard parameters. This classical parameter formulation of the model is summarized in Section 4.2.13. It is emphasised that the classical
parameter formulation is exactly equivalent to the coupled-circuit formulation provided (i) that
the unequal coupling between the d-axis rotor windings is neglected, and (ii) that the same method for representing magnetic saturation is employed in the two models.

Sec. 4.2

Relationship between classical and coupled-circuit parameters

143

Table 4.10 Classical definitions of the principal short- and open-circuit time constants
and dynamic inductances in terms of the coupled-circuit parameters for the d- or q-axis
represented with na = 0, 1, 2 or 3 rotor windings. [CC2CS]
Inputs: Coupledcircuit parameters

Outputs:
Classically-defined standard parameters
Synchronous parameters (na 0)

L aa , L l

L a = L aa + L l
Transient parameters (na > 0)
L aa L l
1
T a' = -------------- L 1a + --------------
b r 1a
La

r 1a , L 1a

L aa + L 1a
T a0' = ---------------------- b r 1a
T a'
1
L a' = ------------------------------------- + L l = --------- L a

1 L aa + 1 L 1a
T a0'
Subtransient parameters (na > 1)
L aa L l L 1a
1
T a'' = -------------- L 2a + -----------------------------------

b r 2a
L aa L l + L a L 1a

r 2a , L 2a

L aa L 1a
1
T a0'' = T a011 = -------------- L 2a + -----------------------

b r 2a
L aa + L 1a
T a''
1
L a'' = --------------------------------------- + L l = ---------- L a'

1
1 - ------1
T a0''
-------- + ------+ L aa L 1a L 2a
Sub-subtransient parameters (na > 2)

1
-
1 L 3a + -------------------------------------------------1
1 - ------1
1
T a''' = --------------
---- + ------+ - + --------
b r 3a
L l L aa L 1a L 2a

r 3a , L 3a

1
-
1 L 3a + -------------------------------------1
1 - ------1
T a0''' = --------------
-------- + ------+ -
b r 3a
L aa L 1a L 2a

T a'''
1
L a''' = ------------------------------------------------------- + L l = ----------- L a''
T a0'''
1
1 - ------1
1
-------- + ------+ - + -------L aa L 1a L 2a L 3a

144

Generators, FACTS devices & system models

Table 4.11 Derivation of coupled-circuit parameters from the classically-defined opencircuit time constants and dynamic inductances for the d- or q-axis represented with
na = 0, 1, 2 or 3 rotor windings. [CS2CC]
Inputs:
Classically-defined
standard
parameters

Outputs:
Coupled-circuit parameters
na 0

La , Ll

L aa = L a L l
na > 0

T a0' , L a'

L a L l L a' L l
L 1a = ------------------------------------------- L a L a'
2

1 La Ll
r 1a = ---------------- ----------------------- b T a0' L a L a'
na > 1

T a0'' , L a''

L a' L l L a'' L l
L 2a = ---------------------------------------------- L a' L a''
2

1 L a' L l
r 2a = ------------------ ------------------------- b T a0'' L a' L a''
na > 2

T a0''' , L a'''

L a'' L l L a''' L l
L 3a = ------------------------------------------------ L a'' L a'''
2

L a'' L l
1
r 3a = ------------------- --------------------------- b T a0''' L a'' L a'''

Use L a' = T a' T a0' L a , L a'' = T a'' T a0'' L a' and

L a''' = T a''' T a0''' L a'' to derive the principal dynamic
inductances if the principal short- and open-circuit time
constants are given; or use T a0' = L a L a' T a' ,
T a0'' = L a' L a'' T a'' and T a0''' = L a'' L a''' T a''' if
the principal dynamic inductances and short-circuit time
constants are given.

Sec. 4.2

Classical parameter model of generator

145

4.2.13 Electromagnetic model in terms of classically-defined standard parameters

Some widely used rotor angle stability analysis software packages employ generator models
which are formulated directly in terms of the classically-defined standard parameters. It is
shown in [12] that the Classical Parameter Model is exactly equivalent to the coupled-circuit
formulation of the electromagnetic equations provided (i) unequal coupling between the d-axis
rotor windings is neglected; and (ii) the same representation of magnetic saturation is employed.

The transfer-function block-diagram representation of the classical parameter model shown

in Figure 4.10 is often presented in the literature [34, 13 (vol 3)]. The associated state- and
algebraic-equations for the model, which are rigorously derived from the underlying coupled-circuit formulation of the model in [12], are summarized in Section 4.2.13.1. The relationships between the classically-defined standard parameters and the underlying coupledcircuit parameters are given in Tables 4.10 and 4.11 of Section 4.2.12.2.

Please turn over to Figure 4.10.

nd 2
(4.133)

E fd

1 '
------------ d
sT d0'

I fd
(4.134)

L d L d'

(4.136)
s1d

i 1d

I sq

nq 1

1
-----------sT q0' '
q
(4.143)

i 2q

Note: i 2d = 0

i 3q i q

and

L d'' L d'''
1
-----------------------------------sT d0''' 2d
L d'' L l
(4.139)

(4.148)
q''

i 1d = 0

if

L d'' L l

(4.141)

s3q
(4.150)

if

nq 3

d'''
(4.142)

d' L d'i d if n d = 1

d = d'' L d''i d if n d = 2

d''' L d'''i d if n d = 3

L q'' L l

q' L q'i q if n q = 1

q = q'' L q''i q if n q = 2

q''' L q'''i q if n q = 3

(4.149)
1 3q L q'' L q'''
-----------------------------------sT q0'''
L q'' L l
L q''' L l
------------------L q'' L l

and

iq

nq 3
n d 2 ; i 3q = 0

d-axis

id

(4.151)

L q'' L q'''
------------------------2
L q'' L l

(4.145)
1 2q L q' L q''
-------------------------------sT q0''
L q' L l

nd 3

i 3q

L q' L l

L q'' L l
-----------------L q' L l

if

i 2d

i 2d i d

nq 2

(4.140)
s2d

L d'' L d'''
------------------------2
L d'' L l

(4.147)

s2q
(4.146)

d''
(4.138)

L d' L l

(4.137)

L q' L q''
-----------------------2
L q' L l

L d''' L l
------------------L d'' L l

i 2q = 0

if

(4.152)
q'''

q-axis

nq 2

Figure 4.10 Transfer-function block diagram representation of the d- and q-axis rotor-winding equations in terms of the classicallydefined standard parameters. The effects of magnetic saturation is represented by the demagnetizing I sd and I sq in the respective axes.

Generators, FACTS devices & system models

I 1q
(4.144)

L q L q'

L d' L d''
1
-------------------------------sT d0'' 1d
L d' L l
(4.135)

L d' L d''
-----------------------2
L d' L l

I sd

nd 3

L d'' L l
-----------------L d' L l

146

nd 1

Sec. 4.2

Summary: Classical Parameter em Model Equations

147

4.2.13.1 Summary of classical parameter electromagnetic model equations

The d-axis rotor winding equations in terms of the classically-defined standard parameters

are summarized below. The equations are applicable to models with up to three rotor windings (i.e. nd = 3). Note that the field voltage and current are expressed in the non-reciprocal
per-unit system as is the demagnetizing current, I sd , representing the effects of magnetic
saturation in the d-axis.
Field Winding Equations
Note: i 2d = 0 if n d 3 and i 1d = 0 if n d 2
d d '
1
---------- = --------E I
dt
T d0' fd fd

(4.133)

I fd = d' + L d L d' i d i 1d + i 2d + I sd

(4.134)

First d-axis Damper Winding Equations (include if nd > 1)

d 1d
1
------------ = ---------- s1d ,
T d0''
dt

(4.135)

s1d = d' 1d + L d' L l i 2d i d ,

(4.136)

L d' L d''
i 1d = ------------------------- s1d ,
L d' L l 2

(4.137)

L d'' L l
L d' L d''
d'' = ------------------- d' + --------------------- 1d .
L d' L l
L d' L l

(4.138)

Second d-axis Damper Winding Equations (include if nd > 2)

d 2d
1
------------ = ---------- ,
dt
T d0''' s2d

(4.139)

s2d = d'' 2d L d'' L l i d ,

(4.140)

L d'' L d'''
i 2d = -------------------------- s2d ,
L d'' L l 2

(4.141)

L d''' L l
L d'' L d'''
d''' = -------------------- d'' + ----------------------- 2d .
L d'' L l
L d'' L l

(4.142)

The q-axis rotor winding equations in terms of the classically-defined standard parameters
are summarized below. The equations are applicable to models with up to three q-axis
damper windings (i.e. nq = 3). By analogy with the field winding, the current in the first qaxis damper winding, I 1q , is expressed in the non-reciprocal per-unit system such that

148

Generators, FACTS devices & system models

I 1q = L aq i 1q . Similarly, the q-axis saturation demagnetizing excitation current, I sq , is in

u

per-unit in the non-reciprocal base system (i.e. I sq = L aq i sq ).

u

First q-axis Damper Winding Equations

Note: i 3d = 0 if n q 3 and i 2q = 0 if n q 2 .
d q'
1
---------- = -------- I ,
dt
T q0' 1q

(4.143)

I 1q = q' + L q L q' i 2q + i 3q i q I sq .

(4.144)

Second q-axis Damper Winding Equations

d 2q
1
------------ = --------- ,
dt
T q0'' s2q

(4.145)

s2q = q' 2q + L q' L l i 3q i q ,

(4.146)

L q' L q''
i 2q = ------------------------- s2q ,
L q' L l 2

(4.147)

L q'' L l
L q' L q''
q'' = ------------------- q' + --------------------- 2q .
L q' L l
L q' L l

(4.148)

Third q-axis Damper Winding Equations

d 3q
1
------------ = ----------- s3q ,
dt
T q0'''

(4.149)

s3q = q'' 3q L q'' L l i q ,

(4.150)

L q'' L q'''
i 3q = -------------------------- s3q ,
L q'' L l 2

(4.151)

L q''' L l
L q'' L q'''
q''' = -------------------- q'' + ----------------------- 3q .
L q'' L l
L q'' L l

(4.152)

It is noted that the d-axis transient flux linkages d' are related to the field winding flux linkages fd in the underlying coupled circuit model as follows:
L d L d'
d' = ------------------- fd .
Ld Ll

(4.153)

Furthermore, it is common practice to define the transient q-axis voltage E q which is in

quadrature with d' such that E q = d' . Since, for the purpose of small-signal modelling,

Sec. 4.2

Summary: Classical Parameter em Model Equations

149

perturbations in the rotor-speed are neglected in the formulation of the electromagnetic

equations it follows that E q = 0 d' and, in the usual case when 0 = 1 then E q = d' .
L L '

q
q
-
By analogy, in the q-axis q' = -----------------and E d = q' .
L q L l 1q

To facilitate the formulation of the stator-winding d-axis flux linkage equations the followk

ing d-axis kth-transient flux-linkage ( d ) and kth-transient inductance ( L d ) quantities are

defined depending on the number of d-axis rotor-windings:
k
d

d' if n d = 1
L d' if n d = 1

k
= d'' if n d = 2 and L d = L d'' if n d = 2

d''' if n d = 3
L d''' if n d = 3

(4.154)

- and similarly for the q-axis.

The d- and q-axis stator winding flux linkage equations are respectively:
k

d = d L d i d and q = q L q i q .

(4.155)

The following expressions for the d- and q-axis airgap mutual flux linkages are derived in
terms of the kth-transient flux-linkages and inductances from (i) the coupled-circuit equations for the airgap flux linkages in (4.12) on page 105 with s m = 2 (i.e. Method 2 for representing magnetic saturation); and (ii) from the equations for the d- and q-axis flux linkages
in (4.23) on page 106 and from (4.155).
k

ad = I fd + L ad i 1d + i 2d i d I sd = d + L l i d = d L d L l i d and
u
k

aq = I 1q + L aq i 2q + i 3q i q I sq = q + L l i q = q L q L l i q .
u

(4.156)
(4.157)

From the coupled-circuit equations for the stator-voltage components in (4.25) on page 107
and the preceding equations for the airgap flux linkages the following equations for the dand q-axis stator voltage equations are derived.
k

vd = rs id 0 q = rs id + 0 Lq iq 0 q
and
= r s i d 0 I 1q + L aq i 2q + i 3q L aq + L l i q + I sq
u
u
k

(4.158)

vq = rs iq + 0 d = rs iq 0 Ld id + 0 d
.
= r s i q + 0 I fd + L ad i 1d + i 2d L ad + L l i d + I sd
u
u

(4.159)

150
4.2.13.2

Generators, FACTS devices & system models

Magnetic saturation in the classical parameter model

Saturation Method 2 in Section 4.2.8.2, in which the demagnetizing effect of magnetic saturation is represented by deducting a component of excitation current from each axis, is
used in the classical parameter model. In that section the resultant airgap flux linkages were
chosen to represent the level of flux-linkages on the saturation characteristic. This choice
can also be applied to the classical parameter model. However, some widely used commercial software packages choose other flux quantities to represent the level of saturation. These
alternative choices change both the initial steady-state operating condition of the machine
as well as the dynamic response. Two such alternative choices are considered below.
Resultant kth-transient flux-linkages as the saturation level indicator

4.2.13.2.1

When saturation of both the d- and q-axis is to be represented the resultant kth-transient fluxk

linkages is sometimes used as the saturation level indicator, particularly when representing round-rotor machines.
k 2

k 2

d + q ,

(4.160)

and the demagnetizing currents in the respective axes are

k

I sd = L ad i sd = d S d and I sq = L aq i sq = q S q .
u
u

(4.161)

The initial steady-state values of the saturation level indicator and the saturated d- and q-axis
airgap inductances are now derived. Once these values are determined the procedure outlined in Section 4.2.13.3 can be used to calculate the initial steady-state values of the other
generator variables.
The following expression for the initial steady-state value of the d-axis saturation demagnetk

izing current is obtained by substituting for the yet unknown value of d from(4.161) into
0

(4.156) and recognizing that under steady-state conditions the damper winding currents are
zero.
k

Sd 0
k
I sd = -------------------------- I fd L ad + L l L d i d .
k
0
0
u
0
1 + S d 0

(4.162)

Similarly, the steady-state q-axis saturation demagnetizing current is given by:

k

Sq 0
k
I sq = -------------------------- L aq + L l L q i q .
0
u
0
1 + S q k0

(4.163)

The following expressions for the steady-state values of the d- and q-axis stator voltages are
obtained by substituting the saturation demagnetizing current components from equations
(4.162) and (4.163) into equations (4.158) and (4.159) to yield:

Sec. 4.2

Transient flux-linkages as the saturation level indicator

151

I fd

0
- ,
v d = r s i d + 0 L q i q and v q = r s i q 0 L d i d + 0 ------------------------k
0
0
0
0
0 0
0 0
1 + S d 0

(4.164)

in which
L + L k L S k
L + L k L S k
ad u
d
l d 0
aq u
q
l q 0

- + Ll
L d = -------------------------------------------------------- + L l and L q = ------------------------------------------------------k
k

0
0
1 + Sd 0
1 + Sq 0

(4.165)

From equations (4.82) on page 125 and (4.164) the initial steady-state voltage phasor is:
j 0
V t = V t0 e = v d0 + jv q0 = r s + j 0 L q I + jE q ,
0
0

(4.166)

j +
in which I = I 0 e 0 0 = i d0 + ji q0 is obtained from (4.83) in terms of the initial steady-

state real and reactive power output and stator voltage magnitude and
I fd

0
- L d L q i d .
E q = 0 ------------------------k
0
0
0
0
1 + Sd 0

(4.167)

Equation (4.166) is rearranged into the same form as equation (4.97) on page 127
V + r + j L I e j 0 e j 0 = E e j 2 ,
s
0 q0 0
q0
t0

(4.168)

from which the initial steady-state value of the angle by which the voltage phasor leads the
d-axis, is given by:

0 = --- atan 2 I 0 r s sin 0 + 0 L q cos 0 V t + I 0 r s cos 0 0 L q sin 0 .

0
2
0
0

(4.169)

This equation is identical in form to the calculation of 0 in equation (4.98) on page 128 for
the coupled-circuit formulation of the model. However, the value of L q is different be0

tween the coupled-circuit and classical parameter formulations because different representations of saturation are used in the respective formulations.
In order to calculate the initial steady-state value of the q-axis synchronous reactance L q in
0

th

equation (4.165) it is necessary determine the initial value of the resultant k -transient fluxk

linkages 0 and thence the value of the q-axis saturation characteristic S q 0 . In the case
k

where L q = L d it can be shown that:

k
0

r s P 0 + 0 L d Q 0
0 L d P 0 r s Q 0
1
= ------ V t + ------------------------------------- + ------------------------------------- .
0 0
Vt
Vt

0
0

(4.170)

152

Generators, FACTS devices & system models

However, when the kth-transient inductances in the respective axes are unequal a closed
k

form solution for 0 is not possible. In this case the following set of three simultaneous
non-linear equations must be solved iteratively to determine the initial steady-state values
k

of L q , 0 and 0 .
0

L + L k L S k
aq u
q
l q 0
- + Ll
L q = ------------------------------------------------------k

0
1 + Sq 0

0 = --- atan 2 I 0 r s sin 0 + 0 L q cos 0 V t + I 0 r s cos 0 0 L q sin 0

0
2
0
0

(4.171)

j 0
j 0 + 0

k
k
k
k
1
+ j 0 L d L q I 0 cos 0 + 0
0 = ------ V t e + r s + j 0 L q I 0 e
0
0

Having solved for the initial value of the saturation level indicator 0 from equation (4.170)
or (4.171) the initial values L d and L q are obtained from (4.165).
0

4.2.13.2.2 Transient d-axis flux-linkages as saturation level indicator

When saturation only of the d-axis is to be represented the transient flux-linkages d' are
sometimes used as the saturation level indicator. This approach is sometimes used to represent saturation in salient pole machines. In this case the d- and q-axis saturation demagnetizing currents are respectively:
I sd = d'S d d' and I sq = 0 .

(4.172)

Since it is assumed that the q-axis is unsaturated it follows that:

L q = L aq + L l ,
0
u

S q d' = 0

(4.173)

and the initial steady-state value of 0 is calculated according to (4.98) on page 128; the initial steady-state values v dq and i dq are calculated from (4.99).

The objective of the following steps is to calculate the initial steady-state value d ' taking
0

account of saturation. The following expression for the initial steady-state value of the d-axis
saturation demagnetizing current is obtained by substituting for d ' from (4.134) into
0

(4.172) and recognizing that under steady-state conditions the damper winding currents are
zero.
S d d0'
- I fd L ad + L l L d' i d .
I sd = ---------------------------0
0
u
0
1 + S d d 0'

(4.174)

Sec. 4.2

Resultant airgap flux-linkages

153

The following expressions for the steady-state values of the d- and q-axis stator voltages are
obtained by substituting for the above expression of the d-axis saturation demagnetizing
current component, together with I sq = 0 , into equations (4.158) and (4.159) to yield:
0

v d = r s i d + 0 L aq + L l i q ,
0
0
u
0

I fd

0
-
v q = r s i q 0 L d i d + 0 ---------------------------0
0
0 0
1 + S d d0'

L ad + L d' L l S d d '
1 + Sd d '

u
0
- + L l .
and L d = ----------------------------------------------------------0

(4.175)

(4.176)

Following a procedure similar to that in equations (4.166) to (4.168) the following relationship is obtained
I fd

0
- L d L q i d .
v d + jv q + r s + j 0 L q i d + ji q = j 0 ---------------------------0
0
0
0
0
0
0
0
1 + S d d 0'

(4.177)

By equating the imaginary components of the above expression the following equation for
the field current is obtained:
1 + S d d 0'
- v q + r s i q + 0 L d i d .
I fd = ---------------------------0
0
0 0
0

(4.178)

By substituting for I fd and I sd from equations (4.178) and (4.172) into equation (4.134)
0

on page 147 the following expression for d ' is derived:

0

v q0 + r s i q 0 + L d'i d 0
d ' = -------------------------------------------.
0
0

(4.179)

The value of d ' is now back-substituted into equation (4.176) to obtain the initial saturated
0

value of the d-axis synchronous inductance L d . This value of L d together with the value
0

of L q in (4.173) are used to calculate the initial values in (4.99) on page 129.
0

4.2.13.2.3 Resultant airgap flux-linkages expressed in terms of kth-transient flux-linkages

and stator-current components
As explained in Section 4.2.8.2, when the resultant airgap flux-linkages are chosen to represent the saturation level, the saturation demagnetizing current components are given by:
I sdq = S dq ag adq ,

in which S dq ag = D S d ag S q ag and ag =

(4.180)
T

adq adq .

(4.181)

154

Generators, FACTS devices & system models

In the present context of directly representing the machine in terms of the classically-defined
standard parameters it is convenient to express the airgap flux-linkages in terms of the kthtransient flux linkages and stator winding current components. Thus, from equations (4.156)
and (4.157) on page 149 it follows that:
k

(4.182)
adq = dq L aidq i dq in which L aidq = D L d L l L q L l .

Substituting for adq from the preceding equation in (4.180) yields the following expression

for the saturation demagnetizing current components:

k

I sdq = S dq ag dq S dq ag L aidq i dq .

4.2.13.3

(4.183)

Balanced steady-state operating-conditions of the classical parameter model

As in the case of the coupled-circuit representation of the generator model it is assumed that
the steady-state generator stator voltage magnitude, V t0 , and the real and reactive power output P 0 = P e and Q 0 of the generator are given in per-unit of the generator base quantities.
0

The procedure for calculating the generator steady-state initial conditions for the classical
parameter model is the same as that described for the coupled-circuit formulation of the
generator model in Section 4.2.9 except that the initial saturated values of the d- and q-axis
synchronous inductances L ad and L aq in equation (4.86) on page 126 are modified de0

pending on the method used to represent the level of magnetic saturation as summarized in
Table 4.12.
Table 4.12 Initial values of the saturation level and the corresponding saturated values
of the d- and q-axis synchronous inductances depending on the method of representing the
level of magnetic saturation.
#

ag

Calculation of m
0

Calculation of L ad and L aq
0
0

Calculated according to (4.68) pg. 121

k

Calculated according to (4.86) pg. 126

If L q = L d then 0 is calculated
k

according to (4.170); otherwise 0 is Calculated according to (4.165)

obtained from the iterative solution
of (4.171).

d ' Calculated according to (4.179)

0

Calculated according to equations

(4.176) and (4.173) respectively.

The initial values of the saturation demagnetizing currents in the non-reciprocal per-unit
system are

Sec. 4.2

Linearization of the classical parameter equations

I sd = S d m m
0
0
0

155

and I sq = S q m m ,
0

(4.184)

where m is the initial value of the saturation level indicator from Table 4.12. The initial
0

values of the following d- and q-axis rotor winding variables are deduced from equations
(4.133) on page 147 to (4.152) on page 148.
d ' = I fd I sd L d L d' i d
0
0
0
0

I 1q = 0

1d = I fd I sd L ad i d
0
0
0
u 0

q ' = I sq L q L q' i q
0
0
0

s1d = 0

2q = I sq L aq i q
0
0
u 0

s2q = 0

i 1d = 0
0

d '' = I fd I sd L d L d'' i d
0
0
0
0

i 2q = 0

2d = 1d

q '' = I sq L q L q'' i q

3q = 2q

s2d = 0
0

i 2d = 0

i 3q = 0

d ''' = I fd I sd L d L d''' i d
0

(4.185)

q ''' = I sq L q L q''' i q
0

4.2.13.4

Linearization of the classical parameter formulation of the electromagnetic equations

In this section the generator d- and q-axis rotor-winding equations (4.133) on page 147 to

(4.152) on page 148, the generator stator voltage equations (4.158) and (4.159) on page 149
and equations in Section 4.2.13.2 representing the effects of magnetic saturation are linearized about the initial steady-state operating point determined in Section 4.2.13.3. The objective is to reduce the linearized electromagnetic equations of the machine, expressed in
terms of the classically-defined standard parameters, to the following matrix form. This formulation is structurally similar to that for the coupled-circuit model equations in (4.41) on
page 110.
A r B ri 0
p dqc
=

C vr D vi I
0

dqc

b
i dq + re E fd .

0
v dq

(4.186)

The differences between the above classical parameter formulation and that of the coupledcircuit parameter formulation in (4.41) are (i) the vector of rotor-winding state-variables
dqc in the classical parameter formulation is different to that in the coupled-circuit formu

lation rdq ; (ii) the field-voltage input in the classical parameter formulation is in per-unit

in the non-reciprocal base system (i.e. E fd ) whereas in the coupled-circuit formulation it is

in per-unit in the reciprocal base system (i.e. e fd ); (iii) although the coefficient sub-matrices
in the two formulations have the same names their values are different. Since the two alter-

156

Generators, FACTS devices & system models

native formulations of the generator electromagnetic equations are structurally the same
they can be used interchangeably within the complete model of the generator which is depicted in Figure 4.1 on page 92. The linearized rotor-equations of motion developed in Sections 4.2.5 and 4.2.6 are applicable to the classical model as are the equations in
Section 4.2.10 which provide the interface between the dq reference frame and the synchronously rotating network RI reference frame.
The rotor d- and q-axis state-variables in the classical formulation are respectively:
cd = d' 1d 2d

and cq = q' 1q 2q .

(4.187)

The detailed derivation of the coefficient matrices in the linearized electromagnetic equations of the generator in (4.186) is given in Appendix 4I. Once these equations are formed
the linearized state- and algebraic equations of the generator including the rotor equations
of motion and interface with the network are formulated as in (4.117) on page 133 but with
x r and u m redefined as follows:

cd
E fd
and u m =
.
x r =

cq
T m

(4.188)

4.2.14 Generator parameter conversions

Conventionally synchronous generator parameters are provided in the form of standard parameters. These are based on manufacturers design calculations or the results of actual transient response tests, such as sudden short-circuit or load-rejection tests.
As explained in Section 4.2.12 the standard parameters are ambiguous if they are specified
without also specifying whether they conform to their Classical or Exact definitions. Furthermore, it is important to know if the standard parameters are saturated or unsaturated.
If saturation is being modelled explicitly then the unsaturated machine parameters should
be used. Conversely, if saturation is not being modelled explicitly, the saturated machine
parameters should be used, though this approach to the representation of saturation in
small-signal stability analysis is not recommended.
It is frequently the case that data supplied by manufacturers is based on the Classical definitions, although this is not always so. While the difference may be insignificant in the case
of salient pole machines, it may not be so when round rotor machines are modelled with
two q-axis damper windings. In the latter case it is possible that significant errors in predicted system responses will occur if the software assumes the parameters conform to the Classical definition when the standard parameters input to the program are actually specified
according to the Exact definitions or vice-versa.

Sec. 4.3

Small-signal models of FACTS Devices

157

Some software packages represent the machine internally using the coupled-circuit formulation although the user is permitted to input standard parameters. The software package
then internally converts the user-supplied standard parameters to coupled-circuit parameters. Again, to avoid ambiguity, it is important that the user is aware whether the software
requires that the standard parameters conform to the exact or classical definitions. Some
software packages internally formulate machine electromagnetic equations directly in terms
of the classically-defined standard parameters as summarized in Section 4.2.13. In such a
case, if the user has standard parameters for the machine which conform to the exact definitions then they should first be converted to conform with the classical definitions.
Figure 4.11 shows a parameter conversion roadmap. It is important to note that direct conversion between the classical and exact definitions of the standard parameters is unnecessary. Rather, if for example, conversion from exact to classical parameters is required then
the ES2CC transformation is applied to the exact standard parameters to yield the coupledcircuit parameters. These are then converted to the classical parameters using the CC2CS
transformation.

CC2ES

Exact
Standard Parameters

ES2CC

Classical
Standard Parameters

CS2CC

Coupled-Circuit
Parameters

CC2CS

Figure 4.11 Generator parameter conversion roadmap.

(Note: The conversions CS2CC and CC2CS are given in tables 4.11 and 4.10 respectively.
The conversions ES2CC and CC2ES are developed in [12] and software implementations
are provided in [24]).

4.3 Small-signal models of FACTS Devices

Equipment and plant which incorporate power electronics are referred to as components in
Flexible AC Transmission Systems or as FACTS devices. The linearized equations will be formulated for the small-signal models of five commonly-used FACTS devices, namely, the
Static VAR Compensator (SVC), the Static Synchronous Compensator (STATCOM), a
High Voltage Direct Current link with Voltage Source Converters (VSCX), a HVDC link
with line-commutated thyristor controlled converters (TCCX), and Thyristor-Controlled

158

Generators, FACTS devices & system models

Series Capacitors (TCSC). The core of many modern FACTS devices is the Voltage Sourced
Converter (VSC) for which the linearized equations are also summarized. The HVDC link
may comprise overhead lines, or underground or submarine cables. Details of these devices
and their operation, items which not directly relevant to the following sections, are described
in [10, 25, 26, 27, 29, 30, 31].
The following assumptions are made which are consistent with the requirements for rotorangle small-signal analysis. (i) The FACTS device is not operating at a limiting condition in
the steady state. (ii) The losses in power electronic converters are constant for small perturbations about the steady-state operating condition. (iii) Only the fundamental-frequency
components of the AC voltages and currents are relevant to the analysis. (iv) DC currents
are free of ripple. (v) Depending on the device, the DC voltage may contain some ripple the average value of the ripple voltage is then used. (vi) When analysing any device, all AC
related parameters are in per-unit on the specified base quantities of the device; at the AC
interface between the device and the network, conversions between the device and network
base quantities may be required. (vii) All DC related quantities are in SI units. As a consequence of these assumptions, (i) the very fast switching processes at the heart of power electronic devices are approximated by algebraic equations; and (ii) many of the sophisticated
limiting controls which are required under large-disturbance conditions are omitted from
the linearized model of the device.
The linearized DAEs for the devices are summarized in this section. However, to integrate
the DAEs of these device models with those of the AC network and the control systems to
which they are connected they must first be rewritten in the form described in Section 4.4.
The details of how such reformulation can be performed are provided in [12].
As shown in Figure 4.12 the general form of a linearized representation of a FACTS device
comprises n t AC terminals that are connected to AC network buses. For generality it is assumed that the base values of the interface quantities between the network buses and the
corresponding device terminals differ, as will be discussed in Section 4.3.1. For the kth terminal the perturbations in the real and imaginary components of the network bus voltage
n k

v RI

are applied to the device terminal through the base conversion factor 1 K V ; the
d k

current outputs i RI

from the device are injected into the network, again through a base
k

conversion factor of K I . The steady-state operating condition about which the model is
linearized is determined from the power flow solution for the buses to which the device terminals are connected.

Sec. 4.3

Linearized equations of quantities at AC terminals of FACTS Devices

159

Device initial steadystate values, SI units

or pu on device base
quantities

Figure 4.12

d 1

i RI

n 1

1
KI

d k

Terminal k.
Base quantities MB(k)

v RI

Terminal nt.
Base quantities MB(nt)

d n t

i RI

v RI

n k

KI

d n t

v RI

i RI

Bus to which
terminal 1 is
connected. Base
quantities SB(1)

n k

1 KV

d k

i RI

nt
KV

nt
KI

i RI

n n t

v RI

n n t

i RI

Bus to which
terminal k is
connected. Base
quantities SB(k)

Bus to which
terminal nt is
connected. Base

, 0k , P 0n k , Q0n k , k = 1 n t

Device outputs, SI
units or pu on device
base quantities

1 KV

n k

ties MB(1)

V0

Device inputs, SI
units or pu on device
base quantities

Terminal 1.
Base quanti-

n 1
v RI

Initial steady-state values of device terminal quantities from power flow solution

Device equations and parameters on device

base quantities, MB(k), k = 1,...,nt.

NETWORK
d 1
v RI

quantities SB(nt)

Interface between the network and a general FACTS device with nt

AC terminals.

Prior to considering a particular device, some results generally applicable to the AC terminals of FACTS devices are first derived.
4.3.1 Linearized equations of voltage, current and power at the AC terminals of
FACTS Devices: general results
As shown in Figure 4.12, a general FACTS device has multiple points of connection to the
AC network. In the following the connection of just one terminal, k, of the FACTS device
to its network bus is considered. The results are applicable to the connection of the other
terminals of the device to their respective network buses. As noted in Section 4.2.3.5 the results in this section are also applicable to calculating the quantities at the stator terminals of
generator models.
Normally the principal base quantities for the kth terminal of the device and the network bus
to which the terminal is connected are the RMS line-to-line voltage, the three-phase apparent power and the fundamental frequency as listed in Table 4.13. These base quantities are
specified by the user (as indicated by the subscript usb). The base value of time, for both
the device and network, t b , is chosen to be one-second.

160

Generators, FACTS devices & system models

Table 4.13

Principal base quantities for the kth device terminal and the network bus to
which it is connected.

Base Quantity
Device Network
Terminal
Bus
d k

V usb

Units

Description

The base value of the fundamental-frequency positivephase sequence component of the RMS line-to-line voltkV (RMS,
th
line-to-line) age of the k terminal of the device. Normally
d k
n k
V usb = V usb .

n k

V usb

The base value of the three-phase apparent power of the

d k

d k
S usb

n k
S usb

MVA

kth terminal of the device. Normally, S usb

is related to

n k

the device rating. The value of S usb is normally the

apparent power base of the power flow analysis.
The base value of fundamental frequency of the kth termi-

d k

f usb

d k

n k

Hz

f usb

nal of the device. Normally f usb

n k

that f usb
work.

n k

= f usb . It is assumed

is the nominal operating frequency of the net-

The set of base quantities for the kth terminal of the device is denoted as MB and the corresponding set of base quantities for the network bus to which the kth terminal is connected
is denoted as SB

The per-unit values of voltage, real and reactive power and current in the base system of the
of the kth terminal of the device (denoted by the superscript (d,k)) are related to the corresponding quantities in the network base system (denoted by the superscript (n,k)) as follows:
n k
k d k
k
d k
n k
V
= K V V
in which K V = V usb V usb ,

n k

+ jQ

n k

= KS P

d k

+ jQ

d k

in which K S

d k

(4.189)
n k

= S usb S usb ,

I n k = K I d k in which K k = K k K k .
I
I
S
V
n k

It is assumed that the initial steady-state values V 0

n k

, P0

(4.190)
(4.191)

n k

and Q 0

at the bus to

which the kth terminal of the device is connected are obtained from the power flow solution.

Sec. 4.3

Linearized equations of quantities at AC terminals of FACTS Devices

161

For the purposes of analysis the voltages and currents are represented by the real and imaginary components of their phasor quantities in the synchronously rotating network frame of
reference. Thus, in per unit on the base quantities of the kth terminal of the device,
d k
d k
d k
d k j
V
= vR
+ jv I
= V
e

= V

d k

cos

+ j sin

(4.192)

in which
V
d k

vR

d k

= V

d k

d k 2

vR

cos

d k 2

+ vI

d k

vI

= V

d k

d k

= atan 2 v I

sin

d k

and v RI

d k

The initial steady-state value of the voltage magnitude is V 0

d k

d k

V usb ; the initial value v RI

0

= 0

d k

vR

, and

(4.193)

= v d k v d k
R
I
n k

= V0

is then obtained by substitution of V

KV

d k

. (4.194)

per-unit on
d k

= V0

and

in (4.194). By substituting superscript d with n in equations (4.192) to (4.194)

the voltage magnitude and RI components are obtained in per-unit of the base voltage of
the network bus to which the kth terminal is connected.
The linearized forms of the voltage magnitude V
of the device are written, respectively, as

d k

and angle
T

V = v R V 0 v R + v I V 0 v I = 1 V 0 v RI v RI
0
0
0
= v I V 20 v R + v R V 20 v I = 1 V 20 v I
0
0
0

of the kth terminal

(pu), and

v R v RI , (rad)
0

(4.195)
(4.196)

in which the superscript (d,k) is omitted from the voltage quantities.

The terminal current phasor is expressed in terms of the terminal voltage and the real and
reactive power injected by the device from its kth terminal into the network as follows. All
quantities are in per-unit on the base quantities of either the kth terminal of the device (d,k)
or those of the network bus to which the kth terminal is connected (n,k).
+ jQ
I = i + ji = P
----------------
R
I

j
Ve
1
= --- P cos + Q sin + j P sin Q cos
V
1
= ------ Pv R + Qv I + j Pv I Qv R
2
V

(pu)

(4.197)

162

Generators, FACTS devices & system models

Equating the real and imaginary components of the preceding equation results in the following matrix relationship:
iR

1
1 v v
= --- cos sin P = ------ R I P .

2
V
iI
vI vR Q
sin cos Q
V
n k

The initial values of the current components i RI

n k

(4.198)

, in the network per-unit system, are

n k

n k

found by substituting the initial values V

, , P0
and Q 0
obtained from the
power flow solution into equation (4.198) and then in the per-unit system of the device ter d k

minal by setting i RI

n k

= i RI K I .
0

The perturbations in the current components about their initial values are obtained by linearizing equation (4.198) and by eliminating perturbations in the voltage magnitude using
(4.195) to give:
i RI = J is S + J iv v RI

(pu), in which S = P Q ,

1 vR0 vI0
1 P 0 2i R 0 v R0 Q 0 2i R0 v I 0
and
.
J is = ------
J iv = ------
V 20 v I v R
V 20 Q 0 2i I v R P 0 2i I v I
0
0
0
0
0 0

(4.199)
(4.200)

The current magnitude:

I =

(4.201)

iR + iI

is linearized about the initial steady-state operating point to yield:

T

I = i R I 0 i R + i I I 0 i I = 1 I 0 i RI i RI
0
0
0

(pu);

(4.202)

I 0 is obtained from (4.201) by substitution of the initial values of the current components
i .
RI0

The apparent power is P + jQ = V I , where all quantities are in per-unit on the bases of
either (i) the kth terminal of the device (d,k); or (ii) the network bus to which the terminal is
connected (n,k). From this relationship it follows that:
P = v R i R + v I i I (pu) and Q = v I i R v R i I (pu).

The perturbations in the real and reactive power are thus given by:

(4.203)

Sec. 4.3

Model of a Static VAR Compensator

163

P = v R0 v I0 i R + i R0 i I 0 v R
Q
v I v R i I
i I i R v I
0
0
0
0

(pu).

(4.204)

Consider the general case of an AC current flow I from terminal 1 to terminal 2 through a
reactance X connected between the terminals, the terminal voltage phasors being V 1 and
V 2 ; all quantities are in pu on the appropriate bases. Thus, V 1 = V 2 + jXI from which the

components of the terminal 1 voltage are:

v 1R = v 2R Xi I , and v 1I = v 2I + Xi R .

(4.205)

Because the equations (4.205) are linear, the linearized equations are formed by replacing the
variables by their perturbed quantities.

4.3.2 Model of a Static VAR Compensator (SVC)

The purpose of the SVC, through its control system, is to maintain the voltage of a busbar
constant by acting as a source or sink of reactive power. From small-signal considerations,
the SVC model comprising a controllable susceptance B , as shown in Figure 4.13, is the
simplest to construct. Since a wide variety of SVC control systems exist in practice the control system model is omitted from the following.
d
V

d
I

Figure 4.13 Model of a SVC

(All quantities are shown in per-unit of the SVC base values)
Since the SVC is a single-terminal device the terminal descriptor k is redundant and is thered
d
d
fore omitted from the following equations. The SVC equations are I = jB V ,

= 0 and Q

= B

d 2

each in per-unit of the SVC base quantities. The positive

direction of current and reactive power flow is from the SVC and into the network. The base

164

Generators, FACTS devices & system models

d

value of admittance on the SVC base quantities is Y b

n

bus base quantities is Y b

d 2

= S usb V usb and on the network

n 2

= S usb V usb . Thus, the SVC susceptance in per-unit of the

network bus base admittance is related to the per-unit value on the SVC base admittance by
B

= KY B

in which K Y = K S K V and where K V and K S are defined in (4.189) and

d

(4.190) respectively. The initial value of the SVC susceptance is B 0

n

where Q 0

and V 0

n 2

= Q0 V0 KY

are the initial steady-state values obtained from the power-flow solu-

tion.
d

The real and imaginary components of the SVC current are respectively i R = B
d

iI

= B

d d
vR

d d
vI

and

which upon linearization yield the following perturbations of the SVC cur-

rent components:
d

i R

i I

0
d

B0

B0

v R

v I

vI

vR

(4.206)

The perturbed output equations for the voltage magnitude and angle are given by equations
(4.195) and (4.196) respectively; the current magnitude is given by (4.202) and the reactive
power by (4.204).
The quantity Q V is the signed current generated by the SVC and is commonly used as
an input signal to the SVC control system for representing the current droop feedback. Noting that Q V = BV it follows that the perturbed variable is:
Q V = B 0 V + V 0 B = B 0 v R V 0 v R + B 0 v I V 0 v I + V 0 B
0
0

(4.207)

in which all quantities are in per-unit on the SVC base values.

The linearized equations of the SVC are written in the following matrix form in preparation
for interconnection with the transmission network, as described in Section 4.4. Note that
the model does not have any state-variables.
z

J
J
J
0 = gz gi gv i n + J gu u
RI
0
J iz J ii J iv
J iu
n
v RI

in which

(4.208)

Sec. 4.3

Model of a Voltage Sourced Converter

165

z = i d i d v d v d V d Q d Q V d I d
R
I
R
I

, u = B . (4.209)

B0

0 0 0 0

B0

0 0 0 0

KV

0 0 0 0

KV

0 0 0 0

d d
vR0 V0

d d
v I0 V 0

0 0 0 0

d
d 2
vI0 V0

J gz =

vI

vR

iI

d d d d
i R0 I 0 i I0 I 0

J gi = 0 , J gv = 0 0 1 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0

d
d 2
v R0 V 0 0 1 0 0 0

iR

0
d

0 1 0 0

B0

0 0 1 0

0 0 0 1

, J iz =

KI 0 0 0 0 0 0 0 0
0 KI 0 0 0 0 0 0 0

J iv = 0 , J gu = v d v d 0 0 0 0 0 V d 0
I0
R0
0

, J ii = 1 0 ,
0 1

, and J iu = 0 .

4.3.3 Model of a Voltage Sourced Converter (VSC)

The Voltage Sourced Converter (VSC) is a core component of a number of FACTS devices
including the Static Synchronous Compensator (STATCOM), VSC based HVDC transmission links, etc. [25, 26, 27, 28]. A generic VSC model, as depicted in Figure 4.14, is developed based on concepts in [28]. While this model is suitable for small-signal rotor-angle
stability analysis it is not applicable when the dynamic behaviour of a fast acting Pulse Width
Modulation (PWM) scheme, switching controls, and such-like is required.
The VSC acts as an AC voltage source where both magnitude and phase angle of the source
are controllable. As shown in Figure 4.14 the VSC model has two terminals, c1 and c2 on
the AC side. This allows the model to be used as a series connected element as well as a
d c2
shunt connected element. In the latter case the voltage V
of terminal c2 is constrained
to zero.

166

Generators, FACTS devices & system models

I ds

Id

c
Cc

c
V dc

I d c1

Vd

Rp

I dp

c1

d c1
V

Rc

I dc

c2

d c2
V
d c2
I
c

The interconnecting network between the VSC V n cr

AC terminals (c1) and
(c2) and the phase reference bus (cr) is reprecr
sented in the power

u1

u2

Figure 4.14

Generic VSC model.

The VSC base voltages and apparent power for terminals c1 and c2 are assumed to be,
d c1

V usb

d c2

= V usb

d c

= V usb

d c1

(kV), S usb

d c2

= S usb

n c1

the corresponding network bus base values are V usb

n c2

V usb

n c2

, S usb

d c

= S usb

(MVA);

n c1

and S usb

(4.210)

for terminal c1, and

for terminal c2.

From equations (4.189) and (4.190) on page 160 the conversion factors between the network and VSC base values for the c1 terminal voltage and apparent power are
c1

KV

d c

n c1

= V usb V usb

c1

= S usb S usb

c2

= S usb S usb

and K S

d c

n c1

d c

n c2

(4.211)

- and similarly for terminal c2:

c2

KV

d c

n c2

= V usb V usb

and K S

(4.212)

The voltage magnitude and phase angle at terminals c1 and c2 are found from (4.193) on
page 161 in terms of the real and imaginary voltage components of terminals c1 and c2,
d c1

v RI

d c2

and v RI

The source voltage magnitude and phase are controlled by the modulation ratio m

and

phase angle control both of which are inputs to the VSC. The magnitude of the source
voltage, which is the magnitude of the voltage difference between the two AC terminals (c1)
and (c2), is related to the modulation ratio as follows:
c

Vs

= kc m

c c
Vd

V usb .

(4.213)

Sec. 4.3

Model of a Voltage Sourced Converter

167

In the latter equation V d is the DC voltage across the converter in kV and k c

is the AC

to DC transformation ratio, which for a linear PWM scheme is 6 4 ; V usb in kV is the

internal AC base voltage of the voltage source [28].
The voltage angle at bus cr,
signal input

cr

rad, is the reference angle for the VSC phase-angle control

rad. Thus, the phase-angle of the voltage source is:

c

cr

(rad).

(4.214)

The internal voltage base V usb is usually chosen to equal the base voltage of the VSC ter d c

minals, V usb , when the VSC is shunt connected. However, for a series connected VSC the
internal base voltage may be selected in relation to the series voltage rating of the device.
The VSC internal and terminal voltage bases are thus related by the following,
c

d c

K C = V usb V usb .

(4.215)

The internal VSC apparent power base, S b , is defined to be equal to the VSC terminal VA
d c

base, S usb MVA. Thus, the VSC internal base value AC current, I b A, is defined as,
c
Ib
d c

where I b

d c

d c

d c

S usb 10
S usb 10 V usb
c d c
= ---------------------------- = ---------------------------- -------------- = K C I b
c
d c
c
3 V usb V usb
3 V usb

(A),

(4.216)

(A) is the VSC base current at its respective terminals.

The voltage source phasor is:

c

c
c j s
V s = V s e
c

where v sRI = v c v c
sR sI

= Vs

cos s sin s

d c1

The initial steady-state values v RI

= v sR + jv sI pu on V usb ,

d c2

and v RI

(4.217)
d c1

= K C v RI

d c2

v RI

; (4.218)

are calculated from the power flow solu-

tion as described in Section 4.3.1. These values are then substituted in (4.218) to calculate
c

the initial values of the voltage source components, v sRI .

The voltage source magnitude V s

ponents of the source voltage are:

and phase s , in terms of the real and imaginary com-

168

Generators, FACTS devices & system models

Vs

c 2

c 2

v sR + v sI =

c T c

v sRI v sRI , s

= atan 2 v sI v sR ,

(4.219)

which are linearized to yield the following expressions for these quantities about their initial
steady-state values:
c

V s

1 c T c
c
1
= --------- v sRI v sRI (pu),
= ----------------- v c
c 2
sI 0

V s c 0
V s0
0
c

where V s

c
v sR v sRI (rad), (4.220)

is obtained by substituting v sRI in (4.219).

It is assumed that the initial steady-state value of the DC voltage, V d , is a specified quantity.
0

Thus from (4.213) it follows that the initial value of the modulation ratio is,
c

m0

c c

c c

= V s V usb k c V d .
0

(4.221)

The perturbations in the VSC source voltage magnitude are obtained by linearizing equation
(4.213) to give:
c
V s

V c
V c
s0
c s0
c
c

pu on V usb .
= --------- V d + ---------- m
c
c
Vd
m0

(4.222)

The VSC source current components i sRI = i c i c , in per-unit on the internal VSC
sR sI

value of base current I b A, are related to VSC terminal currents on the terminal base cur d c

rent I b

, by the following relationships,

c

d c1

i
= i RI
sRI

d c2

K C = i RI

KC .

(4.223)

Furthermore, in accordance with (4.191) on page 160 the terminal c1 and c2 current components in the respective network bus per-unit systems are related to the corresponding values on the VSC terminal base current as follows:
n c1

i
RI

n c2

i
RI

c1 d c1
i RI

in which K I

c2 d c2
i RI

in which K I

= KI
= KI

The net real and reactive power, P s

c

P s + jQ s

c1

c2

c1

KV

c2

KV

= KS
= KS

c1

c2

and
.

(4.224)
(4.225)

and Q s , generated by the VSC are given by:

c c
c
d c
= V s I s pu on S b = S usb MVA

(4.226)

Sec. 4.3

Model of a Voltage Sourced Converter

169

which are calculated in terms of the source voltage and current components in accordance
c

with equation (4.203) on page 162 and the associated perturbations in P s

rived from (4.204).

and Q s

are de-

The often substantial switching losses in a VSC are represented by a user supplied fixed
c

d c

quantity P sl pu on S usb . Thus, the DC power output from the VSC is,
c

Pd

d c

= P s + P sl S usb

c c

= V d I d 10

MW,

(4.227)

where I d is in A and, as mentioned earlier, V d is in kV.

The initial DC power and current are respectively:
c

Pd

d c

= P s + P sl S usb
0

MW and I d

= P d V d 10
0
0

A.

(4.228)

Linearizing equation (4.227) gives the perturbation in the DC power output from the VSC:
c

P d

d c

= S usb P s

= P d V d V d + P d I d I d
0
0
0
0

MW.

(4.229)

As shown in Figure 4.14 there are two shunt elements across the DC terminals (i) a resistc

ance R p across the terminals, (ii) a capacitance C c

F and a resistance R c

in series

across the terminals. In the most general case the relevant equations are:
c

capacitor state equation:

pV dc = I dc C c 10

voltage drop across shunt resistor:

Vd

voltage drop across RC element:

Vd

Kirchoffs Current Law:

Id

= R p I dp 10

= V dc + R c I dc 10

c c

kV,

c c

kV/s,

= I dp + I dc + I ds A.

(4.230)
(4.231)

kV,

(4.232)
(4.233)

Because (4.230) to (4.233) are linear equations the variables can be replaced by their perturbed values. There are some special cases for which the above equations can be modified,
namely,
(a) no shunt elements are modelled;
(b) only the resistance across the terminals is connected;
(c) only the resistance and series capacitance combination is present.
c

It is convenient to define a variable U d representing the external input to the DC terminal

of the VSC which is normally defined as:

170

Generators, FACTS devices & system models

c

= I ds (A) or U d

Ud

(kV)

= Vd

(4.234)

depending on whether the source on the DC side of the VSC is represented as a current or
voltage source.
c

This VSC model has two control input signals denoted by u 1

and u 2 . From a detailed

modelling perspective these two variables are typically (i) the VSC phase-angle measured
with respect to the phase reference bus cr,
m

, and (ii) the VSC voltage modulation ratio

resulting in the following constraints:

c

u1

(rad) and u 2

= m

(4.235)

It is often desirable to provide a simplified functional representation of the VSC control systems, for example, (i) in scoping studies before the details of the VSC control systems are
known, or (ii) in studies where detailed representation of the VSC controls have an insignificant effect on the dynamic performance of the system. For example, if the VSC is equipped
with a fast acting control system whose objective is to maintain constant DC power flow
c

then control input u 1 can be set to P d rather than

. By doing so it is unnecessary to

represent the details of the DC power control loop of the VSC.

4.3.3.1

Summary of linearized VSC equations

Listed below is a consolidated list of the linearized DAEs of the VSC.

DC side equations
Note that the DC side equations can be modified depending on the representation of the
c

DC shunt network. In all such alternative representations the variables P d , V d , I d

c

and I ds are retained.

c
c
1
pV dc = ------------------------------ I dc (kV/s) from (4.230),
c
3
C c 10
c

d c

0 = P d + S usb P s
c

(MW), power conservation from (4.229),

c

0 = P d V d V d + P d I d I d P d
0
0
0
0

(MW) from (4.229),

(4.236)
(4.237)
(4.238)

c
c c
3 c
0 = V dc V d + R c 10 I dc (kV) from (4.232),

(4.239)

c c
3 c
0 = V d + R p 10 I dp (kV) from (4.231),

(4.240)

Sec. 4.3

Summary, linearized VSC equations

c

171

0 = I d + I ds + I dc + I dp (A) from (4.233).

(4.241)

VSC AC side converter equations

c

d c1

d c2

0 = v sRI + K C v RI

K C v RI

(pu) from (4.218),

c

0 = v sR V s v sR + v sI V s v sI V s
0
0
0
0

(4.242)

(pu) from (4.220),

(4.243)

c
c
c
c
c 2
c
c 2
0 = v sI V s v sR + v sR V s v sI s (rad) from (4.220),
0

0
0
n cr

0 = v sI
0

n cr 2

Vs

n cr

v sR

n cr

+ v sR
0

n cr 2

Vs

n cr

v sI

cr

(4.244)

(rad) (4.245)

from (4.196),
0 =
0

v sR

v sI

c
v sI
0

i sR

c
v sR
0

c
i sI

i sR

i sI

i sI i sR
0

v sR

0
0

c
v sI

P s

c
Q s

(pu)

(4.246)

from (4.204) on page 163,

0 =

c
s

cr

(rad) from (4.214),

(4.247)

V c
Vc
s0
c
c s0
c

(pu) from (4.222).

0 = --------- V d V s + ---------- m
c
c
V
m
d
0

(4.248)

VSC terminal 1 / network bus interface

d c1

0 = i sRI + 1 K C i RI

c1

0 = KI
c1

0 = KV

d c1

i RI

d c1

v RI

n c1

i RI

n c1

v RI

(pu) from (4.223),

(4.249)

(pu) from (4.224),

(4.250)

(pu) from (4.189) pg. 160.

(4.251)

VSC terminal 2 / network bus interface

Note that if the VSC is shunt connected equations (4.252) to (4.254) are omitted from the
d c2

model as are variables i RI

(4.242).

n c2

and v RI

d c2

0 = i sRI + 1 K C i RI

c2

0 = KI

d c2

i RI

n c2

i RI

n c2

, including the coefficients of v RI

in

(pu) from (4.223),

(4.252)

(pu) from (4.225),

(4.253)

172

Generators, FACTS devices & system models

c2

0 = KV

d c2

v RI

n c2

v RI

(pu) from (4.189) pg. 160.

(4.254)

Input to / Output from the DC side of the VSC

0 =

c
I ds

+ U d

(A) and 0 = V d + y d
c

or conversely, 0 = V d + U d

(kV) from (4.234),

c

(kV) and 0 = I ds + y d

(A).

(4.255)
(4.256)

VSC control inputs

c

As mentioned earlier, in a flexible VSC model, it is desirable for the control inputs u 1
c

and u 2 to be constrained to two independent variables depending on the users requirements. For detailed VSC modelling the inputs are set to the VSC phase angle and voltage
modulation respectively:
0 =

+ u 1

0 = m

(rad) from (4.235),

c

+ u 2

(4.257)

from (4.235),

(4.258)

Examples of useful alternative constraints on the first control input are:

c

0 = V d + u 1 , or 0 = P d + u 1 , or 0 = y 1 + u 1

(4.259)

in which y 1 is a remote AC power system signal such as the power flow in a transmission
element. Similarly, examples of alternative constraints on the second control input are:
c

0 = Q s + u 2 , or 0 = y 2 + u 2
c

in which y 2

(4.260)

is a remote AC power system signal such as the reactive power flow in a

transmission element or the voltage magnitude of remote bus.

The linearized VSC equations (4.236) to (4.260) can be assembled into a compact matrix
form which is suitable for integration with the linearized DAEs of the power system as outlined in Section 4.4 and elaborated on in [12].
4.3.4 Simplified STATCOM model
The purpose of the STATCOM is similar to that of the SVC, that is, to control voltage by
absorbing or generating reactive power. The STATCOM model in this section is based on
a simplified version of the VSC model described in Section 4.3.3. The AC terminal (c2) of
the VSC in Figure 4.14 on page 166 is connected to the zero voltage reference plane and the
AC terminal (c1) is renamed to (c). That is, the VSC at the heart of the STATCOM is shunt
connected. The basic concepts on which this model are based are found in [25, 26] , for example. The simplified VSC model, which is depicted in Figure 4.15, makes the following idealizing assumptions.

Sec. 4.3

Simplified STATCOM model

173

The VSC phase control system is sufficiently fast and accurate that the generated AC
current phasor is assumed to always be in quadrature with the voltage phasor at the
phase reference bus as depicted in Figure 4.15(b). This has the consequence that the
VSC input u 1 =

= 0 (see Section 4.3.3, page 170).

c

The VSC fixed losses, P sl , are negligible.

The VSC DC capacitor is sufficiently large that the DC voltage is assumed to be constant during small-disturbances. Thus the VSC capacitor is depicted as a fixed DC
source in Figure 4.15(a).

A lossless reactor is connected between the converter AC terminal (c) and the phase
reference bus as shown in Figure 4.15(a). This reactor is included in the model and
therefore the VSC phase reference bus is AC terminal (s) in the simplified VSC model
of the STATCOM.
(a) Below, power-circuit schematic
showing VSC with constant DC voltage
and reactor between the converter (c)
and phase-reference (s) terminals.
c

jX

I d s

d s
V

d s
V

(b) Above, quadrature relationship between the VSC terminal

voltage and current phasors.

u1 u2
c
(c) Right, equivalent control-system
Vs
block-diagram representation of the VSC

for the simplified STATCOM model.

I d s if V c V d s
s

Vd
c
V s

d s
d s
c
I
if V s V

1X

d s

abs
I

d s

---
2

d s
V

I d s

Figure 4.15 Simplified VSC model for a STATCOM: (a) power-circuit schematic, (b) relationship between terminal voltage and current phasors, (c) equivalent control-system
block-diagram of the VSC.
Based on the above assumptions the equivalent control-system block-diagram representation of the simplified VSC model for the STATCOM shown in Figure 4.15(c) is now formulated.

174

Generators, FACTS devices & system models

In accordance with Table 4.13 on page 160 the base values of voltage and apparent power
for the AC terminal of the VSC are specified respectively as
d s

V usb

d s

(kV) and S usb

(MVA). The corresponding quantities for the network bus to

n s

which the STATCOM VSC terminal (s) is connected are V usb

n s

(kV) and S usb

(MVA).

The voltage phasor at the phase reference terminal (s) is,.

d s
d s j
V
= V
e

d s

in per-unit on V usb ,

(4.261)

and is regarded as an input to the VSC model derived from the voltage at the network bus
to which the terminal is connected.
The voltage phasor at the converter terminal (c) is:
c

d c
c
c j s
V
= V s = V s e

c j

c j

d s

in per-unit on V usb . (4.262)

= Vs e
The equality of the phases of the voltages at the (s) and (c) terminals is derived from equa= Vs e

tion (4.214) on page 167 with the assumption, stated earlier, that the VSC phase-angle control input,

, is zero.

Since it is assumed that the DC voltage is constant it follows from (4.213) on page 166 that
the magnitude of the voltage at VSC AC terminal (c) is equal to the VSC modulation ratio
multiplied by a constant gain factor, i.e.
c

Vs

= V

d c

where the constant

d c

= k c V d V usb .

(4.263)

Thus, for modelling purposes, it is permissible and convenient to eliminate the modulation
ratio and instead treat the VSC AC terminal voltage as the second control input signal to the
c

VSC, i.e. u 2 = V s .
The current output from the simplified VSC model is,
s
j ---
s
2
d s j
d s
1 c
I d s = ----------- V V
e
= I
e
,
d s
jX

(4.264)

where the magnitude of the injected current, which is in quadrature with the voltage, is
I

d s

= Vs V

in per-unit on the VSC current base of

d s
Ib

d s

d s

S usb
3
= ---------------------- 10 (A).
d s
3V usb

(4.265)

Sec. 4.3

Simplified STATCOM model

175

Thus, equations (4.264) and (4.265) which represent the current injected by the simplified
VSC model into the network, are equivalent to the control-system block-diagram in
Figure 4.15(c).
A voltage control system can then be combined with the above simplified VSC model in
order to regulate the voltage at the STATCOM terminal, (s). This results in a simplified
model of a STATCOM such as that depicted in Figure 4.16. In this model the STATCOM
d s

current droop feedback is explicitly represented by V dr = K d I

. In common with the
SVC, the STATCOM voltage control system usually incorporates current droop designed to
reduce the voltage at the regulated bus in proportion to the current injected by the STATCOM into the network. Typically, the droop setting, K d , is between 0.01 and 0.05 per-unit
on the STATCOM base quantities. It is also possible to add a supplementary control signal
to the AVR summing junction from, say, a Power Oscillation Damper (POD).
Voltage control system

V ref

V dr

abs

AVR(s)

d s

Vs

1X

Kd

d s

---
2

d s
V

I d s

Simplified VSC model.

Figure 4.16 Simplified STATCOM model formed by combining the simplified VSC model in Figure 4.15(c) with a voltage control system.
The small-signal representation of the simplified VSC model is now derived. As described
d s

in Section 4.3.1 the initial steady-state values of the voltage v RI

d s

and current i RI

com-

ponents in the VSC per-unit system are obtained from that the initial steady-state values of
n s

V0

n s

, P0

n s

and Q 0

obtained from the power flow solution. It is noted that in

the power flow the STATCOM is assumed to be represented as a PV bus 1 in which, for
consistency with the assumption that the STATCOM current is in quadrature with its ter n s

minal voltage, P 0

= 0.

1. In power flow terminology the term PV bus denotes a bus at which the generated
power and voltage magnitude are specified; the generated reactive power and voltage
angle are the unknowns to be determined by the power flow solver.

176

Generators, FACTS devices & system models

With appropriate changes in notation equation (4.193) on page 161 is used to derive the
STATCOM voltage magnitude V
V

d s

d s

from the voltage components v RI

d s

d s 2

vR

as

d s 2

+ vI

which is linearized about the operating point, in accordance with (4.195) on page 161, to
yield:
V

d s d s d s d s d s d s
.
= vR
V 0 v R
+ vI
V 0 v I
0

d s

(4.266)

From (4.196) on page 161, with appropriate changes in notation, the perturbation in the
STATCOM terminal voltage phase angle is:

v d s
v d s
I0
R0
d s

d s .
= --------------------- v R
+ -------------------- d s 2
d s 2 v I
V0
V0

(4.267)

The perturbation in the STATCOM current magnitude is obtained by linearizing equation

(4.265) to yield,
I

d s

= V s V

d s

(4.268)

From equation (4.264) the STATCOM current phasor can be rewritten as,
s
I d s = i d s + ji d s = I d s cos s
--- + j sin --- ,

R
I

2
2

(4.269)

from which it follows that

d s

iR

= I

d s

sin

d s

and i I

= I

d s

cos

(4.270)

The perturbations in the current components output from the STATCOM terminal are derived by linearizing the preceding equations at the operating point to yield,
d s

i R

= sin 0 I

d s

d s

iI

d s

and i I

= cos 0 I

d s

d s

+ iR

. (4.271)

From (4.204) on page 163, with appropriate changes in notation, it follows that the perturbation in the reactive power output from the STATCOM is:
Q

d s

= v d s
I0

d s

vR

d s

i RI

+ i d s
I0

d s

iR

d s

v RI

(4.272)

In accordance with equations (4.189) and (4.190) on page 160 the perturbations in the
STATCOM terminal voltage and current components, in per-unit on the base values of the
network bus to which the STATCOM is connected, are respectively,
n s

v RI

d s

= K V v RI

n s

and I RI

d s

= K I i RI

(4.273)

Sec. 4.3

Modelling of HVDC Transmission Systems

177

As for the SVC model, the linearized equations (4.266) to (4.268) and (4.271) to (4.273) for
the simplified STATCOM model can be rewritten in the following matrix form suitable for
interconnection with the transmission network model. Note that the model does not have
any state-variables. The linearized equations are of the form:
z

J
J
J
0 = gz gi gv i n + J gu u
RI
0
J iz J ii J iv
J iu
n
v RI

in which

z = i d s i d s v d s v d s V d s s I d s Q d s
R
I
R
I

(4.274)

and u = V s .

Note that these equations do not include the model of the STATCOM voltage regulator or
any supplementary control system they represent only the simplified VSC component of
the STATCOM.
4.3.5 Modelling of HVDC Transmission Systems
Two models of HVDC transmission systems are considered with the general structure in
Figure 4.17. The first is based on Voltage Sourced Converters (VSCXs) and the second on
line-commutated Thyristor Controlled Converters (TCCXs). The HVDC transmission system may comprise overhead lines or cables; the system can be either mono-polar or bipolar.
For both the rectifier and inverter it assumed that the number of bridges in series, NB, is the
same as the number of converter transformers operating in parallel on the AC side.
r

d r
V

Rectifier

(r)
I d r

I ds

I dR

r
Vd

L
V dR

HVDC
Transmission System

(L)

I dI

I ds

L
V dI

i
Vd

Inverter

d i
V

(i)
I d i

Figure 4.17 Structure of HVDC system model showing the interface between the DC side
components. The directions of positive voltage and current-flow at the external interfaces
of the components are indicated by the arrows.
Because the models of both VSCX and TCCX systems employ the same model of the
HVDC link, the model of the link is analysed first. Models of the VSCX and TCCX systems
are then considered in turn.
The HVDC models described in the following sections assume that the following information is provided by the power flow solution: the steady-state voltages, the real and reactive
power flows at the AC terminals of the rectifier and inverter, and the initial steady-state values of the DC voltages of each converter. All plant on the AC side of the converter terminals, e.g. three-winding transformers, reactors, is assumed to be modelled in the power flow

178

Generators, FACTS devices & system models

analysis. The nature of the interface between the AC terminals of the rectifier and inverter
and the buses to which they are connected in the AC system is shown in Figure 4.12.
4.3.6 Model of a distributed-parameter HVDC transmission line or cable
In the small-signal model of a long, HVDC transmission line or cable it is not possible to
represent the strictly distributed-parameter nature of the circuit. Let us assume that the dynamics of the line is adequately represented by NT T-sections, of which the kth section is
shown in Figure 4.18. Clearly, the greater the number of sections the closer to a distributedparameter system the model becomes. Let:
RL
LL
CL
RCL

- total link resistance in (ohm),

- total link inductance in (Henry),
- total link capacitance (Farad),
- total cable sheath resistance (ohm).

The parameters of each T-section in SI units are therefore:

L = LL/(2 NT), R = RL/(2 NT), C = CL/NT, RC = RCL/NT.
IL(k-1)

Figure 4.18

IL(k)

VC(k)
Vd(k-1)

IC(k)

RC

Vm(k)

Vd(k)

The kth T-section of a long, HVDC transmission line or cable.

The circuit diagram of the HVDC line / cable depicted in Figure 4.19 comprises the series
connection of NT of the above T-sections. Note that:
1.

The series resistance R r (ohm) and inductance L r (H) connected to the rectifier end
r

of the link (R) includes the resistance R r

(ohm) and inductance L r

(H) of the

series smoothing reactor, if any, connected to the rectifier, i.e.

r

Rr = R + Rr

2.

and L r = L + L r .

(4.275)

Similarly, the series RL branch connected to the inverter node (I) includes the series
smoothing reactor, if any, connected to the inverter, i.e.
i

Ri = R + Rr

and L i = L + L r .

(4.276)

Sec. 4.3
3.

Modelling distributed-parameter HVDC transmission links

179

The parameters of the internal series RL elements are 2R and 2L because the right
arm of the kth T-section is connected in series with the left arm of the (k+1)th T-section.
R

Rr

Lr

2R

I dR

V dR

RC

2L

IL 1

2L

k-1

RC

RC

2R

IL k 1

2R

2L

2L

k+1

2R

NT

Li

Ri

ILk

RC

RC

RC

RC

I dI

V dI

Figure 4.19 Circuit diagram of a HVDC link model comprising the series
connection of NT of the T-sections in Figure 4.18 and the rectifier and inverter smoothing
reactors.
L

The rectifier- and inverter-end voltages, V dR and V dI , are treated as model inputs. The recL

tifier- and inverter-end currents, I dR = I L 0 and I dI = I L N

+ 1 ,

are outputs from the

link model which are to be input to the devices connected to the respective DC terminals.
The state-equations for the inductor currents are written first, followed by the state-equations for the capacitor voltages and finally the algebraic nodal voltage and current equations.
L

All currents are in Amperes (A) and all voltages are in Volts (V) - except V dR and V dI
which are in kV.
Inductor current state-equations
L

pI L 0 = R r L r I L 0 + V dR 10 V m 1 L r ,

(4.277)

pI L k 1 = R L I L k 1 + V m k 1 V m k 2L , k = 2 N T ,

(4.278)

pI L N = R i L i I L N + V m N V dI 10 L i .
T
T
T

(4.279)

Capacitor voltage state-equations

pV C k = I C k C , k = 1 N T .

(4.280)

Nodal voltage and current equations

0 = V C k V m k + R C I C k , k = 1 N T ,

(4.281)

0 = I L k 1 + I L k + I C k , k = 1 N T .

(4.282)

180

Generators, FACTS devices & system models

Note that the equations (4.277) to (4.282) are linear. Each variable can therefore be replaced
L

by its perturbed value at the steady-state value, e.g. the variable V dR is replaced by V dR at
L

the steady-state value V dR . Given the initial steady-state values of the rectifier- and inverter0

end voltages, the initial values of the other variables are obtained by solving the aforementioned set of linear equations with the rates of change in the inductor currents and capacitor
voltages set to zero. The initial values of the rectifier and inverter end currents are given by:
I dI = I dR = V dR V dI R L 10
0
0
0
0

A.

(4.283)

Simplified modelling of the HVDC transmission line / cable is appropriate when the link is
short or when detailed modelling such as that described above is otherwise unnecessary or
infeasible. For example, the link may be represented as a series RL branch incorporating the
rectifier and inverter smoothing reactors, or simply as a series resistance.
4.3.7 Model of HVDC transmission with Voltage Sourced Converters (VSCX)
In the model of a VSCX the rectifier (r) in Figure 4.17 is represented by a shunt connected
VSC model as described in Section 4.3.3. The linearized equations for the rectifier are listed
in Section 4.3.3.1 in which the superscript (c) is replaced by (r). The linearized model of the
inverter is similarly represented - but with (c) replaced by (i). For both converters the DC
input and output are the perturbations in the DC current and voltage respectively, i.e. equation (4.255) on page 172 applies to both converters. Since the converters are shunt connected the equations and variables associated with the second terminal, (r2) and (i2), of the
respective converters are omitted from the model equations as described in Section 4.3.3.
The HVDC link is represented by the equations (4.277) to (4.282). The interconnections between the linearized equations for the converters and HVDC link are represented by the following linear constraint equations:
r

0 = I ds + I dR (A), 0 = I ds + I dI (A), 0 = V d V dR (kV), 0 = V d V dI (kV) (4.284)

Typically, one of the converters is used to control the power transferred by the link and the
other converter is used to control its DC voltage. Each converter is normally used to control
either its reactive power output or the AC voltage of an adjacent bus. The control systems
are specific to the application and are omitted from the VSCX model.
The linearized model equations of the VSCX system can be written in the following form
which is suitable for integration with the linearized DAEs of the power system as described
in Section 4.4. The superscript (T) denotes that the quantity is associated with the integrated
VSCX system.

Sec. 4.3

HVDC transmission with Voltage Commutated Converters, TCCX

px

0
0

J fx J fz

0
x
0

n T
T

= J
+ J
v RI + J T u c
z
gu
gx J gz J gi
gv

T T i n T
0
0
0 J iz J ii
RI

181

(4.285)

The state ( x ), algebraic ( z ) and control-input, ( u c ) variables in the preceding

equation combine the corresponding variables from the HVDC link model and the rectifier
and inverter models. The current components injected by the VSCX into the AC buses - to
n T

which the rectifier and inverter AC terminals are connected - are i RI

n T

and the associated

bus voltages are v RI . The latter current and voltage quantities are in per unit on the base

values of the buses to which the converter terminals are connected.

4.3.8 Model of HVDC transmission with Voltage Commutated Converters
A small-signal model of a HVDC transmission system with voltage-commutated thyristorcontrolled converters is now presented. The resulting TCCX model has the structure shown
in Figure 4.17. A general purpose converter model, depicted in Figure 4.20, is formulated
on the basis of [10, 25, 29, 30, 31, 32, 33]; it is then used in a modular fashion to represent
the converter at one end of the link operating in rectifier mode and at the other end as an
inverter. The distributed parameter model of the HVDC link in Section 4.3.5 is used to connect the rectifier and inverter.
k

Lr

Rr

Smoothing reactor
included in model of
HVDC link. See
Section 4.3.6

I dx

d k
V
k

V dx

d k

AC network
represented in
power flow

Artificial network for calculating the converter com-

d k

mutation voltage V C
d k
V
d k

j XC
I d k

d k
V C

NB
d k
IC

Figure 4.20 Voltage-commutated thyristor-controlled converter. The artificial network

for calculating the converter commutation voltage is also shown.
In the following the converter identifier indicated by the superscript (k) is substituted either
with (r) for the rectifier or (i) for the inverter. The AC terminal quantities are defined in per-

182

Generators, FACTS devices & system models

unit on the device base values as specified in Table 4.13 on page 160. The DC side quantities
are in SI units as specified below.
k

As mentioned in Section 4.3.5, the converter is assumed to comprise N B identical bridges

connected in series on the DC side and in parallel on the AC side.
4.3.8.1

Internal DC voltage and current variables

For the purpose of developing the voltage-commutated thyristor-controlled converter modk

el, internal values of the DC current, I dx (A), and DC voltage, V dx (kV), are employed. In
Section 4.3.8.9 the relationships are established between these internal DC quantities and
their corresponding external interface values shown in the general HVDC link model of
Figure 4.17.
It is important to note that the DC current must flow in the direction of the thyristor valves
k

and by definition this is the positive direction of I dx . Consequently I dx is positive whether

the converter is operating as a rectifier or as an inverter. The direction of positive DC voltk

age V dx is defined to be in the positive direction of I dx . Thus, as will be seen in the followk

ing, V dx is positive if the converter is operating as a rectifier and is negative when operating
as an inverter.
4.3.8.2

Relationship between the AC and DC quantities of the converter

In the analysis of voltage-commutated converters it is assumed that an ideal, sinusoidal,

three-phase source is connected to the converter through a reactance, which is referred to
d k

as the commutating reactance, X C

per bridge. The AC voltage source is referred to as the

commutating voltage. Assuming a thyristor firing-angle delay,

(rad), and an overlap an-

gle, (rad), due to the commutating reactance, it is shown in the analysis of [29] that the
average DC voltage at the converter is,
k

V dx = V d0 cos

R C 10 I dx , (kV),

(4.286)

in which the DC current is I dx (A) and where the no-delay, no-load DC output voltage V d0
d k

(kV) is given in terms of the magnitude of the commutating voltage magnitude, V C

(per-

d k

unit on V usb kV):

k

k d k

V d0 = c vd V C

k d k

(kV), where c vd = --- 2 N B V usb (kV), and

(4.287)

Sec. 4.3

HVDC transmission with Voltage Commutated Converters, TCCX

k
RC

k d k
c rd X C

k
c rd

(ohm), in which

d k 2
3 k V usb

= --- N B ---------------------
d k
S usb

(ohm).

183

(4.288)

The so-called commutation resistance, R C , is an artefact that accounts for the DC voltage drop due to commutation overlap; it does not generate power losses.
k

By substituting for V d0 from (4.287) into equation (4.286) the DC voltage is obtained in
terms of the commutating voltage magnitude, the DC current, firing-angle delay and commutation resistance as follows:
k

k d k

V dx = c vd V C

cos

R C 10 I dx (kV).

(4.289)

In using this converter model the value of the commutation reactance must be selected, ideally it is the Thvenin impedance looking into the AC system from the converter AC terminals. The use of the commutation bus is therefore an artefact for the calculation the DC
k

voltage, V dx . Importantly (i) the transformers and associated components connected to the
AC terminals of the converter are represented in the AC network model; (ii) the commutation reactance and commutation bus are not included in the AC network model.
The converter switching losses are assumed to be fixed and are represented by a user sup d k

plied quantity P sl

d k

pu on S usb . Thus, due to the conservation of energy, the DC power

output from the converter is related to the AC power output P

k

Pd

= P

d k

d k

+ P sl

d k

S usb

k k

= V dx I dx 10

d k

by:

MW.

(4.290)

As mentioned earlier, it is assumed that the initial steady-state values of the following quantities are obtained from the power flow solution: (i) the real and reactive power output from
n k

the converter AC terminal, P 0

angle at the AC terminal,

n k
V0

initial steady-state values

d k
P0
,

n k

and Q 0

respectively; (ii) the voltage magnitude and

n k
0

respectively; and (iii) the DC voltage V dx . The

and

d k
Q0
,

d k
v RI
0

and

d k
i RI ,
0

are calculated from the power

flow solution as described in Section 4.3.1. The initial steady-state values of the DC power
and current are then obtained from (4.290) as follows:
k

Pd

d k

= P0

d k

+ P sl

d k

S usb

(MW) and I dx = P d V dx 10
0

(A).

(4.291)

The perturbation in the DC power output of the converter is now obtained by linearizing
equation (4.290) about the above operating point values.

184

Generators, FACTS devices & system models

k

P d
4.3.8.3

d k

= S usb P

d k

= V dx 10 I dx + I dx 10 V dx .
0

(4.292)

Perturbations in commutation voltage in terms of converter terminal quantities

From Figure 4.20 the commutating voltage phasor is obtained from the conditions at the
converter AC terminals as follows.
d k

k
X C d k
d k
d k
d k j C
d k
d k
d k
pu on V usb ,
V C
= VC e
= v CR + jv CI = V
j -------------- I
k
NB

(4.293)

from which,
d k
v CRI = v d k v d k
CR
CI

d k

= VC

cos C sin C
T

d k
X C d k

= v
+ -------------- i I

R
N Bk

d k

The initial steady-state values v RI

d k

and i RI

(4.294)

d k
d k X C d k
-------------- i R
vI
N Bk

calculated in the previous section are then

substituted in (4.294) to calculate the initial values of the commutating voltage components,
k

v CRI .
0

d k

Based on (4.193) on page 161, the commutating voltage magnitude V C

d k

VC

d k 2

d k 2

v CR + v CI

d k T d k

v CRI v CRI ,

is:
(4.295)

which is linearized according to (4.195) to yield the following expression for the perturbation in the commutation voltage about its initial steady-state value,
d k

V C

1 d k T d k
= -------------- v CRI v CRI (pu),
0

V Cd k

(4.296)

d k

where V C

d k

is obtained by substituting v CRI in (4.295).

The initial steady-state firing-angle delay 0 is obtained by back substituting into (4.289),
k

d k

V dx from the power flow solution, I dx from (4.291) and V C

0
0
0

(calculated above), to give:

V k + R k 10 3 I k
dx 0
C
dx 0
.
= acos --------------------------------------------------------- k d k

c vd V C

(4.297)

k
0

Sec. 4.3

HVDC transmission with Voltage Commutated Converters, TCCX

185

The perturbation in the DC voltage about its initial steady-state value is obtained by linearizing equation (4.289) to yield:
k

d k

V dx = c vd cos 0 V C
4.3.8.4

k d k

c vd V C

sin 0

R C 10 I dx . (4.298)

Perturbation in real and reactive power supplied from the commutation bus

d k
The current output from the converter terminal, I
, is equal and opposite to that output
d k
from the commutation bus, I C , i.e.

I d k = I d k .
C

(4.299)

The real and reactive power supplied from the commutation bus is:
d k

PC

d k

+ jQ C

d k d k
= V C I C .

(4.300)

d k
Substituting in the preceding equation for V C
in terms of the converter terminal voltage
d k
and current from (4.293) and then substituting for I C
from (4.299) yields:
d k

PC

d k

+ jQ C

d k

d k X C d k d k
= V
j -------------- I

N Bk
d k

X C d k 2
d k d k
= V
I
+ j -------------- I

N Bk
= P

d k

d k

d k X C d k 2
jQ
-------------- I

N Bk

From the preceding equation the real and reactive power output from the commutation bus
is expressed in terms of the real and reactive power output from the converter AC terminal,
P

d k

and Q

d k

d k

PC

, and the magnitude of the AC converter current, I

= P

d k

d k

and Q C
d k

The initial steady-state value I 0

d k

, i.e.:

d k

d k X C d k 2
= Q
-------------- I
.

N Bk

(4.301)

is obtained from (4.201) on page 162 using the previous d k

ly determined current components i RI

d k

. Then P C
d k

ing the previously determined values of P 0

d k

, Q0

d k

and Q C

d k

and I 0

are derived by substitutin (4.301).

186

Generators, FACTS devices & system models

Let

phasor
cos

(rad) be the angle by which the fundamental-frequency commutation bus current

I d k
C

d k
lags the rms phase-neutral commutation voltage phasor V C . The factor

is the power factor - or displacement factor - of the fundamental waveforms and

is a function of the converter firing-delay and extinction-delay angles,

(rad), respectively.

(rad) and

An approximate value of the power factor is

cos

= cos

+ cos

2 .

(4.302)

However, an exact expression for the power factor angle is provided in [29]:
tan

= T

2 + sin 2 sin 2
G
= ---------- = ------------------------------------------------------------------------------------------------- .
k
k
k
H
cos 2 cos 2

(4.303)

The reactive power output from the commutation bus is related to the real power output
and the power-factor angle,
d k

QC

, by:
d k

= PC

tan

d k k

= PC

(4.304)

Equations (4.303) and (4.304) are now used to calculate the initial steady-state value of the
d k

extinction-delay angle. The initial steady-state values P C

d k

and Q C

were determined

earlier in this section and 0 was determined in (4.297). Thus, from (4.304) it follows that
k

d k

= atan 2 Q C

d k

PC

k
k
1. Substituting for the known values of T 0 = tan 0 and

in (4.303) results in the following non-linear equation for the initial value of the extinction-delay angle:
k

k
k
k
= c 0 + sin 2 0 T 0 cos 2 0 2 , where

(4.305)

c 0 = 2 0 sin 2 0 + T 0 cos 2 0 is known.

1. Note that the atan2 function is used to ensure that the power-factor angle is located in
the correct quadrant. This is important when the converter is operating as an inverter.
See the footnote on page 126 for the definition of atan2.

Sec. 4.3

HVDC transmission with Voltage Commutated Converters, TCCX

187

The preceding equation can be solved using any one of a number of methods including a
fixed-point iteration method or the Newton-Raphson method with an initial estimate of
k

determined from (4.302).

Let

(rad) be the overlap angle, then

(4.306)

It is noted that inconsistencies can arise between the steady-state solution of converter variables in the power flow and the initial conditions determined for the purpose of linearizing
the converter model. Such inconsistencies are frequently attributable to the way in which the
power-factor angle is expressed as a function of the thyristor firing and extinction angles.
The linearized form of (4.303) is:
T

= 1 H 0 G
k

= c t
0

where
and

T 0 H 0 H

+ c t

(4.307)

k
k
k
k
k
c t = 2 H 0 cos 2 0 + T 0 sin 2 0 1 ,
0

k
k
k
k
k
c t = 2 H 0 cos 2 0 + T 0 sin 2 0 1 .
0

(4.308)

The perturbation in the reactive power output from the commutation bus is thus obtained
by linearizing equation (4.304) about the operating point to give:
d k

Q C
4.3.8.5

d k

= T 0 P C

d k

+ PC

(4.309)

Perturbations in converter AC current and apparent power output

From equations (4.290) and (4.301) the following expressions for the real and reactive power
output from the converter terminals are found:
P

d k

k
Pd

d k
S usb

d k
P sl

and Q

d k

d k
QC

d k

X C d k 2
+ -------------- I
.
N Bk

(4.310)

The perturbations in these quantities about the operating point are obtained by linearizing
the preceding equations to yield:
P

d k

= 1

d k
k
S usb P d

and Q

d k

d k
Q C

d k d k

X C I 0 d k
. (4.311)
+ 2 ---------------------------- I
k

NB

188

Generators, FACTS devices & system models

The perturbations in the AC converter current components about the operating point are
obtained in terms of the perturbation in the converter apparent power output
( S

d k

= P d k Q d k

) and the perturbations in the AC terminal voltage compo-

nents by application of equations (4.199) and (4.200) on page 162 to yield:

d k

i RI

d k

= J is

d k
J is

d k

J iv

d k

d k

+ J iv

1 2
= --------------
V 0d k

d k

d k

v RI

d k

vR

d k
vI
0

(pu), in which

(4.312)

d k

vI

d k
vR
0

, and

d k

d k d k

d k d k

2i R v R Q 0
2i R v I

1 2 P0
0
0
0
0
.
= --------------
V 0d k Q d k 2i d k v d k P d k 2i d k v d k
0
I
R
0
I
I
0

(4.313)

The perturbations in the converter AC current required in equation (4.311) are obtained by
applying equation (4.202) on page 162 to the converter AC terminal (k) as follows:
I
4.3.8.6

d k

d k

= 1 I0

d k T

d k

i RI i RI
0

(pu).

(4.314)

Modifications of the general converter model for inverter operation

The voltage-commutated thyristor-controlled converter model developed above is general

in that it applies whether the converter is operating as a rectifier or inverter. When the conk

verter is operated as an inverter the firing-angle delay must be greater than 2 rad in
order to produce a DC voltage that opposes the flow of the DC current through the thyristor valves. The DC voltage produced by the rectifier forces the DC current through the inverter valves against the opposing inverter DC voltage. When describing the operation of
the inverter it is common practice to refer to the firing-angle advance,
extinction-angle advance,

(rad), and the

(rad) which are related to the corresponding delay angles by:

and

(4.315)

Thus, although it is mathematically unnecessary, in the case of inverter operation the above
equations of the advance angles are added to the general converter equations in accordance
with conventional practice.
4.3.8.7

Converter control input

The converter model has a single control input signal denoted by u

. From a detailed modk

elling perspective this input is usually the thyristor firing-angle delay, . However, it is
also common when representing the inverter control system to employ either the firing-an-

Sec. 4.3

HVDC transmission with Voltage Commutated Converters, TCCX

k

189

gle advance, , or the extinction-angle advance, , as the control signal. Thus, one of
the following constraints on the converter input signal is typically employed for detailed
modelling of the converter controls:
u

(rad) or u

(rad) or u

(rad).

(4.316)

It is often desirable to provide a simplified functional representation of the converter control systems. For example, if the converter is equipped with a fast acting control system
whose objective is to maintain constant DC power flow then the control input u
set to

k
Pd

4.3.8.8

rather than

can be

Summary of the linearized converter algebraic equations

A consolidated list is presented below of the linearized algebraic equations for the voltagecommutated thyristor-controlled converter model. The initial steady-state operating condition is determined from the power flow solution as described in the preceding sections.
DC side equations
k

d k

0 = c vd cos 0 V C

k d k

c vd V C

sin 0

R C 10 I dx V dx (4.317)

from (4.298),
k

0 = V dx 10 I dx + I dx 10 V dx P d
0

from (4.292).

(4.318)

Commutation bus equations

T

d k
X C d k
0 = v d k + ------------- i I

R
N Bk

d k
d k X C d k
-------------- i R
v I
N Bk

d k

v CRI

(4.319)

from (4.294),
1 d k T d k
d k
0 = -------------- v CRI v CRI V C
d k
0

VC

from (4.296),

(4.320)

0 = P

d k

d k

+ P C

0 = c t
0

from (4.301),

+ c t

from (4.307) and (4.308) where T

k

d k

0 = T 0 P C

d k

+ PC

d k

Q C

(4.321)

= tan

(4.322)
k

from (4.309).

(4.323)

190

Generators, FACTS devices & system models

Converter AC terminal equations
k

d k

0 = P d + S usb P
0 =

d k
Q C

d k

0 = J is

d k

from (4.311),

d k d k

X C I 0 d k
d k
+ 2 ---------------------------- I
Q
k

NB

d k

(4.324)

d k

+ J iv

d k

0 = 1 I0

d k

d k

v RI

d k T

i RI

d k

i RI i RI
0

from (4.311),

(4.325)

from (4.312) and (4.313),

d k

(4.326)

from (4.314).

(4.327)

Advance angle equations

0 =

and 0 =

from (4.315).

(4.328)

Input control signal equation

As mentioned in Section 4.3.8.7 it is desirable in a flexible converter model for the control
input signal to be constrained to an independent variable depending of the users specific
requirements. For detailed modelling the input control signal will typically be defined by one
of the following equations:
0 =

or 0 =

or 0 =

, from (4.316). (4.329)

Alternatively if the functional behaviour of the converter control system is to be represented

k

then the input is constrained to some other user-selected system variable y c

k

0 = y c u

such that:

(4.330)

4.3.8.9 Relationship between the internal model and external interface values of the DC voltage and current in the general HVDC transmission model.
k

As described in Section 4.3.8.1 the positive direction of the DC current variable I dx used in
the internal formulation of the converter model is in the direction of the thyristor valves.
k

The positive direction of the DC voltage variable V dx used in the formulation of the model
k

is in the positive direction of I dx . The rectifier (r) and inverter (i) define external interface
variables for the DC voltage and current for the respective converters in the general model
of the HVDC transmission system shown in Figure 4.17 on page 177. Figure 4.21 shows the
relationship between the internal and external interface DC current and voltage variables for
two scenarios:
1.

The DC current flow is from the rectifier to inverter for which the relationships
between the internal and external variables are:

Sec. 4.3

HVDC transmission with Voltage Commutated Converters, TCCX

r

for the rectifier: I ds = I dx

i

and

Vd

= V dx ,

(4.331)

for the inverter: I ds = I dx and V d = V dx .

2.

191

(4.332)

The DC current flow is from the inverter to rectifier for which the relationships
between the internal and external variables are:
r

for the rectifier: I ds = I dx

i

for the inverter: I ds = I dx

and V d = V dx ,
and

(4.333)

V d = V dx .

(4.334)

Of course, in both cases the DC power flow is from the rectifier to the inverter; in the first
case the DC current and power flow in the same direction, whereas in the second case the
direction of DC power flow is opposite to that of the DC current
r

I dx

I ds

V dx
Rectifier
(Internal)

Vd

I ds

HVDC Link

Rectifier
External
Interface

Vd

Inverter
External
Interface

I dx
i

V dx

Inverter
(Internal)

(a) DC current flow from the rectifier to the inverter.

r

I dx

I ds
r

V dx
Rectifier
(Internal)

Vd

Rectifier
External
Interface

I ds

HVDC Link

Vd

Inverter
External
Interface

I dx

V dx

Inverter
(Internal)

(b) DC current flow from the inverter to the rectifier.

Figure 4.21
4.3.8.10

Relationship between converter internal and external DC variables.

Assembly of the TCCX model

In the small-signal model of a TCCX the rectifier (r) in Figure 4.17 is represented by a voltage-commutated thyristor-controlled converter model in accordance with the linearized
equations listed in Section 4.3.8.8. In the rectifier model the superscript (k) in the general
equations is replaced by (r). The relationship between the internal values of the rectifier DC
voltage and current - and the corresponding external interface values of these quantities - is
defined by either equation (4.331) or (4.333) depending on the direction of DC current flow.
The linearized model of the inverter is similarly represented but with (k) replaced by (i). The

192

Generators, FACTS devices & system models

HVDC transmission line/cable is represented by the equations (4.277) to (4.282) on

page 179. The interconnections between the linearized equations for the converters and
HVDC line/cable are represented by the following linear constraint equations:
r

0 = I ds + I dR (A), 0 = I ds + I dI (A), 0 = V d V dR (kV), 0 = V d V dI (kV). (4.335)

It is possible to implement a wide range of control strategies with this general purpose model. The principal control strategies are: (i) the rectifier operates in constant-current control
and the inverter operates in constant extinction angle control, or (ii) the rectifier operates in
constant firing-angle control and the inverter operates in constant-current control. Supplementary control strategies to achieve a range of objectives are possible, including regulation
of power flow, regulation of frequency and damping control. The control systems are specific to the application and are omitted from the model.
The linearized model equations of the TCCX system can be written in the following form
which is suitable for integration with the linearized DAEs of the power system as described
in Section 4.4. The superscript (T) denotes that the quantity is associated with the integrated
TCCX system.
px

0
0

J fx J fz

0
x
0

= J T J T J T z T + J T v RI + J T u c
gu c
gx gz
gi
gv

n T
T T
0
0
0 J iz J ii i RI

The state, the algebraic, and the control-input variables ( x

, z

(4.336)

, u c ) in the preced

ing equation combine the corresponding variables from the rectifier and inverter models and
the HVDC line/cable model. The current components injected by the TCCX into the AC
n T

buses to which the rectifier and inverter AC terminals are connected are i RI
n T
v RI .

and the as-

sociated bus voltages are

These latter current and voltage quantities are in per-unit

of the base values of the network buses to which the converter terminals are connected.
4.3.9 Thyristor Controlled Series Capacitor (TCSC)
A model for the TCSC suitable for small-signal rotor-angle stability analysis is shown in
Figure 4.22.
It is assumed that under steady-state conditions the TCSC is represented in the power flow
n

as a series susceptance, b 0 , between buses j and k as shown in Figure 4.22(a). The superscript (n) denotes that the susceptance is in per-unit on the network base quantities. The
n 1
n 2
steady-state voltages at buses j and k, V j0 = V 0
and V k0 = V 0
respectively, are also

obtained from the power flow solution.

Sec. 4.3

Thyristor Controlled Series Capacitor (TCSC)

193

j
Remainder of power-system

n 1
V j 0 = V 0

j
jb 0

n 2
V k0 = V 0

I d 1

d 1
V

k
I d 2

d 2
V

(a) Steady-state representation of

TCSC in power-flow.

(b) Dynamic representation of TCSC.

Figure 4.22 Representation of the TCSC (a) under steady-state operating conditions in
the power flow, and (b) as a dynamic device with controllable series reactance.
Thus in the complex network nodal current equations in (4.337) below the TCSC is repren

sented by the admittance y jk = y kj = jb 0

I
1

Y 11

Y 1j

of the branch between buses j and k.

Y 1k

V 1

|
|
|
|
|
|
|
I = Y Y + + Y +

V
j1
j1
jk
jk
j
j
I
Y k1
Y kj
Y k1 + + Y kj +
V k
k
|
|
|
|
|
|
|

(4.337)

where V m and I m , m = 1 N , are respectively the voltage and current injected into bus
m, each in per-unit on the network base quantities.
In order to represent the TCSC in a dynamic model of the system, those terms associated
with the steady-state representation of the TCSC are deleted from the nodal admittance matrix. At bus j, these are respectively the series element y jk and term y jk in the self admittance term Y jj . The same approach is adopted at bus k for the terms y kj and y kj . In the
dynamic model depicted in Figure 4.22(b) the series branch is replaced by equivalent current
injections from terminals 1 and 2 of the TCSC into the network buses j and k. It is assumed
d

that the control input is the reactance X of the series capacitor. The superscript (d) denotes that the quantity is in per-unit on the TCSC base quantities.
The methodology outlined below can be adapted to derive the relationships of greater complexity for the current flows through a series branch whose impedance (or admittance

194

Generators, FACTS devices & system models

y = g + jb ) is to be controlled dynamically; there may also be shunt elements at terminals 1

and 2 in Figure 4.22(a).

It is assumed that the following network base quantities are specified for buses j and k to
which terminals 1 & 2 of the TCSC are connected:
n

n 1

V usb = V usb

n 2

= V usb

n 1

(kV) and S usb = S usb

n 2

= S usb

(MVA).

(4.338)

Correspondingly, it is assumed that the TCSC device base quantities are:

d

d 1

V usb = V usb

d 2

= V usb

d 1

(kV) and S usb = S usb

d 2

= S usb
d

The base value of admittance in the TCSC per-unit system is Y b

n

(MVA).
d

(4.339)

d 2

= S usb V usb ohm and

n 2

= S usb V usb ohm. Thus, the susceptance in the

in the network per-unit system it is Y b

network and TCSC per-unit systems are related by

b

= KY b

(4.340)

in which K Y = K S K V and where K V and K S are defined in accordance with (4.189) and
d

(4.190) on page 160. Thus, b 0

= b0 KY .

d
d
d
Let I = i R + ji I be the phasor current flow from terminal 2 to 1 of the TCSC and let
d
d 2 d 1
d
d
d 2
d 1
d 2
d 1
V
= v R + jv I = V
V
= vR
vR + j vI
vI

(4.341)

be the voltage difference between the terminals of the TCSC. The TCSC terminal currents
are:
I d 1 = i d 1 + ji d 1 = I d
R
I

d 2
d
d 2
d 2
and I
= iR
+ ji I
= I .

(4.342)

d
d d
d d
d d
The TCSC current/voltage characteristic is I = jb V
= b v I + jb v R from
which it follows that:
d

iR

= b

d d
vI

and i I

= b

d d
vR .

(4.343)
d

The initial steady-state value of the TCSC voltage difference is V 0

n 2 n 1
= V 0
V0
KV .

The TCSC control input is the series reactance rather than the series susceptance. The two
quantities are related by b
b

= 1 X

d 2

for which the linearized form is:

= 1 X 0 X

d 2

= b 0 X

(4.344)

Sec. 4.4

Linearized power system model

195

The perturbation in the TCSC current components about their initial steady-state values are
found by linearizing (4.343) and substituting for b
d

i R

= b 0 v I

d 2

v I b 0 X
0

and i I

from (4.344) to yield:

d

d 2

= b 0 v R + v R b 0 X
0

. (4.345)

The linearized form of the relationships between the TCSC current i RI = i d i d

R
I

and its terminal currents are found from (4.342) to be,

d 1

i RI

d 2

= i RI

and i RI

= i RI

(4.346)

The linearized form of the equations describing the interconnection between the device terminals and the network buses to which they are connected are:
n 1

i RI

n 1

v RI

d 1

= K I i RI

n 2

= K I i RI

n 2

= K V v RI

i RI

d 1

, v RI

= K V v RI

d 2

for the current and

d 2

for the voltages.

(4.347)
(4.348)

The linearized equations (4.345) to (4.348) are sufficient to represent the TCSC. However,
it is also desirable to provide supplementary output equations for perturbations in the following quantities at one or both of the TCSC terminals:

the voltage magnitude and angle by application of (4.195) and (4.196) on page 161;

the real and reactive power flow by application of (4.204) on page 163;

the magnitude of the current flow by application of (4.202) on page 162.

The TCSC linearized model equations (4.345) to (4.348) together with the supplementary
output equations can be written in a form suitable for integration with the linearized DAEs
of the power system as described in Section 4.4.

4.4 Linearized power system model

A fundamental assumption in this work is that during steady-state operation all generators
and loads produce balanced three-phase fundamental-frequency voltages and currents. It is
assumed the interconnecting AC network is three-phase and balanced. Consequently, for
steady-state modelling purposes, a fundamental-frequency positive phase sequence representation of the generators, loads and AC transmission system is employed. In this representation the balanced three-phase system voltages and currents are represented by
fundamental-frequency positive phase sequence stationary complex phasors. For the purpose of small-disturbance modelling the stationary assumption is relaxed to permit small
perturbations in the current and voltage phasor magnitudes and phases. The resulting con-

196

Generators, FACTS devices & system models

cept of quasi-stationary, fundamental-frequency, positive phase sequence phasors [34]

forms the basis for rotor-angle small-signal stability analysis of power systems in this section.
A high-level mathematical description of how the linearized differential and algebraic equations (DAEs) of the interconnected system are assembled from (i) the DAEs of each of the
devices and their associated control systems, and (ii) the network nodal admittance equations which are provided below. The resulting equations have a modular and extremely
sparse structure which must be exploited to ensure computationally efficient analysis of
large power systems.
4.4.1

General form of the linearized DAEs for a device and its controls

It is assumed that the i

th

device has an ordered list of n t AC terminals which are connected

i

to a corresponding list of network buses b i k , k = 1 n t . It is also assumed that the

i

device and its controls do not have inputs from any other device. (Note that the equations
of devices that are interconnected through their control systems are combined to form a sinth

gle super-device). The i device and its associated controls is represented by a set of linearized DAEs of the following form:
nt

px i =

i
J fx x i

i
J fz z i

0 =

i
J gx x i

0 =

i b i m
J ix
x i

nt

i
J gz z i

l=1
nt
i

l=1

i b i m
J iz
z i

l=1

u i

i b i l

J gi

i b i m b i l

i b i m

i b i l

J fi

J ii
J iu

i b i l

i RI

i b i l

+ J fv

i b i l

+ J gv

bi l

v RI

bi l

v RI

+ J fu u i

+ J gu u i

(4.349)

i b i l

i RI

i b i l

i RI

i b i m b i l

+ J iv

for m = 1 n t

bi l

v RI

in which x i , z i and u i are respectively vectors of n x states, n z algebraic variables and

nu

external

i b i l

i RI

input-variables,

= i i b i l i i b i l
R
I

injected by the l
bi l

v RI

th

associated
T

the

th

device.

Moreover,

are the real and imaginary components of the current

terminal of the i

= v bi l v b i l
R
I

with

th

device into bus b i l , l = 1 n t

and

are the real and imaginary components of the voltage at bus

Sec. 4.4

Network nodal current equations

197

b i l , l = 1 n t . The voltage and current components are in the synchronously-rotating

i

network frame of reference, and are each in per-unit on their respective network base quani

tities. The constant coefficient matrices J fx , J fz , etc. are typically sparse and depending on
the device some of the coefficient matrices may be zero.
4.4.2 General form of the network nodal current equations
It is assumed that bus k is connected to a list d k of n d current injecting dynamic devices
k

and to a list c k of n c immediately adjacent buses through series admittance elements. Each
k

element of d k contains the identifier of both the device and the terminal within the device
which is connected to the bus. Most buses in large sparse networks do not have any dynamic
devices connected to them and so in most cases d k is empty. Applying Kirchoffs Current
Law to bus k results in the following nodal current equations, one for the real component
of the current and the other for the imaginary component.
nc

0 =

k
Y kk v RI

in which Y kk =

ck l
Y kl v RI

l=1

G kk B kk
B kk G kk

nd

d i k

k
i RI

(4.350)

i=1

and Y kl =

G kl B kl
B kl G kl

(4.351)

correspond respectively to (i) the sum of all admittance elements connected to bus k, and (ii)
the negated total series admittance between buses k and l (i.e. admittances of parallel branches between two nodes are summed). These equations are sparse in the sense that they involve the voltages at a very small subset of the buses in the network.
The network nodal current equations for all of the buses k = 1 n b are now expressed
in the following matrix form in which the buses connected to dynamic devices are partitioned from the internal passive buses. It is emphasised that computationally sparse matrix
storage and analysis methods are used in which only the non-zero admittance blocks in each
nodal current equation are stored and analysed. Furthermore, for many purposes distinguishing between the dynamic and passive buses is unnecessary. The network nodal current
equations are:
d

0 = Y dd Y dp v RI J di i ,
RI
0
Y pd Y pp v p
0
RI

in which:

(4.352)

198

Generators, FACTS devices & system models

v RI =

v RI

and v RI

n db

v RI

are respectively vectors of voltage components (i) of the n db buses to which dynamic devices (d) are connected; and (ii) the remaining set of passive buses (p). Furthermore,
1

i RI

i RI =

nd

i RI

is the vector of the current components injected by the n d dynamic devices into the buses
to which they are connected. For the ith device the current injection vector is composed of
n t elements, one for each of its AC terminals, such that:
i

i b i 1

i RI =

i RI

(4.353)

i b i n t
i

i RI

4.4.3 General form of the linearized DAEs of the interconnected power system
The equations for each dynamic device with the form in (4.349) are interconnected through
the network nodal current equations (4.352) to yield the following general form of the linearized DAEs of the interconnected power system.
px
J fx J fz J fi J fv

0
J gx J gz J gi J gv
0 =
J ix J iz J ii J iv
0
0 0 J di Y dd
0
0 0 0 Y pd
0

J fu
x

z
J gu
0

i RI
+ J iu u
0

d
0
Y dp v RI

0
Y pp v p
RI
0

x 1
u 1
z 1

In equation (4.354) x = , z = , u =

x n
z n
u n
d
d
d

(4.354)

are respectively the system

state-variables, the internal algebraic-variables and the external system input variables of all

Sec. 4.4

Linearized DAEs of the power system

199

the dynamic devices connected to the system. The coefficient matrix J fx has the following
block diagonal structure:
nd

J fx = D J fx J fx ,

(4.355)

and the coefficient matrices J fz , J fu , J gx , J gz and J gu similarly have a block diagonal

structure. The structure of the coefficient matrices associated with the current and voltage
components such as J fi , J gv and J iz depend on the relative ordering of the dynamic devices and the buses to which they are connected. In any event, for large systems, these matrices
are also very sparse.
4.4.4 Example demonstrating the structure of the linearized DAEs
For illustrative purposes the structure of the linearized DAEs of the small power system in
Figure 4.23 is given in Figure 4.24. The system model in Figure 4.23 incorporates a device
(1) with two terminals connected to two buses (1 and 2); buses 1 and 2 are each connected
to two devices (1 and 2) and (1 and 3) respectively; and finally the bus (3) is connected to
one dynamic device (4). This interconnection of dynamic devices and buses represents the
range of possibilities usually encountered in practice.

1
2 1

1 1

i RI

1 2

i RI

3 2

i RI

v RI

i RI

v RI

v RI

v RI

v RI

4 3

i RI

Figure 4.23 Example network to illustrate the modular and sparse structure
of the linearized DAEs of the interconnected power system.

J fx

(2,1)

J fz
3
4

(4,3)

(1,1)

J gi

(2,1)

J gz
3
4

(1,1)
(1,2)

(1,1)

(2,1)

(2,1,1)

(4,3)

J iv
(3,2,2)

J iz
J ix

(2,1,1)

J ii

(3,2,2)

J ii
(4,3)

J iv
(4,3,3)

J iz

(4,3,3)

J ii
I
I

J iv

P Y
11

I
I

Y 13
Y 22

Y 31

Y 14

R
Y 25

Y 33

Y 35

T
S Y
44 Y 45

Y 41
Y 52

Y 53

J fu

J fu

J fu
1

Y 54 Y 55

i RI

(1,2)
i RI

(2,1)
i RI

(3,2)
i RI

(4,3)
i RI

1
v RI

2
v RI

3
v RI

4
v RI

5
v RI

u 1

u 2

u 3

u 4

J gu

J gu

J gu

J gu
1 1

(1,1)

J iv

(3,2)

J ix

(1,1,1) (1,1,2)
J iv

J iv

(1,2,1) (1,2,2)
J iv

J ii
J iz

(3,2)

(1,1,1) (1,1,2)
J ii

J ii

J gv

J fu

Figure 4.24 Structure of the linearized DAEs of the system in Figure 4.23.

J iu

1 2

J iu

2 1

J iu

3 2

J iu

4 3

J iu

Generators, FACTS devices & system models

(2,1)

(4,3)

J gi

(1,2,1) (1,2,2)
J ii

J iz
J ix

(3,2)

J gv
(4,3)

J gz
J iz

(1,2)

J gv

J gv
J gi

(1,2)

J ix

(1,1)

J gv

(3,2)

J gz
J gx

(4,3)

J fv

(2,1)

J gi
3

J gx

J ix

(1,2)

J gi

J gx

J fv
J fi

x 1

x 2

x 3

x 4

z 1

z 2

z 3

z 4

(3,2)

J fi
J fz

F J1
gz

(1,2)

J fv

J fv
(3,2)

J fz
J fx

(1,1)

J fv

(2,1)

J fi
3

J fx

J gx

(1,2)

J fi

J fx

(1,1)

J fi

200

A J1
fz

px 1

px 2

px 3

px 4

Sec. 4.5

Load models

201

In Figure 4.23 blank space indicates the corresponding matrix entries are zeros. The matrix
block labelled A corresponds to the sub-matrix J fx in equation (4.354), block B corresponds to J fz , etc.

4.5 Load models

4.5.1 Types of load models
In large scale power system stability studies loads are typically aggregated at bulk supply substations. In rotor-angle stability studies a static representation of loads is commonly employed, as reported in a recent international survey [37]. In such a representation the real and
reactive power consumed by the load at a point in time is dependent on the voltage and frequency at the same instant [35, 36]. Composite load models, such as those described in [38],
which incorporate an equivalent representation of the distribution network as well as a variety of dynamic and static load model components are not considered in this book. The following is a general representation of a static load model.
P0
V mp1
V mp2
V mp3
P = ------ a 1 ------
+ a 2 ------
+ a 3 ------
1 + af f f0
V 0
V 0
V 0
as
Q0
V m q1
V mq2
V mq3
+ b 2 ------
+ b 3 ------
1 + bf f f0
Q = ------ b 1 ------

V0
V0
V 0
bs

(4.356)

in which P and Q are the real and reactive power consumed by the load in per-unit on the
base MVA of the system; V is the terminal voltage of the load in per-unit of the base voltage
of the bus to which the load is connected; and f is the frequency of the load bus voltage in
per-unit of the system base frequency. P 0 , Q 0 , V 0 and f 0 = 1 are the corresponding initial
steady-state values. The load model parameters are a i , m pi , b i , m qi for i = 1 3 , a f and
b f . Furthermore, in (4.356)
3

as =

i=1

a i and b s =

bi .

(4.357)

i=1

The formulation in (4.356) can be used to represent a range of commonly employed load
models such as the ZIP (composite constant impedance, current and power) representation
and the exponential load model.
Three basic types of static loads can be represented by the following:
1.

A constant impedance load: a1 = 1 and a2 = a3 = 0, mp1 = 2; b1 = 1 and b2 = b3 = 0,

mq1 = 2, af = bf =0;

202
2.

Generators, FACTS devices & system models

A constant current load: a1 = a3 = 0 and a2 = 1, mp2 = 1; b1 = b3 = 0 and b2 = 1,
mq2 = 1, af = bf =0;

3.

A constant power load: a1 = a2 = 0 and a3 = 1, mp3 = 0; b1 = b2 = 0 and b3 = 1,

mq3 = 0, af = bf =0.

Note that the load can be any linear combination of the above and the type of the real and
reactive parts may differ. For example, a load may have a constant-current real component
and a constant-impedance reactive component; this combination is the most commonly
used static load model in the absence of any further information or measurement on the load
characteristics of the system under study [35]. The frequency dependence of loads is not
modelled.
Work, such as [38, 40], shows that loads may have a significant impact on the damping of
rotor modes and therefore accurate dynamic modelling, whenever possible, of loads is highly desirable in both transient and small-signal stability studies.
4.5.2 Linearized load models
For the purposes of small signal analysis, equations (4.356) are readily linearized about the
steady-state operating point to yield:
P = C pv V + C pf f

and Q = C qv V + C qf f ,

(4.358)

P
1
in which C pv = -----0- n p , C pf = P 0 a f , and n p = ----- a i m pi
as
V0
i=1

(4.359)

and C qv , n q and C qf are calculated analogously.

The above small-signal model of a load is connected to the network by relating the real and
reactive power, voltage magnitude and bus voltage frequency to the voltage components,
v RI , of the bus to which the load is connected and the current components, i RI , injected by

the load into the bus. The reference for the voltage and current components is the synchronously rotating network frame of reference; the voltage and current are each in per-unit on
their respective network base quantities. It is assumed that the initial values of the real and
reactive power, P 0 and Q 0 , and the voltage magnitude and angle, V 0 and 0 are given by
the power flow solution. From these initial values the following are derived:
v R = V 0 cos 0 , v I = V 0 sin 0 .
0
0

The voltage magnitude is V =

point to yield:

(4.360)

v R + v I which is linearized about the steady-state operating

Sec. 4.5

Linearized load models

203

vR
vI
V = -------0 v R + ------0 v I .
V0
V 0

(4.361)

The perturbation in the frequency of the load bus frequency is given by:
1 d
f = ------ ---------- ,
b dt

(4.362)

in which b is the base value of system frequency in rad/s and is the perturbation in
the bus voltage angle in radians and t is time in seconds. It is important to recognise that
is not a state-variable and therefore that rate-of-change of bus voltage angle is approximated by means of a highpass (i.e. washout) filter with a very short time constant, for which
2
T f = f ------ with f = 0.1 is a reasonable choice1. Thus, in the s-domain the bus frequen b

cy perturbation is approximated by:

1
s
f s = ------ ---------------- s .
b 1 + sT f

(4.363)

Transformation of the preceding equation to the time-domain results in the following stateand algebraic equation:
1
1
px f = ---- x f -------------
T f
T 2
b f

(4.364)

1
f = x f + ------------
b T f

The bus voltage angle is = atan 2 v I v R which upon linearization yields the following expression for :
vI
vR
= ------0 v R + -------0 v I .
V 20
V 20

(4.365)

The current injected by the load into the bus to which it is connected is given by:
P + jQ
1
i R + ji I = ------------------ = ------ Pv R + Qv I + j Pv I Qv R .
2
v R + jv I
V

(4.366)

By linearizing the real and imaginary components of the preceding equation about the initial
steady-state operating point the following equation for the perturbation of the current components injected by the load into the network are obtained:
1. Some time domain analysis programs utilize integration algorithms which require integration time-steps to be shorter than the shortest model time-constant. For use in such
programs a larger value of f may be required.

204

Generators, FACTS devices & system models

i RI = J iz z J iv v RI ,

(4.367)
T

in which z = P Q V f ,

(4.368)

1 v R 0 v I0 2V 0 i R0 0 0
1 P Q
J iz = ------
and J iv = -----2- 0 0 .
2
V 0 v I v R 2V 0 i I 0 0
V 0 Q 0 P 0
0
0
0

(4.369)

The initial steady-state values of the current components are obtained from (4.366). The linearized state- and algebraic-equations (4.358), (4.361), (4.364), (4.365) and (4.367) representing the load are now amalgamated into the following matrix equation which is in the general
form of the device equations in (4.349) on page 196.
px f
0
0

J fx J fz 0

1 0 C pv C pf
0 1 C qv C qf
0 0

0 0

(4.370)

1
J fz = 0 0 0 0 ------------2- ,
b Tf

0 0 1

x f

z
= J gx J gz 0 J gv

i RI

0 J iz I J iv
v RI

in which J fx = ---- ,
Tf

J gz =

0
0

0
, and J gv =
1

1 ----------- b T f
0

J gx = 0 0 0 1 0
0
0

0
0

v R 0 v I0
------- ------
V 0 V 0 ;
0
v
I
------0
V 20

(4.371)

0
v
R 0
------2-
V0

J iz and J iv are defined in (4.369).

If the frequency dependence of loads is neglected then equation (4.370) is simplified by

omitting the state-variable x f , the algebraic-variables f and , and the associated equations. In this case it is also possible to omit the explicit equations for the load and instead
implicitly include the effect of the voltage dependence of the load by modifying the self-admittance of the bus to which the load is connected as follows. It can be shown that the perturbation of the current injected by the load is given by:
i RI = Y L v RI .

(4.372)

Sec. 4.6

References

v R0
P 0 + c R -------
1
V0
in which Y L = -----2-
V 0

vR
Q 0 c I -------0

V0

205

v I 0
Q 0 + c R ------

V 0

v I0
P 0 + c I ------

V 0

vR
vI
c R = P 0 -------0 n p 2 + Q 0 ------0 n q 2 and
V0
V 0
vI
vR
c I = P 0 ------0 n p 2 Q 0 -------0 n q 2 .
V 0
V0

where

(4.373)

(4.374)

4.6 References
[1]

Siemens PTI PSSE, developed and distributed by Siemens Power Technologies

International (PTI), See URL (correct at 4 March 2015): http://w3.siemens.com/
smartgrid/global/en/products-systems-solutions/software-solutions/planningdata-management-software/pti/pages/power-technologies-international-(pti).aspx

[2]

GE PSLF, developed and distributed by General Electric Company, See URL

(correct at 4 March 2015): http://www.geenergyconsulting.com/practice-area/software-products/pslf

[3]

W. G. Heffron and R. A. Phillips, Effect of a Modern Amplidyne Voltage Regulator

on Underexcited Operation of Large Turbine Generators [includes discussion],
Power Apparatus and Systems, part iii. Transactions of the American Institute of Electrical Engineers, vol. 71, 1952.

[4]

F. P. de Mello and C. Concordia, Concepts of Synchronous Machine Stability as Affected by Excitation Control, Power Apparatus and Systems, IEEE Transactions on,
vol. PAS-88, pp. 316-329, 1969.

[5]

A. E. Blondel, C. O. Mailloux and C. A. Adams, Synchronous Motors and Converters:

Theory and Methods of Calculation and Testing. McGraw-Hill Book Company, 1913.

[6]

R. H. Park, Two-reaction theory of synchronous machines generalized method of

analysis-part I, American Institute of Electrical Engineers, Transactions of the, vol. 48,
pp. 716-727, 1929.

[7]

P. C. Krause, F. Nozari, T. L. Skvarenina and D. W. Olive, The Theory of Neglecting Stator Transients, Power Apparatus and Systems, IEEE Transactions on, vol. PAS98, pp. 141-148, 1979.

[8]

A. W. Rankin, Per-unit impedances of synchronous machines, American Institute of

Electrical Engineers, Transactions of the, vol. 64, pp. 569-573, 1945.

206

Generators, FACTS devices & system models

[9]

A. W. Rankin, Per-Unit Impedances of Synchronous Machines - II, American Institute of Electrical Engineers, Transactions of the, vol. 64, pp. 839-841, 1945.

[10]

P. Kundur, Power system stability and control. New York: McGraw-Hill, 1994.

[11]

IEEE Recommended Practice for Excitation System Models for Power System Stability Studies, IEEE Std 421.5-2005 (Revision of IEEE Std 421.5-1992), pp. 1-85, 2006.

[12]

D.J. Vowles, M.J. Gibbard, Small-Signal Modelling for the Analysis of Rotor-Angle Stability and Control of Large Power Systems, AUPress In preparation.

[13]

Determination of synchronous machine stability study constants, vol. 1-3, Electric Power

Research Institute, 1980.

[14]

IEEE Guide: Test Procedures for Synchronous Machines Part I-- Acceptance and
Performance Testing Part II-Test Procedures and Parameter Determination for Dynamic Analysis, IEEE Std 115-2010.

[15]

IEEE Guide for Synchronous Generator Modeling Practices and Applications in

Power System Stability Analyses, IEEE Std 1110-2002 (Revision of IEEE Std 11101991), pp. 1-72, 2003.

[16]

F. P. de Mello and L. H. Hannett, Validation of Synchronous Machine Models and

Derivation of Model Parameters from Tests, Power Apparatus and Systems, IEEE
Transactions on, vol. PAS-100, pp. 662-672, 1981.

[17]

I. M. Canay, Causes of Discrepancies on Calculation of Rotor Quantities and Exact

Equivalent Diagrams of the Synchronous Machine, Power Apparatus and Systems,
IEEE Transactions on, vol. PAS-88, pp. 1114-1120, 1969.

[18]

I. M. Canay, Modelling of alternating-current machines having multiple rotor circuits, Energy Conversion, IEEE Transactions on, vol. 8, pp. 280-296, 1993.

[19]

G. Shackshaft and P. B. Henser, et. al. Model of generator saturation for use in power-system studies, Electrical Engineers, Proceedings of the Institution of, vol. 126, pp.
759-763, 1979.

[20]

C. Concordia, Synchronous Machines: Theory and Performance. John Wiley & Sons,
Inc., New York, 1951.

[21]

B. Adkins and R. G. Harley, The General Theory of Alternating Current Machines: Applications to Practical Problems. Chapman and Hall, Ltd., London, 1975.

[22]

S. D. Umans, J. A. Mallick and G. L. Wilson, Modeling of Solid Rotor Turbogenerators Part I: Theory and Techniques, Power Apparatus and Systems, IEEE Transactions on, vol. PAS-97, pp. 269-277, 1978.

[23]

J. A. Mallick, G. L. Wilson and S. D. Umans, Modeling of Solid Rotor Turbogenerators Part II: Example of Model Derivation and Use in Digital Simulation, Power
Apparatus and Systems, IEEE Transactions on, vol. PAS-97, pp. 278-291, 1978.

Sec. 4.6

References

207

[24]

Website for the electronic version of this book:

http://www.adelaide.edu.au/press/titles/small-signal/

[25]

N. G. Hingorani and L. Gyugyi, Understanding FACTS: Concepts and Technology of

Flexible AC Transmission Systems, IEEE Press, New York, 2000.

[26]

X. P. Zhang, C. Rehtanz, and B. Pal, Flexible AC Transmission Systems: Modelling and

Control, Springer-Verlag, 2012.

[27]

Y. H. Song and A. T. Johns, Editors, Flexible ac transmission systems (FACTS), The

Institution of Electrical Engineers, Stevenage, UK, 1999.

[28]

S. Arabi and P. Kundur, A versatile FACTS device model for powerflow and stability simulations, Power Systems, IEEE Transactions on, vol. 11, pp. 1944-1950, 1996.

[29]

E.W. Kimbark, Direct Current Transmission, Wiley-Interscience, New York, 1971.

[30]

E. Uhlmann, Power Transmission by Direct Current, Springer-Verlag, Berlin/Heidelberg, 1975.

[31]

J. Arrillaga, High Voltage Direct Current Transmission, 2nd Edition The Institution of
Electrical Engineers, Stevenage, UK, 1998.

[32]

J. Arrillaga and B. Smith, AC-DC Power System Analysis, The Institution of Electrical
Engineers, Stevenage, UK, 1998.

[33]

S. Arabi, G. J. Rogers, D. Y. Wong, P. Kundur, and M. G. Lauby, Small signal stability program analysis of SVC and HVDC in AC power systems, Power Systems,
IEEE Transactions on, vol. 6, pp. 1147-1153, 1991.

[34]

M. Ili and J. Zaborszky, Dynamics and Control of Large Electric Power Systems, John
Wiley and Sons, Inc., 2000.

[35]

IEEE Task Force on Load Representation for Dynamic Performance, Load Representation for Dynamic Performance Analysis, Power Systems, IEEE Transactions
on, vol. 8, pp. 472-482, 1993.

[36]

IEEE Task Force on Load Representation for Dynamic Performance, Standard

Models for Power Flow and Dynamic Performance Simulation, Power Systems,
IEEE Transactions on, vol. 10, pp. 1302-1313, 1995.

[37]

J. V. Milanovic, K. Yamashita, S. Martinez Villanueva, S. Z. Djokic, and L. M. Korunovic, International Industry Practice on Power System Load Modeling, Power Systems, IEEE Transactions on, vol. 28, pp. 3038-3046, 2013.

[38]

D. Kosterev, A. Meklin, J. Undrill, B. Lesieutre, W. Price, D. Chassin, R. Bravo, and

S. Yang, Load modeling in power system studies: WECC progress update, in Power
and Energy Society General Meeting - Conversion and Delivery of Electrical Energy in
the 21st Century, 2008 IEEE, 2008, pp. 1-8.

208

Generators, FACTS devices & system models

[39]

Load representation for dynamic performance analysis [of power systems], Power
Systems, IEEE Transactions on, vol. 8, pp. 472-482, 1993.

[40]

I. A. Hiskens and J. V. Milanovic, Load modelling in studies of power system damping, Power Systems, IEEE Transactions on, vol. 10, pp. 1781-1788, 1995.

Appendices

209

Appendix 4I Linearization of the classical parameter

model of the generator.
The objective is to develop the linearized equations for the generator model in terms of the
classically-defined standard parameters in the form described in Section 4.2.13.4. The formulation proceeds assuming that there are three rotor windings in each axis after which it is
demonstrated that the equations are readily reformulated for models with fewer rotor windings.
The rotor d- and q-axis state-variables are respectively:
cd = d' 1d 2d

and cq = q' 1q 2q

(4.375)

The algebraic-variables associated with the d-axis rotor windings are grouped as follows:
z fd = I fd , z 1d = i 1d s1d d''

and z 2d = i 2d s2d d''' .

(4.376)

Similarly for the q-axis rotor windings:

z 1q = I 1q , z 2q = i 2q s2q q''

and z 3q = i 3q s3q q'''

(4.377)

Based on the above groupings of state- and algebraic-variables the d-axis rotor-winding
equations in (4.133) to (4.142) on page 147 are linearized about the initial operating condition derived in Section 4.2.13.3 as follows.
0 c fd C 1d C 2d cd
0
p cd

c
c fd 1 c f1d c f2d z fd
fid
0
=
+
i d +

c
C 1d 0 C 11d C 12d z 1d
1id
0

c 2id
0
C 2d 0 C 21d C 22d z 2d

b ced
0
1 I + 0 E
sd
fd
0
0
0
0

(4.378)

in which the coefficient sub-matrices are:

1
c fd = --------- 0 0
T d0'

0 0
0 0 0
1
0 0 0
, C 1d = 0 ---------0 , C 2d =
T d0''
1
0 ----------- 0
T
0 0 0
d0'''

c fid = L d L d' , c 1id = 0 L d' L l 0

, c 2id = 0 L d'' L l 0

(4.379)

(4.380)

210

Generators, FACTS devices & system models

0
0
0
0 0
0
1
1
0
0 0
1
c fd = 1 0 0 , C 1d =
, C 2d =
L d'' L l L d' L d''
L d'' L d'''
------------------ --------------------- 0
0 0 -----------------------
L d' L l L d' L l
L d'' L l

(4.381)

c f1d = c fid 0 0 , c f2d = c f1d

(4.382)

L d' L d''
1 ------------------------- 0
L d' L l 2
, C 12d = c 1id 0 0
C 11d =
0
1
0
0
0
1

(4.383)

L d'' L d'''
0
1 -------------------------- 0
1
L d'' L l 2
, C 22d =
C 21d =
L d''' L l
0
1
0
0 0 --------------------
L d'' L l
0
0
1

(4.384)

0 0
0 0

1
and b ced = --------0 0

T d0'

(4.385)

The d-axis rotor-winding algebraic variables z rd are defined as:

z fd

z rd = z 1d

z 2d

(4.386)

The coefficient sub-matrices in (4.379) to (4.384) are combined to form the following consolidated set of sub-matrices:
c fd

1 c f1d c f2d
C zd = c fd C 1d C 2d , C zd = C 1d , C zzd = 0 C 11d C 12d and
C 2d
0 C 21d C 22d
c fid
c zid = c 1id
c 2id

and the following sub-matrix is defined:

(4.387)

App. 4I

Linearization of the classical parameter model

211

1
c zsd = 0
0

(4.388)

The consolidated vector of algebraic variables in (4.386) and associated consolidated coefficient sub-matrices in equations (4.387) and (4.388) are substituted in the d-axis rotor-winding equations in (4.378) to yield the following compact formulation:
0 C zd cd
p cd
0
0
b
=
+
i d +
I sd + ced E fd

c
c
C
C
0
zid
zsd
0
zd zzd z rd

(4.389)

The q-axis rotor-winding equations (4.143) to (4.152) on page 148 following compact formulation of the linearized q-axis rotor-winding equations is similarly developed to yield,
0 C zq cq
p cq
0
0
,
=
+
i q +
I

c ziq
c zsq sq
C zq C zzq z rq
0

(4.390)

in which the following variable and coefficient sub-matrix definitions apply:

z 1q

z rq = z 2q ,

z 3q

(4.391)

c 1q

1 c 12q c 13q
C zq = c 1q C 2q C 3q , C zq = C 2q , C zzq = 0 C 22q C 23q ,
C 3q
0 C 32q C 33q
c 1iq

1
c ziq = c 2iq and c zsq = 0 .
0
c 3iq

(4.392)

in which, analogously with the d-axis sub-matrix definitions in equations (4.379) to (4.384),
the following q-axis sub-matrix definitions are obtained from the q-axis rotor-equations.
1
c 1q = --------- 0 0
T q0'

0 0
0 0 0
1
0 0 0
, C 2q = 0 ---------,
0 , C 3q =
T q0''
1
0 ----------- 0
T q0'''
0 0 0

c 1iq = L q L q' , c 2iq = 0 L q' L l 0

(4.393)

, c 3iq = 0 L q'' L l 0 ,

(4.394)

212

Generators, FACTS devices & system models

0
0
0
0 0
0
1
1
0
0 0
1
c 1q = 1 0 0 , C 2q =
, C 3q =
, (4.395)
L q'' L l L q' L q''
L q'' L q'''
-----------------

- --------------------- 0
---------------------0 0
L q' L l L q' L l
L q'' L l
c 12q = c 1iq 0 0 , c 13q = c 12q ,

(4.396)

L q' L q''
1 ------------------------- 0
L q' L l 2
, C 23q = c q2i 0 0 ,
C 22q =
0
1
0
0
0
1

(4.397)

L q'' L q'''
0
1 -------------------------- 0
1
L q'' L l 2
, C 33q =
.
C 32q =
L q''' L l

0
1
0
0 0 ------------------- L q'' L l
0
0
1

(4.398)

0 0
0 0

The d- and q-axis rotor-winding equations (4.389) and (4.390) are now combined to give,
0 C zdq cdq
p cdq
0
0
b
=
+
i dq +
I sdq + ce E fd .

C
C
C zdq C zzdq z rdq
0
zidq
zsdq
0

(4.399)

The following variable and coefficient matrix definitions apply to the preceding equation.

z
i
I
cdq = cd , z rdq = rd , i dq = d and I sdq = sd

cq
z rq
iq
I sq

(4.400)

C zdq = D C zd C zq , C zdq = D C zd C zq , C zzdq = D C zzd C zzq

(4.401)

C zidq = D c zid c ziq , C zsdq = D c zsd c zsq and b ce =

b ced
0

(4.402)

In order to readily generalize the formulation of the equations to generator models with fewer than three rotor windings in each of the axes it, is convenient to define the kth-transient
flux-linkages in terms of the rotor winding state- and algebraic variables as follows. The coefficient matrices in the following equation change depending on the number of rotor windings:

App. 4I

Linearization of the classical parameter model

213

cdq
k
dq = C kdq C kzdq

z rdq

(4.403)

in which C kdq = D c kd c kq and C kzdq = D c kzd c kzq and where

(4.404)

c kzd = 0 c k1d c k2d , c kzq = 0 c k2q c k3q .

(4.405)

For the case of three rotor-windings in each axis the constituent matrices have the following
values:
c kd = c kq = c k1d = c k2q = 0 and c k2d = c k3q = 0 0 1 .

(4.406)

The d- and q-axis voltage equations (4.158) and (4.159) on page 149 are consolidated to give:
k

v dq = 0 W dq dq Z dq i dq

k
in which W dq = 0 1 and Z dq =

1 0

(4.407)
k

0 Lq

rs
k

0 Ld

(4.408)

rs

Three alternative methods of representing the saturation level have been presented: (i) the
resultant airgap flux-linkages ( ag ) in Section 4.2.13.2.3; (ii) the resultant kth-transient fluxk

linkages ( ) in Section 4.2.13.2.1; and (iii) the transient d-axis flux-linkages ( d' ) in
Section 4.2.13.2.2. For each of these methods the perturbations in the saturation demagnetization currents can be expressed as follows. The definitions of the coefficient matrices in
the equation are given in Table 4.14 for each of the three methods for representing the saturation level.
k

I sdq = C sdq cdq + C skdq dq + C sidq i dq

(4.409)

In Table 4.14 the saturation characteristics for the d- and q-axes are combined into a single

diagonal matrix,
S dq m = D S d m S q m

(4.410)

where m is the saturation level indicator for the method chosen to represent the effects of
magnetic saturation. The saturation demagnetizing current components in the d- and q-axes
in terms of the selected saturation level indicator m are, in matrix form,

in which mdq = md mq

tion level indicator.

I sdq = S dq m mdq

(4.411)

is the vector of the d- and q-axis components of the satura-

214

Generators, FACTS devices & system models

Table 4.14 Coefficient matrices of the linearized, saturation demagnetizing current

equations for the three methods of representing the saturation level.
#

Coefficient matrices C sdq , C skdq and C sidq

See Section 4.2.13.2.3, m = ag and mdq = ad aq

C sdq = 0
1

ag

S dq ag
T
1
C skdq = S dq ag + ----------- ------------------------- adq adq and

0
ag
ag 0
0
0
0
C sidq = C skdq L dqai in which
k

adq = dq L aidq i dq where L aidq = D L d L l L q L l and

ag =

adq adq

See Section 4.2.13.2.1, m = and mdq = k k

d q

T
k
k
1 S dq k
C sdq = 0 C skdq = S dq 0 + ------ ---------------------- dq dq and
k
k

0 0 0
0
k

C sidq = 0 in which S dq = D S d S q & =

See Section 4.2.13.2.2, m = d' and mdq = d' 0

3

d'

dq dq

S dq d'
C sdq = S dq d ' + ------------------------ cdq , C skdq = 0 and C sidq = 0
d' 0 0
0

in which S dq d' = D S d d' 0 0 0

Equations (4.403), (4.407) and (4.409) are incorporated with the rotor-winding equations in
(4.399) to yield the following system of DAEs:
0

C zdq

p cdq
0
C zsdq C zidq
C zdq C zzdq

0
0
0
= C kdq C kzdq I
0
C skdq I C sidq
C sdq 0
0
k
0
0
0 0 W dq 0 Z dq

cdq

z rdq
b ce

0
k
0
0 dq +
0 E fd

0 I sdq
0

i dq
0
I

v dq

(4.412)

App. 4I

Linearization of the classical parameter model

215

In order to simplify the structure of the above equations it is convenient to define the following consolidated vector of algebraic variables:
z rdq

z e = k
dq

I sdq

(4.413)

and the associated consolidated matrix coefficients:

C zdq

C zzdq

C ce = 0 C zdq 0 0 , C ec = C kdq , C ee = C kzdq I

C sdq
0 C skdq

C zsdq
0

C zidq

, C ei =

and

C sidq

C ve = 0 0 W 0 .
dq

(4.414)

Substitution of the quantities in (4.413) and (4.414) into (4.412) results in the following compact form of the linearized electromagnetic equations:
0 C ce 0 0
p cdq

= C ec C ee C ei 0
0
k
0
0 C ve Z dq I

rdq

b re
z e
+

0 E fd
i dq

0
v dq

(4.415)

The algebraic variables z e are now eliminated from (4.415) to yield the generator electro

magnetic equations in the required form:

A r B ri 0
p cdq
=

C vr D vi I
0

cdq

b
i dq + re E fd

0
v dq

(4.416)

in which A r = C ce J ec , B ri = C ce J ei , C vr = C ve J ec and D vi = C ve J ei Z dq
1

and where J ec = C ee C ec and J ei = C ee C ei

(4.417)
(4.418)

The perturbations in the algebraic variables are given by:

z e = J ec cdq + J ei i dq

(4.419)

216

Generators, FACTS devices & system models

As outlined in Section 4.2.13.4 the electromagnetic equations in (4.416) are combined with
the rotor equations of motion and the network interface equations as in (4.117) on page 133.
However, x r and u m are redefined as follows:

cd
E fd
and u m =
.
x r =

cq
T m

(4.420)

Appendix 4II Forms of the equations of motion of the

rotors of a generating unit
App. 4II.1

Introduction

In the literature various forms of the equations of motion are employed, often depending
on the application. However, it is important to understand the nature of any approximations
used and their relevance to the application. When attempting to simulate small-signal events
using a large-signal (transient stability) software, it has been found that the damping of variables does not match that derived from a small-signal analysis software. The reason may be
associated with the form of the shaft equations employed in the large-signal simulation software. Consequently, several forms of the large-signal equations in per-unit form are derived
in this appendix, followed by the associated small-signal versions.
Direct
axis

Tg

Tm
Prime
mover

Generator

Coefficient of viscous damping, B

(a)

Synchronously
rotating vector

= s t

(b)

Reference

Figure 4.25 (a) Rotating mechanical system, prime mover and generator
(b) Rotating speed vectors and associated angles
For a given generating unit consider the mechanical system of Figure 4.25(a) rotating at an
angular velocity t electrical rad/s (the term speed is synonymous). Let us assume (i) a
two-pole generator (i.e. n pp = 1 ), (ii) the total moment of inertia of the rotating system is J
(kg-m2), (iii) the effective coefficient of viscous friction for small disturbances in speed in
the vicinity of synchronous speed is B (N-m/rad/s), and (iv) the shaft is infinitely stiff. The

App. 4II.1 Forms of the generator equations of motion

217

equation of motion of the shaft in SI units, due to a net accelerating torque T a t (N-m) acting on it, is:
d
J t + B t s = Tm t Tg t = Ta t ,
dt

(4.421)

where T m t is the prime mover torque (N-m); T g t is the airgap torque of electro-magnetic origin developed by the generator; time is in seconds.
Under perturbed, stable conditions the instantaneous angular velocity ( t ) of the unit varies about the synchronously rotating speed reference ( s , rad/s) of the system. As shown
in Figure 4.25(b) the rotor angle t of the unit with respect to a stationary system reference frame is:
t = s t + t (rad),

where t (rad) is the rotor angle of the unit with respect to the synchronously rotating
speed reference. The rotational speed of the shaft is thus:
d t
d t
t = ------------- = s + ------------ (rad/s), or
dt
dt

d
t
-----------= t s .
dt

(4.422)

For clarity at the present, let us denote per-unit quantities by the subscript p and let base
speed be b rad/s, thus the per unit speed is
p t = t b ,

and

d p t
d t
-------------- = b ---------------.
dt
dt

(4.423)

Since we are considering a two pole machine base electrical and mechanical speeds are identical. Let us define the per-unit synchronous speed as,
0 = s b .

(4.424)

Note that normally, but not necessarily, b = s rad/s.

Let us define base power S b (VA) and base torque T b (N-m) by the relationship:
S b = b T b (see Tables 4.2-4.4 on page 102).

(4.425)

Dropping the time dependency of the variables and dividing (4.422) through by b , we find
that in per-unit
d
----- = b p 0
dt

d
1 d
and --------p- = ------ 2 .
dt

b d t

Let us now consider several forms of the per-unit shaft equations.

(4.426)

218

Generators, FACTS devices & system models

App. 4II.2 Shaft equations expressed in terms of per-unit angular

speed and torques
Dividing through (4.421) by the base torque T b of the generating unit, and expressing the
equation in terms of per unit speed p based on (4.423), we find
Tm Tg
b J d p B b
--------- --------- + ---------- p 0 = ------ ----Tb
Tb Tb
Tb
dt

= T mp T gp per unit.

(4.427)

Tb

2S b

b
b
- = 2 -------- = 2H , where
Firstly, let us rearrange in (4.427) the term, --------

2
b J
--------- S .
H =
b
2

(4.428)

H is defined as the inertia constant of the unit, and is the ratio of the stored energy in Joules

of the rotating system at base speed to the base apparent power, S b , of the unit in VA. The
rotating system normally consists of the prime mover, and the generator and exciter.
Secondly, consider in (4.427) the term, B b T b = D where D is a damping torque coefficient. It is the damping torque for a speed difference equal to base speed per unit of the
base torque.
Thirdly, the terms on the right hand side of (4.427) are the per unit mechanical and airgap
electro-mechanical torques, Tmp and Tgp respectively.
The per-unit equations of motion of the rotor resulting from (4.426) and (4.427), respectively, are
d
----- = b p 0
dt

and 2H --------p- + D p 0 = T mp T gp ,
dt

(4.429)

which are the equations of motion listed in equations (4.56) and (4.57) on page 114.
Note that there are no approximations made in the derivation of the above equations. However, the relationship between damping torque and speed perturbations about synchronous
speed, characterized by the coefficient D, is generally unknown at any operating condition.
Typically, the damping torque coefficient D is small, and being unknown, it is often set to
zero or some low value. Importantly, damping effects of electromagnetic origin are due to
losses in the resistances of the damper windings of the generator models. Such losses should
therefore not be accounted for in damping torque coefficient of the shaft acceleration equation.

App. 4II.3 Per-unit shaft acceleration equation

219

The per-unit equations of rotor motion given by (4.429) are linear in the speed and torque
variables. These variable can therefore be replaced by the perturbed variables to form the
set of linearized equations, i.e.
d
---------- = b p
dt

and 2H ------------p- + D p = T mp T gp ;
dt

(4.430)

these are the linearized equations of motion listed in (4.58) and (4.59) on page 114.

App. 4II.3 Per-unit shaft acceleration equation in terms of rotorspeed and power
In a rotating system, the power, torque and speed of rotation - all in SI units - are related
through the product non-linearity
P = T (W).

(4.431)

The mechanical power delivered to the shaft by the turbine and the power transferred across
the airgap of the machine are respectively,
P m = T m (W) and P g = T g (W).

(4.432)

Note that because the shaft is assumed to be rigid the mechanical and electrical rotor-speeds
are identical, observing that in this analysis a two pole machine is assumed.
Based on equation (4.432) it can be shown that the per-unit shaft acceleration equation can
be expressed in terms of the per-unit accelerating power P ap = P mp P gp , i.e.
d p
2H --------- + D p 0 = P mp P gp p
dt

(4.433)

for which the linearized form is,

d p
2H ------------- + D p = P mp P gp 0 .
dt

(4.434)

Furthermore, it can be shown that in the linearized acceleration equations the terms in the
torque perturbations in (4.430) and in terms of the power perturbations (4.434) are exactly
equivalent, i.e. T mp T gp = P mp P gp 0 .
It is important to note that no approximations have been made in the formulation of this
non-linear equation. Specifically, perturbations in the rotor-speed in calculating the relationship between the accelerating torque and power have been retained.
As mentioned earlier normally, but not necessarily, synchronous speed s rad/s is equal to
the base speed b rad/s in which case 0 = s b = 1 per-unit. If this is the case it follows
that the per-unit perturbations in torque and power are identical.

220

Generators, FACTS devices & system models

App. 4II.4 Shaft acceleration equation neglecting speed perturbations in the torque/power relationship
It is commonly assumed, for the purpose of calculating the generator d- and q-axis stator
voltages, that the perturbations of the rotor-speed from synchronous speed are negligible.
As stated in Sections 4.2.2 and 4.2.5 this assumption is adhered to in this book, i.e. the airgap
power is related to the airgap torque by:
P gp = 0 T gp

(repeat of (4.52) on page 113).

Given the above approximation it is essential that the relationship between the mechanical
power and torque also neglect perturbations in the rotor-speed, that is,
P mp = 0 T mp .

(4.435)

The consequence of ignoring the above approximation but rather setting P mp = p T mp is

discussed in Appendix 4II.5.
With the assumption that perturbations in the rotor-speed are negligible for the purpose of
calculating the shaft accelerating power, it is shown that the linearized shaft equations expressed in terms of torque perturbations (4.430) are identical to the equations expressed in
terms of power perturbations(4.434).

App. 4II.5 A common misunderstanding in calculating the accelerating torque and power
The consequences of a misunderstanding that sometimes occurs in the formulation of the
shaft acceleration equation are now discussed. The misunderstanding is that the rotor speed
perturbations are neglected in the relationship between the airgap torque and power (as discussed above) but erroneously the speed perturbations are retained in the relationship between the mechanical power and torque. That is to say, the error is to define the per-unit
mechanical and airgap power differently as follows:
Error: P mp = p T mp and P gp = 0 T gp .

(4.436)

From the preceding equation the following inconsistent expression for the accelerating
torque is derived,
T mp T gp = P mp p P gp 0 ,

(4.437)

which when substituted into the per-unit acceleration equation (4.429) results in,
d

Error: 2H --------p- + D p 0 = P mp p P gp 0 .
dt

(4.438)

Linearizing the above inconsistent expression for the acceleration equation about the operating point yields,

App. 4II.5 A misunderstanding in formulating the acceleration equation

221

d p
2
2H ------------- + D p = P mp p P mp p p P gp 0
0
0
0
dt

which, upon substitution of p = 0 , results in the following incorrect form of the linearized acceleration equation.

d p
2
2H ------------- + D + P mp 0 p = P mp P gp 0 = T mp T gp .
0
dt

(4.439)

Comparison of the correct formulation of the linearized acceleration in (4.430) with the
above expression (4.439) shows that the effect of the misunderstanding is erroneously to in2

crease the damping coefficient of the generator model by P mp 0 .

0

Chapter 5
Concepts in the tuning of power system stabilizers
for a single machine system

5.1 Introduction
Although this chapter is concerned with the application of a power system stabilizer (PSS)
to a single-machine system, the concepts for the most part are applicable to multi-machine
systems: such applications will be discussed in Chapters 9 and 10. Various important aspects
of the tuning of the PSS can therefore considered in some detail because the analysis involves a simple system only.
The reasons for the wide-spread deployment of PSSs in power systems today are twofold,
(i) to stabilize the unstable electro-mechanical modes in the system, (ii) to ensure that there
is an adequate margin of stability for these modes over a wide range of operating conditions
and contingencies, that is, the electro-mechanical modes are adequately damped. Some systems, such as the Eastern Australian grid, would be unstable without the use of both PSSs
and stabilizers installed on certain FACTS devices.
A marginally stable electro-mechanical mode is oscillatory in nature and is very lightly
damped. The frequency of rotor oscillations is typically between 1.5 to 15 rad/s, and the 5%
settling time may be many tens of seconds. Typically a mode of a lengthy duration would
not satisfy the system operators criterion for modal damping. A stable mode is said to be
positively damped, whereas an unstable mode is referred to as being negatively damped.

223

224

Tuning PSSs for SMIB system

Ch. 5

With the growth of power systems, and the need to transmit power over long distances by
means of high-voltage transmission lines, the problems of instability following a major fault
or disturbance have increased. Instability in such events is typically the result of a generator
falling out of step due to insufficient synchronizing torques being available to hold generators in synchronism. In order to increase the synchronizing torques between generators,
high-gain fast-acting excitation systems were developed with the objective of increasing field
flux linkages rapidly during and following the fault. However, such high-gain excitation systems may introduce negative damping on certain electro-mechanical modes.
In linear control systems design, rate feedback is employed not only to stabilize an unstable
system but also to enhance the systems damping performance. A PSS that uses generator
speed (i.e. the rate of change of rotor angle) as a stabilizing signal is such a rate-feedback
controller. However, to introduce on the shaft of the generator a torque of electromagnetic
origin that is purely a damping torque requires that the compensation transfer function provided in the PSS is properly designed. Pure damping occurs when the induced electrical
torque is in phase with speed; this is an essential function of the PSS. Such a torque opposes a
change in rotor-speed.
The main role of the PSS is to provide damping of the electro-mechanical modes for small
disturbances on the system. Therefore, in order to analyse the dynamic performance of the
system and to tune stabilizers, the non-linear equations describing the dynamic behaviour of
the generator and system are linearized about a steady-state operating condition. As outlined
in Section 2.1.1 a set of linear equations in terms of the new set of perturbed variables results.
A significant consequence is that the powerful methods and techniques provided by linear
control systems theory become available both for the analysis of dynamic performance and
for controller design.
This chapter is concerned with illustrating the concepts associated with the design and tuning of a PSS using small-signal analysis techniques. The performance of the system under
large-magnitude disturbances - such as fault conditions as mentioned above - is treated briefly in Chapter 10. However it should be mentioned that, following clearance of a fault, the
system may appear to be stable following the second or subsequent swings in the rotor angles but becomes unstable as the steady-state is seemingly approached. Such instability is due
to the existence of one or more unstable electro-mechanical modes in the post-fault operating condition. For example, transmission lines may have been switched out of service in order to clear the fault and instability is a consequence of network voltages falling during the
post fault period. The PSSs must be designed to ensure small-signal stable operation in the
steady-state that follows the worst-case contingencies.
The benefits of small-signal analysis in complementing large-signal (or transient stability)
analysis are described in Section 10.9.1.1.

Sec. 5.1

Introduction

225

The paper by de Mello and Concordia in 1969 provided the basis for the design of many
Power System Stabilizers in operation today [1]. Based on the concept of damping torques
developed on the generator shaft, a technique is presented in the paper for the design and
tuning of a speed-input PSS for a single-machine infinite-bus (SMIB) system. The PSS transfer function is designed to provide phase compensation for the transfer function between
the voltage reference of the AVR and the electrical torque. Ideally, any perturbations in shaft
speed produce pure damping torques on the shaft.
In a set of papers by Larsen and Swann in 1981 [2] the concepts in [1] were extended and
applied to the tuning of PSSs and their tuning on site. Firstly, frequency response measurements between the voltage reference of the AVR and the terminal voltage yield a transfer
function which, because the speed perturbations are assumed negligible, is equivalent to the
phase response between the voltage reference and electrical torque. The transfer function is
called the generator (G), excitation system (E) and power system (P) transfer function,
GEP(s). Secondly, the PSS compensation is then designed to offset the phase-lags in
GEP(s) by means of phase-lead transfer function blocks. Finally, the PSS gain is raised until
prolonged oscillations are observed, i.e. the generator is on the brink of modal instability.
The PSS gain is then set to 1/3rd of the limiting value, providing a gain margin of 3:1 or 10
dB. This approach to PSS tuning is considered in more detail in Chapter 6.
An alternative approach which is applicable to single- and multi-machine systems is based
on the Method of Residues and is also described in Chapter 6. Some of the features of the
these methods, and approaches to the tuning and implementation of PSSs, are described in
an IEEE Tutorial Course [3].
Using the damping torque concepts developed for the single machine system [1], [2] and [4],
a procedure called the P-Vr approach for the tuning of PSSs in multi-machine systems was
proposed by Gibbard in 1988 [5]. This procedure is in part an extension of the GEP Method, however, specific and meaningful information is derived concerning both the phase compensation for, and the gain setting of, the PSS. (See Section 1.2, item 3)
The P-Vr tuning procedure is described in this chapter for a single machine system to illustrate in some detail the concepts in the tuning of PSSs, however - as stated earlier - the procedure is readily applied to multi-machine systems and is the subject of later chapters.
The literature on PSSs, and their associated stabilizing signals, for single- and multi-machine
systems is fairly extensive and is discussed in more detail in Chapter 8, Types of Power System Stabilizers. The purpose of this chapter, however, is to provide an understanding of the
fundamentals of PSS design and tuning.
In this chapter the preliminary tuning of a PSS will be based on the Heffron and Phillips
model of the SMIB system [6]. This model, being fourth order, is amenable to simple analysis and thus it is possible to derive simple, closed-form solutions for certain transfer func-

226

Tuning PSSs for SMIB system

Ch. 5

tions and torque-related expressions which are applicable to a range of operating conditions.
These features provide a more detailed insight and understanding of the relevant concepts. However, to partially bridge the gap between the low-order Heffron and Phillips model and the
practical models used in multi-machine systems, the tuning of a PSS for a higher-order machine model will be employed later in the chapter. For higher-order models it is not practical
to derive similar closed-form solutions and therefore other approaches must be employed.
In addition, a somewhat more realistic representation of the external system will be adopted
so that system quantities, such as voltage levels on busbars, lie within normally acceptable
ranges.
Some of the earlier material in this Chapter, Section 5.2 to 5.7, is also covered in the IEEE
Tutorial Course [3].
In this and subsequent chapters we will employ the term range of operating conditions. It
is assumed that, for the subsequent analysis, a set of steady-state conditions are selected
which encompass those conditions for which the stabilizer is to be tuned. The latter conditions
should include normal operation and contingencies such as line and generator outages, etc.
By judiciously selecting the encompassing set should result in a reduction of the number of
conditions that need to be studied and result in an acceptable stabilizer design.
In the following sections the excitation system 1, which includes the automatic voltage regulator (AVR), is modelled by a simple first-order transfer function. In practice for the tuning
of the PSS an accurate model of the excitation system and associated parameters are required. The excitation system should be properly tuned and the model validated by measurements. Because the excitation system is in the PSS control loop the resulting
performance of the PSS is likely to be poor if the that model is inadequate.

5.2 Heffron and Phillips Model of single machine - infinite bus system
De Mello and Concordia based their analysis on a linearized model, developed by Heffron
and Phillips [6], of a single machine connected to an infinite bus through an external impedance, r e + jx e , representing the sum of the impedances of the generator transformer and the
Thvenin equivalent impedance of the system. The Heffron and Phillips model includes a
third-order representation of the generator and a first-order model of the excitation system
as shown in Figure 5.1. The constants K1 to K6 are defined in [1] and are given in
Appendix 5I.1.

1. According to [7] the excitation system is comprised of that equipment providing field
current for a synchronous machine, including all power, regulating, control, and protective elements. The regulation of terminal voltage is a function of the AVR.

Sec. 5.3

Synchronizing and damping torques

P1
_

+_

K1

P3

Pm

Pa

227

/s

/(sM)

P2
K2

Eq

Figure 5.1

K4

K3
1+sK3Tdo

K5
K6
Kex
+ 1+sTex

+
_

Vt
+

Vr

Heffron and Phillips Model of a SMIB system.

All variables are in the Laplace (or s) domain, although much of the later analysis is conducted in the frequency domain with s = j f , where f is the frequency in rad/s. The per-unit
perturbations in the variables are defined below:
P 1 Electrical torque, a function of rotor angle
P 2 Torque of electro-magnetic origin
P 3 Damping torque associated with windage, friction, and losses in the damper
windings
P a Accelerating torque acting on the shaft of the generator-turbine unit
P m Prime mover torque
E q Voltage proportional to direct axis flux linkages
V t

Terminal voltage

V r

Reference voltage

Rotor speed
Rotor angle

5.3 Synchronizing and damping torques acting on the rotor of a synchronous generator
In the context of a linearized model of a single-machine infinite-bus system, de Mello and
Concordia [1] developed the concept of a complex torque P s of electro-magnetic origin
acting on the shaft of a generator. For reasons discussed shortly, the component of this
torque in-phase with speed was called a damping torque and that component in-phase with
rotor angle was called a synchronizing torque. Both these torques are braking torques, that is,
they act to oppose changes in speed or rotor angle, respectively.

228

Tuning PSSs for SMIB system

Ch. 5

In the context of the Heffron and Phillips model of Figure 5.1, let us consider the torque
of electro-magnetic origin, P 0 , developed on the rotor of the generator. If the perturbations in prime-mover torque are negligible, P m = 0, the sum of the electrical torques acting
on the rotor can be expressed as
P 0 s = P a = P 1 s + P 2 s + P 3 s ,

(5.1)

or P 0 s = K 1 s + K 2 E q s + D s .

(5.2)

Based on (4.58) the relation between the rotor angle (rad) and the per-unit speed can be expressed as
s = 0 s s ,

(5.3)

in the s-domain. The dependency on the Laplace operator s, i.e. (s), will not be shown
when the variables are clearly in the s-domain.
The terms in (5.2), other than K 2 E q , are expressed in terms of and . Let the transfer function between and E q be E q , and let that between and E q be
E q where, based on (5.3),
E q
s - E
q
------------ = ---- ------------ .

(5.4)

The component of electrical torque K 2 E q s in (5.2) can be split into components, a real
component in phase with rotor angle and a real component in phase with speed. The expression for P 0 s can then be written in the form:
P 0 = K 1 + K 2 E q + D + K 2 E q .

(5.5)

The paths in the Heffron and Phillips model encompassed by transfer function E q
are shown in Figure 5.2. This transfer function can be found by block diagram manipulation
to be
E q
K 3 K 4 1 + sT ex + K 5 K ex
-.
------------ = -----------------------------------------------------------------------------------------------------------------------2

s K 3 T ex T d0 + s T ex + K 3 T d0 + 1 + K 3 K 6 K ex

(5.6)

Referring to (5.5), let us define the coefficients of and as k s s and k d s respectively, i.e.
k s = K 1 + K 2 E q

hence (5.5) becomes

and k d = D + K 2 E q ,
P 0 = k d + k s .

(5.7)
(5.8)

The terms k d s s and k s s s in (5.8) are damping and synchronizing torques, respectively. Consequently k d s is known as a damping torque coefficient and k s s as a synchronizing torque coefficient.

Sec. 5.3

Synchronizing and damping torques

P1
_

+_

K1

P3

Pm

Pa

229

/s

/(sM)

P2
K2

Eq

Figure 5.2

K4

K3
1+sK3Tdo

K5
K6
Kex
+ 1+sTex

+
_

Vt
+

Vr

Evaluation of the transfer function E q

in the Heffron and Phillips model.

As mentioned earlier, for analysis in the frequency domain, let s = j f where f is the frequency in rad/s. From the shaft relation (5.3), we note that in the frequency domain the rotor angle and speed are related by j f = j 0 f j f . Equation (5.8) can thus
be written in the form:
P 0 j f = k d j f j 0 f k s j f = a + jb j f ,

(5.9)

where a = k d and b = 0 f k s are torque coefficients and are respectively the real and
imaginary parts of the transfer function P 0 j f j f . Thus any positive torque coefficient in phase with speed produces a positive damping torque on the machine shaft. Correspondingly, any positive torque coefficient in quadrature lagging on speed implies a positive
synchronizing torque.

A useful expression for each of the torque coefficients can be derived from (5.9), i.e.
P 0 j f
f P 0 j f
k d = ----------------------- and k s = ------ ----------------------- .
j f
0 j f

(5.10)

Later in Section 5.10.2 reference is made to the phrase disabling the shaft dynamics of the machine for the purpose of calculating the torque coefficients. The shaft dynamics are disabled
if the speed signal is completely isolated, for analysis purposes, from the accelerating torque
acting on the shaft. This achieved by opening the output of the block 1 sM in
Figure 5.2 1. We can now treat the speed signal as an input signal to the transfer function P 0 in (5.10), and hence calculate the torque coefficients.
1. Since M = 2H, setting the machine inertia constant, H, to infinity serves the same purpose [4].

230

5.4

Tuning PSSs for SMIB system

Ch. 5

The role of the Power System Stabilizer - some simple concepts

Before further examining the Heffron and Phillips model in detail, and in order to provide
some insight, some simple concepts about the role, function and effects of a PSS can be
demonstrated by simplifying the Heffron and Phillips model shown in Figure 5.1. Let us
assume that (i) there are no perturbations in the reference voltage Vr, (ii) the exciter and
open-circuit time constants, T ex and K 3 T d0 (with K 3 0.3 ), respectively, are very short,
and (iii) any disturbances to the system occur through mechanical torque perturbations,
P m . The model of the generator in Figure 5.1 can then be simplified to the second-order
model shown in Figure 5.3 in which the torques acting of the shafts and the resulting speed
and angular deviations are shown 1.
ks
Ps
Pm +

Pd

Figure 5.3

P0

/(sM)
kd

o/s

A very simplified model of the generator in a SMIB system.

In Figure 5.3 the torque of electro-magnetic origin P 0 acting on the rotor is the sum of
the synchronizing and damping torques, P s and P d , respectively, i.e.
P 0 s = P s + P d = k s + k d .

((5.8) repeated)

The transfer function between the mechanical torque perturbation and the rotor angle response of the generator shaft is
0 2H
---------= ---------------------------------------------------------------------- .
2
P m
s + k d 2H s + 0 k s 2H

(5.11)

Typically the damping of such a system is light and its response to a step input is oscillatory.
Thus there are a pair of complex poles that lie at:
kd 2
kd
kd
0 ks
1 4 0 k s
,
s 1 s 2 = ------- j --- --------------- ------- = ------- j ----------4H 2 2H 2H
4H
2H

(5.12)

1. The model in Figure 5.3 has the same form as that which is obtained with a classical generator model. In the latter it is assumed that the voltage proportional to d-axis flux-linkages are constant during the study period (i.e. E q = 0 ).

Sec. 5.4

The role of the Power System Stabilizer

231

when k d 2 0 k s 2H . The frequency of the damped rotor oscillations (the imaginary part
of the pole) is d 0 k s 2H . From a consideration of the pole locations we can conclude that as the damping torque coefficient kd is increased (hypothetically) from some initial
value:

The poles shift more-or-less directly to the left 1 in the complex s-plane at a constant
frequency of oscillation as long as the associated damping ratio is less than about 0.2 0.3.

The damping constant, defined in Section 2.8.2.1, is = k d 4H , i.e. it is inversely

proportional to H, the inertia constant 2. For example, consider two generating units
with identical per-unit parameters on machine base but their inertia constants are in
the ratio of 2:1, say; of the two, the lighter unit is the better damped.

The frequency of damped oscillation is proportional to the square root of ratio

0 k s 2H . Thus, if switching lines out of service halves the synchronizing torque
coefficient - or if the inertia constant of a unit is doubled - the frequency of oscillation
is decreased by about 30% (to 1 2 ).

If the damping ratio exceeds ~0.2 as kd is increased, the trajectory of the complex
poles move along a semi-circle of constant radius in the complex s-plane (see
Figure 2.8).

If the value of = k d 4H is positive (i.e. the damping torque coefficient k d is

negative) then, according to (5.12), a pair of poles lie in the right-half of the complex
s-plane; the system is therefore unstable.

Let us assume that we can add a feedback loop from rotor speed to the torque signal
P dp - as shown in Figure 5.4 - such that P dp = k . It is clear that increasing the gain
k has the same effect as increasing the damping torque coefficient k d , that is, enhancing the
damping of rotor oscillations. A PSS is a device that ideally induces on the rotor a torque of
electro-magnetic origin proportional to speed perturbations. The ideal PSS gain k, which is
a damping torque coefficient, we shall call the damping gain of the PSS. The ideal PSS will
produce a direct left-shift in the rotor mode, as manifested in (5.12), from k d 4H to
k + k d 4H . The gaol in the tuning of a practical PSS is to achieve the same result, the
damping gain of the PSS being adjusted to meet the specifications on damping for the rotor
modes of oscillation.
1. By direct left-shift is implied that the eigenvalue / mode shift is j0 , 0 .
2. Some additional significance of this result is derived from the analysis in Section 13.2.2.

232

Tuning PSSs for SMIB system

Ps
Pm +
_

Pdp

Ch. 5

K1
_

Pa
Pd

/(sM)
kd

o/s

Generator
model

k
Ideal PSS

Figure 5.4 The ideal PSS introduces a pure damping torque P dp

on the rotor of the generator.
As noted Section 5.1, the PSS feeds back a signal proportional to speed, i.e. the rate of
change of position, a technique known as rate feedback. The technique is commonly used
in conventional control system design to control speed of a rotating device 1, to improve the
damping - or to stabilize - a closed-loop system. Under ideal steady-state conditions the perturbations in speed are zero in a stable system - and thus rate feedback acts only when the
system is disturbed in some way, e.g. by noise - which is always present.
Note that speed appears to be the ideal stabilizing signal because the damping torque induced on the shaft of the generator by the associated PSS is related to speed through a simple gain. Moreover, this gain being a damping torque coefficient has practical significance,
e.g. a damping gain in the range 20 - 30 pu on machine MVA rating is a moderate damping
gain setting for a speed-PSS. Practical forms of the speed-PSS, such as the integral-ofaccelerating-power PSS, are discussed in Chapter 8.
In practice the ideal speed-PSS, consisting simply of a damping gain k, is replaced by a
transfer function H PSS s = kG s , where G s consists of a compensating transfer function and relevant filters. In this chapter the design of the transfer function kG s is outlined
for a SMIB system to illustrate a number of concepts which are also applicable to multi-machine systems.

5.5 The inherent synchronizing and damping torques in a SMIB system

Prior to attempting to discuss or tune the PSS for a generator it is instructive to ascertain the
levels of the inherent damping of the rotor modes. One approach is to calculate the eigenvalues of the system. A second approach is to determine the natural or inherent damping
torque coefficient over the range of rotor frequencies for the system (although in this case
1. In order to control speed the feedback signal (from the ideal PSS) must negated at the
summing junction as shown in Figure 5.4.

Sec. 5.5

Inherent synchronizing and damping torques

233

of a SMIB system there is only one mode of rotor oscillation). In order to provide a complementary point of view, the second approach will be adopted.
Based on the Heffron and Phillips model of Figure 5.1 and equations (5.7) and (5.8) the synchronizing and damping torque coefficients can be determined for the SMIB system. In the
absence of a PSS, k s and k d in (5.7) are the synchronizing and damping torque coefficients
inherently produced by the generator. Using a simple example let us calculate these torque

coefficients to examine not only their values in the vicinity of the frequency of the single
rotor mode but also how they vary with the generator loading.
5.5.1 Example 5.1
For present purposes a third-order model of the generator and a first-order model of the
excitation system are used. The unit is connected to an infinite bus through transformer and
transmission line reactances representing the external system.
The parameters of a SMIB system and the steady-state operating conditions are:
Generator: D = 0, H = 3.0 MWs/MVA, ra = 0, xd = 1.9 pu, xq = 1.8 pu, x d = 0.30 pu,
T do = 6.5 s, rating 250 MVA.
Exciter: Kex = 200 pu, Tex = 0.02 s.
Transformer and line reactance: xt = 0.15 pu and xL = 0.225 pu, respectively.
The generator is under closed-loop voltage control, terminal voltage Vt = 1.0 pu.
Operating Conditions: System frequency = 50 Hz. Rated real power output is P = 0.9 pu,
and reactive power outputs are Q = -0.20, 0.0, 0.2 and 0.4 pu.

The machine and system base is 250 MVA. The constants K1 to K6 are defined in [1] and
Appendix 5I.1.
Note. For present purposes it is assumed that this SMIB system is a representation of a generator
and transmission system within a multi-machine system in which the electro-mechanical modes
may range from 1.5 - 15 rad/s. While there is a single rotor mode in this example, the PSS is to
be tuned for the latter frequency range.

For each operating condition at 0.9 pu real power output the rotor angle, the terminal voltage angle, the infinite-bus voltage, the K-constants, the eigenvalues for the rotor mode, and
the inherent synchronizing and damping torque coefficients are given in Table 5.1. The eigen-analysis reveals that all the real parts of the rotor mode are positive and thus the system
is unstable for the range of reactive power outputs of the generator; the frequency of the
unstable rotor modes is between 7.9 and 9.3 rad/s. The damping and synchronizing torque
coefficients for all operating conditions are given at a frequency of 8.69 rad/s, a value close
to the mid-range of the frequencies of oscillation of the rotor modes. The negative damping
torque coefficients at 8.69 rad/s (-16.9 to -2.3 pu) are consistent with the unstable rotor

234

Tuning PSSs for SMIB system

Ch. 5

modes. Relative to the damping torque coefficients the variation of the synchronizing
torque coefficients (1.6 to 1.2 pu) is much less over the range of operating conditions.
Table 5.1

K-constants, rotor modes and inherent torque coefficients kd , ks ,

for P = 0.9 pu,

Q
pu

Rotor
Angle
deg.

Angle Infinite
Term.
Bus
Voltage Voltage
deg.
pu

K1

K2

K3

K4

K5

K6

Eigenvalues,
rotor mode

kd

ks

at
at
8.69 8.69
rad/s rad/s

-0.2

85.9

17.4

1.127

1.355 1.665 0.297 2.664 -0.121 0.204 1.15 j 9.23

-16.9

1.62

0.0

77.0

18.6

1.055

1.289 1.523 0.297 2.437 -0.072 0.292 0.514 j 8.70

-6.56

1.44

0.2

70.0

20.0

0.985

1.202 1.371 0.297 2.194 -0.051 0.357 0.267 j 8.25

-3.22

1.29

0.4

64.9

21.6

0.915

1.122 1.227 0.297 1.964 -0.048 0.404 0.198 j 7.94

-2.34

1.20

For later reference it is instructive to consider the frequency response of the transfer function for the inherent torque coefficients. Referring to Figure 5.2 and (5.6), the transfer function is
P a
P 1 + P 2
---------= -------------------------- P = 0

m
K 2 K 3 K 4 1 + sT ex + K 5 K ex
0
= ------ K 1 ------------------------------------------------------------------------------------------------------------------------2
s
s K T T + s T + K T + 1 + K K K
3 ex

d0

ex

d0

(5.13)

3 6 ex

where P a is the accelerating torque. An examination of (5.13) reveals that at both low and
high frequencies ( s j0 or s j ) the transfer function rolls off at 20 dB/decade at a
phase angle of 90 , i.e. the damping torque coefficients at these frequencies are zero.
There is phase variation from 90 in the intermediate frequency range as shown in the responses in Figure 5.5(a).
Note from Figure 5.5(a) the damping torque coefficient, kd = gain x cosine(phase angle), is
negative for all the selected outputs; the associated inherent damping torques in
Figure 5.5(b) are therefore destabilizing. This result is consistent with those revealed in
Table 5.1.
Clearly for this system the degree of instability (as revealed by both the negative inherent
damping torque coefficient and the real part of the eigenvalue of the rotor mode) increases
as the generator operates at increasingly leading power factors. The potential for small-signal
instability is a characteristic of generator operation at leading power factors. A device called
an under-excitation limiter is normally fitted to a machine to prevent the steady-state operating point from drifting too far into the leading power factor region.

Excitation system gain and stability

Phase Response (deg)

Magnitude (dB)

60
50
40
30
20
10
0
10

10

10

90
95

100
105
110
0
10

(a)

10
Frequency (rad/s)
Q=0.2
Q= 0.2

Q=0.0
Q=0.4

10

235

Synch. Torq. Coeff. Damp. Torq. Coeff.

Sec. 5.6

(b)

0
10
20
30
40
0
10

10

10

2.5
2
1.5
1
0.5
0
10

10
Frequency (rad/s)
Q=0.20
Q= 0.20

10

Q=0.0
Q=0.4

Figure 5.5 (a) Frequency response of the transfer function P a for the inherent
torque coefficients for real power output P = 0.9 pu as reactive power Q (pu) varied.
(b) The associated damping and synchronizing torque coefficients.
We have examined only a few operating conditions at rated real power output of 0.9 pu. For
present purposes, however, this simple set of studies reveals that a PSS - when installed should possess a damping gain, i.e. a positive damping torque coefficient, greater than
k = 16.9 pu to ensure stability under the most onerous operating condition, P = 0.9 and
Q = -0.2 pu.
Note from Table 5.1 that the voltage at the infinite-bus end of the high voltage lines is outside a practical range of 0.95 to 1.05 pu for three of the four operating conditions. Later in
Section 5.10.4 - when a more practical system is employed - we will ensure that a more appropriate set of operating conditions is studied.

5.6 Effect of the excitation system gain on stability

In practice it is common for excitation system gains to lie in the range of 200 to 400 pu 1, or
more. High gains may be desirable to satisfy requirements for steady-state voltage regulation, or to boost generator field flux linkages during the fault period thereby enhancing syn1. For example, subject to certain conditions, a requirement in a set of Rules [8] requires
that generators must have an excitation control system that regulates voltage at an agreed
location to within 0.5% of the set-point. This implies an excitation system gain of at least
200 pu.

236

Tuning PSSs for SMIB system

Ch. 5

chronizing torques between generation in the early post-fault period. In the case of smallsignal dynamic performance, the effect of the excitation system gain on damping torques,
and therefore stability, is analysed in detail in [1].
From Table 5.1 it is observed that the system is unstable for the range of steady-state operating conditions covered in the studies. Because the excitation system gain of 200 pu in this
application is considered fairly high, it is instructive to assess the effect of lower and higher
gains on the rotor modes.
In Figure 5.6 are shown the loci of the rotor mode as the excitation system gain is increased
from zero to 300 pu for two of the operating conditions included in Table 5.1. Without a
PSS this SMIB system is stable only at very low excitation system gains, i.e. less than 30 pu.
As the power factor becomes less lagging, i.e. at Q = 0, the damping of the rotor mode tends
to degrade further as the gain increases.
In this simple model of the generator and excitation system the higher-order dynamics are
ignored. Such a model may be satisfactory at low excitation system gains, but at high gains
the effect of the unmodelled dynamics is to degrade stability. Care therefore should be taken
when analysing simplified low-order models in high-gain excitation systems.

Imaginary Part (rad/s)

x 300
x 250
x 200

Locus 1

x 150

8.5

x 100
Kex=0 x
8

7.5
-1

x
25

x 50

xx 300
x 200
x
Locus 2
x
100
Kex=0 x x x
25 50
-0.5
0
0.5
Real Part (Np/s)

Q=0

Q=0.4
1

1.5

Figure 5.6 Variation of the rotor mode with increasing excitation system gain Kex for two
steady-state operating conditions, P = 0.9 and Q = 0, Q = 0.4 pu. No PSS in service.

5.7

Effect of an idealized PSS on stability

In Section 5.4 it was noted for the very simplified model of the SMIB system that the leftshift in the rotor mode due to an idealized speed-PSS is k 4H , where k is the damping
gain of the PSS and H is the inertia constant. Somewhat analogous to the treatment in
Section 5.4 of the effect of the PSS on the rotor mode, let us consider the performance of
an idealized PSS on the SMIB system represented by the Heffron and Phillips model of
Figure 5.1 on page 227.

Sec. 5.8

Tuning concepts for a speed-PSS

237

Because the element D in Figure 5.1 is a pure damping gain, D = P dp , it can be replaced - or augmented - by an idealized speed-PSS of gain k. Setting k = D = 20 pu and by
calculating the poles of the closed-loop transfer function V t s V r s ((5.69)
Appendix 5I.2), the effect on the rotor mode of such a PSS can be assessed.
Let us consider in Table 5.1 on page 234 the steady-state condition P = 0.9, Q = 0 pu for
which the rotor mode is 0.514 j8.70 with no PSS in service. Based on the above-mentioned calculation the rotor mode with the idealized PSS is 1.040 j8.57 . The mode shift
due to the action of the idealized PSS is 1.554 j0.13 which, for practical purposes, is a
direct left-shift of 1.554 Np/s. This is in good agreement with the shift of k 4H = 1.667
Np/s for an inertia constant of H = 3 MWs/MVA. The result will be reviewed later when
a practical PSS is employed with the Heffron and Phillips model of the SMIB system.

5.8 Tuning concepts for a speed-PSS for a SMIB system

The simple Heffron and Phillips model for the linearized SMIB system, shown in Figure 5.1
on page 227, will used to illustrate the concepts employed in the tuning of a PSS based on
the so-called P-Vr method.
As foreshadowed in Section 5.4, the form of the transfer function of the speed-PSS is
H PSS s = kG s where G s = G c s G W s G LP s ;

(5.14)

the right-hand-side transfer functions are defined below. Let us now consider a five-step
procedure for the tuning of the PSS transfer function H PSS s .
1.

Determine the compensating transfer function Gc(s) such that, over a selected range
of modal frequencies, a torque of electro-magnetic origin proportional to speed is
induced by the PSS on the shaft of the generator.

2.

Select the value of the damping gain k.

3.

Select the parameters of the PSS washout filter G W s . This filter blocks steady-state
offsets (i.e. DC signals) and significantly attenuates low-frequency signals below the
range of rotor modal frequencies. Ideally, its transfer function over the range of
rotor modal frequencies is 1 0 .

4.

Select the parameters of the low-pass filter G LP s that significantly attenuates highfrequency signals above the range of rotor modal frequencies. In the range its transfer function is ideally 1 0 . The filter may also attenuate the higher-frequency shaft
torsional modes to ensure that they are not excited by the PSS. If a true-speed input
PSS is employed, which is rare in practice, it may require specialized filters to attenuate the torsional modes.

238
5.

Tuning PSSs for SMIB system

Ch. 5

Implement the PSS tuning, assess its performance, and confirm the validity of the
design over an encompassing range of operating conditions.

These steps are discussed in the following sections.

5.8.1 Determination of compensating transfer function
The determination of compensating transfer function G c s is the first step in the PSS tuning procedure outlined above.
The output of the speed-PSS is injected at the summing junction of the voltage reference
and terminal voltage signals - or at some point in the AVR of low signal-power level. The
PSS forms a feedback loop as shown in dashed lines in Figure 5.7. Note that the sign on the
output signal of the PSS is (maybe unexpectedly) + V s . However, comparing Figure 5.7
with Figure 5.4, we note that the negation already occurs at the output P 2 of the internal
path P 2 in Figure 5.7 and hence a positive sign must arise the input summing
junction.
From Steps 3 and 4 above it follows that, over the range of frequencies of the electro-mechanical modes (i.e. 1.5 to 15 rad/s), the product of the washout and the low-pass filter
transfer functions G W s G LP s should be close to 1 0 . Under these conditions the PSS
transfer function of (5.14) is of the ideal form
H PSS s = kG c s ,

(5.15)

where k is the desired damping torque coefficient (or PSS damping gain), and G c s is a
compensating transfer function.
From an examination of Figure 5.1 on page 227 the equation for the torque of electromagnetic origin is, in (5.1),
P 0 s = P 1 s + P 2 s + P 3 s , P m = 0 .
The action of the speed-input PSS is to induce a torque of electromagnetic origin proportional to speed on the shaft of the generator through the electro-magnetic components of
torque P 1 and P 2 , shown in Figure 5.7. The system being linear, the principle of superposition can be employed to derive expressions for each of the torque components in terms
of the relevant variables, and then combine them appropriately.

Sec. 5.8

Tuning concepts for a speed-PSS

P1

K1

P3

Pa

+_

/(sM)

Pm

239

/s

P2
K2

K4

K6
Kex
+ 1+sTex

K3
1+sK3Tdo

HPSS(s)
= k G(s)

K5

Eq

Speed-input
PSS

Vt _

Vs _ _

Vr

Figure 5.7 Inclusion of a speed-stabilized PSS in the SMIB model.

The paths in the Heffron and Phillips model encompassed by transfer function P 2
are shown by the solid line in Figure 5.8(a). In the absence of perturbations in reference voltage V r (or stabilizing signal V s ) the latter transfer function can be shown by block diagram manipulation to be:
P

K 2 K 3 K 4 1 + sT ex + K 5 K ex

= -------------------------------------------------------------------------------------------------------------------------
2
V = 0
s K 3 T T d0 + s T ex + K 3 T d0 + 1 + K 3 K 6 K ex
r
ex

: ---------2-

P1
_

Pm
+_

K1

P3
Pa

P1
_

P2
K2

Eq

K4

K3
1+sK3Tdo

Kex
+ 1+sTex
(a)

Pa

/s

/(sM)

P2
K2

K5
K6

+_

Pm

/s

/(sM)

K1

P3

K4

K5

+
_

Vt Eq
+

Vr

K6

K3
1+sK3Tdo

Kex
+ 1+sTex
(b)

Figure 5.8 Evaluation of transfer functions,

(a) P 2
(b) P 2 V r
V r = 0

(5.16)

= 0

+
_

Vt
+

Vr

240

Tuning PSSs for SMIB system

Ch. 5

The transfer functions (5.16) and P 1 = K 1 can be combined into a single transfer
function, H P s , i.e.
P
P 2
.
H P s = ---------= K 1 + --------- V = 0
V = 0
r
r

(5.17)

Let us now examine the path in Figure 5.8(b) from the voltage reference V r to the torque
component P 2 . The blocks enclosed by the dashed line form the transfer function 1:
P 2
H PVr s = ---------V r

K 2 K 3 K ex
= ------------------------------------------------------------------------------------------------------------------------- .
2
s K 3 T ex T d0 + s T ex + K 3 T d0 + 1 + K 3 K 6 K ex
= 0

(5.18)

A form of this transfer function that will be useful later is:

kc
H PVr s = ---------------------------------- .
1 + c1 s + c2 s 2

(5.19)

It has been emphasized that Figure 5.8(b) and (5.18) represent the transfer function from
the voltage reference to a component of the torque of electromagnetic origin for no perturbations in the steady-state rotation of the rotor, i.e. = = 0 ; this, in effect, implies
that the shaft dynamics are disabled. Importantly, the transfer function H PVr s , which is calculated with the shaft dynamics disabled, will be referred to in the following chapters as the
P-Vr transfer function or the P-Vr characteristic of the generator.
A useful modified form of Figure 5.7 is represented by the block diagram of Figure 5.9. The
P-Vr transfer function H PVr s is formed as P 2 V r Vs = 0 or P 2 V s Vr = 0 .
The blocks in Figure 5.9 representing the dynamic behaviour of the shaft are shown as
J s = 1 sM and N s = o s . Turbine / governor action can be included - if desired
- in the block HGOV(s). The transfer function of the speed-PSS is represented in its ideal
form as
V s
H PSS s = --------- = kG c s .

(5.20)

Observe that in Figure 5.9 there are three distinct feedback paths, namely, the path through
the PSS and the transfer function H PVr s , the path through the rotor angle, and the path
through the governor and turbine. We shall find the formation of the first two of the separate paths is useful and revealing in the analysis of both single- and multi-machine systems.

1. This transfer function is related to the GEP(s) function as explained in Section 6.4

Sec. 5.8

Tuning concepts for a speed-PSS

HPSS(s)
= k Gc(s)

Feedback path
through PSS

Vs
Vr

241

P2_V

HPVr(s)

Generator and
power system

(see (5.18))

HGOV(s)
Pm

Pa

HP(s)

Feedback path through

rotor angle

J(s) =
1/(sM)

N(s) =
o/s

(see (5.17))

Figure 5.9 Modified form of Heffron and Phillips model including PSS and turbine/governor transfer functions. (X is the point at which the speed path is opened to disable the
shaft dynamics.)
From Figure 5.9 we note that the output of block H PVr s comprises the superposition of
two components P 2 = H PVr V r + V s . For no disturbances at the reference input,
P 2

P 2 V s
= ---------- ---------
V s
V r = 0

(5.21)

= H PVr s kG c s

when each input-output relation is replaced by its transfer function. By definition, the term
H PVr s kG c s in (5.21) is a damping torque coefficient (see Section 5.3). It is clear that a
torque of electromagnetic origin in phase with speed is induced by the PSS if this coefficient
is set to the scalar PSS damping gain, k, i.e.
H PVr s kG c s = k ,

thus

G c s = 1 H PVr s .

(5.22)
(5.23)

Note that G c s is chosen to compensate for the transfer function H PVr s . Following the
substitution of (5.23) in (5.20), the basic transfer function for the speed-PSS becomes
H PSS s = kG c s = k H PVr s .

(5.24)

Let us determine, for the Heffron and Phillips model, the compensating transfer function
to be implemented by the PSS. Substitution of (5.18) in (5.23) yields:
2

s K 3 T ex T d0 + s T ex + K 3 T d0 + 1 + K 3 K 6 K ex
G c s = ------------------------------------------------------------------------------------------------------------------------- .
K 2 K 3 K ex

In an alternative form this compensating transfer function can be expressed as:

(5.25)

242

Tuning PSSs for SMIB system

K 3 T ex T d0
T ex + K 3 T d0
1 + K 3 K 6 K ex
2
G c s = ------------------------------------- 1 + s -------------------------------- + s -------------------------------- .
1 + K 3 K 6 K ex
K 2 K 3 K ex
1 + K 3 K 6 K ex

Ch. 5

(5.26)

As has now been established for the single-machine system, we will find that in the multimachine case the compensation provided by the PSS is simply the inverse of the P-Vr transfer function H PVr s ; it will form the basis for the tuning of PSSs in the following sections
and chapters. Note:

We are proposing to cancel the set of poles in the transfer function P 2 V r of

(5.18) with a corresponding set of zeros; this is allowable as long as the poles of (5.18)
are stable.

The transfer function G c s in (5.25) is not proper (see definition in Section 2.2),
however, this is remedied later.

The basis for the calculation of the P-Vr transfer function can be established from Figure 5.9
with the PSS and governor blocks removed from the diagram. The torque of electromagnetic origin is P 2 V + P . If, in the figure, the speed output from the block J(s) is opened
at X then the speed signal - and thus the shaft dynamics - are disabled ( = = 0 ). Under this condition the output of the block H P s , P , is zero and thus the P-Vr transfer
function can be calculated directly from the transfer function P 2 V V r
.
= 0

The inherent damping and synchronizing torques of the generator are supplied through the
block H P s . As was revealed in Example 5.1, Section 5.5.1, the inherent damping torques
may augment or degrade the damping torque induced by the PSS over the frequency range
of the rotor modes of interest.
For this simple SMIB system, with a third-order model of the machine, the compensating
transfer function G c s of the PSS can be calculated directly from (5.26) for a given operating condition. We shall see in Chapter 10 for higher-order generator models and multi-machine systems that such a direct closed-form calculation is not practical and alternative
methods will be employed for determining the P-Vr characteristic and the associated compensating transfer function G c s . However, for present purposes, the P-Vr transfer function is calculated directly from (5.18); this is illustrated in the following example.
5.8.2 The nature of the P-Vr characteristic
In a theoretical analysis [10] it is shown that the P-Vr transfer function of a machine consists
of two components. The first component depends on the parameters of the generator and
its AVR/exciter and is independent of the external system. On the other hand, the second
component depends on both the parameters of the generator and the dynamics of all other
generators in the system; however, this second term is dominated by the Thvenin equiva-

Sec. 5.8

Tuning concepts for a speed-PSS

243

lent impedance seen from the terminals of the generator looking into the rest of the system.
This result provides a theoretical basis for the observation in [5] that the P-Vr transfer function is relatively robust to changes in the system operating conditions. Typically the frequency response of the P-Vr transfer function for both the gain (for real power outputs greater
than 0.7 pu) and the phase shift - in particular - do not vary appreciably over a wide range
of operating conditions and system configurations. Consequently, PSSs designed based on
the synthesized P-Vr transfer function are also robust over a wide range of operating conditions.
It must be emphasized for future reference: The PSS must be tuned to be robust 1 to a full range
of N and N-1 operating conditions. For this purpose it is necessary to select a set of operating
conditions which encompass, and therefore include, the range of conditions. Be examining the
bordering conditions this approach reduces the number of cases for which the P-Vr characteristics must be evaluated.

5.8.3 Example 5.2: Evaluate the P-Vr characteristics of the generator and determine the PSS compensating transfer function.
The P-Vr characteristics for the generator in the SMIB system of Example 5.1 are to be calculated using the P-Vr transfer function (5.18) when the real power output is P = 0.9 pu and
the reactive power outputs are Q = -0.2, 0.0, 0.2 and 0.4 pu. The P-Vr frequency response
plots are calculated using the machine and system parameters listed in Example 5.1, together
with the K-constants found in Table 5.1 on page 234.
The P-Vr frequency response plots which are shown in Figure 5.10 cover the range of modal
frequencies of 7.9 to 9.3 rad/s for the four operating conditions, Cases A to D.
Based on a mid-range modal frequency of 8.7 rad/s 2 the phase of the P-Vr characteristics
are, from Figure 5.10, -58.9 , -47.7 , -41.3 and -37.5 for values of Q = -0.2, 0, 0.2 and
0.4 pu, respectively. We will select the P-Vr characteristic for Case B, Q = 0, as its phase of
-47.7 lies close to the middle of the band of characteristics at the modal frequency, assuming for the present that the PSS is to be tuned for a real power output of P = 0.9 pu. This
means the phase of the characteristic for any other value of Q, 0.2 Q 0.4 , will be within
about 11 of that of the selected P-Vr characteristic. We shall find later, for a range of more
practical operating conditions, the latter phase variation is typically within 5 to 10 of
the selected characteristic 3.
1. See item 3 of Section 1.2
2. This is the frequency of the unstable rotor mode for P = 0.9, Q = 0.
3. A wider spread exceeding 10 may not be unusual depending on the generator parameters and if the range of leading and lagging power outputs of the machine is wide. The
rated reactive power range of the generator may be wide but not necessarily achievable
due to voltage constraints, transformer tap ranges, etc.

244

Tuning PSSs for SMIB system

Ch. 5

Magnitude (dB)

20
10
0
10
20
0
10

10

10

Phase (deg)

50

100

150
0
10

10
Frequency (rad/s)
A: Q=0.2
C: Q= 0.2

Figure 5.10

10

B: Q= 0.0
D: Q= 0.4

P-Vr characteristics of SMIB System for P = 0.9, Q = -0.2, 0.0, 0.2 and 0.4 pu.

The steady-state or static gain of the P-Vr transfer function is, from (5.18),
K 2 K 3 K ex 1 + K 3 K 6 K ex . The static gain range is within 4.5 dB of the gain of the P-Vr
characteristic for Case B, Q = 0. Later, in Chapter 10 for the multi-machine system, we shall
find that over the range of operating conditions the gain variation is typically within 3 dB
of the selected characteristic. However, in this example the range of reactive power outputs
in Cases A and D may be considered to represent the more extreme conditions and may be
weighted accordingly. We shall refer to a selected or representative P-Vr characteristic such
as Case B as the Design Case.
We have now determined the P-Vr characteristic for which the compensating transfer function of the PSS is to be evaluated. Substitution in (5.26) of the relevant parameters given in
Example 5.1, together with the K-constants for Q = 0 from Table 5.1 on page 234, yields the
compensating transfer function:
G c s = 0.20263 1 + s0.10638 + s 2 0.0021058 .

(5.27)

As noted earlier the above transfer function is not proper; this will be remedied in
Section 5.8.6. It has also been noted that the required form of the PSS transfer function is
kG c s . We will now determine the damping gain k.

Sec. 5.8

Tuning concepts for a speed-PSS

245

5.8.4 Determination of the damping gain k of the PSS

The determination of the damping gain k is the second step in the PSS tuning procedure
outlined in Section 5.8.
In Section 5.4 and in the associated Figure 5.4 it is shown that an ideal PSS introduced a
damping torque of electro-magnetic origin proportional to speed, P dp = k , where k is
both a damping torque coefficient and the damping gain of the ideal PSS. Likewise, we find
from consideration both of the loop through the PSS in Figure 5.9 and of equations (5.21)
- (5.23) that a torque proportional to speed is produced, i.e.
P 2 V

Vr = 0

= H PSS s H PVr s = kG c s 1 G c s = k .

(5.28)

The damping gain k of the PSS should be large enough to:

swamp any inherent negative damping torques over the range of rotor modes for the
range of operating conditions,

ensure that the most lightly-damped rotor mode satisfies the criteria for system damping.

As pointed out earlier, a damping gain of 20 to 30 pu on machine rating is considered to be

a moderate gain.
5.8.5 Example 5.3. Calculation of the damping gain setting for the PSS
For illustrative purposes it will be assumed that, for the most poorly damped rotor mode,
the damping performance criterion as represented by its 2% settling time should be better
than 8 s; this implies that the real part of the rotor mode should be more negative than
= 4 8 = 0.5 Np/s (see Section 2.8.2.1)
According to Table 5.1 on page 234, without a PSS the most poorly damped mode is that
for the operating condition P = 0.9, Q = -0.2, i.e. 1.146 j9.23 . To achieve the specified
damping target of -0.5 Np/s the PSS must cause a left shift in the mode of at least
= 0.5 1.146 = 1.646 Np/s. From Section 5.4 and (5.12) we note, for a rough
guide, that for the very simplified model of the SMIB system the left-shift in the mode
associated with a damping gain kd is = k d 4H . For an inertia constant H = 3 the value
of k d required to achieve the specified left shift is k d = 4H = 4 3 1.646 = 19.75 pu.
Thus, we anticipate that a moderate PSS damping gain of k = 20 pu will achieve the specified damping performance. Moreover, the value for the damping gain of 20 pu will swamp
the inherent damping torque coefficient of -16.9 pu for Q = 0.2 pu.

246
5.8.6
5.8.6.1

Tuning PSSs for SMIB system

Ch. 5

Washout and low-pass filters

The washout filter

The selection of parameters of the washout filter is the third step listed in the PSS tuning
procedure in Section 5.8.
A washout filter is incorporated in the PSS to ensure that steady-state or slow changes in
system frequency - and thus shaft speed - do not offset, in the steady state, the terminal voltage of the generator from its reference value. The transfer function of the washout filter is
of the form:
sT W
G W s = ------------------1 + sT W

(5.29)

s = j f

The frequency response of the filter is shown in Figure 5.11(a). Note from (5.29) that as
f 0 , G W j f 0 , i.e. steady-state offsets or d.c. signal levels are blocked. Signals below the corner frequency 1 T W are attenuated, and the phase shift introduced by the filter
tends to 90 as f 0 . For frequencies much greater than the corner, G W j f 1 0 , i.e.
unity gain. A basis for the selection of the washout time constant T W is to place its corner
frequency (i.e. 1 T W rad/s) about a decade below the lowest frequency of the rotor modes
of oscillation, normally an inter-area mode. At a frequency a decade above the corner frequency the phase lead introduced by the washout filter is about 5 , i.e. almost negligible for
tuning purposes. In particular, this basis for the selection of the washout time constant mitigates against any excessive phase lead at low inter-area modal frequencies.
Two washout filters in cascade are often employed in practice to block slow ramp-like
changes in system frequency that occur in system operation. The washout filter will be discussed more fully in Chapter 8.
5.8.6.2

The low-pass filter

The selection of the parameters of the low-pass filter is the fourth step in the PSS tuning
procedure of Section 5.8.
Low-pass filters are added to attenuate high-frequency signals that would otherwise be amplified by the PSS, and to ensure that the PSS transfer function is proper. Attenuation of
shaft torsional mode components is typically accomplished by specialized filtering of the
speed input signal. The simplest form of the filter is
1
G LP s = --------------------------------------------------------------------------- .
1 + sT 1 1 + sT 2 1 + sT p

(5.30)

The corner frequencies associated with the time constants T 1 T 2 are usually placed a
decade above the highest frequency of the rotor modes of oscillation, normally a local area

Sec. 5.8

Tuning concepts for a speed-PSS

247

20
30
40
2

Phase (deg)

10

(a)
(a)

Magnitude (dB)

10

10

10

20
40
60
80
0
10

10

100

80

45

Phase (deg)

Magnitude (dB)

or an inter-machine mode of oscillation. The order p of the filter is selected to provide adequate high-frequency attenuation and / or to ensure a proper PSS transfer function. The
frequency response plot of a second-order low-pass filter G LP s = 1 1 + s c 2 is
shown in Figure 5.11(b).

60
40
20
0
2
10

10
10
Frequency (rad/s)

10

10

10

10

90
135
180
0
10

(b)

10
10
Frequency (rad/s)

10

Figure 5.11 Frequency response of (a) a first-order washout filter and (b) a second-order
low-pass filter. The corner frequencies are normalized to 1 and 10 rad/s, respectively. For
another corner frequency, c rad/s, scale frequency axes to c or c 10 .
Assume that the two corner frequencies of the second-order low-pass filter are set to 200
rad/s 1. Based on Figure 5.11, a decade below the corner, i.e. at 20 rad/s, the phase lag is
11.4 . Ideally, for tuning purposes the phase lags due to the low-pass corners should be
small to negligible over the range of the modes of rotor oscillation, typically 1.5 to 15 rad/
s. Of course, in a specific application the relevant modal frequency range might be much
less. Often, in practice, the maximum corner frequency of the low-pass filters may be limited
by the PSS manufacturer to values in the range from about 20 to 50 rad/s.
Note that torsional modes may not be insignificant and must be attenuated. The frequencies
of these modes may be as low as 8 Hz (50 rad/s). For example, if 40 dB attenuation at 50
rad/s is required, the corner frequency of a second-order filter is 5 rad/s. Such a filter introduces a phase lag of 90 at 5 rad/s - possibly in the mid-range of modal frequencies; more1. For the purposes of this study the corner frequencies of 200 rad/s are chosen so that the
effects of the low-pass filter over the modal frequency range are minimal.

248

Tuning PSSs for SMIB system

Ch. 5

over, the damping gain of the PSS amplifies the torsional modes if insufficiently attenuated.
This conflict between the requirements of the PSS design and attenuation of the torsional
frequencies is overcome either by the adoption of specialized filtering of the PSS speed-input signal 1 or by using alternative PSS structures such as the integral-of-accelerating-power
PSS; this is discussed in more detail in Chapter 8.

5.9

Implementation of the PSS in a SMIB System

The implementation of the PSS design, assessment of its performance, and confirmation of
the validity of the design is the fifth step listed in the PSS tuning procedure of Section 5.8.
5.9.1 The transfer function of the PSS
In summary, it has been noted that the following are among the requirements for the implementation of a practical PSS:

A washout filter (5.29) is required to eliminate the offset resulting from the steadystate level of the input signal to the PSS. Its response to frequencies in the range associated with rotor modes should ideally be 1 0.

The elemental PSS transfer function in (5.26) is not proper; the selection of a low-pass
filter (5.30) of an appropriate order can overcome this deficiency. The response of the
filter to frequencies in the range associated with rotor modes should ideally be 1 0 .

The PSS must not excite the torsional modes of the rotors of the generator - primemover system. Depending on the magnitude and frequency of the torsional modes the
higher-order low-pass filter having an order 2 or greater may serve this purpose though separate specialized filtering of the PSS speed-input signal may be required.

The damping gain setting k (pu on generator rating) is selected to have a moderate
value 20 - 30 pu and / or to satisfy the criterion for system damping performance though very high damping gain settings should be avoided.

The practical form of the PSS should include or account for the above set of requirements
over the range of modal frequencies. Two general forms of the speed-PSS, kG s , which
includes the washout and low-pass filters, therefore become
1
H PSS s = kG s = k -------------------- G W s G LP s ,
H PVr s

(5.31)

H PSS s = kG s = kG c s G W s G LP s ,

(5.32)

or

where the filter transfer functions are given by (5.29) and (5.30), respectively.

1. Notch filters are used for the same purpose [9].

Sec. 5.9

Implementation of the PSS in a SMIB System

249

The compensating transfer function G c s of the speed-PSS (5.26) is of the form:

1 + c1 s + c2 s 2
G c s = ---------------------------------- .
kc

(5.33)

Substitution of the latter equation and the filter transfer functions in (5.32) yields an equation containing all the relevant parameters for the practical PSS for this single machine case,
i.e.
1 + c1 s + c2 s 2
sT W
1
H PSS s = kG s = k ------------------- ----- ---------------------------------------------- .
1 + sT W k c 1 + sT 1 1 + sT 2

(5.34)

It is important to note that, in the context of (5.34), the gain k has been referred to as the
damping gain of the PSS. If the washout filter is ignored, the DC gain of the PSS is k k c ;
conventionally this is referred to as the PSS Gain. The choice of the parameters and the im-

plementation of the practical PSS is now examined.

5.9.2 Example 5.4. The dynamic performance of the speed-PSS
In Examples 5.1 to 5.3 a SMIB system has been analysed for the operating conditions
P = 0.9 pu, Q = -0.2, 0.0, 0.2 and 0.4 pu. The single rotor mode of oscillation to which the
criterion for damping must be applied is given in Table 5.1; these modes lie in the range 7.9
to 9.3 rad/s and all are unstable. The components of the tuning procedure will be now be
reviewed.
The PSS compensating transfer function. This was evaluated in Example 5.2. Because the
PSS is a fixed parameter device, i.e. its parameters are not changed with changes in operating
conditions, the compensating transfer function is based on a P-Vr characteristic that best
represented the phase responses of the characteristics over the set of operating conditions.
The appropriate P-Vr characteristic was shown to be that for the operating condition
P = 0.9, Q = 0 (Case B); the associated compensating transfer function is given by (5.27), i.e.
G c s = 1 + s0.10638 + s 2 0.0021058 4.935 .
The washout filter. The corner frequency of the filter is placed a decade or more below the
lowest rotor modal frequency, 7.9 rad/s. For a washout time constant of TW = 5 s the filters
corner frequency is 0.2 rad/s, a value that satisfies the latter requirement. The frequency response of this washout filter is shown in Figure 5.11(a). (Note that a washout time-constant
of 5 s will ensure that the PSS contributes a pure damping torque at a frequency of about
10/5 = 2 rad/s.)
The low-pass filter. The corner frequency of the filter should ideally be a decade or more
above the highest rotor modal frequency, 9.3 rad/s. The selection of a second-order filter
with corners at 200 rad/s results in a small phase lag of 5 at 9.3 rad/s; this selection ensures
that the PSS transfer function is proper.

250

Tuning PSSs for SMIB system

Ch. 5

(If connected to multi-machine system, and if modal frequencies other than that at 9 rad/s
and other issues are not of concern, the washout and low-pass filter time constants could be
modified to 2 and 0.01 s, say.)
The PSS transfer function. On insertion into (5.34) of the parameters determined above, together with the damping gain of k = 20 pu on machine rating, the practical PSS transfer
function becomes
s5
1
1 + s0.1064 + s 2 0.002106
H PSS s = 20 --------------- ------------- ------------------------------------------------------------------- .
1 + s5 4.935 1 + s0.005 1 + s0.005

(5.35)

A plot of the PSS transfer function is shown in Figure 5.12 in which the following features
are observed. (i) Below the range of modal frequencies, 1-15 rad/s, the washout filter becomes effective. (ii) Over the range of modal frequencies the responses are the mirror image
of those of the P-Vr characteristic for Design Case B, Figure 5.10. The phase lag of ~ 45
in the P-Vr characteristic at the modal frequency of 8-9 rad/s is cancelled by the phase lead
introduced by the PSS. (iii) Over the range of modal frequencies the PSS gain is close to
20 4.935 , or 12.2 dB. (iv) Above the modal frequency range the second-order low-pass filter becomes effective. As mentioned earlier, to reduce the high frequency gain the corner
frequencies of the low-pass filter should be reduced.

Magnitude (dB)

40
30
Range of modal
frequencies

20
10
0
1
10

10

10

10

120
Phase (deg)

100
80
60
40
20
0
1
10

10
10
Frequency (rad/s)

10

Figure 5.12 Frequency response plot of PSS transfer function, (5.35). Over the modal
frequency range the responses are the mirror image of those of the P-Vr characteristic for
Design Case B

Sec. 5.9

Implementation of the PSS in a SMIB System

251

Note that in the transfer function of the compensation in (5.35), i.e.

1
1 + s0.1064 + s 2 0.002106
G Hc s = ------------- ------------------------------------------------------------------- ,
4.935 1 + s0.005 1 + s0.005

(i) the zeros are real and lie at -12.5 and -38.0, and (ii) the low-pass filter time constants (5
ms) may be too short to implement in a PSS in practice 1. However, the 5 ms time constants
are retained in the present analysis so that their effect over the range of modal frequencies,
1.5 - 15 rad/s (~0.25 - 2.5 Hz), is minimal.
The rotor modes of the SMIB system with the PSS in service can now be calculated assuming that the SMIB and PSS parameters are those provided in Example 5.1 and (5.35). The
resulting rotor modes and the associated shifts in the modes for the selected operating conditions are listed in Table 5.2; the same information is displayed in Figure 5.13 on page 252.
Table 5.2

Effect on rotor modes with PSS in service (P = 0.9 pu), k = 20 pu.

Case /
Q pu

Eigenvalues, rotor mode

with PSS out of
service

with PSS
in service

Mode shift

A / -0.2

1.148 j9.23

0.684 + j9.54

1.832 + j0.305

B / 0.0

0.514 j8.70

1.072 + j8.67

1.586 j0.025

C / 0.2

0.267 j8.25

1.008 + j8.08

1.275 j0.166

D / 0.4

0.198 j7.94

0.836 + j7.75

1.034 j0.193

The table reveals that, with the PSS in service, the 2% settling time of the rotor mode for all
operating conditions is shorter than the criterion of 8 s, i.e. all modes are better damped than
a mode with a real part of -0.5 Np/s. The PSS design therefore satisfies this performance
specification.
Note that the left-shift in modes varies over a range of generator reactive output. The reasons for this and the variations in modal frequencies are now examined.
5.9.3 Analysis of the variation in the mode shifts over the range of operating conditions
Upon an examination of Table 5.2 and Figure 5.13 the following questions arise: Why does
the extent of the left shift of the modes increase as the reactive power output of the generator changes from 0.4 lagging to 0.2 leading? Bearing in mind that the PSS tuning is based
on the Design Case B (P = 0.9, Q = 0 pu), why does the modal frequency increase when the
reactive power output decreases from that for the Case B, and decrease as the reactive out1. The realization of an alternative transfer function to overcome both this and the high
gain at higher frequencies is considered in an example in Section 5.12.

252

Tuning PSSs for SMIB system

Ch. 5

put increases from that for the design case? The following addresses these and some other
questions.

Imaginary Part (rad/s)

10
A

9.5

A
9
B

B
8.5
PSS On
8

C
C

PSS Off

D
D

7.5
7
2

1.5

1
0.5
0
0.5
Real Part (Np/s)

A Q=0.2, B

Q=0.0, C

Q=0.2, D

1.5

Q=0.4

Figure 5.13 Rotor modes of oscillation for the PSS in and out of service for operating
conditions P = 0.9 pu: Q = -0.2 (A), 0 (B), 0.2 (C) and 0.4 (D) pu; damping gain k = 20 pu.
In Example 5.3 of Section 5.8.5 it was pointed out that, for the very simplified SMIB system
of Figure 5.4, an estimate of the left-shifts of the rotor modes due to the action of the ideal
PSS is = k d 4H = 1.667 . Because the left-shifts of the rotor modes in Table 5.2 differ
significantly from the latter value, let us investigate the reasons for these discrepancies.
For a SMIB system it will be shown in Chapter 13 that, due to an increment in the damping
gain k of the PSS, the shift in the complex rotor mode h is given by
h
h = -------------- H PVr h G c h k ,
2H

(5.36)

where h is the complex participation factor of the generators speed state in the mode
h , evaluated with the PSS in service with the damping-gain setting, k = ko. [Equation (5.36)
is derived from (13.8)]. The left-shift in mode h is h . Ideally, the compensating
transfer function of the PSS is given by (5.23), i.e. G c h = 1 H PVr h , thus (5.36) becomes
h
h = -------------- k .
2H

If the participation factor h of the generator is complex, the mode shift

(5.37)

Sec. 5.9

Implementation of the PSS in a SMIB System

h = r h + j i h k 2H ,

253
(5.38)

is complex. However, for cases in which the participation of the generator is relatively high
the participation factor h is real or almost real, the mode shift h is directly to the left
in the complex s-plane 1; this left-shift is a prime aim of stabilization using PSSs.
Note that equations (5.23), (5.36) and (5.37) strictly apply at the complex rotor mode
h = h j h . However, because the modes are relatively lightly damped, with damping
ratios typically less than 0.15 to 0.20, it is assumed that conventional frequency response analysis can be applied with s = h j h . The affect of this assumption will be assessed below.
Strictly, equation (5.37) applies only to the case when the compensating transfer function
G c s is based on the P-Vr characteristic for the operating condition selected to be the Design Case, i.e. P = 0.90, Q = 0 as determined in Example 5.2. However, it is illustrative to
examine the effect on the mode shift of employing the P-Vr characteristic at some other operating condition p than that for the Design Case, and attempt to account for the contributions to any differences in the respective mode shifts. For the Design Case let
G co s = G c s . The P-Vr characteristic at operating condition p is given by,
k cp
H PVrp s = ---------------------------------------- .
1 + c 1p s + c 2p s 2

(5.39)

Substituting for G co s and H PVrp s from (5.33) and (5.39), respectively, in (5.36) the
mode shift for operating condition p is:
k cp
p s
1 + c 1o s + c 2o s 2
p = ------------ -------------------------------------------- -------------------------------------------- G W s G LP s k , (5.40)
2H s =
k co
1 + c 1p s + c 2p s 2
h

where subscript o refers to the coefficients in the compensating transfer function which is
implemented in the PSS. In Section 3.10 it is shown that the participation factor is a function
of the eigenvectors and thus it must be calculated for the mode hp = hp j hp . The other terms in (5.40) will be calculated for s = hp and for the frequency s j hp . The expression for the mode shift in (5.40) reduces to:
p s Kp s
hp ---------------- -------------- k K filt 180 + p hp + p s o s + filt s , (5.41)
2H
Ko s
where the gains Ko, Kp and the phase shifts o , p are calculated from the P-Vr character-

istics (5.18); p is the phase angle of the participation factor at hp . The gain and phase
contributed by the combination of the washout and low-pass filters are Kfilt and filt .

1. For this case of a single rotor mode h 0.5 in (5.37).

254

Tuning PSSs for SMIB system

Ch. 5

Based on (5.41), for the modal frequency s = h and the frequency s = j h , we can calculate and account for all contributions to the mode shift and provide an estimation of that
shift. This then allows us to:

compare the estimated value with the actual mode shift (which employs the complex
modal frequency h );

examine the contributions to the mode shift for operating conditions other than the
condition used as basis for tuning purposes, and hence

gain an understanding of the nature of the complex mode shifts observed in

Figure 5.13 in terms of the P-Vr characteristics for the various operating conditions
versus the condition used as the basis for tuning purposes.

For each operating condition in Table 5.3 the estimated mode shifts evaluated from (5.41)
and their components using s = hp are compared to those shifts which are calculated with
s = j hp . The operating conditions and closed-loop rotor modes are those listed in
Table 5.2. In columns 4 to 10, Table 5.3, are incorporated the components or elements that
determine not only the nature and extent of the left-shift but also the change in the modal
frequency due to the action of the PSS. Note, however, the incremental gain k in (5.41)
assumes a value of 20 pu based on the damping gain setting, ko = 0 (because the PSS is initially out of service). The significance of using what may be considered to be large value of
incremental gain and its effect on the participation factor is discussed below.

From Table 5.3 the following are preliminary insights into the PSS design are derived:
1.

For this example and for the operating condition p, the estimated left-shifts in column 10 based on P-Vr characteristic for p agree within one percent for two methods
of calculation, s = h and s = j h . For tuning purposes the use of the conventional frequency response method in the analysis does not lead to significant errors.
For damping ratios of the rotor modes exceeding 0.15 to 0.2 the accuracy of the calculation may decrease significantly and should be verified.

2.

Consider in Figure 5.13 and Table 5.2 a selected operating condition - say Case C with the PSS off; the mode for this condition is 0.267 j8.25 . When the PSS is in
service with a damping gain of 20 pu the mode is shifted by 1.275 j0.166 to
1.008 + j8.08 . With a small imaginary component in the mode shift, the compensating transfer function G c h has effected a more-or-less pure left shift of the mode while
the damping gain has determined the extent of the shift. For the selected operating condition the components in columns 3 to 9 in Table 5.3 which contribute to the mode
shift do not alter significantly as the damping gain is increased. Therefore, as implied
by (5.36), h k , i.e. the incremental mode shift is proportional to the increment in

Sec. 5.9

Implementation of the PSS in a SMIB System

255

damping gain; it will be noted in Figure 10.26 that this applies to relatively large
changes in gain of 5 to 20 pu.

Table 5.3

Case
/
Q pu
(P =
0.9
pu)
col.1
A/
-0.2
B/
0.0
C/
0.2
D/
0.4

Components of estimated mode shift (5.41), s = h or s = j h ,

P = 0.9, Q pu

Average
Product
Modal
participActual
Response
Washout Low-pass H PVrp s .
Estimated
ation factor
Mode
Shift based
Filter
filter
s = h ,
Mode Shift using
G co s
on P-Vr for
av
or Freq.
P-Vr for operating Design Case B
Response
condition p
mag, phase mag (abs), mag (abs), mag, phase
(see Table 5.2)
s = j h
(abs), ( )
phase
(abs), ( ) phase
col.2

h
j h
h
j h
h
j h
h
j h

col.3
0.450, 5.7

0.475, 4.7

0.483, 4.2

0.484, 3.5

cols.4-5

cols.6-7

cols.8-9

col.10

col.11

1.00, 1.2

1.00, -5.5

1.00, 1.2

1.00, -5.5

1.240, -12.1 1.838 + j0.347 1.832 + j0.305

1.241, -11.2 1.831 + j0.315

1.00, 1.3

1.01, -5.0

1.000, 0

1.00, 1.3

1.00, -5.0

1.000, 0

1.00, 1.4

1.01, -4.6

0.792, 7.0

1.00, 1.4

1.00, -4.6

0.797, 6.2

1.00, 1.4

1.01, -4.4

0.644, 10.7

1.00, 1.5

1.00, -4.4

0.651, 9.7

1.601 j0.028 1.586 j0.025

1.579 j0.028
1.276 j0.179 1.275 j0.166
1.268 j0.161
1.028 j0.204 1.034 j0.193
1.030 j0.186

Eigenvalues, rotor modes, PSS in service (see Table

5.2)

h for Cases A: 0.684 + j9.54 ; B: 1.072 + j8.67 ; C: 1.008 + j8.08 ; D: 0.836 + j7.75 .
h = Im h .
Note: H=3.0 MWs/MVA; k = 20 pu on machine MVA rating; K-constants from Table 5.1 on page 234;
K p K o p o represents H PVrp s G co s at s = or s = j .
h
h

Consider now the contributions to the complex mode shift as revealed in the columns of
Table 5.3.
3.

The participation factor is non-linearly dependent on the complex modal frequency,

h , through its eigenvectors, p hk = w hk v kh . The participation factors in column 3
are therefore taken as the average of the complex participation factors for the PSS
in- and out-of-service. For case A the latter two factors are, respectively,
0.463 11.4 at a damping gain of k = 20 pu and 0.442 0.2 at k = 0 pu. Because
the participation factors change as the gain k increases from 0 to 20 pu, the average
value is assumed to be a best estimate for present purposes. Note from (5.38) that a

256

Tuning PSSs for SMIB system

Ch. 5

negative phase angle for the participation factor leads to an increase the modal frequency h , while a positive angle produces a decrease in frequency. The latter observa-

tions explain in part the shapes of the eigenvalue plots in Figure 5.13 for the range
of operating conditions. Note the magnitude of the participation factors are less
than 0.5; the reasons for the slightly lower values are explained in Chapter 9.
4.

The phase angles introduced by the washout and low-pass filters in columns 5 and 7
contribute to the change in modal frequency, i.e. a negative phase angle so introduced
produces an increase in modal frequency and vice-versa.

5.

For the P-Vr characteristic on which the PSS compensation is based (i.e. Case B) the
product H PVro h G co h = 1 0 (see columns 8-9). As shown in Figure 5.10
on page 244 the phase of the P-Vr characteristic H PVrp h for Case A (Q = -0.2)
lags (is more negative than) that of the selected P-Vr (case B, Q = 0). Consequently
the phase of the product H PVrp h G co h in (5.36) is negative, thus leading to
an increase in modal frequency. For Cases C and D (Q = 0.2, 0.4) there is a decrease
in modal frequency corresponding to their P-Vr phase characteristics being more
positive than that of Case B at the modal frequency.

6.

The low-frequency gain of the P-Vr characteristic (Figure 5.10 on page 244) for
Case A being greater than that for the design Case B results in the magnitude of the
product H PVrp h G co h in (5.36) being greater than 1. The magnitude of the
resulting mode shift is thus greater for those P-Vr characteristics whose low frequency
gains are greater than that of the design Case B and vice-versa.

7.

For the Design Case B the actual mode shift (column 11) at the modal frequency
s = h should ideally be purely real. However, the non-zero phase shifts in the
washout and low-pass filters as well as the participation factor introduce a small
imaginary component into the actual mode shift.

8.

The main factors which cause the left-shift to deviate from the ideal, or from that
for the Design Case, are highlighted in (5.36), i.e. the variations with operating condition of both the participation factor from 0.5 0 and the product
H PVrp h G co h from 1.0 0 .

Note that the observations and concepts introduced in the items above are particularly relevant to the design of PSSs in multi-machine systems that are considered in Section 10.4.
For implementing the compensation based on the P-Vr characteristic selected for the tuning
of PSSs, it may be that the manufacturer of the PSS has not provided sufficient number of
blocks to supply the phase lead required at the modal frequency h . The phase of the product H PVr h G h (of unity magnitude) will then be negative (due to the net phase lag)
and will result in an increase in the frequency of the mode shift according to (5.36). If the

Sec. 5.10

Tuning of a PSS for 6th order generator model

257

net phase lag is large, say 30 , the increase in modal frequency is equal to sin 30 = 0.5
times the magnitude of the mode shift; correspondingly the left-shift is reduced to
cos 30 = 0.87 times the magnitude.

5.10 Tuning of a PSS for a higher-order generator model in a SMIB

system
In the previous sections of this chapter there are several noteworthy features:

The tuning of the PSS was based on a closed-form relation for the compensating transfer function of the PSS, given by (5.26), for the third-order generator model and its
first-order excitation system. In practice such an expression is not easily derived for a
higher-order generator, particularly in a multi-machine system.

Because of the simplicity of the SMIB system model the P-Vr characteristics and other
relations can be presented as closed-form expressions in an in-depth analysis. Insightful information about the process of PSS tuning and its performance can then follow.

The operating conditions in the SMIB system, in which the external system is represented by a simple series impedance, resulted in bus voltage levels outside the range of
0.95 to 1.05 pu for the more extreme operating conditions as shown in Table 5.1.
Results are more meaningful if practical operation within voltage limits is observed;
this is the case in the following analysis for a much wider range of conditions including line outages.

The real power output of the generator was confined to its rated value. The performance of the system with the PSS in service at rated and at lower real power outputs is
also of interest.

For practical purposes the phase characteristics of the P-Vr about the design case were
revealed to be more-or-less invariant over a range of operating conditions. The variation in the P-Vr gain characteristics was as much as 4 dB, this value however
depends on the choice of the Design Case P-Vr characteristic. Do these observations
apply to higher-order generator models with operational constraints applied?

In order to introduce further operational and modelling considerations, the P-Vr characteristics will thus be based on both a sixth-order generator model and its excitation system and
a wider, practical and encompassing set of normal and contingency operating conditions. In
normal practice, higher-order generator models and higher-order excitation systems are
used in the simulation of the dynamic performance of power systems.
The purpose of the following sections is to examine again (i) the synchronizing and damping
torque coefficients of the sixth-order generator and its excitation system in a SMIB context,
(ii) its P-Vr characteristics, and (iii) the tuning of the PSS. The procedure will form the basis

258

Tuning PSSs for SMIB system

Ch. 5

for the tuning of PSSs in multi-machine power systems. In addition, it is of interest to examine
the effect of the d- and q-axis windings of the generator on the synchronizing and damping
torque coefficients.

The more practical SMIB power system and the model of the generating unit are described
in Section 5.10.1. The tuning of the associated PSS is illustrated by means of an example in
which both normal and line-outage conditions are considered for a range of generator real
and reactive power outputs. The P-Vr characteristics are calculated and the performance of
the PSS - the tuning of which is based on these characteristics - is then examined. From such
an examination the implications for the tuning of PSSs in a multi-machine system are assessed.
5.10.1 The power system model
A more practical SMIB system now considered consists of the generator connected to an
infinite bus through a step-up transformer and a pair of transmission lines as shown in
Figure 5.14. However, in this case the transformer is fitted with a tap changer, shunt capacitance is included in the model of the transmission lines, and a constant impedance load is
connected to the high voltage terminals of the generator transformer. Post-fault contingencies are represented by one or both of the circuits a and b being out-of-service. This arrangement represents more closely a practical configuration in the vicinity of a generating
station that feeds into a large system. Clearly, it not intended to model a tightly-meshed system of generating stations, loads and interconnecting transmission lines.
b

Constant Impedance Load

Figure 5.14

Infinite Bus

A more practical SMIB system.

For a given set of machine and system parameters, the real and reactive power outputs (P,
Q) of the generator together with the complex power of the load are selected as input quan-

tities together with a terminal voltage setting of 1 pu.

The Thvenin equivalent at the generator terminals of the system between the infinite bus
and the generator terminals is calculated. The model of the generating unit, together with
the Thvenin-equivalent voltage source and the associated impedance, form the model of
the SMIB system.
For convenience, the coupled-circuit form of the sixth-order generator model described in
[9] is adopted for the analysis in the remainder of this chapter. Closed-loop control of the

Sec. 5.10

Tuning of a PSS for 6th order generator model

259

generator terminal voltage is again implemented by means of a simple first-order model of

the excitation system.
The linearized model of the generator and excitation system used in this application is also
that described in [9]. The equations for the sixth-order generator in a SMIB system, together
with those of the excitation system, are given for completeness in Appendix 5I.3. Because
a model of the stabilizer is required in a later section, the equations for the complete generator-excitation system-PSS are also included in the Appendix. These equations are in statespace form, rather than the transfer function form employed in the earlier sections of this
chapter.
5.10.2 Calculation of the synchronizing and damping torque coefficients
In Section 5.3 it was noted that the synchronizing and damping torque coefficients can be
calculated from the transfer function P 0 given by (5.10) with the shaft dynamics disabled. In the context of the state equations given by (3.9), this implies removing the row and
column associated with the speed state in the A matrix and treating shaft speed as an
input signal. The latter column then becomes a column associated with shaft speed in the
input matrix, B. The column associated with the speed output in the output matrix C is also
moved to the column of the D matrix corresponding to the speed input; the torque of electro-magnetic origin P 0 remains an output signal.
For illustrative purposes and simplicity, assume that (i) a set of third-order state equations
describes the dynamic performance of the system, (ii) x 2 t is the speed state , and (iii)
y t is the output P 0 . The state equations are:
a 11 a 12
x 1 t
x 2 t = a 21 a 22
x 3 t
a 31 a 32

x1 t

a 13

x1 t

b1

a 23 x 2 t + b 2 u t ,
x3 t
a 33
b3

y t = c 11 c 12 c 13 x 2 t .
x3 t

(5.42)

Setting the input u t to zero and eliminating the rows of the A, B & C matrices associated
with the speed state as well as the corresponding column in the A matrix, we find:
a
x 1 t
= 11
a 31
x 3 t

a 13
a 33

x1 t
x3 t

a 12
a 32

x 2 t , y t = c 11 c 13

x1 t
x3 t

+ c 12 x 2 t . (5.43)

with x 2 t being the speed input signal. From the modified state equations (5.43) the frequency response P 0 j f j f is calculated. The calculation of the frequency responses of the torque coefficients is outlined in Section 5.3 and is given by:
P 0 j f
f P 0 j f
k d = ----------------------- and k s = ------ -----------------------
j f
0 j f

[(5.10), repeated].

260

Tuning PSSs for SMIB system

Ch. 5

5.10.3 Calculation of the P-Vr characteristics for a SMIB system with high-order
generator models
As mentioned in the Introduction, Section 5.1, the theoretical basis for the use of the P-Vr
characteristic for the tuning of PSSs is given in [10]; in this paper the application of P-Vr
characteristic in PSS designs, described in earlier papers [5] and [11], is confirmed. The analysis in [10] is applied to multi-machine power systems - of which the SMIB system is a special case; the results of this analysis are reviewed in the multi-machine context in Section 9.4.
The tuning of the speed-PSS described in Section 5.8 is based on determining the compensating transfer function of PSS, G c s , given by (5.23), i.e.
G c s = 1 H PVr s .
In the case of the Heffron and Phillips model of the SMIB system a closed-form expression
for the second-order P-Vr transfer function, H PVr s - and thus G c s , was derived. For
the seventh-order generator-excitation system model, the P-Vr characteristic is sixth-order
(the shaft dynamics being disabled). For this transfer function, and for multi-machine systems
in particular, the derivation of a closed-form solution for G c s is not only tedious but is
unnecessary. Unnecessary, because

the calculation of the P-Vr frequency responses is straight-forward and is based on a

set of state equations relating the torque of electromagnetic origin to the reference
voltage with the speed state disabled;

from the set of P-Vr frequency responses - such as those in Figure 5.10 - the P-Vr
characteristic is selected that best represents the family of such characteristics over the
range of operating conditions;

typically, only a low-order representation of the P-Vr characteristic is required in order

to provide a suitable PSS compensating function, G c s ;

it is usually relatively easy to synthesize a low-order transfer function representation of

the selected P-Vr characteristic.

Because the P-Vr transfer function is that from the voltage reference input to the torque of
electro-magnetic origin as output with the shaft dynamics disabled, its calculation is similar to
that for the torque coefficients in (5.42) and (5.43). The torque of electromagnetic origin
P 0 remains an output signal. Retaining the third-order system of (5.42) for illustrative purposes for this case, we note that y t remains the torque of electromagnetic origin P 0 and
x 2 t the speed state , however, the input u t is now the voltage reference signal, V r .

On elimination of the speed state, (5.42) reduces to:

Sec. 5.10

Tuning of a PSS for 6th order generator model

a a
b
x 1 t
x t
x t
= 11 13 1
+ 1 u t , y t = c 11 c 13 1
.
x t
a
a

x
x
3
31 33
3
3
3

261

(5.44)

The P-Vr characteristic P 0 j f V r j f is calculated from the state equations, modified

as illustrated by (5.44).
5.10.4 Example 5.5: Tuning and analysis of the performance of the PSS for the
higher-order generator model
5.10.4.1

Parameters of the SMIB system and generator

The generator is connected to the infinite bus through a step-up transformer and two 330
kV transmission lines; the lines are connected to a common bus at their midpoint as shown
in Figure 5.14. Nominal system frequency is 50 Hz.
Parameters:

Transmission lines. For each of the four 330 kV line sections, length 290 km, the
parameters are given on the generator MVA rating and base of 500 MVA. The series
impedance each line is Z L = 0.0225 + j0.225 pu, shunt susceptance b = 0.11 pu. For
the outage of transmission line section a in Figure 5.14 the series impedance from
the generator HV bus to the infinite bus is Z L = 0.03375 + j0.3375 pu, the associated
shunt susceptance is b = 0.33 pu. For the outage of two lines, both sections a and
b are out of service and Z L = 0.045 + j0.450 pu, b = 0.22 pu.

Transformer. Series impedance is Z t = j0.15 pu on 500 MVA. The off-nominal tap

setting t is the per-unit turns ratio (= 1 pu at nominal taps); the taps are assumed to be
on the low-voltage side and adjustable in steps of 1%. (In practice, taps are located on
the high voltage side of the transformer, however, the convention in [9] is adopted.)

Constant shunt admittance load located at the stations high voltage bus. The per unit
complex load power at 1 pu voltage is Po = 0.09, Qo = 0, 0.02, 0.03, 0.04 (or Qo = 0.29
pu when a shunt reactor is brought into service).

Generator. Rating 500 MVA.

D = 0, H = 3.0 MWs/MVA, ra = 0, xd = 1.9 pu, xq = 1.8 pu, x d = 0.30 pu,

xl = 0.20 pu,
x q = 0.55 pu, x d = 0.26 pu, x q = 0.26 pu, T do = 6.5 s, T qo = 1.4 s,
T do = 0.035 s, T qo = 0.04 s, on the generator rating.

The per-unit parameters on generator rating for the corresponding sixth-order coupled-circuit model are:

262

Tuning PSSs for SMIB system

Ch. 5

Xd = 1.9, Xad = 1.7, Rfd = 0.00088453, Xfd = 0.106250, R1d = 0.0227364, X1d = 0.15,
Xq = 1.8, Xaq = 1.6, R1q = 0.0046564,

X1q = 0.4480,

R2q = 0.0336146,

X2q = 0.0724138.

Excitation system: Kex = 200 pu, Tex = 0.02 s.

The state-equation model for the generator, which is under closed-loop voltage control, is
given in Appendix 5I.3.1.
The ranges of operating conditions for the generator are P = 0.9 and 0.7 pu, 0.2 Q 0.4
pu, and P = 0.1, 0.3, 0.5, 0.7 and 0.9 pu at unity power factor. Normal conditions and one
line and two lines out-of-service are analyzed. All conditions are subject to the 330 kV busbar voltages lying in the range 0.95 to 1.05 pu. Note that the above selection of operating conditions encompasses not only a wide range of normal and outage conditions but also reactive
power outputs at 0.9 and 0.7 pu real power.
5.10.4.2

Steady-state and dynamic performance of the system, no PSS in service

The steady-state voltages and rotor angles, together with the associated eigenvalues, are calculated for the range of operating conditions using the system and generating-unit parameters listed above. The relevant results are given in Table 5.4 for selected operating
conditions.
Without a PSS at a generator real-power output of 0.9 pu (450 MW), the system is unstable
for all the selected operating conditions. For all real power outputs and a given system configuration (e.g. a single line outage) shown in Table 5.4 it is significant that:

the rotor angle increases as the power factor becomes more leading - and can exceed
90 ;

correspondingly, the stability of the system degrades.

Furthermore, at a selected complex power output (e.g. S = 0.9 + j0.3 pu), stability degrades
with the increased series impedance of the transmission system associated with the outages
of line sections a then a and b.
5.10.4.3

Inherent synchronizing and damping torques coefficients, 6th order generator model

As was the case for the third-order Heffron and Phillips model in Section 5.3, it is again revealing to examine the inherent synchronizing and damping torque coefficients for the rotor
before the tuning of the PSS is considered. In this case the generator terminal voltage is controlled by a first-order model of the excitation system.

Sec. 5.10

Tuning of a PSS for 6th order generator model

Table 5.4

263

Steady-state operating conditions and eigenvalues for the SMIB system; no

PSS in service (P, Q in pu on 500MVA base)

Generator Lines
Case Output,
out of
P, Q pu service

Voltage pu ( )
Load
Trans- Rotor
angle
pu
former
Gen.Ter- Gen.
Inf.
Po, Qo # taps %
minals HV Bus Bus

Eigenvalues,
Rotor Mode

0.9, -0.1

none

0.09, 0

80.7

1 17.6

1.024

1.020

0.773 j9.16

0.9, 0

none

0.09, 0.04

-1

76.0

1 17.7

1.019

1.002

0.552 j9.12

0.9, 0.2

none

0.09, 0.04

-4

67.6

1 17.6

1.021

0.959

0.261 j9.02

0.9, 0.4

none

0.09, 0.29

-6

60.8

1 17.6

1.015

0.967

0.113 j8.98

0.9, -0.1

one

0.09, 0

+5

87.7

1 24.5

0.978

1.017

0.936 j8.48

0.9, 0.3

one

0.09, 0.03

-8

68.2

1 21.7

1.053

0.952

0.245 j8.44

0.9, -0.07

two

0.09, 0.03

+6

91.7

1 30.1

0.965

1.049

0.927 j7.98

0.9, 0.3

two

0.09, 0.03

-7

74.0

1 27.6

1.041

0.951

0.322 j7.84

0.7, -0.2

none

0.09, 0.04

76.2

1 13.1

1.035

1.049

0.323 j9.06

0.7, 0

none

0.09, 0.04

-3

64.6

1 13.0

1.036

1.004

0.065 j8.92

0.7, 0.2

none

0.09, 0.04

-5

55.9

1 13.1

1.029

0.954

0.198 j8.76

0.7, -0.2

one

0.09, 0.03

+5

81.3

1 18.2

0.990

1.044

0.493 j8.41

0.7, 0.2

one

0.09, 0.03

-6

59.6

1 16.8

1.040

0.950

0.096 j8.19

# Load at 1 pu voltage on generator HV bus

The modified state equations for the calculations of the inherent torque coefficients are derived as outlined in Section 5.10.2. From these equations the frequency response of the synchronizing and damping torque coefficients are calculated, that is, from the transfer function
P 0 j f j f .
With reference to Figure 5.15 and Table 5.4 it is noted that the inherent damping torque coefficient is markedly negative for many of the cases over the range of the rotor modal frequencies, 7.8 to 9.2 rad/s. This observation applies at the more heavily stressed conditions,
particularly (i) as the series impedance of the transmission system increases with the line outages, and (ii) at the more leading power factors. The least damped condition is that for case
G with a coefficient of -13 pu, only slightly poorer than that for case A. Note that: (i) as the
damping torque coefficients become more negative for the cases shown in the figures so the
degree of instability increases for the corresponding cases in Table 5.4; (ii) at the lower interarea modal frequencies, potentially 1.5 to 6 rad/s, the inherent damping torque coefficients
tend to be negative. Such degradations in damping must be remedied by the action of the
PSS.

(a)

50
40
30
20
10
0
10
20
0
10

10

10

2
1.5
1
0.5
0
10

10
Frequency (rad/s)

A: 0.10
C: 0.20
G: 0.07

10

B: 0.0 All in
D: 0.4 All in
H: 0.3 Two out

Synch. Torq. Coeff. Damp. Torq. Coeff.

Tuning PSSs for SMIB system

Synch. Torq. Coeff. Damp. Torq. Coeff.

264

(b)

Ch. 5

60
50
40
30
20
10
0
10
0
10

10

10

2
1.5
1
0.5
0
10

10
Frequency (rad/s)

J: 0.20
L: 0.20
M: 0.20

10

K: 0.0 All in
All in
N: 0.2 One out

Figure 5.15 Inherent synchronizing and damping torque coefficients for SMIB system at
a generator real power output (a) P = 0.9 pu, and (b) P = 0.7 pu. Reactive power output varies between 0.2 pu leading and 0.4 pu lagging with all lines in service, or the outage of one
or two lines.
As observed in Section 5.5.1 the damping gain k of the PSS - when installed - should be
greater than k = 13 pu to ensure stability under the most onerous operating condition, i.e.
in Case G: P = 0.9, Q = -0.07 pu, with two lines out of service.
Note from Table 5.4 that the magnitudes of damping ratios of the rotor modes are generally
less than 0.1. The torque coefficients are therefore again calculated using conventional frequency response methods with s = j f .
5.10.5 The P-Vr characteristics for a SMIB system with a 6th order generator
model
The basis of the P-Vr characteristics and their calculation are discussed in Section 5.10.3.
The P-Vr characteristics are shown in Figure 5.16 for generator real power outputs of 0.9
and 0.7 pu. These characteristics mainly relate to the more extreme ends of the range of reactive power outputs and system configurations listed in Table 5.4.

Sec. 5.10

Tuning of a PSS for 6th order generator model

20
Magnitude (dB)

Magnitude dB)

20
10
0
10

10

10
0
10
20
0
10

10

50

50

Phase (deg)

Phase (deg)

20
0
10

100
150
200
0
10

(a)

265

10
Frequency (rad/s)

A: Q=0.1
C: Q= 0.2
G: Q=0.07

10

B: Q=0
All in
D: Q=0.4 All in
H: Q=0.3 Two out

10

10

100
150
200
0
10

(b)

10
Frequency (rad/s)

J: Q=0.2
L: Q= 0.2
M: Q=0.2

10

K: Q= 0

All in
All in
N: Q= 0.2 One out

Figure 5.16 P-Vr characteristics for the SMIB system at generator real power outputs
(a) P = 0.9 pu, and (b) P = 0.7 pu; all lines in service, or outage of one or two lines.
Note that, for constant real power outputs, the low-frequency gains of the P-Vr characteristics increase as the reactive power changes from lagging to leading.
As is the case for the lower-order generator model of Section 5.8, the P-Vr characteristics
in the latter figures form the basis for the tuning of the speed-PSS.
5.10.6 Tuning a speed-PSS for a SMIB system with a 6th order generator model
The same form kG s of the transfer function of the speed-PSS is employed as in
Section 5.8. The tuning of the PSS transfer function is considered in the same five steps.
Step 1 of Section 5.8.

The determination of the compensating transfer function G c s such that, over a selected
range of modal frequencies, a torque of electro-magnetic origin proportional to speed is induced by the PSS on the rotor of the generator.
5.10.6.1

Selection of the P-Vr characteristic for compensating transfer function G c s

For the selected set of operating conditions listed in Table 5.4 the frequencies of the rotor
modes cover a range from 7.8 to 9.2 rad/s. Based on a mid-range modal frequency of about

266

Tuning PSSs for SMIB system

Ch. 5

8.7 rad/s, the phase of the P-Vr characteristics for P = 0.9 pu varies over the range 53.5
(Case A) to 35.8 (Cases J & H). Let us choose the P-Vr characteristic for P = 0.9, Q = 0.2
pu (Case C) because its phase of 43.5 lies close to the middle of the band of characteristics
at the modal frequency. The phase of any characteristic associated with the operating conditions listed in Table 5.4 is thus within 10 of that of the selected characteristic. Case C
will be called the Design Case.
At low frequencies (1 rad/s), the extent of the range of gains of the P-Vr characteristics
shown in Figure 5.16(a) is from 7.8 dB (Cases H) to 14.6 dB (Case A). These differ from the
selected P-Vr characteristic (Case C) by -3.9 and +2.9 dB, respectively.
An inspection of the P-Vr characteristics for P = 0.7 pu in Figure 5.16(b) reveals that these
characteristics are suitably represented by Design Case C.
5.10.6.2

Synthesis of the compensating transfer function G c s

We wish to synthesize the P-Vr transfer function, H PVrS s , for Design Case C in order to
form the PSS compensating transfer function G c s by means of (5.23). In order to provide
some useful insights the process is illustrated using some basic concepts in frequency response analysis.
Imagine, on the P-Vr characteristic of Case C in Figure 5.16(a), the straight-line asymptotes
as f 0 and as f are drawn on the magnitude plot for the Design Case C. These
asymptotes roll off at 0 and -40dB/decade, respectively, and intersect at 19 rad/s. Because
the slope of the magnitude plot monotonically decreases with increasing frequency it is assumed that there are no zeros in the transfer function (or, if any exist, they are cancelled by
closely-located poles). Furthermore, the phase response tends to 180 as f . A second-order form of the synthesized transfer function H PVrS s possessing a pair of complex
poles is thus assumed, i.e.
2

H PVrS s = k co s + 2 n s + n

2s
s
or H PVrS s = k c 1 + --------- + ------ ,
n
n

(5.45)

where k c = k co n . The frequency response characteristics for such a transfer function are
shown in Figure 2.16. From a comparison of the P-Vr response for Case C in Figure 5.16(a)
with that of the latter figure it is observed that:

The intersection of the asymptotes occurs at the undamped natural frequency

n = 19 rad/s; the associated magnitude response in Figure 2.16 is 6 dB down at this
frequency. Assume that the damping ratio = 1; the corresponding gain from the PVr plot is about 4 dB down. The damping ratio is thus less than 1.0 and therefore the
transfer function possesses a pair of complex poles.

Sec. 5.10

Tuning of a PSS for 6th order generator model

267

There is no resonance apparent in the magnitude response, i.e. the damping ratio is
greater than 0.707.

A trial-and-error method, a curve fitting process, or some other convenient method can now
be adopted to determine the damping ratio . From the eigenvalues of the A matrix for the
P-Vr characteristic in (5.45) it may be possible for this simple SMIB system to isolate a pair
of poles for the second-order transfer function of (5.45). These are found to be located at
16 j10.2 - from which = 0.85 and n = 19 rad/s. The low-frequency gain of the PVr characteristic for Case C in Figure 5.16(a) is 11.7 dB or 3.84 pu. Inserting the relevant
values into (5.45), the synthesized P-Vr transfer function for Case C becomes:
2

H PVrS s = 3.84 1 + 0.0895s + 0.00277s .

(5.46)

Thus, based on (5.23), Step 1 for the determination of the compensating transfer function
G c s can be completed, i.e.
2

G c s = 1 H PVrS s = 1 + 0.0895s + 0.00277s 3.84 .

(5.47)

Steps 2, 3 and 4 of Section 5.8:

These steps cover the selection of: the value of the damping gain k; the parameters of the
PSS washout filter; and the high-pass filter parameters that ensure G c s is proper. It is assumed that attenuation of the torsional modes is not required.

As stated earlier a practical form of the speed-PSS for the SMIB system includes the damping gain k, together with the washout and low-pass filters, as explained in Section 5.9. Incorporating the PSS compensating transfer function 1 H PVr s from (5.45), the practical PSS
transfer function described in (5.34) becomes:
2s
s 2
1 + --------- + -----n
n
sT W
1
H PSS s = kG s = k ------------------- ----- ---------------------------------------------- .
1 + sT W k c 1 + sT 1 1 + sT 2

(5.48)

For the range of modal frequencies of rotor oscillation shown in Table 5.4 (i.e. 7.8 - 9.2 rad/
s) an examination of Figure 5.15 for the damping torque coefficients reveals that a damping
gain of about 13 pu is required to overcome the inherent negative damping. A value of
k = 20 pu is adopted on a trial basis. The same parameters as in Section 5.9 are adopted for
the washout and low-pass filters because the range of modal frequencies is essentially the
same. With the insertion of the latter parameters together with those from (5.47) into (5.48)
the transfer function of the speed-PSS for the SMIB system becomes:
s5
1
1 + s0.0895 + s 2 0.00277
H PSS s = 20 --------------- ---------- ------------------------------------------------------------- .
1 + s5 3.84 1 + s0.005 1 + s0.005

(5.49)

268

Tuning PSSs for SMIB system

Ch. 5

Note that (i) the zeros in the transfer function of the compensation are complex and (ii) the
low-pass filter time constants of 5 ms are very short. These matters are considered in more
detail in Section 5.12.
Prior to considering Step 5 in the tuning procedure of Section 5.8 (i.e. the assessment if the
dynamic performance of the PSS) it is instructive to view the frequency characteristic of the
damping torque induced by the PSS. Step 5 is discussed in Section 5.11.
5.10.6.3

Damping and synchronizing torque coefficients induced by the PSS

Let us assess the coefficients of the synchronizing and damping torques induced on the generator through the action of the PSS. Not only does this serve as a check and partial validation of the analysis and tuning of the PSS represented by (5.49), but it also provides
additional insights into the action of the PSS.
Referring to Figure 5.9 on page 241, the procedure for examining the synchronizing and
damping torques developed through the action of the PSS involves disabling the shaft dynamics of the machine, injecting a speed perturbation into the PSS+excitation-system+machine loop only, and calculating the complex torque P a , i.e.
P a

P m = 0

= H PVr s H PSS s .

(5.50)

Note that the inherent torque coefficients associated with the path P are excluded
in formulation of (5.50).
The product H PVr s H PSS s in (5.50) is by definition a complex torque coefficient, s .
This coefficient can be expressed in terms of the PSS compensating transfer function G c s
and the washout and low-pass filters ( G W s and G LP s ) by:
s = H PVr s G s k = H PVr s G c s G W s G LP s k .

(5.51)

For the operating condition which forms PSS Design Case C, the compensating transfer
function of the PSS is ideally given by (5.23), i.e. G c s = 1 H PVr s , thus (5.51) becomes
s = G W s G LP s k .

(5.52)

Assuming the transfer functions of the washout and low-pass filters are each real numbers,
i.e. 1 0 , at the modal frequency, equation (5.52) represents the already-stated purpose of
the speed-PSS, i.e. it should induce a damping torque coefficient equal to k as outlined in
Section 5.8.1.
For operating conditions other than the Design Case, (5.51) becomes:
p s = H PVrp s G co s G W s G LP s k ,

(5.53)

where the subscripts p refer to the P-Vr transfer function which changes with operating
condition; subscripts o refer to the compensating transfer function which is implemented

Sec. 5.10

Tuning of a PSS for 6th order generator model

269

in the PSS. At the modal frequency s = h j h the transfer functions of the washout and
low-pass filters are close to 1 0 , thus (5.53) becomes
Kp h
p j h = ------------------ k p h o h ,
Ko h

(5.54)

where Kp, p and Ko, o are defined following (5.41).

As shown in Section 5.8.3 and Section 5.10.6.1, at the modal frequency h the P-Vr phase
responses typically lie in a narrow band of some 5 10 about that of the Design Case.
The torque coefficient in (5.54) is essentially a real number, i.e. a damping torque coefficient,
given by
Kp h
p h = ------------------ k .
Ko h

(5.55)

The significance of the above result is that the damping torque coefficient induced by the
PSS at any operating condition and the modal frequency h is the damping gain k modified
by the ratio of the P-Vr gain at the operating condition to that of the Design Case.
Firstly, consider now the calculation of the torque coefficients induced through the PSS path
only in Figure 5.9 on page 241. The complex coefficient s is expressed in (5.51) in terms
of the synthesized and PSS transfer functions H PVrS s and H PSS s , given by (5.46) and
(5.49), respectively. For the SMIB system the frequency responses of the synchronizing and
damping components are derived from s using the result from (5.10) as employed in
Section 5.10.2.
In Figure 5.17 are shown the synchronizing and damping torque coefficients induced by the
PSS for two operating conditions, the Design Case C (P = 0.9, Q = 0.2 pu), and the most
poorly-damped condition Case G (P = 0.9, Q = -0.07 pu, two lines out-of service). The details of the steady-state operating conditions for these cases are given in Table 5.4.
Two important features of the responses of the damping torque coefficients are seen in the
figure. Firstly, they are more-or-less flat over the range of modal frequencies of interest, 7
to 9 rad/s. Secondly, the damping torque coefficient of 20 pu required by the design is
achieved in the Design Case C, however, for Case G a value slightly less results. In
Figure 5.16(a) the respective P-Vr gains at a modal frequency in the vicinity of 7 to 9 rad/s
are 3.7 and 3.3 pu. According to (5.55) the damping torque coefficient induced by the PSS
for Case G is 3.3 3.7 20 = 17.8 , a value which closely agrees with that found from
Figure 5.17.

Tuning PSSs for SMIB system

Synchronizing Torque Coeff. Damping Torque Coeff.

270

Ch. 5

25
20
15
10
5
0
5
1
10

10

10

10

10

10
8
6
4
2
0
2
1
10

10

10
10
Frequency (rad/s)

C: Q=0.2
G: Q=0.07

10

Design Case
Two lines out

Figure 5.17 Coefficients of the per unit generator synchronizing and damping torques induced through the PSS path (i.e. excludes the inherent torques); P = 0.9 pu.
Secondly, consider inherent torque coefficients that are induced through the rotor angle
path P in Figure 5.9 on page 241. The frequency response P j f j f for
Cases C and G are shown in Figure 5.18 and possess some interesting features. Firstly, the
magnitude responses roll off fairly consistently at 20 dB/decade and the phase responses
vary about an angle of 90 . This is consistent with integration in the transfer function
= o j f (as revealed in (5.3) and Figure 5.9). Secondly, the damping torque coefficient is related to the cosine of the phase angle and is therefore positive when the phase
angles are more positive than 90 .
However, as foreshadowed in Section 5.5.1, the positive contribution of the PSS to the generator torque coefficients will be reduced if the inherent torque coefficients are negative. As
established in Section 5.10.4.3 and Figure 5.15 the inherent damping torque coefficients are
mainly negative at the modal frequencies listed in Table 5.4 on page 263; it is therefore necessary to examine the net effect on the torque coefficients of the PSS and rotor-angle paths.

Sec. 5.10

Tuning of a PSS for 6th order generator model

271

Magnitude (dB)

80
60
40
20
0
20
1
10

10

10

10

10

80
Phase (deg)

85
90
95
100
105
110
115
1
10

10
10
10
Frequency (rad/s)
Case C

10

Case G

Figure 5.18 Frequency response of the inherent torque coefficients for Case C (the PSS
Design Case); P = 0.9, Q = 0.2 pu, and Case G (Outage of two lines); P = 0.9, Q = -0.07pu.
In Figure 5.19(a) and (b) the frequency responses of the inherent and PSS-induced torques
coefficients are shown separately and summed together. Figure 5.19(a) applies to the Design
Case C and Figure 5.19(b) to the worst case operating condition, Case G. The responses of
the PSS-induced torques coefficients in the two figures are the same as those in Figure 5.17.
However, due to the very different nature of the inherent torque coefficients in the two cases, the combined responses are markedly different. While the combined responses are
somewhat flat over the modal frequency range 7 to 9 rad/s, the values of the sum of the
damping torque coefficients are significantly different, i.e. about 17 and 5 pu in
Figure 5.19(a) and (b), respectively. Although these coefficients are both positive, the eigenvalue of the rotor mode for the Design Case will be characterized by a significantly greater
shift into the left-half s-plane than that for Case G. This is to be examined in Section 5.11
in assessing the performance of the PSS over the range of operating conditions.
Due to the inclusion of the damper windings to represent rotor eddy-current losses in the
model of the sixth-order generator the frequency response characteristics of the inherent
damping torque coefficients are seen in Figure 5.19 to vary greatly over the selected range
of operating conditions. In Case C at low frequencies (less than 1 rad/s) the damping torque
coefficient is significant, being of the order of 50 to 60 pu. For case G, on the other hand,
the damping torque coefficient is negative and destabilizing.

272

Tuning PSSs for SMIB system

60
40
20
0
20
1
10

10

10

10

8
6
4
2
0
1

10

(a)

10

10
10
10
Frequency (rad/s)

Inher. & PSSinduced torq. coeffs

Torq. coeffs induced by PSS only
Inherent torque coefficient only

Damping Torque Coeff.

20

15
10
5
0
5
10
15
1
10
7

Synch. Torque Coeff.

Synch. Torque Coeff.

Damping Torque Coeff.

80

Ch. 5

10

(b)

10

10

10

10

6
5
4
3
2
1
0
1
1
10

10
10
10
Frequency (rad/s)

10

Inher. & PSSinduced torq. coeffs.

Torq. coeffs induced by PSS only
Inherent torque coefficient only

Figure 5.19 Components of the torque coefficients induced on the generator for the operating condition (a) Case C (the PSS Design Case); P = 0.9, Q = 0.2 pu,
(b) a worst case operating condition, Case G (outage, two lines); P = 0.9, Q = - 0.07 pu.
As an aside, it is of interest to ascertain what are the components of, or the contributions to,
the inherent torque coefficients by individual rotor windings given a set of generator and system parameters and selected operating conditions. With shaft dynamics disabled the inherent torque and its components are calculated from (5.77) in Appendix 5I.3.2, i.e.
P 0 = K 1 + K 2 fd + K 21 d1 + K 22 q1 + K 23 q2 ,
the rotor angle and flux states being given in terms of the input speed signal by (5.76). The
coefficient K1 is a synchronizing torque coefficient whereas K2, K21, K22 and K23 are torque
coefficients due to flux linkages associated with the field, direct-axis and the two quadratureaxis windings. The inherent synchronizing and damping torque coefficients are shown in
Figure 5.20 for two operating conditions with all lines in service, P = 0.9 pu and (a) Q = 0.2,
and (b) Q = -0.07 pu (outage, two lines); these conditions are the same as in Figure 5.19.
In the leading power factor operating condition (Q = -0.07 pu, Figure 5.20(b)) the contributions to the damping torque coefficients by the q-axis windings are negligible, however, in
Figure 5.20(a) for the lagging power case (Q = 0.2 pu) they are significant. The field and daxis windings tend to be the dominant contributors to the damping torque coefficients.
While it is possible to assign to each of the damping windings a contribution to the torque
coefficients, concerns over the validity of the results are likely to arise due to questions re-

Sec. 5.11

Performance of the PSS for a 6th-order generator model

273

(a)

60
40
20
0
1

10

10

10

10

2
1
0
1
1
10

10
10
Frequency (rad/s)

Total
Field
qaxis 1

10

Rotor Angle
daxis
qaxis 2

Sync. Torq. Coeff. Damp. Torq. Coeff.

Sync. Torq. Coeff. Damp. Torq. Coeff.

lating to the accuracy of both the model and the parameters values attributed to the windings. The frequency dependence of the damping contribution of the damping windings
emphasizes the importance of employing higher-order generator models to adequately represent the damping performance of power systems.

(b)

5
0
5
10
15
1
10

10

10

10

2
1
0
1
1
10

10
10
Frequency (rad/s)

Total
Field
qaxis 1

10

Rotor Angle
daxis
qaxis 2

Figure 5.20 Components of the inherent synchronizing and damping torque coefficients
for SMIB system for generator real power output (a) P = 0.9, Q = 0.2 pu, all lines in service,
and (b) P = 0.9, Q = -0.07 pu, outage of two lines.

5.11

Performance of the PSS for a higher-order generator model

Step 5 in the tuning procedure outlined in Section 5.8 is the confirmation of the validity of
the design, and assessment of the performance of the PSS.

The improvements in the damping of the rotor modes of oscillation due to the operation of
the PSS are revealed in Table 5.5 and Figure 5.21.

274

Tuning PSSs for SMIB system

Table 5.5

Case

Ch. 5

Effect on rotor modes with PSS in service; 6th order generator model.
PSS damping gain k = 20 pu on machine rating.
Generator
Output
P, Q pu

Lines
out of
service

Eigenvalues, rotor mode

with PSS out of
service

with PSS
in service

Mode shift

0.9, -0.1

none

0.773 j9.16

1.156 j9.51

1.929 j0.35

0.9, 0

none

0.552 j9.12

1.271 j9.31

1.823 j0.20

0.9, 0.2

none

0.261 j9.02

1.338 j9.03

1.599 j0.02

0.9, 0.4

none

0.113 j8.98

1.305 j8.93

j0.05
1.418

0.9, -0.1

one

0.936 j8.48

0.632 j8.64

1.568 j0.16

0.9, 0.3

one

0.245 j8.44

1.042 j8.34

1.687
j0.10

0.9, -0.07

two

0.927 j7.98

0.409 j8.04

1.336 j0.06

0.9, 0.3

two

0.322 j7.84

0.774 j7.72

1.096
j0.11

0.7, -0.2

none

0.323 j9.06

1.727 j9.46

2.050 j0.40

0.7, 0

none

0.065 j8.92

1.770 j8.94

1.705 j0.02

0.7, 0.2

none

0.198 j8.76

1.579 j8.66

1.381
j0.10

0.7, -0.2

one

0.493 j8.41

1.171 j8.57

1.664 j0.15

0.7, 0.2

one

1.240 j8.03

1.144
j0.15

0.096 j8.18

It is informative to compare these results for the 6th order generator model with those in
Figure 5.13 on page 252 and Table 5.2 for the 3rd order model. The left-shift of the mode
in Table 5.5 for Design Case C is 1.60 Np/s, which agrees well both with that the Design
Case B in Table 5.2 and with the predicted left-shift k d 4H = 1.67 Np/s for a SMIB system in Section 5.9.3. As expected, the shift in the associated frequency ( h = 0.02 rad/s)
in Design Case C is minimal for the reasons discussed in the latter section.
The same explanations that are given in Section 5.9.3 for the nature and magnitude for the
mode shifts are also applicable to the 6th order generator in a SMIB system. In the results
for the latter, however, there are some additional issues to consider; these are:

At a lower real power output, i.e. P = 0.7 pu, and a selected reactive power output, the
rotor mode when the PSS is either in- or out-of-service is better damped than for
P = 0.9 pu, e.g. comparing Cases B and K in Table 5.5 for all lines in service. This is to
be expected as the corresponding steady-state rotor angle is smaller (see Table 5.4 on
page 263).

Sec. 5.11

Performance of the PSS for a 6th-order generator model

275

10

Imaginary Part (rad/s)

All lines in; AD

A

9.5
B
9
8.5

C
D
One line out;
EF

D C
E

F
PSS On

PSS Off
G
H

8
H
7.5

B A

Two lines out; GH

P=0.9 pu

7
2.5

1.5

0.5

0.5

1.5

1.5

Imaginary Part (rad/s)

10
9.5

All lines in; JL

J
J
PSS Off

9
PSS On
L
8.5

L
M

M
N

N
One line out; MN

7.5
P=0.7 pu
7
2.5

1.5

1
0.5
0
Real Part (Np/s)

0.5

A P=0.9 Q=0.1, B Q=0.0, C Q=0.2, D Q=0.4

E P=0.9 Q=0.1, F Q=0.3; G P=0.9 Q=0.1, H Q=0.3
J P=0.7 Q=0.2, L Q=0.2; M P=0.7 Q=0.2, N Q=0.2

Figure 5.21 Rotor mode for the PSS on and off for the set of operating conditions given
in Table 5.5, Cases A to N. The Design Case is Case C. The upper plot is for P = 0.9 pu, the
lower for P = 0.7 pu.

The modal frequencies decrease as one and then two lines are taken out of service with the result that the series impedance of the transmission system increases. This
matter was observed in Section 5.4 based on a simple analysis of the factors that
determine the real and imaginary parts of the rotor mode.

It is observed in Figures 5.13 and 5.21 that, with PSSs off, the most poorly damped
rotor mode is associated with the most leading power-factor condition. However, a
characteristic of the P-Vr tuning approach is that the gain in the P-Vr characteristics is
greatest for this condition. As illustrated in Figure 5.21 the net result with the PSS in

276

Tuning PSSs for SMIB system

Ch. 5

service is that the left-shift in the rotor mode is greatest for the leading power factor
condition - a desirable and beneficial outcome.
The discussion so far has considered real power outputs at or near rated values. The question arises: how does the PSS perform at lower values of real power outputs? To illustrate
the answer, the P-Vr characteristics and the rotor modes of oscillation are also calculated at
real power outputs at P = 0.9, 0.7, 0.5, 0.3 and 0.1 pu, all at unity power factor. The P-Vr
characteristics are shown in Figure 5.22 together with that for the Design Case C (P = 0.9,
Q = 0.2 pu). The associated rotor modes with and without the PSS in service are listed in
Table 5.6.
Table 5.6

Case

Shifts in rotor modes at lower real power outputs with PSS in service.
PSS damping gain k = 20 pu on machine rating.
Generator
Output,
pu
P, Q pu

Eigenvalues, rotor mode

Lines
out of
service

with PSS out of

service

with PSS
in service

Mode shift

0.9, 0

none

0.552 j9.12

1.271 j9.31

1.823 j0.20

0.7, 0

none

0.065 j8.92

1.770 j8.94

1.705 j0.02

0.5, 0

none

0.539 j8.50

1.874 j8.36

1.335
j0.14

0.3, 0

none

0.779 j8.04

1.610 j7.89

j0.15
0.831

0.1, 0

none

0.816 j7.77

1.121 j7.70

0.305
j0.07

1.338 j9.03

1.599 j0.02

C (Ref) 0.9, 0.2

none

0.261 j9.02

Based on Table 5.6 let us consider the performance of the system with the PSS out-of-service as the real power output is decreased at unity power factor. The rotor mode is stable
for real power outputs of 0.7 pu or less. With further reduction in real power the damping
improves; for power outputs less than 0.5 pu the damping performance criterion stated in
Section 5.8.5 is satisfied, namely, that the real part of the rotor mode should be more negative than 0.5 Np/s. In fact, at the lower power levels in this example, PSS action may not
be required - but this depends on the encompassing range of operating conditions.
For real power outputs between 0.9 and 0.5 pu, Figure 5.22 reveals that the magnitude and
phase responses in the P-Vr characteristic lie in a relatively narrow band, the centre of which
is the design characteristic, Case C. The shifts in the rotor modes due to the action of the
PSS are listed in Table 5.6. The explanation for the extent of the mode shifts in the power
range is similar to that given in Section 5.9.3. For real power outputs less than 0.5 pu the left
shift of the mode reduces significantly with reduction in power output. This is due to the
gain of the associated P-Vr characteristics reducing significantly below that of the Design
Case characteristic 1. Nevertheless, the damping of the rotor mode for lower power levels

Sec. 5.11

Performance of the PSS for a 6th-order generator model

277

below 0.5 down to 0.1 pu is markedly enhanced. Typically, in practice, the PSS may be
switched out of service for real power levels less than ~0.3 pu.

Magnitude (dB)

20
10
0
10
20
0
10

10

10

Phase (deg)

0
50
100
150
200
0
10

10
Frequency (rad/s)
B: P=0.9
R: P=0.5
T: P=0.1

10

K: P=0.7
S: P=0.3
C: Reference

Figure 5.22 P-Vr characteristics for the SMIB system for real power outputs from 0.1 to
0.9 pu at unity power factor.
While the effect of the variation of the real power output on the P-Vr characteristic is noted,
information concerning the effect of reactive power output on the characteristic may be deduced from Figure 5.16 on page 265. From the figure for the system in this example and
loading conditions (P = 0.9 & 0.7 pu) it can be deduced that, at constant real power output,
(i) the magnitude responses of the P-Vr transfer function typically lie within a band of 2
dB about the Design Case C, (ii) the phase responses lie within a band of 10 over the
range of modal frequency. It may also be observed that the magnitude responses at constant
real power decrease as the power factors change from maximum leading to maximum lagging. The two effects on the P-Vr characteristics of the variations in real and reactive outputs
are explained in the context of multi-machine systems in Chapters 9 and 10. It is interesting
to examine the relation between the P-Vr characteristics of Figure 5.16 and the plots in
Figure 5.21.

1. The reason for the nature of the P-Vr gain variation is discussed in Section 9.4.1.

278

Tuning PSSs for SMIB system

Ch. 5

Due to the basis of selection of the synthesized P-Vr for Design Case the PSS transfer function is robust - for practical purposes - to changes in operating conditions at the higher values of generator real power outputs, together with variations in reactive power, was pointed
out in Section 5.9.3 1. The results of further studies as revealed in Figures 5.16 and 5.22, also
demonstrate the robustness of the PSS design 2 to wide variations in operating conditions.
The two effects on the P-Vr characteristics of the variations in real and reactive outputs are
explained in the context of multi-machine systems in Section 9.4.1; the robustness of fixedparameter PSSs and an associated theoretical basis for robustness are discussed further.

5.12 Alternative form of PSS compensation transfer function

In Section 5.10.6.2 the compensation transfer function for the PSS included a pair of complex zeros to represent the inverse of the P-Vr transfer function. Referring to (5.49) and ignoring the gain kc = 3.84 for present purposes, the relevant transfer function is
1 + s0.0895 + s 2 0.00277
G c s = ------------------------------------------------------------- .
1 + s0.005 1 + s0.005

(5.56)

The frequency response plot for this PSS transfer function is shown in Figure 5.23.
A practical PSS may not be capable of accepting complex zeros, moreover, time constants
of 5 ms may be too short to implement in either a digital or analog PSS. Alternatively, let us
assume the transfer function may be represented by a set of lead blocks of the form:
k

Gc s =

1 + sT

nk
- , in which the zeros are real and T nk T dk .
----------------------- 1 + sT

i=1

(5.57)

dk

The alternative transfer function in (5.57) is required not only to increase the values of the
time constants Tdk in the denominator of the transfer function in (5.56) but also to provide
the phase lead determined from the P-Vr characteristic over the range of modal frequencies,
1.5 - 15 rad/s. A third-order lead transfer function is found which closely matches the frequency response over the desired modal frequency range as shown in Figure 5.23:
1 + 0.0438s 3
1 + s 22.8 3
G c s = ---------------------------- -------------------------- .
1 + 0.0158s
1 + s 63.3

(5.58)

1. For operating conditions, P = 0.9 and 0.7 pu, the magnitude of the P-Vr transfer functions consistently lie within a band of 2 dB, that is, by factors of 1.26 and 0.79 (see
Figure 5.16). In the case of the SMIB system this provides confidence that for practical
purposes the left-shifts of the rotor mode over the encompassing range of operating conditions will lie in the range 1.26:1 and 0.79:1.
2. See item 3 of Section 1.2

Sec. 5.13

Tuning an electric power PSS

279

Magnitude (dB)

50
40
30
20
Range of modal
frequencies

10
0
1
10

10

10

10

10

Phase (deg)

120
100
80
60
40
20
0
1
10

10

10
10
Frequency (rad/s)

Complex zeros

10

Lead blocks

Figure 5.23 Frequency response of the PSS second-order transfer function

when its zeros are complex or, alternatively, when the transfer function is
implemented by three lead blocks.
The maximum deviation of the latter frequency response from that of that derived from
(5.56) is [0.8 dB, 1.3 ] at 10 rad/s and [1.1 dB, 5.9 ] at 15 rad/s. Although the denominator
time constants are increased from 5 ms to 16 ms, a further increase in these time constant
may impact deleteriously on the desired frequency response of the compensation in the
range of modal frequencies. It is also noted in Figure 5.23 that the gain at high frequencies
in the alternative compensation is reduced for torsional frequencies above 100 rad/s (~16
Hz). The concern that the high frequency gain in the PSS might excite such modes is overcome in the integral-of-accelerating-power PSS by means of a pre-filter, a purpose of which
is to significantly attenuate the torsional modes (see Section 8.5).

5.13

Tuning an electric power-PSS based on the P-Vr approach

In [14] the authors highlight an electric-power based PSS that does not require the implementation of a phase-lead transfer-function network. The power-based PSS may have some
attraction (i) in cases where only the damping of local modes is of concern; (ii) in a PSS for
which the number of blocks may be restricted 1. It is sometimes the case too that a unit is
provided with a power input PSS even in situations where this would not have been the preferred choice. The P-Vr approach of Sections 5.8.1 to 5.8.5 can also be applied to the tuning
1. An alternative approach to tuning a power-based PSS is described in Section 8.3.

280

Tuning PSSs for SMIB system

Ch. 5

of this type of PSS with the advantages revealed earlier, namely that the procedure is systematic and the value of the PSS gain is a meaningful quantity - it is also the damping gain defined in Section 5.4.
In the following the transfer function of the power-PSS is derived and assumes that the
number of blocks in the PSS is restricted. The analysis commences with equation (5.21)
which is employed in development of the P-Vr-based speed-PSS in Section 5.8.1. The latter
equation is based on an ideal PSS developing a damping torque proportional to rotor speed
and takes the form:
P 2
P 2
V s
---------= k = --------- --------- pu,
V = 0 sdd
V s

sdd
r

(5.59)

assuming the shaft dynamics are disabled (sdd) on all generators; k is a damping torque
coefficient. (It is seen from Figure 5.9 that P 2 V r and P 2 V s are the same transfer
functions.)
V

is identified as the P-Vr characteristic H PVr s and ---------s the transfer

The term ---------2V s

sdd

function of the speed-PSS; k is also the damping gain, G c s is the compensation transfer function of the speed-PSS. That is:
P 2

V r = 0 sdd

= H PVr s k G c s

((5.21) repeated)

(5.60)

Based (4.59) the equation of rotor motion derived for the case when the perturbations in
mechanical power are zero is
1
2Hs = P e = P 2 pu or = ---------- P 2 .
2Hs

(5.61)

The substitution for from (5.61) in (5.59) yields:

P 2
P 2
V s
---------= 2Hs --------- 1 ---------- .
V = 0 sdd
V s
P 2
sdd
r
P

In the above equation the term ---------2V s

may be identified as H PVr s , the P-Vr transfer

sdd

function, and ---------s- identified as the transfer function of the power-PSS.

P 2

(5.62)

Sec. 5.13

Tuning an electric power PSS

281
P

Strictly, over the range of modal frequencies of interest we require ---------2-

V = 0 sdd
r

= k .

Furthermore, in order to attenuate low and high frequencies responses outside the range,
washout and the low-pass filters are added, i.e.
P 2
sT 1
sT 2
1
---------= k ------------------ ------------------ --------------------- .

V = 0 sdd
1 + sT 1
1 + sT 2 1 + sT LP
r

(5.63)

By equating (5.62) and (5.63), the transfer function of the power PSS is determined, namely:
V s
sT 1
sT 2
P 2

1
PSS power = ---------- = k ------------------ ------------------ --------------------- 1 2Hs --------- . (5.64)

P 2
1 + sT 1
1 + sT 2
1 + sT LP
V s

sdd
P

For the Design Case (e.g. Case C in Section 5.10.6) the P-Vr transfer function ---------2V s

is
sdd

calculated as in Sections 5.8.1 to 5.8.3. The extended P-Vr transfer function,

P 2
2Hs ---------V s

, is then derived by curve-fitting. Let us assume that synthesized transfer funcsdd

tion is of the simple form As 1 + Bs . The resulting transfer function of the fixed-parameter power-PSS can be then be formed, i.e.:
T1 A
sT 2
1 + sB
PSS power s = k ------------------ ------------------ --------------------- .
1 + sT 2 1 + sT 1 1 + sT LP

(5.65)

Note that, because of negation in the power-PSS path of (5.64), the signal V s is fed into
the summing junction with a positive sign 1.
5.13.1 Example 5.6: Tuning of a power-based PSS
Case C (the Design Case), and the more extreme cases, Cases A, G and H in the SMIB system of Section 5.10.5 and Table 5.4 are used to demonstrate the tuning of a power-PSS.
The P-Vr characteristic for the Design Case C, shown in Figure 5.24(a), is combined with
P

the transfer function 2Hs to form the extended P-Vr, XPVr(s), 2Hs ---------2V s

; the P-Vr
sdd

and the XPVr are also shown in Figure 5.24(a). The synthesized transfer function of
XPVr(s) - which results from curve fitting - is shown in the same figure and found to be
XPVr(s) = As 1 + Bs = 27.1s 1 + 0.108s .

1. A similar phenomenon occurs with the speed-PSS. See Section 5.8.1.

(5.66)

282

Tuning PSSs for SMIB system

Ch. 5

Assuming the damping gain is k = 20 pu on generator rating, and that the time constants
of the washout and low-pass filters are 5 s and 0.01 s respectively, the transfer function of
the power-PSS based on Design Case C is:
s5
5 27.1
1 + s0.108
PSS power = 20 --------------- ----------------- ------------------------- .
1 + s5 1 + s5 1 + s0.01

(5.67)

For comparison, the transfer-function of speed-PSS derived in Section 5.10.6.2 is

s5
1
1 + s0.0895 + s 2 0.00277
H PSS s = 20 --------------- ---------- ------------------------------------------------------------- . ((5.49) repeated)
1 + s5 3.84 1 + s0.005 1 + s0.005

(5.68)

40
20
0
20
1
10

10

10

10

Phase (deg)

100
0
100
200
1
10

10
10
Frequency (rad/s)
Extended PVr: XPVr
Curve fitted XPVr
PVr

(a)

10

Sync. Torq. Coeff. Damp. Torq. Coeff.

Magnitude (dB)

The eigenvalues for the four cases with the speed- and power-PSSs out- and in-service are
calculated; these are listed in Table 5.7.

30
25
20
15
10
5
0
5
1
10

10

10

10

0.5

0
1

10

10
10
Frequency (rad/s)

10

A: Q=0.1
C: Q=0.2
G: Q=0.07
H: Q=0.3
SpeedPSS, Case C: Q=0.2

(b)

Figure 5.24
(a) Case C: P-Vr, the extended P-Vr (XPVr), and the synthesized characteristic of XPVr(s)
(b) Synchronizing and damping torque coefficients induced by the power-PSS.
Cases A and C: All lines in service; Cases G and H: Two lines out. For comparison the
torque coefficients of the speed-PSS for Case C from Figure 5.17 are also plotted.

Sec. 5.13

Tuning an electric power PSS

Table 5.7

Case

283

Rotor modes with speed- and power-PSS in service; 6th order generator
model. PSS damping gain k = 20 pu on machine rating.,
Generator
Output
P, Q pu

Lines
out of
service

Eigenvalues, rotor mode

with PSS out of
service

with speed-PSS
in service

with power-PSS
in service

0.9, -0.1

none

0.773 j9.16

1.156 j9.51

1.443 j9.53

0.9, 0.2

none

0.261 j9.02

1.338 j9.03

1.524 j8.95

0.9, -0.07

two

0.927 j7.98

0.409 j8.04

0.465 j7.98

0.9, 0.3

two

0.322 j7.84

0.774 j7.72

0.809 j7.65

Note in Table 5.7 that the rotor mode shifts for both types of PSS are directly to the left.
However, it is observed that the left-shifts due to the power-PSS are greater than those of
the speed-PSS. The differences arise due to (i) mismatches between the XPVr characteristic
based on Case C and the fitted transfer-function XPVR(s), (ii) differences between the P-Vr
characteristics of Design Case C and the other (more extreme) cases as seen in Figure 5.16.
In the case of the speed-PSS the induced damping torque coefficients in the PSS path,
shown in Figure 5.17 for Cases C and G, are 20 and 18 pu over a modal frequency range 2
to 15 rad/s. The question arises: what are the corresponding characteristics of the damping
torque coefficients for the power-PSS?
The characteristics of the synchronizing and damping torque coefficients for the power-PSS
are illustrated in Figure 5.24(b) for the four cases when the damping gain is 20 pu. At the
rotor modal frequencies of 7.5 to 9.5 rad/s the damping torque coefficients range from 14
to 26 pu. The reasons for the differences between the damping torque coefficients for Case
C and the other cases are as explained above for the greater rotor-mode shifts by the powerPSS. The torque coefficients for a speed-PSS based on Case C are shown in Figure 5.24(b)
for comparison.
However, in a multi-machine scenario with inter-area modal frequencies above 2 rad/s,
damping torque coefficients exceeding 11 pu are potentially induced by the power-PSS for
the worst-case operating condition (H). Adjusting the damping gain setting from 20 pu will
raise or lower the damping torque coefficients and likewise modify the left-shifts of the rotor
mode.
Note that: (i) In synthesizing the power-PSS difficulties may occur in finding a simple curvefitted transfer function for the XPVr due to the shape of P-Vr and the modal frequencies of
interest. (ii) As will be discussed in Chapter 8, power-PSSs may cause terminal voltage and
reactive power swings due to changes and ramping of mechanical power [15]. Reducing the
washout time constant from 5 s in the latter example may alleviate this problem.

284

Tuning PSSs for SMIB system

Ch. 5

5.14 Summary: P-Vr approach to the tuning of a fixed-parameter PSS

The P-Vr approach provides a systematic and consistent method for tuning PSSs for a selected set of encompassing operating conditions. It possesses the following features.
1.

The aim of the PSS tuning procedure is to introduce on the generator shaft a damping torque (a torque proportional to machine speed); this causes the modes of rotor
oscillation to be shifted to the left in the complex s-plane.

2.

The PSS compensation transfer function G c s is tuned to achieve a direct left-shift

in the complex s-plane of the relevant modes of rotor oscillation.

3.

The damping gain k (pu on machine rating) of the PSS determines the extent of the
left-shift.

4.

The damping torque coefficient contributions induced by the PSS can be designed
to be constant (flat) over a desired range of local- and inter-area modal frequencies
(e.g. as in Figure 5.17). The damping gain should be selected to ensure that the
damping torque coefficients swamp any inherent negative contributions by the generator over the range of operating conditions.

5.

If the above features are realized the PSS transfer function kG c s is said to be
robust (see item 3, Section 1.2).

The analysis of the P-Vr approach to the tuning of fixed-parameter speed-PSS and its implications for a single-machine infinite-bus system have demonstrated the following.
1.

For practical purposes, the phase response of the P-Vr is more-or-less invariant over
an encompassing range of operating conditions. Similarly, at the higher real power
outputs, typically 0.5 to 1 pu of rated power, the magnitude response retains its
shape and consistently lies in a band of 2 dB of the Design Case P-Vr characteristic. The PSS based upon the generators P-Vr characteristics is robust because it
induces positive damping torque coefficients on the shaft of the generator over: (i)
the defined range of modal frequencies, (ii) the encompassing range of N and N-1
operating conditions, (iii) a range of leading and lagging power factors.

2.

At higher levels of generator real power output the magnitude and phase of the P-Vr
frequency response characteristics lie in relatively narrow bands for a wide range of
operating conditions. This permits the selection of a Design Case whose magnitude
and phase response are within the band.

3.

In the following chapters it is recommended that system studies be conducted over a

wide range of operating conditions. The selection of the Design Case should be
representative of the P-Vr characteristics for a set of operating conditions which
encompasses the extreme range of operation, including the most leading and lagging
reactive power output conditions. That is, by a prudent choice of an encompassing,

Sec. 5.14

Summary: P-Vr approach to PSS tuning

285

representative set of N and N-1 operating conditions a fewer number of studies may
be required to determine the selection of the Design Case.
4.

The selection of the Design Case may be biased by experience, e.g. power factors
less than 0.98 leading are very unlikely to occur.

5.

The formulation of a PSS damping gain (in pu on machine MVA rating) has the
advantages, (i) it is the damping torque coefficient induced on the generator over the
design range of modal frequencies; (ii) it is a meaningful number - a moderate damping gain is 20-30 pu, high values exceed 30 to 50 pu. The term PSS gain lacks meaning unless it is clearly defined on such a basis.

6.

As the operating conditions change the inherent damping torque coefficients can
vary from significant negative to large positive values over the modal frequency
range of concern. By adjustment of the damping gain such negative torques must be
swamped by the positive damping coefficients induced by the PSS.

7.

An alternative PSS tuning approach is the GEP(s) method (see Chapter 6). It
employs the phase characteristics of the P-Vr method but provides no information
or guidance on the selection of the PSS gains. The extension of the analytical
approach to the GEP(s) to include the additional information available from the PVr method is a simple further step.

8.

It is shown in Section 5.10.6.2 that the compensation transfer function for the
speed-PSS may include a pair of complex zeros to represent the inverse of the P-Vr
transfer function. A practical PSS may not be capable of accepting complex zeros,
however, it is shown the transfer function may be represented by a set of lead blocks
of the form 1 + sT nk 1 + sT dk , in which the zeros are real and T nk T dk .

9.

In the tuning of speed-PSSs a trade-off may need to be found between the modal
frequency range of interest, the possible high gain of the PSS compensation transfer
function at higher frequencies, the low-pass filter parameters, and the attenuation of
the torsional modes. Use of notch filters [9], or employing integral-of-acceleratingpower PSSs (Section 8.5), may overcome some of these issues. In the case of hydrogenerators no adverse interaction between the generator and the network at the torsional modal frequencies have been reported [9].

10.

A comprehensive evaluation of PSS performance must include consideration of the

response of the system to large disturbances (i.e. transient stability analysis). This
aspect has been omitted from this chapter which has focussed on fundamental concepts but is examined briefly in Section 10.9. The settings of limiters on the output
of the PSS are not considered.

11.

It is shown in Section 5.13 that the P-Vr approach for tuning a speed-PSS can be
adapted to the tuning of a power-based PSS 1. The same advantages of the speedPSS apply to power-PSS; for example, the procedure is systematic, the value of the
PSS gain is a meaningful quantity - it is also a damping gain. An advantage of this

286

Tuning PSSs for SMIB system

Ch. 5

power-PSS is that it does not require the implementation of a phase-lead transferfunction network.The power-PSS may be more relevant, say, to cases for which a
simple PSS is required for damping a local mode, when the number of blocks provided for implementing the PSS is limited, a speed-stabilizing signal is not readily
available, etc.
Other benefits of the P-Vr approach will become evident when (i) the GEP and the Method
of Residues are discussed in Chapter 6, and (ii) multi-machine systems are analysed in
Chapter 10.

5.15 References
[1]

F. P. de Mello and C. Concordia, Concepts of Synchronous Machine Stability as Affected by Excitation Control, Power Apparatus and Systems, IEEE Transactions on,
vol. PAS-88, pp. 316-329, 1969.

[2]

E.V. Larsen and D.A. Swann, Applying Power System Stabilizers: Part I-III, Power
Apparatus and Systems, IEEE Transactions on, vol. PAS-100, pp. 3017-3046, 1981.

[3]

IEEE Power & Energy Society, IEEE Tutorial Course: Power System Stabilization Via
Excitation Control, 09TP250, 2009.

[4]

F. P. de Mello, J. S. Czuba, P. A. Rushe, and J. R. Willis, Developments in application

of stabilizing measures through excitation control, in CIGRE, 1986, 3805.

[5]

M. J. Gibbard, Co-ordinated design of multimachine power system stabilisers based

on damping torque concepts, IEE Proceedings C Generation, Transmission and Distribution, vol. 135, pp. 276-284, 1988.

[6]

W. G. Heffron and R. A. Phillips, Effect of a Modern Amplidyne Voltage Regulator

on Underexcited Operation of Large Turbine Generators, Power Apparatus and Systems, IEEE Transactions on, Part III. Transactions of the American Institute of Electrical
Engineers, vol. 71, pp. 692-697, 1952.

[7]

IEEE Standard 421.1, IEEE Standard Definitions for Excitation Systems for Synchronous Machines, 2007.

[8]

Australian Energy Market Commission, National Electricity Rules, Vers. 55, March
2013.

[9]

P. Kundur, Power System Stability and Control, McGraw-Hill, Inc., 1994.

[10]

D.M. Lam and H. Yee, A Study of Frequency Responses of Generator Electrical

Torques For Power System Stabilizer Design, Power Systems, IEEE Transactions on,
vol. 13, pp. 1136-1142, 1998.

1. An alternative approach to tuning a power-based PSS is described in Section 8.3.

Sec. 5.15

References

287

[11]

M. J. Gibbard, Robust design of fixed-parameter power system stabilisers over a

wide range of operating conditions, Power Systems, IEEE Transactions on, vol. 6, pp.
794-800, 1991.

[12]

D. Simfukwe and B. C. Pal, Robust and Low Order Power Oscillation Damper Design Through Polynomial Control, Power Systems, IEEE Transactions on, vol. 28, pp.
1599-1608, 2013.

[13]

D. D. Simfukwe, B. C. Pal, R. A. Jabr, and N. Martins, Robust and low-order design

of flexible ac transmission systems and power system stabilisers for oscillation damping, Generation, Transmission & Distribution, IET, vol. 6, pp. 445-452, 2012.

[14]

G. R. Berube, L. M. Hajagos, and R. Beaulieu, Practical utility experience with application of power system stabilizers, in Power Engineering Society Summer Meeting,
1999. IEEE, 1999, pp. 104-109 vol. 1.

[15]

J.C.R. Ferraz, N. Martins, N. Zeni Jr, J.M.C. Soares, G.N. Taranto, Adverse Increase
in Generator Terminal Voltage and Reactive Power Transients Caused by Power System Stabilizers, Proceedings of IEEE Power Engineering Society Winter Meeting, 2002.

288

Appendices

Tuning PSSs for SMIB system

Ch. 5

Appendix 5I
App. 5I.1 K-coefficients, Heffron and Phillips Model of SMIB System
The steady-state values of variables are designated with the subscript 0. All quantities are
in per-unit except angles are in degrees. Computation of the initial conditions is based on
solution of the SMIB model equations given in the Appendix of [1].
Vb

Source voltage of the Thvenin equivalent of the system connected

at the generator terminals.

et0

Generator terminal voltage.

ed0, eq0

Direct and quadrature components of the terminal voltage.

id0, iq0

Direct and quadrature components of the generator current.

Eq0

Voltage behind quadrature-axis reactance of the generator.

Angle between the quadrature axis of the generator and the infinite bus.

re, xe

Resistance and reactance of the Thvenin equivalent of system connected

at the generator terminals.

A = r e + x e + X d X q + x e
K 1 = E q0 V b r e sin 0 + x e + X d cos 0 A +
i q0 V b X q X d x e + X q sin 0 r e cos 0 A
K 2 = r e E q0 A + i q0 1 + x e + X q X q X d A
K 3 = 1 + x e + X q X d X d A

K 4 = V b X d X d x e + X q sin 0 r e cos 0 A
K 5 = V b e d0 e t0 X q r e sin 0 + x e + X d cos 0 A +
V b e q0 e t0 X d r e cos 0 x e + X q sin 0 A
K 6 = e q0 e t0 1 X d x e + X q A + r e X q e d0 e t0 A

App. 5I.2 Transfer function of the SMIB system

289

App. 5I.2 Transfer function of the SMIB system with closed-loop

control of terminal voltage
Based on Figure 5.1 for the Heffron and Phillips model for the SMIB system with closedloop voltage control, the transfer function for the terminal voltage response to perturbations
in reference voltage is:
2

s MK 3 K 6 + sDK 3 K 6 + 0 K 3 K 1 K 6 K 2 K 5
V
---------t = -------------------------------------------------------------------------------------------------------------,
4
3
2
V r
s a + s a + s a + sa + a
4

(5.69)

where the damping torque coefficient D = 0.

a 4 = c 3 T do T ex , a 3 = c 3 T do + c 2 T ex , a 2 = c 2 + T ex c 1 + MK 3 K 6 K ex ,
a 1 = c 1 + 0 T ex c 0 + DK 3 K 6 K ex , a 0 = 0 c 0 + K 3 K ex K 1 K 6 K 2 K 5 ,
c 3 = MK 3 , c 2 = M + DK 3 T do , c 1 = 0 K 1 K 3 T do , c 0 = K 1 K 2 K 3 K 4 .

The K-coefficients are listed in Appendix 5I.1

App. 5I.3

Model of the 6th-order generator and excitation system

App. 5I.3.1 State-space model

In Section 12.6 of [9] the state-space model of a 6th-order generator model, having one and
two damper windings on the direct and quadrature axes, respectively, is developed. It is of
the form:
x t = A x t + B u t , y t = C x t + D u t . Thus,

r
a 11 a 12 a 13 a 14 a 15

a 21 0 0 0 0

0 a 32 a 33 a 34 a 35
fd
=

0 a 42 a 43 a 44 a 45
1d
0 a 52 a 53 a 54 a 55

1q
0 a 62 a 63 a 64 a 65

2q

a 16

a 36 fd
+
a 46 1d
a 56 1q
a 66 2q

b 11 0
0 0
0 b 32 P m .
0 0 E fd
0 0
0 0

(5.70)

The elements of the A- and B-matrices are defined in Section 12.6 of [9].
Elements a 12 to a 16 are functions of K 1 , K 2 , K 21 , K 22 and K 23 , respectively (see equations
12.171 to 12.183, [9]).

290

Tuning PSSs for SMIB system

Ch. 5

The elements a 3i , a 4i , a 5i , a 6i are functions of the generator parameters. Note, in equation

12.179 of [9]
0 R 1d
L ads
a 44 = ---------------- 1 -------------- + m 3 L ads .

L 1d
L 1d

The terminal voltage is given by equation 12.185 of [9]:

V t = K 5 + K 6 fd + K 61 1d + K 62 1q + K 63 2q ,
e

E t0

E t0

(5.71)

d0
q0
- R a m 1 + L l n 1 + n 1 L aqs + ------- L l m 1 R a n 1 + m 1 L ads ,
where K 5 = -------

e q0
e d0

1
K 6 = -------- R a m 2 + L l n 2 + n 2 L aqs + -------- L l m 2 R a n 2 + L ads ------- m 2 ,
L fd

E t0
E t0

e d0
e q0

1
K 61 = -------- R a m 3 + L l n 3 + n 3 L aqs + -------- L l m 3 R a n 3 + L ads -------- m 3 ,
L 1d

E t0
E t0

e d0
e q0
1
K 62 = -------- R a m 4 + L l n 4 L aqs -------- n 4 + -------- L l m 4 R a n 4 m 4 L ads ,
L 1q

E t0
E t0
e d0
e q0
1
K 63 = -------- R a m 5 + L l n 5 L aqs -------- n 5 + -------- L l m 5 R a n 5 m 5 L ads .
L 2q

E t0
E t0

The coefficients m 1 m 5 n 1 n 5 are defined in Section 12.6 of [9]. The transfer function
of a first-order excitation system of the form shown in Figure 5.7 on page 239 is
K ex
E fd = -------------------- V r V t + V s ,
1 + sT ex

(5.72)

where E fd is the field voltage (pu), and V s in the output from a PSS, if fitted.
The time-domain form of (5.72) is
K ex
1
E fd = -------- E fd + -------- V r V t + V s , or
T ex
T ex
E fd = a 77 E fd + a 72 + a 73 fd + a 74 d1 + a 75 q1 + a 76 q2 +
+ b 72 V r + V s

(5.73)

App. 5I.3 Model of the generator and excitation system

291

where a 71 = 0 a 72 = K 5 K ex T ex , a 73 = K 6 K ex T ex , a 74 = K 61 K ex T ex ,
a 75 = K 62 K ex T ex , a 76 = K 63 K ex T ex , a 77 = 1 T ex , b 72 = K ex T ex .

(5.74)

Augmenting the state-space model (5.72) with (5.73), the equation for the model becomes:

r
a 11 a 12 a 13 a 14 a 15

a 21 0 0 0 0

fd
0 a 32 a 33 a 34 a 35

1d = 0 a 42 a 43 a 44 a 45

0 a 52 a 53 a 54 a 55
1q
0 a 62 a 63 a 64 a 65

2q
0 a 72 a 73 a 74 a 75
E fd

a 16 0
0

b 11 0

a 36 a 37 fd
a 46 0 1d +
a 56 0

1q

a 66 0

2q

a 76 a 77 E fd

0 0
0 0 P
m
0 0
V r
0 0
0 0
0 b 72

(5.75)

Note, the element a 37 assumes the value of b 32 = 0 R fd L adu in (5.70).

App. 5I.3.2 Calculation of the inherent torque coefficients.
In Section 5.3 it was noted for the SMIB system that the synchronizing and damping torque
coefficients can be calculated from the transfer function P 0 s s given by (5.9) on
page 229 with the shaft dynamics disabled. For the 6th-order generator model a new set of
ABCD matrices is formed by (ii) eliminating the first row in (5.75) and (ii) expressing the
electrical torque P 0 as output and the speed signal r as input, i.e.

0 0

a 32 a 33
fd

a 42 a 43
1d
=

a 52 a 53
1q

a 62 a 63
2q
a 72 a 73
E
fd

0 0 0
a 34 a 35 a 36
a 44 a 45 a 46

0
a 37 fd
0 1d

a 54 a 55 a 56 0
a 64 a 65 a 66 0

1q
2q

a 74 a 75 a 76 a 77 E fd

a 21

0
0
0
0
0
0

(5.76)

The expression for torque, P 0 , is given by Equation 12.170 in [9]:

P 0 = K 1 + K 2 fd + K 21 d1 + K 22 q1 + K 23 q2 .

These K-coefficients are listed in Equation 12.171 of [9].

(5.77)

Chapter 6
Tuning of PSSs using methods based on
Residues and the GEP transfer function

6.1 Introduction
In Section 5.8 the P-Vr method for the tuning of the PSS for a generator in a single-machine
infinite-bus (SMIB) system is described. Several other methods, which will be shown to be
somewhat related to the P-Vr method, are described in the literature. Two other methods
will be discussed here, the first is based on Transfer-Function Residues, the second on the
so-called GEP Method. The P-Vr method, the Method of Residues and the GEP Method
are reconciled for a practical, multi-machine system in [1]. However, for illustrative purposes
in this chapter we will examine only the application of the Residues and GEP Methods to a
generator in a SMIB system.
The background to the Method of Residues is provided in [2] and its application to PSSs is
illustrated in Appendix A of [3]. The method is also used in practice for the design of Power
Oscillation Dampers (PODs) which are fitted to FACTS devices such as SVCs, typically to
enhance the damping of inter-area modes. The design of PODs using the Method of Residues is described in [4], however, this topic is considered in more detail in Chapter 11.

293

294

6.2

PSS tuning based on Residues and GEP(s)

Ch. 6

Method of Residues

6.2.1 Theoretical basis for the Method

The theoretical basis, calculation and significance of the residues of a transfer function are
discussed earlier in Section 2.5. In essence, for a set of distinct poles r i is the residue of the
pole at s = p i . A transfer function G s is described by its partial fraction expansion equation (2.14), or by
r1
r2
ri
rn
P s
G s = ----------- = ------------- + ------------- + + ------------ + + ------------- .
Q s s p1 s p2
s pi
s pn

(6.1)

The derivation of residues from the state equations is outlined in Section 3.7.
Consider a SMIB system for which a PSS is to be designed and installed. The transfer function from the reference voltage input to the speed output signal of the generator is
G S s = V ref . The PSS, with transfer function F s , is a speed-input PSS (although
other stabilizing signals can be employed). When operating in closed-loop the PSS output is
connected to the AVR summing junction, as shown in Figure 6.1.
It is emphasized that the following simple approach to the determination of the compensation transfer function of the PSS is based on the change of the rotor mode of oscillation
when the PSS feedback path is switched from open to closed loop.
AVR, excitation system,
generator, & power system
Vref
GS(s)

+
VS

F(s)

PSS transfer function

Figure 6.1 SMIB system GS(s) and PSS transfer function F(s) on open loop.
Let the PSS transfer function be:
F s = k R H s = k R G c s G W s G LP s ,

(6.2)

where the transfer function G c s of the PSS in this application is designed to provide the
appropriate phase compensation and is assumed to consist of m lead or lag blocks of the
form 1:
1. This form is used in the determination of the order m and time constants Tn and Td in
Appendix 6I.1.

Sec. 6.2

Method of Residues
1 + Tn s m
G c s = ------------------ .
1 + Td s

295

(6.3)

The PSS gain setting in (6.2) is k R in pu on device base (note, this is not the damping gain
associated with the P-Vr method). The washout and low-pass filter transfer functions,
G W s and G LP s , are given by (5.29) and (5.30), respectively. It is assumed that the values
of the time constants in the latter two transfer functions have been appropriately selected
(see Section 5.8.6). The objective of the tuning procedure is to determine the values kR, Tn,
Td and m that satisfy the relevant requirements on damping.
Note that in Figure 6.1 positive feedback is assumed for the following analysis. The transfer
function of the SMIB system and PSS when the loop is closed is therefore:
GS s
GS s
W s = ----------------------------------- = ----------------------------------------- using (6.2).
1 G S s F s
1 k R G S s H s

The poles of the closed-loop transfer function are derived from its characteristic equation:
1 G S s F s = 0 ;

(6.4)

these poles are also the eigenvalues of the system of W s 1. Let us evaluate the shift h
in the pole (eigenvalue) h resulting from the closure of the feedback loop. Assume the
plant and system, G S s , is excited by the eigenvalue h when on open loop, i.e. from (6.1),
GS h = rh s h ,

where r h is the residue of the eigenvalue h of the forward-loop transfer function, G S h .

The associated characteristic equation is, from (6.4),
rh
1 -------------- F h = 0 , or s = h + k R r h H h ,
s h

(6.5)

noting that F s = k R H s in (6.2). Suppose the pole of the closed-loop system is shifted
by a small amount h from the open-loop pole h . The root of the new characteristic
equation is thus s = h + h and (6.5) becomes:
h + h h k R r h H h + h = 0 .

(6.6)

If the mode shift is small then the transfer-function H(s) in the neighbourhood of s = h
can be represented by the first-order Taylor series expansion:
H s

H h + h = H h + -------------- h .
s s = h
1. See Sections 3.7and 2.5

(6.7)

296

PSS tuning based on Residues and GEP(s)

Ch. 6

Substitution of (6.7) in (6.6) leads to

kR rh H h
h = --------------------------------------- ;
H h
1 r h k R ----------------- h

(6.8)

h
- are given in Appendix 6I.2, (6.25) and (6.26).
equations for the evaluation of -----------------

If the gain k R is chosen such that

H h
r h k R ------------------ 1 ,
h

(6.9)

h k R r h H h .

(6.10)

then (6.8) reduces to:

The result in (6.10) is significant for the design of the PSS compensation. As illustrated in
Figure 6.2 the residue of the eigenvalue h of G S s is a complex number, r h h , where
h = arg r h . In order for the mode shift h in (6.10) to be 180 , i.e. a direct left-shift
of h in the complex s-plane is required,
arg r h H h = 180 .

(6.11)

The compensation angle provided by the PSS thus must be

= arg H h = 180 arg r h .

r 1 1
r 2 2

(6.12)

0 1 180 Lead compensation 180 1 required

180 2 360 Lag compensation 2 180 required

Figure 6.2 Examples of phase compensation required for possible residues of G S h .

Based on (6.10) the following is a procedure for determining the PSS parameters.
To effect a left shift in the rotor mode of interest, h , choose the parameters Tn, Td for the
m lead or lag compensator blocks, (6.3), such that arg r h H h = 180 .

Sec. 6.3

Tuning a PSS using the Method of Residues

297

In order to determine a nominal value of an upper limit on the gain kR such that the approximation for the mode shift in (6.10) is acceptable, let us define a nominal limit kRm based on
(6.9) as
H h
r h k Rm ------------------ = 1 , or
h
1
k Rm = -------------------------- pu.
H h
r h ----------------- h

(6.13)
(6.14)

An acceptable gain might lie in the range 0 k R 0.1k Rm , say, but will depend on the nature
of the problem.
It was shown in Section 3.7 that the residue for mode h is the product of an observability
measure c mh and a controllability measure b hq given by
h
r mq = c mh b hq .

(6.15)

Ideally for a selected stabilizing signal these measures should reflect both good observability
of the stabilizing signal and good controllability of the output variable.
From (6.10) it is noted that (i) the larger the magnitude of the residue, the greater is the mode
shift; (ii) for a robust design it is desirable that the magnitude and phase of the residue remain more-or-less unchanged over the range of operating conditions; (iii) the real and imaginary components of the open-loop mode h may also vary over a range of operating
conditions in practice.

6.3

Tuning a speed-PSS using the Method of Residues

6.3.1 Calculation of the compensation transfer function of the PSS

As an example consider the tuning of a true-speed PSS founded on the Method of Residues.
The design is based on a selected set of operating conditions for the SMIB system described
in Table 5.4 on page 263 in which the rotor modes with the PSSs out of service are also listed. To represent a range of operating conditions four normal conditions are selected (Cases
A, B, C, D) together with cases in which one and two lines are out of service (E, F and G,
H respectively).
For the transfer function from the voltage reference input signal to speed-output perturbations the residues, which are calculated using a small-signal dynamics software package, are
displayed in Figure 6.3. The residues lie in a relatively narrow phase-band of approximately
21 with a spread in magnitude between 0.19 and 0.31 units.

298

PSS tuning based on Residues and GEP(s)

Ch. 6

Residues for the transfer function

h V ref h

Figure 6.3

Polar plot of residues for selected cases A to H for the SMIB system.
The residue for Case C is selected as representing the group.
It is now necessary to select from the group in Figure 6.3 a representative residue on which
the compensation is to be based. If the residue of maximum amplitude is selected (e.g. that
for Case A in the figure) the resulting value of nominal limit kRm would be lower than if some
other residue were selected (according to (6.14)). The decision depends on the application
or may be determined by the system criteria which specify the minimum level for damping
(say) for the outage of two lines (Cases G and H here). However, as in Section 5.10.6.1, the
design Case C is again selected to facilitate a comparison of the performance between the
PSSs based on the Residues and P-Vr methods. From the polar plot in Figure 6.3 the residue
for Case C, 0.266 132.4 , is selected as it fairly well represents the group in amplitude and
phase. The PSS is thus required to provide phase-lead compensation of
180 132.4 = 47.6 .
From Table 5.4 for Case C the modal frequency is C = 0.261 j9.02 . Compensation for
the phase shift introduced by the washout and low-pass filters at this complex modal frequency is also required.
Let us assume a single lead transfer function, and washout and low-pass filters with time
constants of TW = 5 s and TLP = 0.0125 s, respectively, are employed. The PSS of (6.2) thus
takes the form:
sT W
F s = k R H s = k R ------------------1 + sT W

1 + Tn s
----------------1 + Td s

1 -------------------.
1 + T LP s

(6.16)

At the complex modal frequency a net phase lag of 5.27 is introduced by the washout and
low pass filters; the total phase compensation required is therefore 47.57 + 5.27 = 52.84
at the rotor mode s = C = 0.261 j9.02. Employing the algorithm outlined in

Sec. 6.3

Tuning a PSS using the Method of Residues

299

Appendix 6I.1 an iterative procedure is used to calculate the parameters of the lead transfer
function for the rotor mode. The resulting time constants are Tn = 0.359 and Td = 0.038 s.
It was suggested earlier that an acceptable gain might lie in the range 0 k R 0.1k Rm , where
the nominal upper gain limit kRm is determined by (6.14). Substitution in the latter equation
H s
with values r h = 0.266 132.4 and --------------s

s = h

= 0.171 j0.221 yields kRm = 13.4 pu on

machine MVA rating. Let us assume an acceptable gain range for kR of 0 to 1.34 pu.
6.3.2 Design Case C. Performance of the PSS with increasing PSS gain
Using the PSS transfer function given by (6.16) let us estimate the values of the rotor mode
as the gain kR is increased from zero (open loop) to 10% of kRm, i.e 1.34 pu.
Three values of the mode are calculated for each value of kR, that is, (i) an approximate value
of the mode based on (6.10), (ii) a corrected value based on (6.8), and (iii) an eigenvalue calculated using the software package Mudpack [5]. The results of these calculations are shown
in Table 6.1.
Table 6.1
Gain

kR (% of
kRm)
0
1
2
3
4
5
6
7
8
9
10#

Rotor modes as PSS gain kR is increased from zero to 10% of kRm.

Approximate *

Corrected+

Eigenanalysis

Real

Real

Real

0.261
0.148
0.035
-0.078
-0.192
-0.305
-0.418
-0.531
-0.644
-0.758
-0.870

Imag
9.018
9.020
9.021
9.021
9.021
9.021
9.025
9.022
9.022
9.022
9.022

0.261
0.148
0.034
-0.080
-0.194
-0.308
-0.422
-0.537
-0.651
-0.765
-0.879

# 0.1kRm=1.34 pu

Imag
9.018
9.019
9.016
9.011
9.003
8.993
8.981
8.966
8.949
8.930
8.909

0.261
0.147
0.030
-0.089
-0.210
-0.334
-0.459
-0.587
-0.716
-0.846
-0.976

* Eqn. (6.10)

Imag
9.018
9.016
9.012
9.004
8.991
8.974
8.953
8.925
8.892
8.851
8.804

+ Eqn. (6.8)

Difference
EigAnal-Approx.
Real
-0.001
-0.005
-0.011
-0.018
-0.029
-0.041
-0.056
-0.072
-0.088
-0.106

Imag
-0.004
-0.009
-0.017
-0.030
-0.047
-0.072
-0.097
-0.130
-0.171
-0.218

Difference
EigAnal-Corrd
Real
-0.001
-0.004
-0.009
-0.016
-0.026
-0.037
-0.050
-0.065
-0.081
-0.097

Imag
-0.003
-0.004
-0.007
-0.012
-0.019
-0.028
-0.041
-0.057
-0.079
-0.105

EigAnal: Eigen-analysis

From the table it is observed that the real parts of the rotor modes based on the approximate
and corrected estimated values start to differ from the eigenvalues by more than 10% for
PSS gains kR greater than 8% to 10% (i.e. 1.1 - 1.3 pu). It should be remembered that more
accurate eigen-analysis takes into account the effect of other zeros and poles on the modal
trajectory as kR is increased whereas the approximate trajectory increases linearly. Engineer-

300

PSS tuning based on Residues and GEP(s)

Ch. 6

ing judgement based on system requirements is needed to decide on the value of the gain
setting - and if it is appropriate for other operating conditions.
6.3.3

Significance of the gain kR and the damping gain k

The gain kR itself has no significance unless it can be related to the damping torque coefficient induced by the PSS, i.e. the damping gain k. Let us relate the results of the simulation
study in Table 6.1 to that based on the P-Vr-based design in Chapter 5 for the same operating condition, Case C in Table 5.4. However, to facilitate the comparison, the following
modifications are made to the PSS based on the P-Vr transfer function in (5.49) on
page 267, (i) the low-pass filter time constants of the PSS transfer function are set to 0.0125
and 0.005 s, and (ii) its damping gain is set to k = 77.4% of 20 pu (15.5 pu); the washout
filter time constant remains at 5 s. The latter gain setting results in the real part of the eigenvalue (the rotor mode) being equal to that produced by simulation for the tuning design
based on Residues Method, i.e 0.976 Np/s with kR = 10% of kRm (see column 6 of
Table 6.1).
Referring to Figure 6.4 we note that for the PSS tuned based on the P-Vr Method the damping torque coefficient is essentially flat at ~15 pu over the frequency range 1 to 10 rad/s for
the setting of the damping gain to 15.5 pu. For the PSS design based on Residues Method
the damping torque coefficient equals that of the P-Vr method at 15 pu at a frequency of
~9.2 rad/s. However, in the residues-based method the damping torque coefficient is seen
to vary markedly over the frequency range 2 to 12 rad/s, a variation which will be shown to
be unsatisfactory if the SMIB system represents an approximate, reduced equivalent of a
larger multi-machine system in which a range of electro-mechanical modal frequencies exist.
Thus, in a multi-machine system to ensure a robust PSS design using the Residues Method
it is necessary to take into account all of the electro-mechanical modes in which the generator participates. A reduced-order system equivalent with a range of modal frequencies is
proposed in [6] to facilitate robust application of the Residues Method. Nevertheless, it is
necessary to verify robustness by closely examining damping performance over an encompassing range of operating conditions.

6.4

Conclusions, Method of Residues

Based on the preceding analysis and the example it is noted that:

In the Method of Residues tuning procedure, the determination of the PSS compensation parameters is based both on the value of the complex rotor mode on open loop
and on the variation of the mode on closed loop as the gain kR is incremented (as in
Table 6.1). In both the P-Vr method and the GEP Method (which follows next) the
analysis is based on frequency response, s = j f . Clearly an analysis based on complex
modal frequencies in the Method of Residues is likely to be more accurate - all else
being equal; this is discussed in Section 5.9.3 with reference to Table 5.3 on page 255.

Sec. 6.4

Conclusions, Method of Residues

301

Damping Torque (pu)

15

Synchronizing Torque (pu)

20

5
1
10

10
5
0

10

10

10

2
1.5
1
0.5
0

0.5
1
10

10
10
Frequency (rad/s)

10

PVr based Method

Residues based Method

Figure 6.4 Case C: Synchronising and damping torque coefficients induced on the generator. PSS tuning is based on (i) P-Vr Method and (ii) Method of Residues.

The relations (6.10) and (6.8) for the approximate and corrected mode shifts provide
useful estimates of the mode shifts with increase in gain kR. However, in the multimachine case, eigen-analysis may reveal that the effect of system poles and zeros may
cause the trajectory of the rotor mode to deviate from a direct left-shift at relatively
low values of the gain kR.

In the example presented the Method of Residues can provide a basis for determining
the PSS transfer function, and yields acceptable values of mode shifts with increasing
PSS gain. However, the performance of the PSS needs to be validated over an encompassing range of operating conditions using a small-signal dynamics software package,
particularly in the application of the method to PSSs in multi-machine systems [4],
[10].

In the P-Vr method, the damping gain k represents the damping torque coefficient
induced by the PSS over a desired range of rotor modal frequencies. In the Method of
Residues the value of the PSS gain kR has no obvious significance.

302

6.5

PSS tuning based on Residues and GEP(s)

Ch. 6

The P-Vr method yields inherently robust PSS designs (see item 3 of Section 1.2).
Moreover, the PSS can be tuned to yield a more-or-less constant positive damping
torque coefficient over a modal frequency range covering the low frequency inter-area
modes to the higher frequency local-area and intra-plant modes (see Figure 6.4). In
comparison, robust PSS design using the Residue Method requires accurate determination of the residues of all rotor modes in which the generator participates significantly.

The GEP Method

Larsen and Swann described in 1981 a practical procedure for tuning PSSs based on measurements taken in the field [7]. The procedure is based on the design approach of de Mello
and Concordia [8]. The transfer function between the voltage reference input to the AVR
and the electrical torque developed on the shaft is called the generator, excitation system and
power system transfer function, GEP(s). GEP(s) can be shown to be proportional to the
transfer function from voltage reference ( V r s ) to terminal voltage ( V t s ); the frequency
response V t j f V r j f is relatively straight-forward to measure in the field. The compensation angle for the PSS transfer function is the negative of the phase shift of the measured frequency response. From this result, a compensating transfer function is synthesized
for the PSS. A further test is performed to determine the gain setting of the PSS. This test
consists of raising the gain until the onset of instability is observed; the PSS gain is then set
to 1/3rd of this value - providing a gain margin of about 10 dB. Further developments of
the GEP approach are reported in [6], [9].
The use of the field measured frequency-response for PSS design relies on the assumption
that, because the generator is connected to a large power system, its speed remains more-orless constant during the frequency response measurements. This is equivalent to assuming
that the inertia constant of the unit is very large or, alternatively, the speed and angle perturbations are negligible. For example, based on the Heffron and Phillips model of a SMIB system under closed-loop voltage control in Figure 5.8 on page 239, the transfer functions for
both the torque of electro-magnetic origin and the terminal voltage can be written with respect to the reference voltage perturbations as:
K 2 K 3 K ex
P 2 s
----------------- = -------------------------------------------------------------------------------------------------------------------- , and

2
V r s
s K 3 T ex T d0 + s T ex + K 3 T d0 + 1 + K 3 K 6 K ex

(6.17)

K 6 K 3 K ex
V t s
---------------- = --------------------------------------------------------------------------------------------------------------------,

2
V r s
s K 3 T ex T d0 + s T ex + K 3 T d0 + 1 + K 3 K 6 K ex

(6.18)

respectively; both these equations are independent of the shaft-dynamics. An inspection of

the two transfer functions reveal that (i) they are simply related by the scalar ratio K 2 K 6
and, (ii) the phase responses of terminal voltage and the P-Vr characteristics are identical.
The PSS tuned according to the GEP method must therefore introduce phase lead to com-

Sec. 6.6

Tuning a speed-PSS using the GEP Method

303

pensate for the lagging phase characteristic of V t j f V r j f . The phase response of

the synthesized compensation therefore corresponds to the negation of the lagging phase
characteristic.

6.6 Tuning a speed-PSS using the GEP Method

Let us consider the tuning of a true-speed PSS based on the GEP method for the sixth-order
generator and SMIB system of Section 5.10.4.1. Because the measurements of the frequency
responses when conducted in the field are likely to be at lower levels of real power output,
operating conditions Cases R and K for the SMIB system listed in Table 5.6 on page 276 are
used as examples. (For reasons of security, field measurements are likely to be carried out at
lower values of real power output.) The generator is under closed-loop voltage control with
power outputs of 0.5 and 0.7 pu at unity power factor; the associated rotor modes are
0.539 j8.50 and 0.065 j8.92 , respectively.
The terminal voltage frequency responses are compared with those of their P-Vr characteristics in Figure 6.5. The presence of the resonance associated with the lightly damped rotor
mode for Case K is particularly evident in the terminal voltage responses of the GEP frequency responses. If the GEP transfer function is calculated with all shaft dynamics disabled
(GEPSDD), the resonances in Cases R and K are eliminated and the phase response of
GEPSDD generally agrees closely with that of the P-Vr transfer function, as seen in
Figure 6.6.
The P-Vr characteristic representing the synthesized GEPSDD phase responses is
2

1 + s0.005 1 + s0.05 .

Thus the associated transfer function of the PSS derived from the negation of the synthesized GEPSDD phase responses takes the form,
5s
1 + s0.05 2
k G --------------- ------------------------- ,
1 + 5s 1 + s0.005

(6.19)

which includes a washout and a first-order a low-pass filter; let kG = 1.

6.6.1 Example 2. Performance of the PSS based on Design Case C
For purposes of comparison with the Method of Residues the performance of the PSS based
on the GEP approach is evaluated for Design Case C of Table 5.4, P = 0.9 Q = 0.2 pu using
the PSS given by (6.19).
The damping torque coefficient for the PSS is calculated as in Section 6.3.3 for the Residue
Method. The frequency response of the damping torque coefficient is shown in Figure 6.7
and reveals that the value of the coefficient at the modal frequency of 9.0 rad/s is 4.1 pu.
Consequently the PSS gain kG = 1 is equivalent to a damping gain of 4.1 pu in the vicinity
of the modal frequency of rotor oscillations.

304

PSS tuning based on Residues and GEP(s)

Ch. 6

Magnitude (dB)

20
10
0
10
20
30
0
10

10

10

Phase (deg)

0
50
100
150
0

10

10
Frequency (rad/s)

Vt/Vr Cases: R, P=0.5

PVr Cases: R, P=0.5

10
K, P=0.7
K, P=0.7

Figure 6.5 Frequency responses for the transfer function V t V r and the P-Vr
characteristics for lower levels of real power output at unity power factor.

Phase (deg)

0
50
100
150
0

10

10
Frequency (rad/s)

10

GEPDD Cases: R, P=0.5,

K, P=0.7
PVr Cases: R, P=0.5,
K, P=0.7
Synthesized PVr from Case C

Figure 6.6 Comparison of the GEP phase response with shaft dynamics disabled (GEPSDD) and those of the associated P-Vr characteristics. The synthesized phase response for
the design Case C of Table 5.4 is also shown.

Sec. 6.6

Tuning a speed-PSS using the GEP Method

305

Damping Torq. Coeff.(pu)

Note that in Figure 6.7 the damping torque coefficient has, relatively, a more level frequency
response over the range 1.5-15rad/s than that in Figure 6.4 for the PSS design based on the
Method of Residues. This is because the phase compensation of (6.19) reflects that of the
P-Vr approach over the frequency range - rather than that designed at a single selected modal frequency.
4.5
4
3.5
3
2.5
2
1.5
1
0
10

10
Frequency (rad/s)

10

PSSinduced damping torque coefficient

Figure 6.7 Case C: PSS-induced damping torque coefficient for a PSS designed according
to the GEP method and with gain kG = 1.
Let us compare the mode shifts based on the GEP Method and the P-Vr method. For
kG =4.88 pu (equivalent to a PSS damping gain of 20 pu on machine MVA rating), the rotor

mode using the GEP method is 1.35 j8.94 for operating condition C. The mode shift
associated with this gain is 1.61 j0.08 , a value which is close agreement with that derived
based on the P-Vr method, i.e. 1.60 j0.02 (refer to Table 5.5).
From a comparison of Figure 6.5 and Figure 6.6, the following observations are offered.

As is to be expected, the magnitude responses of the conventional GEP and the P-Vr
transfer functions differ. (Because the generator is under closed-loop voltage control
the gain of the GEP transfer function tends to unity at low frequencies.)

The conventional GEP phase response may be distorted by a number of resonances

associated with the rotor modal frequencies in a multi-machine system. (These should
be significantly attenuated by the damping introduced by a properly-designed PSS).
Such phase responses and the lack of appropriate magnitude information complicate
the synthesis of the PSS transfer function. (It is usually a matter of finding a number
of lead networks to obtain the desired lead compensation over the range of modal frequencies.)

For measurements in the field, one would be reluctant to proceed with frequency
response measurements approaching a resonance because of an uncertainty concern-

306

PSS tuning based on Residues and GEP(s)

Ch. 6

ing its magnitude. The range of the frequencies measured may therefore be restricted
to lower values when resonances are encountered. Such resonances may not be
encountered at lower real power outputs from the generator.

The magnitude responses for the P-Vr and the GEPSDD transfer functions differ by
a constant gain value. There is close agreement between the P-Vr and GEPSDD
phase responses for the sixth-order model of the generator. The GEPSDD phase
response which is determined by analysis in this exercise is seen to provide a good
smoothed representation of that of the GEP.

6.7

Conclusions, GEP method

Based on the preceding analysis and Example 2 it is noted that:

1.

In a single- or multi-machine system the field-measured GEP responses may be

adulterated by resonances due to lightly damped local or intra-station modes. The
range of the measured frequency responses may be curtailed as the size of the resonance under field conditions is typically unknown.

2.

Field-measured GEPs can assist in the validation of the small-signal system model
of the generator and system used in simulation-based PSS tuning methods. It should
be noted that, for PSS tuning using analytical techniques, accurate models of the
generator and excitation system are highly desirable.

3.

By eliminating the resonances associated with the rotor modes of the conventional
GEP transfer functions, the phase responses of the GEPSDD transfer functions
determined by analysis provide a sounder basis for the determination of phase compensation required for PSS than by field measurements. However, for the evaluation
of the GEPSDD one would need to have confidence in the accuracy of the model of
the generator and system.

4.

If, in the tuning of a PSS based on the conventional analysis of GEP(s), the shaft
dynamics are disabled 1 then the same phase information is available in the frequency responses of both the GEP(s) and the P-Vr methods. However, the associated magnitude responses in the GEP(s) method are ignored. Because the P-Vr
method provides guidelines for the settings of PSS damping gains, the concepts,
analysis and results in Chapters 10, 12 -14 are lost.

5.

Of the three methods which are discussed in this chapter and are complementary,
the P-Vr frequency response provides continuous, consistent information over the
range of rotor modal frequencies and encompassing operating conditions for both
magnitude and phase; this simplifies considerably the synthesis of the PSS transfer
function and its tuning.
1. The transfer function in the case is P s V ref s .

Sec. 6.8
6.

References

307

The PSS damping gain associated with the P-Vr method has particular significance. It

provides a sound and proper basis for the systematic selection of PSS gains. Moreover, its significance as a damping torque coefficient is used in the theoretical and
practical applications demonstrated in Chapters 12 to 14. It is also a meaningful
quantity. For example, 20 pu damping gain on machine rating is a moderate gain
value; neither the PSS gain kR of (6.2), associated with the Residue Method, nor kG
(6.19) of the GEP Method, have any significant meaning when expressed on
machine rating.
7.

Because of the advantages listed above, and other merits, the P-Vr approach is
employed in practice by a number of organizations for the tuning of PSSs in the
multi-machine environment described in Chapter 10.

6.8 References
[1]

M. J. Gibbard and D. J. Vowles, Reconciliation of methods of compensation for

PSSs in multimachine systems, Power Systems, IEEE Transactions on, vol. 19, pp.
463-472, 2004.

[2]

F. L. Pagola, I. J. Perez-Arriaga and G. C. Verghese, On sensitivities, residues and

participations: applications to oscillatory stability analysis and control, Power Systems, IEEE Transactions on, vol. 4, pp. 278-285, 1989.

[3]

CIGRE Technical Brochure no. 166 prepared by Task Force 38.02.16, Impact of Interactions among Power System Controls, published by CIGRE, 2000.

[4]

P. Pourbeik and M. J. Gibbard, Damping and synchronizing torques induced on

generators by FACTS stabilizers in multimachine power systems, Power Systems,
IEEE Transactions on, vol. 11, pp. 1920-1925, 1996.

[5]

D.J. Vowles and M.J. Gibbard, Mudpack User Manual: Version 10S-03, School of
Electrical and Electronic Engineering, The University of Adelaide, July 2014.

[6]

F. De Marco, N. Martins, and J. C. R. Ferrari, An Automatic Method for Power System Stabilizers Phase Compensation Design, Power Systems, IEEE Transactions on,
vol. 28, pp. 997-1007, 2013.

[7]

E. V. Larsen and D. A. Swan, Applying power system stabilizers: Part I III, Power
Apparatus and Systems, IEEE Transactions on, vol. PAS-100, pp. 30173046, June
1981.

[8]

F. P. Mello and C. Concordia, Concepts of Synchronous Machine Stability as Affected by Excitation Control, Power Apparatus and Systems, IEEE Transactions on,
vol. PAS-88, pp. 316-329, 1969.

[9]

P. Kundur, M. Klein, G. J. Rogers, and M. S. Zywno, Application of power system

stabilizers for enhancement of overall stability, Power Systems, IEEE Transactions on,
vol. 4, pp. 614626, 1989.

308
[10]

PSS tuning based on Residues and GEP(s)

Ch. 6

P. Pourbeik, Design and Coordination of Stabilisers for Generators and FACTS devices
in Multimachine Power Systems, PhD Thesis, The University of Adelaide, Australia,
1997.

Appendices

309

Appendix 6I
App. 6I.1

Algorithm for the calculation of stabilizer parameters

The algorithm is based on the general form of the stabilizer transfer function F(s)
TW s w 1 + Tn s m
z
1
F s = k fds H s = k fds ------------------- ------------------ --------------------- ,
1 + T LP s
1 + TW s
1 + Td s

(6.20)

which consists of the compensator, and the washout and low-pass filters [4], [10]. Lead compensation is to be designed such that the compensation angle provided by the stabilizer
is = arg H h = 180 arg r h at the selected complex frequency s = = + j .
For this complex mode let us assume that the washout and low-pass filters introduce a phase
lead of ; the maximum phase lead to be contributed at frequency by the compensator
is then m = . For a compensator consisting of m first-order lead blocks the phase
lead to be contributed by each block is max = m m . Let us therefore consider the firstorder lead compensator described in Section 2.12.1.4 Example 10 1.
G s = 1 + sT 1 + sT .

(6.21)

If the maximum phase lead that can be produced by the above compensator is assumed to
be 60 , then the allowable range for is 0.07 1 . Substitution of the complex frequency
s = = + j in (6.21) results in:
2

1 T 1 + + T + + jT 1
G = ------------------------------------------------------------------------------------------------------------------- .
2 2 2
2
1 2T + T +

The phase lead introduced by the compensator at s = is equal to

m
T 1
= ------ = atan ---------------------------------------------------------------------------2 2
2
m
1 T 1 + + T +

The desired maximum phase lead, max occurs at = c where d d

(6.22)

= c

= 0. Noting

that is a constant and applying the constraints T 0 and 1 , the expression

d d = = 0 is solved to give
c

c =

2
1 - -----
-------- - --- +
2 T T
T

1. Note that for = 0 , i.e. s = j f , c = 1 T and max = asin 1 1 + .

(6.23)

310

PSS tuning based on Residues and GEP(s)

Ch. 6

Substitution of (6.23) into (6.22) yields an expression for the maximum phase lead:
1 T T + 2 T 2 T
--------------------------------------------------------- --- 1
2

2
T
max = atan Z where Z = ----------------------------------------------------------------------------------------2
2
1 T T + T

(6.24)

To design a first-order lead block to provide the required maximum phase lead max at the
modal frequency h = h + j h , the following procedure is proposed to solve for the compensator parameters and T in (6.21).
1.

Select m such that max 60 where max = m m .

2.

max
Calculate 0 = ---------------------------------as an initial estimate of .

3.

Calculate T 0 = 1 0 h as an initial estimate of T.

4.

Set (i) a tolerance level for the iterative calculations, e.g. = 0.001, and (ii) the counter to k = 1.

5.

Given that c = h , solve for T using (6.23) in the form

1 sin

1 + sin max

k 1 h h T h 1 + k 1 T + 1 = 0

and choose the smallest positive value of T. Set Tk = T.

6.

Solve (6.24) for letting X = tan max 2 , i.e. solve the equation
2

4T k X h 4T k X h 1 + 2 + 4X 4X h T k 1 = 0

and choose the smallest positive value of . Set k = .

7.

If k k 1 and T k T k 1 then end, else go to 5.

The required first-order compensator parameters are = k and T = Tk.

Note that if the compensation angle is negative, i.e. max 0 , the parameters of a lag compensation transfer function are calculated.

App. 6I.2 Algorithm for the calculation of stabilizer parameters

311

App. 6I.2 Calculation of the nominal upper limit of the range of stabilizer gains
A nominal measure of the upper limit of range of stabilizer gain is shown to be:
H h
k Rm = 1 r h ----------------- h

(6.14) (repeated)

The residue rh is specified as are the parameters of the washout and low-pass filters. Having
selected the desired order m of the compensator its parameters are calculated using the
Hs
above algorithm. It then remains to calculate -------------and evaluate it at the selected modal fres

quency s = h = h + jh .
Assuming the three blocks in the transfer function are in forms such as those in (6.20), the
derivative can be expressed in a general form:
p1
a + sT p
a + sT 1
T 1 aT 2
J s = -----------------1- = p -------------------------------------------------------------.
1 + sT

p+1
s
s
2
1 + sT 2

(6.25)

From the above expression the derivative of the three transfer functions in (6.20) can derived by setting the following quantities for the:
compensator G c s ,
a = 1, Tn = T1, Td = T2, m = p;
washout filter G W s ,

a = 0, Tw = T1 = T2, w = p;

low-pass filter G LP s ,

a = 1, T1 = 0, TLP = T2, z = p.

Given H(s) in (6.20) is of the form H s = A s B s C s , the expression for the derivative is:
-------------Hs
= A s B s C s + B s C s A s + C s A s B s .
s
s
s
s

(6.26)

Chapter 7
Introduction to the Tuning of
Automatic Voltage Regulators

7.1 Introduction
7.1.1 Purposes
Given a model and the parameters of the generator and its exciter, there is little published
in the literature describing the various methods for the tuning of automatic voltage regulators (AVRs) to achieve certain performance specifications for the generator off- and on-line.
An aim of this chapter is to introduce and provide an analytical basis for various tuning
methodologies, which provide a set of parameters for the particular AVR model. Further
analysis may depend on the type and form of the AVR supplied by a manufacturer. However, even for complex AVR structures, the proposed methodologies may provide an initial
set of parameters based on a simplified model of the AVR. Subsequent fine-tuning, based
on the complex structure, can then yield an appropriate final set of parameters.
It should be emphasized that the tuning methodologies considered here are based on the
concept of transient gain reduction, though various other design approaches are employed
[1]. Depending on the type of AVR, rate-feedback may also be used to essentially effect a
similar behaviour as transient gain reduction. Furthermore, more modern systems which
employ proportional-integral-derivative (PID) controls can be tuned to give a response akin
to transient gain reduction. It is recognized that manufacturers of the equipment have their

313

314

Introduction to the tuning of AVRs

Ch. 7

own, effective procedures for tuning. However, when tuning, it is important in a number of
scenarios to account for the power system characteristics over an encompassing range of
normal and outage conditions. The latter considerations are often of concern to the transmission service provider (TSP) who may be responsible for system security. It is therefore
desirable that staff in such TSPs understand the relevant methodologies and can undertake
or validate, if necessary, the tuning of AVRs.
A further objective in the description of the methodologies is to provide for young engineers
an introductory and a reference text which not only covers the relevant control systems
background but also highlights the power system requirements and performance.
7.1.2 Coverage of the topic
Because powerful methods of analysis are available in linear control systems theory, the tuning of AVRs is based on small-signal analysis and the linearized models of the power system
and associated devices. The performance of the resulting tuned AVR, and the other elements of the power system, should then be subject to simulation studies for an appropriate
set of large-signal disturbances over the range of operating conditions. In such simulations
the limits on AVR and the exciter quantities, as well as saturation, should be modelled.

7.2

The excitation control system of a synchronous generator

The IEEE Standard 421.1 [2] defines the excitation control system (ECS) as the feedback
control system that includes the synchronous generator and its excitation system. Essentially
the excitation control system is the system which excites and controls the rotor field current
of the generator and thus the term ECS includes the generator. The excitation control system as well as the excitation system (ES) are shown in the block diagram of Figure 7.1. The
excitation system is defined as the equipment providing field current for a synchronous generator, including all power, regulating, control, and protective elements. The main power
element is the exciter, however, the regulating, control, and protective elements are referred to in the Standard as the synchronous machine regulator.

Synchronous
machine
regulator

Exciter

Synchronous
generator

Power System

Excitation System
Excitation Control System

Figure 7.1

Block diagram of the excitation control system [2].

Sec. 7.2

The excitation control system of a generator

315

We shall refer to the synchronous machine regulator as the automatic voltage regulator or
AVR 1. We will assume the AVR comprises all the control elements and any lower-power,
power-electronic devices which drive the input to the exciter.
IEEE Standard 421.2 [3] concerning aspects of the testing and evaluation of dynamic performance of excitation control systems is also of interest here.
A component of the AVR is the compensating control provided to ensure that the excitation
control system satisfies certain steady-state and dynamic performance criteria for the unit offand on-line. The main objective in AVR tuning is to determine the parameters of the appropriate compensator which satisfy the criteria. The block diagram of the excitation control
system, which forms the basis for the analysis which follows later, is shown in Figure 7.2.
The element K A 1 + sT A typically represents the simplified dynamics of the AVR power
amplifier and the gain. However, this block diagram does not apply to rate-feedback compensation which is treated in Section 7.10.
When the generator is on-line the transfer function Ggen(s) includes the dynamics associated
with the external power system. When the unit is off-line it is assumed to be under closedloop voltage control and operating at rated voltage at synchronous speed. It should be noted
that the dynamic behaviour of the excitation system and generator may differ significantly
when off- or on-line under closed-loop voltage control.

AVR
Vref

Verr

Gc(s)

KA
1+ sTA

Compensation

Vr

Ef

Vt

Gex(s)

Ggen(s)

Exciter

Generator

Excitation system

Figure 7.2 Excitation control system with compensation in the forward path of the AVR
When forming models of the components of the excitation control system from results of
tests, careful attention must be paid to the per unit definitions of the components and the
per unit relations between components. Commonly-used definitions are listed in Chapter 4.

1. The IEEE Standard 421.1 [2] refers to automatic voltage regulator as a term often used
to designate either the voltage regulator alone or the complete control system comprised
of limiters, etc.

316

7.3

Introduction to the tuning of AVRs

Ch. 7

Types of compensation and methods of analysis

A variety of types of compensation is employed in the AVR and a number of these will be
discussed in the following sections. Each of these are analysed and, where appropriate, a design method or procedure is formulated. The aim in each method is to achieve a desired
transient gain reduction over a selected range of modal frequencies. For each type, illustrative examples are provided of the tuning method to achieve specified dynamic performance
criteria with the generator off- and on-line in a simple power system. The types of compensation considered are:

Transient gain reduction (TGR), the simpler type of compensation 1 (Section 7.6).

Proportional plus Integral plus Derivative (PID) compensation (Section 7.7). A number
of forms of PID compensation is analysed and procedures for the calculation of the
parameters are proposed. In the case of a type of PID, called Type 2B here, a detailed
analysis is undertaken to determine a suitable set of PID parameters for both normal
and N-1 operating conditions for a remote, three-generator power station in which
brushless AC exciters are installed. (Section 7.11).

Proportional plus Integral (PI) compensation (Section 7.9). The concepts in the PI

design procedure are simple, and follow on from the earlier sections. The extension of
PI compensation to PID compensation using a series lead-lag block is illustrated and
the equations for the parameter conversion to the PID structure are formulated.

Rate feedback of the AVR or exciter output (RFB) (Section 7.10). The aim of the analysis
is to determine the gain and time constant in the feedback transfer function that satisfy the performance specifications. A method for calculating the latter parameters is
proposed which is based on a simple model of the excitation system. However, it is
shown that the results can be applied to more complex systems which, for instance,
include PI compensation in the forward path.

The small-signal performance of the system is analysed using software packages such as
Matlab [4] or Mudpack [5]. Several methods of linear system analysis such as frequency
response, root-locus, step response and eigen-analysis are employed. The theory behind
these linear system analysis techniques are described in references [8] and [9].

7.4 Steady-state and dynamic performance requirements on the generator and excitation system
In some types of excitation systems there is a requirement for high values of gain in the forward loop of the excitation system for closed-loop voltage control. Such gains are typically
employed (i) to provide fast response of the generator terminal voltage to disturbances, (ii)
1. Transient gain reduction has not been used in some cases with fast-response, high gain,
static excitations systems [6], [7].

Sec. 7.4

Performance requirements on the ECS

317

to boost field flux linkages following a major disturbance in order to increase synchronizing
power, and (iii) to satisfy requirements on the error in the terminal voltage in the steady-state
(zero frequency and final equilibrium point).
At higher modal frequencies high gains in the forward loop of the closed-loop voltage control system are destabilizing. In order to provide a stable, robust system it is necessary to
reduce, by compensation, the high forward loop gain (KA) to a lower transient gain (KT)
at higher frequencies. This concept is illustrated in Figure 7.3.
KA
Magnitude
(dB)

1
KT
2

Frequency
(rad/s)

Figure 7.3 Straight-line approximation of the magnitude response of the compensation

transfer function showing high gain at low frequency and transient gain reduction at higher
frequencies.
In general, the basis of the approaches is (i) to establish a transient gain reduction over an
appropriate range of modal frequencies, and (ii) to provide a well-damped response of the
generator terminal voltage to a step change in reference voltage. Such a range of modal frequencies depends on the rating and location of the generator and its participation in the inter-area, local-area and intra-station modes. It will be assumed for present purposes that the
range of modal frequencies is 1.5 to 15 rad/s. On the other hand, if local-area modes only
were of concern the relevant frequency range might be 6 to 12 rad/s (see Section 1.5). The
compensation provided should also satisfy the criteria for the dynamic performance of the
generator off-line under closed-loop voltage control (this matter is treated in sections 7.6.3,
7.7.2.3 and 7.11.2.2).
The level of transient gain is determined in the first instance by the closed-loop performance
of the generator and excitation system under voltage control when the unit is off-line and
running at synchronous speed. In order that the terminal voltage response to a step change
in reference voltage is well damped, it is suggested in [10] that the transient gain K T should
be approximately less than T do 2T E . For example, if the generators open-circuit time
constant is T do = 5 s and the exciter time constant is T E = 0.1s, the transient gain should
be less than 25 pu on machine base. It is also stated in [10] that ... the closed-loop response
of this voltage-component regulating-loop (i.e. ignoring the contribution from rotor angle)
under load conditions is not materially different from the closed-loop response on open cir-

318

Introduction to the tuning of AVRs

Ch. 7

cuit. This implies that the approximate value K T T do 2T E is appropriate on-line.

However, it also pointed out in the discussion in [10] that the performance with values of
transient gain exceeding T do 2T E has proved satisfactory. Note that cases are illustrated
in which the closed-loop terminal voltage response to a step change in reference voltage is
well damped but the rotor modes are poorly damped. Such cases may required a coordinated
tuning with the PSS installed on the generator.
The range of the transient gain is typically 25 to 50 pu on machine base. In the examples in
this chapter a transient gain of 32 pu (~30 dB) is usually adopted.
In order to ascertain the characteristics of the closed-loop voltage control system of a generator, either off- or on-line, step changes in the voltage reference of up to 5% of rated terminal voltage are employed in practice. The size of the steps is such that the response is
measurable above the noise level but not so large as to produce over-voltages of the exciter
output. However, for the small-signal (i.e. linear) simulation studies in this book, a step size
of 1% from 1 pu is a more convenient and meaningful measure.
Minimum performance requirements on generators and excitation systems are usually specified by the system operator in a code or a set of rules; the latter will be referred to as the
Rules. As an example - and as applied in this chapter - typical requirements that may be
imposed on the generator and excitation system are:

Under closed-loop voltage control the generating unit must be stable when off-line
and on-line. For planning purposes, when the unit is on-line the halving time of any
inter-regional or intra-regional rotor oscillations should be less than 5 s (e.g. [11]).

Each excitation control system must provide continuous voltage regulation to within
0.5% of the selected set-point value at all operating points within generator capability,
(e.g. [11]). This is interpreted as requiring that the effective DC gain of the terminal
voltage control loop is at least 200 pu for the steady-state terminal voltage error to be
less than 0.5% (see Section 2.10.1) (this is assuming a proportional-only control).

With the generator on-line and under closed-loop voltage control, the settling time
following a disturbance equivalent to a 5% step change in the measured generatingunit terminal voltage must be less than 5 s. This must be satisfied at all operating
points within the generating unit capability. It is assumed here that the step change
does not lead to the activation of limiters in the excitation system. The settling time is
the time for the terminal voltage response to decay to within a prescribed percentage
of the final steady-state level. (In Section 7.11 a 10% settling time employed; unless
otherwise stated a 2% settling time is adopted in this book.)

When the unit is under closed-loop voltage control and is running off-line at rated
speed the corresponding settling time is 2.5 s, or less.

Sec. 7.5

A single-machine infinite-bus test system

319

The objective of the analysis in this chapter is to determine the compensation that must provided by the AVR to satisfy the relevant Rules and, when appropriate, to align with excitation system models in the IEEE Standard 421.5 [12].
Other helpful background material is provided in [7] and in [13] to [15].
As mentioned earlier an objective of this chapter is to provide a theoretical background and
some guidelines to the various approaches to AVR tuning. Each method of compensation
that is analysed is followed by an illustrative example, typically for the generator off- and online under closed-loop voltage control. Because the requirements and performance specifications may vary from application to application, there is flexibility available for fine tuning
of the controls based on any of the approaches.
The following discussions are based on small-signal analysis. In particular, Bode plots of the
open-loop frequency response of a system will be employed to assess the stability and performance of the closed-loop system using the concepts of Phase Margin and the gain-crossover frequency. The open-loop system should therefore have no poles or zeros in the righthalf of the s-plane [8].
As emphasized earlier, the results obtained based on small-signal analysis should be reviewed in the context of appropriate and relevant large-signal (transient stability) studies.
In modelling excitation systems, limiting of the outputs is imposed on certain types of
blocks, namely integrators, first-order blocks, lag-lead and lead-lag blocks. Use of an integrator, say with windup limiting, can result in additional phase shifts.Windup should not be
an issue in strictly small-signal analysis but one should be aware of it occurring following
large-signal disturbances. (See Appendix 7I.4.)

7.5 A single-machine infinite-bus test system

In the examples to illustrate the application of the methods of compensation the single-machine infinite-bus (SMIB) system shown in Figure 7.4 is employed 1. One line (a) of a pair
of parallel lines is out of service, e.g. for maintenance or following a fault on the line. The
generator output at unity power factor is 0.4 pu power (a value above the range 0.1 to 0.3
pu at which the PSSs typically are switched into service). The load is modelled as constant
impedance. For on-line analysis the generator transfer function block, referred to as
G gen s , accounts for the dynamics of both the generator and the system to which it is connected. The sixth-order generator parameters are listed in Appendix 7I.1. Note that the
model of the exciter is a simple first-order lag; a more detailed model is used for a brushless
exciter in Section 7.8.
1. For the studies in Section 7.8 concerning the Type 2B PID, a different but more relevant
power system is chosen.

320

Introduction to the tuning of AVRs

xt =0.15, tap 0.96

Ch. 7

Infinite Bus

V=1.043 6.8

VB= 0.9871 0

Vt = 1.0 10
Constant Impedance
Load

Parameters per line:

Z=0.0225+j0.225, B=0.11

Figure 7.4 SMIB test system, outage of line a; Case W.

Generator output P = 0.4, Q = 0 pu. Load: P = 0.02, Q = 0 pu at 1 pu voltage.
Parameters 1 are in pu on machine base (MVA). System frequency is 50 Hz.
Only this one operating condition (an N-1 condition) is analysed to illustrate some of the
issues that may arise in the design process. Without a PSS installed on the generator in this
case a lightly damped rotor mode is present. In practice an encompassing range of normal
and N-1 operating conditions should be examined in order to determine the appropriate
compensation. For the purposes of this chapter we will deem damping performance to be
adequate if the halving time of the rotor mode of oscillation is shorter than 5 seconds 2. This
is consistent with the requirements in the Australian National Electricity Rules (NER) [11].

7.6

Transient Gain Reduction (TGR) Compensation

7.6.1 Introduction
In classical control terminology Transient Gain Reduction (TGR) is referred to as cascade
lag-lead compensation; such compensation is incorporated in the control system of the AVR
as shown in Figure 7.5.

Excitation system
Vref

Verr

1+ sTC

KA

1+ sTB

1+ sTA

TGR
Compensation

TA = 0

Vr

Ef
Gex

Ggen

Vt

Exciter
Gex = 1/(KE+sTE)

Figure 7.5 Excitation control system with Transient Gain Reduction

1. The generator and excitation system parameters which are given in Appendix 7I.1.1 differ from those in Section 5.10.4. However, the transformer and line parameters are the
same on the generator MVA base as those in the latter section.
2. The stability limit is derived from 1 e

a5

= 0.5 , i.e. a = 0.139 Np/s.

Sec. 7.6

Transient Gain Reduction Compensation

321

The transfer functions for the TGR block and the associated classical lag-lead block are:
1 + sT C
-----------------1 + sT B

or

1 + sT
------------------- , respectively, where T B T C or 1 .
1 + sT

(7.1)

Let f be the frequency of the exciting sinusoidal signal. In the frequency domain at high
frequencies ( f T 1 ) the transfer functions in (7.1) reduce to T C T B or 1 . At such frequencies the gain of the AVR transfer function in the forward loop is
V r j f V err j f = K T = K A T C T B ,
where K T is the per-unit transient gain. TGR compensation thus provides a gain reduction
KT/KA = TC/TB.
The concept of TGR is illustrated in Figure 7.3 on page 317 in which the corner frequencies
are 1 = 1 T B and 2 = 1 T C .
The classical design approach for determining the parameters T and are considered in
Section 2.12.1.5 and in texts on control system analysis [8], [9]. However, let us demonstrate
a somewhat different approach bearing in mind that the relevant performance requirements
of the generation and excitation system have to be satisfied for the generator on-line. In particular, the unit must operate stably at lower power levels before the PSS is switched on, say,
at 0.3 pu power. Furthermore, the requirements should apply for the appropriate range of
operating conditions, particularly at leading power factors, N-1 contingencies, etc.
A method for determining the parameters for TGR compensation and evaluating its performance is demonstrated in the following illustrative examples which consider both the online and off-line cases.
7.6.2
line
7.6.2.1

The performance of the generator and compensated excitation system onPreliminary off-line considerations

When the unit is off-line (i.e. with the generator main breaker open and the unit running
isolated from the power system) and under closed-loop voltage control the transfer function
of the generator is assumed to be G gen s = 1 1 + T d0 , T d0 = 5.0 s. The simple transfer
function of the exciter is G ex s = K E 1 + sT E , where KE = 1.0, TE = 0.1 s. (For the online analysis the generator parameters are listed in Appendix 7I.1.1.) For this unity feedback system, shown in Figure 7.5, the low frequency or DC gain of the forward loop is thus
KA when off-line. In order for the voltage regulation to be better than 0.5% when the unit
is off-line, the AVR gain KA = 250 pu is selected; this results in a steady-state error of
1 1 + 250 = 0.4 % - which is less than the specified value of 0.5%. (see Section 7.4).

322
7.6.2.2

Introduction to the tuning of AVRs

Ch. 7

On-line studies

The approach to the design of the TGR block is demonstrated for an excitation system with
an AVR gain initially set to lower values of KA without compensation. The initial study is
based on the generator on-line connected to the external system shown in Figure 7.4. A
study of the performance of the generator on-line without and with TGR compensation in
its AVR will follow.
For KA = 1 pu, TA = 0 and without compensation the Bode Plot V t V ref with the terminalvoltage feedback loop open is shown in Figure 7.6. If the forward-loop gain KA is increased
to 32 pu (~30 dB) without compensation, we note that the gain cross-over frequency c0
is 3.1 rad/s and the associated Phase Margin (PM) is 78 . With such a PM not only is the
system stable when the voltage feedback loop is closed, but the time response of the terminal voltage to a step change in reference voltage will be over-damped. (A PM > 60 for the
terminal-voltage feedback loop typically provides a well-damped response to disturbances
in that loop. However, the response associated with the lightly-damped rotor mode at about
9 rad/s (evident in Figure 7.6) may be superimposed on the well-damped terminal voltage
response to a step change in reference voltage.)
A transient gain KT = 32 pu will be adopted in this and other examples. As stated in
Section 7.4 it is suggested that the transient gain should be less than 25 pu for the latter time
constants. However, it will be established that the selected value of transient gain is satisfactory for this N-1 operating condition.
The proposed AVR gain is KA = 250 pu or 48 dB. Since the desired transient gain is KT = 32,
the TGR transfer-function must provide attenuation of KT/KA = 32/250 = 0.128 (-18 dB)
at frequencies less than that of gain-crossover frequency ( co = 3.1 rad/s). Furthermore,
the PM of 78 with KT = 32 pu should not be significantly reduced by the TGR transferfunction. As pointed out, such a value for the PM is likely to lead, under closed-loop voltage
control with TGR, to an over-damped terminal voltage response to a step change in reference voltage.
Let us locate the upper corner 1 T C of the TGR transfer function at a decade below co .
The effect of the corner at 1 T C = co 10 will cause the PM to be reduced by about 5 .
Therefore 1 T C = 0.31 rad/s, or T C = 3.125 s.

Sec. 7.6

Transient Gain Reduction Compensation

323

Magnitude (dB)

40
20
0
20
40
60
80
100
2
10

10

10

10

10

Phase (deg)

0
50
100
PM

150
200
2
10
KA=1

10
10
10
Frequency (rad/s)
KA=32

10

KA=250 + TGR

Figure 7.6 Unit on-line: Bode Plots of the open-loop terminal voltage for (i) no TGR
compensation, KA = 1 pu; (ii) no TGR, KA = 32 pu; (iii) with TGR, KA = 250 pu.
The attenuation provided by the TGR transfer function over the frequency range, from DC
to high frequencies ( T 1 ), is shown in Section 7.6.1 to be T C T B or 1 . The log-magnitude attenuation of the TGR transfer function is 18 dB, i.e. 20 log 10 = 18 dB, or = 8 .
The lower corner frequency of the TGR transfer function is then located at 1 T = 0.04
rad/s, i.e. T B = T = 25 s.
1 + s3.125
1 + s25.0

Thus, for KA = 250 pu, the TGR transfer function is ------------------------- . The Bode Plots for the case
KA = 250 pu (with TGR compensation), is shown in Figure 7.6.

The PM for the compensated case is 76 at 3.1 rad/s (a value close to that of 78 for the
uncompensated case with KA = 32 pu). Eigen-analysis conducted on the closed-loop system
(with outage of line a) reveals that the rotor mode of oscillation is 0.27 j8.83; the associated halving time is 2.55 s which satisfies the damping performance requirements. The
closed-loop time responses for perturbations in the terminal voltage, exciter voltage and field
current due to a step increase in the reference voltage of 0.01 pu (i.e. 1%) are shown in
Figure 7.7.

324

Introduction to the tuning of AVRs

0.95

20
15
10

0.9
0.85

25

(%)

If

(%)

1.05

Ch. 7

5
Ef

0.8
0.75
0

4
Time (s)
V

0
0

4
Time (s)
I

Figure 7.7 Unit on-line with closed-loop control of terminal voltage: perturbations in generator terminal voltage (Vt), field voltage (Ef) and current (If ) for a step change in reference
voltage from a steady-state value of 1.0 pu to 1.01 pu (1%).
Increasing the forward-loop gain, or increasing the time constant TC, could reduce the PM
from 76 to a lower value. This would improve the closed-loop terminal voltage response
from over-damped to well damped. However, there may be concern that the compensated
system lacks robustness, say, to 6 dB variation in the loop gain. For example, it may be
deduced from Figure 7.6 that a 6 dB increase in gain increases the gain-cross-over frequency
to about 6 rad/s. This would not only result in a poorly-damped terminal voltage response
to a step change in the reference-input but, due the proximity to the lightly-damped rotor
mode, a damped oscillation of frequency 8.8 rad/s would be superimposed on it. Bearing in
mind that the system is operating in a N-1 condition, the risk associated with the lack of robustness would need to be taken into consideration. It will be shown later it is necessary to
examine a range of encompassing N and N-1 conditions to establish the validity of the selected compensation.
As will be illustrated in Appendix 7I.5, it is desirable for the phase response shown in
Figure 7.6 to be flatter in the vicinity of the gain cross-over frequency to ensure a phase
margin of at least, say of 60 , for a 6 dB variation in the loop gain.
7.6.3 The performance of the generator and compensated excitation system offline
The performance of the generating unit when running off-line at rated speed and under
closed-loop voltage control is now considered.
Typically, the only relevant generator parameter when off-line is its open-circuit time
constant 1. As stated earlier, the generator is modelled in Figure 7.5 by the simple transfer
function 1 1 + sT d0 . The gain K A = 250 pu.

Sec. 7.6

Transient Gain Reduction Compensation

325

The open-loop Bode Plots, V t V ref , (i) with no compensation, and (ii) with the TGR parameters determined above for the on-line case, are shown in Figure 7.8. With no compensation, the phase margin (PM) of 26 at 21 rad/s suggests that under closed-loop voltage
control the responses to step changes in voltage will be lightly damped and oscillatory. With
compensation the phase margin PM is 60 at 5.5 rad/s, thus the closed-loop step response
is adequately damped. The time responses of terminal voltage to a 1% step change in reference voltage are shown in Figure 7.9.

Magnitude (dB)

50

50
2
10

10

10

10

10

Phase (deg)

0
50
100
150
200
2
10

10

10
10
Frequency (rad/s)

TGR Compensation

No

10

Compensation

Figure 7.8 Unit off-line, KA = 250 pu: Bode Plots of the open-loop terminal voltage
V t V ref with and without TGR
(with TGR, PM=60 at 5.5 rad/s; without TGR, PM=26 at 21 rad/s).

1. The effects of generator saturation have been ignored. Saturation may result in a reduction of loop gain, at 1 pu terminal voltage, in the simple model of the generator.
In some types of excitation systems, such as brushless, it may be necessary to model the
demagnetizing effects of the generator field current on the performance of the exciter
and the reduction in generated field voltage due to loading of the excitation system rectifier [12].
These effects are taken into account in Section 7.11.

326

Introduction to the tuning of AVRs

Ch. 7

1.5

1
0.75

(%)

1.25

0.5
0.25
0
0

0.5

1
Time (s)

TGR Comp.

1.5

Uncompensated

Figure 7.9 Unit off-line: Perturbation in terminal voltage on closed loop due to a step
change in reference voltage from a steady-state value of 1.0 pu to 1.01 pu (1%)
(with and without TGR compensation).
7.6.4 Comparison of performance of the excitation control system on- and off-line
The analysis of the performance of the generating unit on- and off-line is summarised in
Table 7.1.
Table 7.1

Performance of the unit on- and off-line with TGR compensation

Performance Measures

Phase Margin from Bode

Plots
KA=250 pu

Uncompensated

On-line
32 at
13.2 rad/s

Off-line
26 at
21 rad/s

TGR compensation

76 at
3.1 rad/s

Value

-0.27 j 8.8 a

Halving time (s)

Peak overshoot (% of
step size)
90% rise time (s)
2% settling time (s)
Steady-state error, (% of
step size)
Effective DC gain (pu)

2.6

none

9% at 0.55 s

0.6
4.3

0.3
0.8

1.5

0.4

67

250

60 at
5.5 rad/s
Performance under closed-loop voltage control with TGR Compensation (KA = 250 pu):
Rotor mode of oscillation

Voltage response to a step

change of 0 to 1% in Vref
Terminal voltage
regulation

Note a: Without compensation the on-line rotor mode is 0.054 j9.35 .

Sec. 7.7

PID compensation

327

The performance of the generating unit off-line satisfies the performance criteria. However,
it is significant that, for the selected operating condition with the unit on-line, the effective
steady-state (DC) gain is reduced from 250 to 67 pu. As a result, the steady-state error exceeds the requirements of the Rules. Furthermore, due to the lower gain, the response of
terminal voltage to a step change in voltage reference is somewhat sluggish (see Figure 7.7)
- although the settling time satisfies the specification of 5 s.
Ideally, it is desirable to adjust the AVR gain of the unit so that the effective gain on-line is
200 pu or more, i.e. by a factor of 200/67 = 3, so that the requirements of the Rules are satisfied. The AVR gain must therefore be increased significantly; however, this increase must
be attenuated at low frequencies by the same factor so that conditions in the vicinity of the
gain cross-over frequency in the Bode Plot remain unchanged. The high-gain solution may
be unacceptable. The resolution of this problem is (i) to review the design in the light of the
Rules and the relevant operating conditions, (ii) to examine alternative methods of compensation - for example, PID compensation which is considered next.

7.7 PID compensation

In considering the use of Proportional plus Integral plus Derivative (PID) Compensation it
should be emphasized that the approach adopted here is aimed specifically to its application
in the tuning of AVRs. The design of PID controllers for other applications, such as process
controls, is covered in texts such as [14], [15].
The structure of the PID Compensation for application to excitation control systems is
shown in Figure 7.10. The block K G 1 + sT G in the AVR is a simple representation of
other dynamic or control elements.

KI/s
Vref

KP

+
+

KDs
1+sTD

Figure 7.10

KG
1+ sTG

Vr

Ef
Gex

Vt
Ggen

Generator and excitation system

Block diagram of PID compensation.

When under closed-loop voltage control the purpose of the integrating block is to integrate
out any steady-state voltage error to zero by providing, in effect, an infinite steady-state
gain. The voltage regulation is therefore zero or, in the steady state, the pu terminal voltage
is equal to the pu reference voltage (see Section 2.10.1.2). This feature applies to both off-

328

Introduction to the tuning of AVRs

Ch. 7

and on-line conditions and therefore it is an attractive approach to compensation to ensure

the relevant requirements of the Rules are always satisfied. As in Transient Gain Reduction
a second aim for the compensator is to provide a relatively low transient gain, KT, say 25 to
50 pu, in the forward path over the relevant frequencies of the rotor modes of oscillation.
The derivative action of the derivative block (i) speeds up what otherwise might be a sluggish response of the unit to disturbances in terminal voltage, (ii) boosts the speed of response of the field voltage following the application and clearance of a major fault, and (iii)
limits the potentially high gain at high frequencies by means of the low-pass filter
1 1 + sT D . During the immediate post-fault transient the associated increase in field flux
linkages enhances the synchronizing power flow between the generator and the system thereby enhancing stability.
Procedures for the general-purpose tuning of PID compensation is covered fairly extensively in the literature. However, a systematic approach to the design of PID compensation for
application to the tuning of AVRs of synchronous generators is to be developed for which
the theoretical background to the method is outlined in the following section. Thereafter,
an example of the method is provided for comparison with the other methods of compensation described in other sections.
We will assume that the general form of the desired characteristics for PID compensation
in the application to AVR tuning is summarized in the straight-line frequency response plot
of the transfer function G c s in Figure 7.11.

1
Magnitude
(dB)

1
-------

TD

1
-------

K
-----2

TG

Frequency (rad/s)

Figure 7.11 Straight-line approximation of the magnitude response of the PID

transfer function G c s .

Sec. 7.7

PID compensation

329

7.7.1 PID Compensation: Theoretical Background

KI/s
+
KP
KDs
1+sTD

KG

+
+

1+sTG

Figure 7.12 Block diagram of the practical PID compensator

The block diagram of a practical form of the PID compensator is shown in Figure 7.12. Its
transfer function is:
KI
KG
sK D
G c s = ----- + K + ------------------- ------------------- .
s
1 + sT G
P 1 + sT D

(7.2)

It is convenient to consider a number of forms of this compensator both from the analysis
and practical applications points of view. We will describe the forms that are to be analysed
as PID Types 1 to 3 as follows:
Type 1:

T D = 0, K G = 1 , T G = 0. The PID transfer function is

G c1 s = K I s + K P + sK D

(7.3)

Type 2A and 2B: T D 0, K G 0 , T G = 0. The PID transfer function is

G c2 s = K I s + K P + sK D 1 + sT D K G .

(7.4)

Type 3:
T D 0, K G 0 , T G 0. The PID transfer function is described by (7.2)
with G c3 s = G c s .
While the forms of the transfer functions for Types 2A and 2B are identical, the locations
of the upper corner frequencies 2 and 1 T D with respect to the range of modal frequencies differ. This is explained in more detail in Section 7.7.1.2.
The analysis of the PID transfer function represented by (7.2) is somewhat complex. Initially
however, we can obtain useful insights and information from an analysis based on the simple
Type 1 PID transfer function of (7.3). We will return to Types 2 and 3 later.

330
7.7.1.1

Introduction to the tuning of AVRs

Ch. 7

Characteristics of Type 1 PID compensation

In Type 1 PID compensation it is assumed that the derivative time constant is zero,
T D = 0 . Three convenient forms of the simple Type 1 PID are:
2

KI
KD
KP KI
KD s + KP s + KI
2
G c1 s = ----- + K + sK D = ------- s + s ------- + ------- = ----------------------------------------- .
s
s
KD KD
s
P

(7.5)

The zeros of this transfer function s = 1 2 can be derived from the numerator of
(7.5) when it is expressed in the following form:
2

s + 1 s + 2 = s + s 1 + 2 + 1 2 = 0 .

Equating

the

coefficients

of

in

(7.6)

with

the

numerator

(7.6)
coefficients

s + K P K D s + K I K D in (7.5), we find that the corner frequencies are

KP
KP
4K D K I
2
1
1 2 = ----------- ----------- K P 4K D K I = ----------- 1 1 ----------------- , moreover,
2
2K D 2K D
2K D
KP

(7.7)

1 + 2 = K P K D and 1 2 = K I K D .

(7.8)

In the sequel it is helpful to express (7.5) in terms of the zeros s = 1 2 :

KD 2
G c1 s = ------- s + s 1 + 2 + 1 2 .
s

(7.9)

Note that if K P 4K I K D the corner frequencies reduce to

KI
KP
K D K I K P
1 ------- , and 2 ------- 1 ------------- ------- .
2
KP
KD
KP KD

(7.10)

If the corner frequencies are well-spaced apart, say 2 1 10 , we observe that the frequency 1 is that associated with the corner of the transfer function K I s + K P , likewise
2 with the corner of K P + sK D .

For the simple Type 1 model of the PID, the condition for the corner frequencies of (7.7)
to be real is K P 2 K D K I . Thus the value of the proportional gain, K Pb , at which the real
zeros evolve from complex values is
K Pb = 2 K D K I .

(7.11)

Sec. 7.7

PID compensation

331

Substitution of this value of K Pb in (7.7) yields the identical corner frequencies

K Pb
1 2 = ----------- . According to (7.8) when 1 = 2 we find
2K D
1 2 =

KI KD

K P = K Pb

rad/s,

(7.12)

at which frequency the zeros are real for increasing values of K P .

The frequency response of the transfer-function G c1 s is obtained by setting s = j f in
(7.5) or equivalently in (7.9):
KD
KI
2
G c1 j f = K P + j K D f ----- = K D 1 + 2 + j ------- f 1 2 .

(7.13)

For real values of the corner frequency the straight-line approximation of the magnitude response of the simple Type 1 PID transfer function (7.5) is shown in Figure 7.13.
Range of modal
frequencies

1
Magnitude
(dB)

2
KK P = -----1

K = KI

Frequency f
(rad/s)

Figure 7.13 Straight-line approximation of the magnitude response of the simple Type 1
PID transfer function G c1 s ( T D = 0 ).
The desired transient gain reduction of the compensator is ideally established over the range
of modal frequencies in the constant gain region of the straight-line response between the
corner frequencies 1 and 2 . At very low and at high frequencies the slope of the magnitude response is 20 and +20 dB/decade, respectively. In the actual magnitude response
the gain is a minimum, K min = K P , at the geometric mean of the two corner frequencies 1,
i.e. at a frequency of min =

1 2 .

Using the result of equation (7.8), the frequency at minimum gain is given by
min =

KI KD ,

(7.14)

332

Introduction to the tuning of AVRs

Ch. 7

and is independent of the value of the proportional gain, K P . Again we note the values of
1 2 from (7.12) and min from (7.14) are identical.

Let us assume K P is to be varied over a range of values consistent with an appropriate level
of transient gain reduction. Substitution of s = j min = j K I K D in (7.5) for G c1 s yields
a value of G c1 j min = K P . In other words, the minimum gain of the simple Type 1 PID over
the frequency range is equal to the setting of the proportional gain, KP. As has been noted - in
terms of the magnitude response of Figure 7.13 - the straight-line segment between 1 and
2 represents the gain KP (dB). The value of the proportional gain KP is thus close to the

value of the transient gain, KT, particularly when the corners are well-spaced apart.
The above analysis provides a procedure for AVR tuning when based on the simple Type 1
PID transfer function (7.5). The steps in the procedure are:
1.

Select a value for the desired minimum transient gain K P over the modal frequency
range. Note that, since this is the minimum value between the corner frequencies,
the effective gains closer to the corners will be somewhat higher.

2.

Select the value for the lower corner frequency 1 which is typically close to the
value K I K P (see (7.10)); hence deduce a value for K I = 1 K P .

3.

In order to achieve the desired transient gain reduction, set min to a value in the
vicinity of the geometric mean of the selected modal frequency range, say min = 3
rad/s for a range of modal frequencies, 1.0 to 10 rad/s. From (7.14)
min =

4.

K I K D ; hence estimate K D = K I min .

From (7.11) check that the value of K P selected ensures that the corner frequencies
assume real values. That is, ensure K P 2 K D K I .

5.

Evaluate the actual corner frequencies of the simple Type 1 PID transfer function
from (7.7), and plot the frequency responses.

6.

Analyse the dynamic performance of the generator and excitation system when offline and on-line over a range over steady-state operating conditions. Adjust the
2

1. Alternatively, for a transfer function of the form of (7.5), i.e. as + bs + c s , a simple

analysis reveals the magnitude of the transfer function is a minimum at min = c a ,
s = j . Furthermore, the phase angle traverses from 90 at low frequencies to + 90
at high frequencies and passes through zero phase at min . This result can also be determined easily from (7.13).

Sec. 7.7

PID compensation

333

parameters of the PID to satisfy the performance specifications for the generating
unit off- and on-line; repeat steps 1 to 6.
It should be noted that the gains, as determined and discussed in the context of this chapter,
are in per-unit on the relevant generator and exciter field voltage and current base values,
and generator stator base voltage. However, the gains of the AVR as identified in the actual
software (or hardware) of the excitation system may be on quite different and varied bases
depending on the manufacturers scaling system in their control design. As such, care needs
to be exercised in the field when translating gains to and from the actual settings in the controls and those which are used in simulation platforms such as that discussed here.
7.7.1.2

Characteristics of Type 2 PID compensation

An expanded form of the Type 2 PID transfer function of (7.4), in which

T D 0, K G 0 , T G = 0 , is
2

G c2 s = s K D + K P T D + s K P + K I T D + K I s 1 + sT D K G .

(7.15)

The zeros of this transfer function are derived from the numerator of (7.15):
KI
KP + KI TD
2
s + s ---------------------------- + ---------------------------- = 0 .
K D + K P T D K D + K P T D

(7.16)

This equation is of the form:

2

s + 1 s + 2 = s + s 1 + 2 + 1 2 = 0 .

(7.17)

Equating the coefficients of s in (7.16) and (7.17), and solving for 1 2 , we find:
2

KP + KI TD
K P + K I T D 4 K D + K P T D K I
1 2 = ------------------------------------ --------------------------------------------------------------------------------------2 KD + KP TD
2 KD + KP TD
4 K D + K P T D K I
KP + KI TD
= ------------------------------------ 1 1 -------------------------------------------
2
2 K D + K P T D
KP + KI TD

thus 1 + 2 = K P + K I T D K D + K P T D ; 1 2 = K I K D + K P T D .

(7.18)

(7.19)

From (7.18) it can be shown that the zeros of G c2 s , 1 2 , are real if

KP KI TD 2 KD KI .

(7.20)

An alternative form of (7.15) - which accommodates real corner frequencies 1 and 2 - is

1 + s 1 1 + s 2
G c2 s = K --------------------------- --------------------------- ,
1 + sT D
s

(7.21)

334

Introduction to the tuning of AVRs

Ch. 7

where K = K I K G . The method for deriving the straight-line approximation of the frequency response of G c2 j f based on (7.21) is outlined in Section 2.12 and texts on control
system analysis. This response is shown in Figure 7.14 for what will be called Type 2A PID
compensation in which
1 range of modal frequencies 2 1 T D .

Type 2B PID compensation, for which 1 2 range of modal frequencies , and

1 2 1 T D will be considered in Section 7.8.
Range of modal
frequencies

1
Magnitude
(dB)

1------

K
-----1

TD

Frequency f
(rad/s)

Figure 7.14 Straight-line approximation of the frequency response

of the Type 2A PID compensator transfer function G c2 s .
In a Type 2A PID the desired transient gain reduction of the compensator is ideally established over the range of modal frequencies in the constant gain region of the straight-line
response between the corner frequencies 1 and 2 . For the Type 2A PID (with K G = 1 )
the transient gain K T is assumed to set to the value of KP, the proportional gain for purposes
of design. At low and high frequencies the slope of the response on either side of these frequencies is 20 and +20 dB/decade, respectively. The effect of the corner 1 T D (> 2 )
results in the gain at high frequencies of the PID transfer function (7.15) tending to
K G K D T D + K P , as s .

(7.22)

This gain may be considered unacceptably high and appropriate adjustments to the parameters in (7.22) may be required.
The frequency at which the magnitude response of the PID transfer function (7.15) is a minimum involves some tedious analysis 1.

1. This involves finding the minimum of the magnitude of the PID transfer function (7.15)
with s = j f and making use of the Symbolic Maths Toolbox in MATLAB.

Sec. 7.7

PID compensation

335

For a non-zero value of T D the corner frequency 1 T D is typically a decade or more greater
than min . In this case the magnitude and phase contribution of the transfer function
1 1 + sT D to the frequency response at min is reduced by a factor of more than 0.995
( 0.043 dB) and by an angle less than 5.7 , respectively. As a result the minimum of the
magnitude of the practical PID transfer function occurs at a slightly lower frequency than
min based on (7.14).
7.7.1.3

Characteristics of Type 3 PID compensation

Based on (7.2) and (7.15), the Type 3 PID Compensator can be expressed as
2

s KD + KP TD + s KP + KI TD + KI
KG
G c3 s = ------------------------------------------------------------------------------------------- ------------------- ,
s 1 + sT D
1 + sT G

(7.23)

which, like (7.21), is of the form:

KG
1 + s 1 1 + s 2
G c3 s = K I --------------------------- --------------------------- ------------------- ,
1 + sT 1 + sT
s
D

(7.24)

where 1 and 2 are real values given by (7.18) and T D 0, K G 0 , T G 0 . Being three
transfer function blocks in series, the frequency response analysis G c3 s based on (7.24)
can be conducted as described in Sections 2.7 and 2.12, and in texts on control system analysis.
With T G 0 in (7.24), and 1 T G 1 T D , the low-pass filter 1 1 + sT G may be employed
to attenuate high frequency signals - such as noise - as illustrated in Figure 7.11 on page 328.
7.7.2 Tuning methodology for PID Compensation Types 1 and 2A
7.7.2.1

Tuning of simple Type 1 PID compensation

Let us adopt the following specifications as the starting point for the tuning of Type 1 PID
compensation based on (7.5).
1.

Over the range of local- and inter-area modal frequencies of 1.5 to 12 rad/s the
effective transient gain, KT, of the AVR / excitation system is to be 25 - 50 pu on
machine base (i.e. 28.0 - 34.0 dB); a value of 32 pu will be selected.

2.

The corner frequencies of PID transfer function are to lie outside the range of
modal frequencies of 1 to 10 rad/s.

3.

For a 0 to 1% step change at the voltage reference input of the closed-loop voltage
control system, off-line or on-line, (a) the terminal voltage overshoot should be less
than 7.5%, (b) the 90% rise time should be less than 1 s, and (c) the 2% settling time
of the terminal voltage should be less than 5s.

336
4.

Introduction to the tuning of AVRs

Ch. 7

The criterion for on-line system damping performance is that the real parts of all
rotor modes should be less than -0.139 Np/s (a halving time of 5 s).

Let us follow the steps in Section 7.7.1.1 for the determination of the parameters of the simple PID transfer function.
1.

Assume the proportional gain setting of the PID, K P , is 32 pu; let KG = 1.

2.

Assume the lower corner frequency, 1 K I K P , is 0.5 rad/s. The value of K I is

thus 16 pu/s.

3.

Based on the straight-line magnitude response in Figure 7.13, select the value of frequency (rad/s) at which the magnitude of the simple PID is to be a minimum. For
the range of modal frequencies, say, a value of min = 3.2 rad/s is close to the
geometric mean ( 10 ) of the specified modal frequency range. According to (7.14)
2

the ratio K I K D = min = 10.24 , thus selecting K I = 16 yields K D = 1.56 pu-s;

let KD = 1.5 pu-s (say) with the result that min = 3.27 rad/s.
4.

From (7.11), the value of the proportional gain K P at which the complex zeros of
the Type 1 PID transfer function assume a real value is K Pb = 2 K D K I = 9.80 pu.
We also know that the value of gain at min is the setting of K P , K P K Pb .

5.

With KP = 32 pu, KI = 16 pu/s and KD = 1.5 pu-s the corner frequencies of the
Type 1 PID are calculated from (7.7). The values are 1 = 0.51 and 2 = 20.8 rad/
s; both values are outside the specified range of modal frequencies.

It is instructive to derive the frequency response of the simple Type 1 PID compensator for
a range of values of KP from 0 to 44 pu, including the selected gain setting of 32 pu. The
responses are shown in Figure 7.15. For KP = 32 pu we note the value is almost constant
over the selected modal frequency range, that is, for all intents and purposes the value of the
transient gain KT is equal to the proportional gain KP .
The high-frequency gain (above the range of modal frequencies) may become excessive. To
limit the gain the time constant T D in the PID transfer function of (7.4) is set to a non-zero
value. This is considered in the tuning of Type 2A PID compensation in the following section.

Sec. 7.7

PID compensation

337

Magnitude (dB)

50
40
30
20
10
0
1
10

10

10

10

Phase (deg)

50

50

10

10
10
Frequency (rad/s)
KP=0
K =16
P
K =32
P

4
20
36

8
24
40

10
12
28
44

Figure 7.15 Frequency responses of the Type 1 PID for K I = 16 pu/s, K D = 1.5 pu-s,
T D = 0 s and values of K P between 0 and 44 pu; the dashed lines apply to complex zeros
of G c1 s .
7.7.2.2

Tuning of Type 2A PID compensation

An approach to the tuning of this compensation for the off- and on-line dynamic performance of the AVR is covered in some detail in Sections 7.7.2.2 to 7.7.2.4. Because a lightly
damped rotor mode arises in the on-line case, an exploratory tuning of a PSS for the generator and system is considered.
As foreshadowed in Section 7.7.1.2 and Figure 7.14 the low-pass filter 1 1 + sT D is introduced to the Type 2A PID transfer function to limit its high-frequency gain.
The same specifications as for the Type 1 PID are assumed for the tuning of Type 2A PID
compensation based on (7.4), however, an appropriate value of the time constant T D must
be determined. For the range of modal frequencies, 1.5 to 12 rad/s, a frequency of
min = 3.27 rad/s is nominally assumed at which the magnitude response of the Type 1
PID transfer function is a minimum; in fact, the minimum will be a lower value for the Type
2A PID.

338

Introduction to the tuning of AVRs

Ch. 7

Steps 1, 2 and 3 of the PID design procedure are the same as those listed in Section 7.7.2.1
for the Type 1 PID, which yielded K P = 32 pu, K I = 16 pu/s and K D = 1.5 pu-s. The corner associated with T D is selected to have a value greater than that of the upper corner frequency, 2 = 20.8 rad/s, calculated in step 5 for the Type 1 PID. Let us consider three
alternative values for TD, namely T D = 0.0125 , 0.025 and 0.0375 s; the associated corner frequencies of 80, 40 and 26.7 rad/s, respectively.
With T D 1 2 and KG = 1 the value of the proportional gain K Pb at which the real corners evolve from complex values is derived from (7.20), i.e. K Pb = K I T D 2 K D K I pu. For
the range of values of T D selected, the corner frequencies are real for values of the gain K P
greater than K Pb = 10.4 pu.
The corner frequencies 1 and 2 of the Type 2A PID are calculated from (7.18); the values
are given in Table 7.2 and are outside the specified range of modal frequencies, 1 to 10 rad/
s.The frequencies at which the magnitude responses of the Type 2A PID transfer function
(7.15) are a minimum, together with other relevant statistics, are also listed in Table 7.2.
The frequency responses of the PID for the selected values of T D are shown in Figure 7.16.
Table 7.2

TD
(s)
0
0.0125
0.0250
0.0375

Characteristics of Type 2A PID (7.15)

Corner Frequencies (rad/s)

At minimum of magnitude
response*:

1 TD

min
rad/s

Magnitude (pu)

Phase

0.51
0.51
0.51
0.51

20.8
16.4
13.6
11.6

80
40
26.7

3.27
2.92
2.73
2.58

32
32.2
32.3
32.5

0
-1.6
-3.0
-4.3

PID Parameters are K P =32 pu, K I =16 pu/s, K D =1.5 pu-s, K G =1 pu;
T D = 1 D is a variable parameter.
* Values read off frequency responses in Figure 7.16

High
frequency
gain
(pu) |(dB)
152 | 43.6
92 | 39.3
72 | 37.1

Sec. 7.7

PID compensation

339

Magnitude (dB)

65
60
55
50
45
40
35

Phase (deg)

30
1
10

10

10

10

50
0
50
1

10

10

10
10
Frequency (rad/s)

T =0
D
TD=0.025

Figure 7.16

10

10

T =0.0125
D
TD=0.0375

Frequency response of the Type 2A PID for a range of values of T D .

The PID parameters are K P = 32 pu, K I = 16 pu/s, K D = 1.5 pu-s, K G = 1 pu, T D .

7.7.2.3

Dynamic performance of the Type 2A PID with the generating unit off-line

The following studies are based on the SMIB system in which one line is out of service as
shown in Figure 7.4. The block diagram of PID compensation and generator are shown in
Figure 7.10 on page 327; the gain K G of the series block in the PID is set to unity and its
time constant T G is zero. The parameters of the sixth-order generator are listed in
Appendix 7I.1.1. When off-line, it is assumed - using first-order transfer functions - that
the relevant generator and excitation parameters are T do = 5.0 s and T E = 0.1 s, respectively.
The purpose of the following analysis is to establish whether the specifications for the offline performance are satisfied. Firstly, from the open-loop frequency response plot of the
PID plus the excitation system and generator, it is desirable to determine information on the
stability of the closed-loop voltage control system and the nature of its dynamic performance. The Bode plot of the open-loop system, V t V ref , is shown in Figure 7.17 for the
range of values of TD listed in Table 7.2.

340

Introduction to the tuning of AVRs

Ch. 7

Magnitude (dB)

50

50

100
1
10

10

10

10

10

Phase (deg)

100
120
140
160
180
1
10

10

T =0
D
TD=0.025

Figure 7.17

10
10
Frequency (rad/s)

10

T =0.0125
D
TD=0.0375

Unit off-line. Bode plot V t V ref of the open-loop voltage control system.

Type 2A PID parameters are K P = 32 pu, K I = 16 pu/s, K D = 1.5 pu-s, K G = 1 pu; T D s.

From Figure 7.17 it is observed that at the gain-crossover frequency of 5.9 rad/s the phase
margin is approximately 71 - 73 for all values of T D . The generating unit when off-line is
therefore stable under closed-loop voltage control and the response of terminal voltage to a
step change in reference voltage is well damped. This is observed in Figure 7.18 which reveals that the value of the time constant T D has little effect on the response.
Because the phase margin shown in Figure 7.17 for all values of T D is adequate, it may be
of interest to increase the gains of the PID by increasing the setting of K G , say, by factors
of two or three (6 and 9.5 dB respectively). By lowering the 0 dB axis in the Bode diagram
of Figure 7.17 by 6 or 9.5 dB the resulting phase margins are shown in Table 7.3 to lie between 69 and 56 .

Sec. 7.7

PID compensation

341

1.1

(%)

1.05

1
0.95
0.9
0.85
0

0.5

1.5
2
Time (s)

TD=0
TD=0.025

2.5

TD=0.0125
TD=0.0375

Figure 7.18 Unit off-line. Perturbations in terminal voltage due to a step change in reference voltage from a steady-state value of 1.0 pu to 1.01 pu (1%). The unit is under closedloop voltage control for the specified range of values of T D , K G = 1 .
(Note scale range on y-axis).
Table 7.3

Phase Margins of the off-line generator for a range of settings of K G and T D

KG

T D = 0. s

T D = 0.0125 s

T D = 0.025 s

T D = 0.0375 s

(pu)

PM

gco

PM

gco

PM

gco

PM

gco

1
2
3

73
69
68

5.7
9.9
13.4

72
67
65

5.7
10.2
14.1

72
65
60

5.8
10.5
14.6

71
62
56

5.9
10.8
14.9

PM: Phase Margin ( ).

gco : Gain cross-over frequency (rad/s)

For the purpose of analysis of the performance of the unit on-line in the next section, it is
desirable to choose an appropriate value of KG and T D . To ensure the closed-loop terminal
voltage response to a disturbance is well-damped, let us choose from the table a value of
phase margin better than (greater than) 65 . Moreover, with higher values of the gain KG
the transient gain and the high frequency gain of the PID may be too high in the particular
application. Let us restrict the evaluation of the closed-loop terminal voltage responses to
those for KG = 1 and 2 pu. To complement the responses shown in Figure 7.18 for KG = 1
pu, the closed-loop voltage responses to a step change in the reference voltage when the unit
is off-line is shown in Figure 7.19 for KG = 2 pu and the range of values of T D .

342

Introduction to the tuning of AVRs

Ch. 7

1.1

(%)

1.05

0.95
0.9
0.85
0

0.5

1.5
2
Time (s)

TD=0
TD=0.025

2.5

TD=0.0125
TD=0.0375

Figure 7.19 Unit off-line. Perturbations in terminal voltage from the initial steady-state
value due to a step change from1.0 to 1.01 pu (1%) in the reference voltage under closedloop voltage control for a range of values of T D , KG = 2 pu. (Note scale range on y-axis).
A comparison between the closed-loop step responses for generator off-line for the gains
KG = 1 and 2 pu is shown in Table 7.4 (step change of +0.01 pu (1%) in reference voltage).
Table 7.4

KG

Characteristics of closed-loop, off-line, 0-0.01 pu step responses

for K G = 1 and 2 pu

Rise Time:
90% of step
size (s)

Peak Overshoot
(% on step size)

Time to Peak
(s)

2% Settling
Time (s) ##

~0.29 s for all 5.6 to 4.6% for all values 0.63 to 0.60 s for all val- ~2.1 s for all
1 values of T
of T D *
ues of T D *
values of T D
D
~7.5% for all values
~0.16 s for all
except for T D =0.0375 s
2 values of T
D
the value is 8.6%

0.33 to 0.30 s for all val- < 0.62 s for

ues - except for
all values of
T D =0.0375 s the value is T
D
0.27s

* Note: First to last values in the range are T D = 0, 0.0125, 0.025, 0.0375 s
## 2% Settling-Time requirement is less than 5 s (see Section 7.7.2.1)

In this application, selecting a value of T D = 0.025 s ensures that the overshoot of the terminal voltage to a step in the reference is less than 7.5% of the step size for both gain K G = 1
and 2 pu; for higher values of T D the peak overshoot increases rapidly. Moreover, for
K G = 1 and 2 pu we note from Table 7.3 that the phase margin for T D = 0.025 s is better

Sec. 7.7

PID compensation

343

than or equal to the specified limiting value of 65 . We also note that the step responses are
markedly faster for K G = 2 than for K G = 1 pu.
7.7.2.4

Dynamic performance of the Type 2A PID with the generating unit on-line

Selecting TD = 0.025 s and gain KG = 1 and 2 pu, let us consider the stability of the generating unit on-line and under closed-loop voltage control.
The SMIB test power system is shown in Figure 7.4. Let us consider two operating conditions,

A line outage condition, Case W: generator output P = 0.4, Q = 0 pu (as in Figure 7.4
on page 320).

All lines in service, Design Case C: generator output P = 0.9, Q = 0.2 pu 1.

As in the previous AVR tuning method (Section 7.6), Case W is analysed initially. The Bode
plot of the open-loop transfer function V t V ref for the generating unit on-line is displayed
in Figure 7.20 together with the response when the PID is replaced by a simple gain element
having the same value as the PID proportional gain ( K P = 32 pu). Over the selected range
of rotor modal frequencies, 1 to 10 rad/s, the magnitude responses in the figure being in
close agreement reveals that the transient gains are practically identical.
At the gain cross-over frequency of 3.2 rad/s the Phase Margin of 79 suggests that for
KG = 1 pu the closed-loop step response V t V ref should be well-damped; however the resonance at ~9 rad/s in the Bode plot may result in a rotor oscillation of about that frequency
being superimposed on the terminal voltage response. By raising the magnitude plot of
Figure 7.20 by 6 dB - for KG = 2 pu - the Bode plot reveals that the Phase Margin is reduced
to 68 at a gain cross-over frequency of 6.6 rad/s. The closed-loop terminal voltage response should again be well-damped, but the oscillatory response of the rotor mode should
be accentuated as the gain cross-over frequency approaches the resonant frequency of ~9
rad/s. These results are demonstrated in the associated closed-loop step responses of
Figure 7.21.
Because of oscillatory nature of the responses, as revealed in the closed-loop performance,
it is clear that the rotor mode is lightly damped and that a PSS is required.

1. Case C is the Design Case in Section 5.11. With all lines in service the steady-state
power flow conditions for Case C are the same as those in Table 5.4. However, the generator and exciter parameters in Chapter 5 differ from those listed in the Appendix 7I.1.1.
Comparison of results between this and Chapter 5 may be misleading.

344

Introduction to the tuning of AVRs

Ch. 7

Magnitude (dB)

40
20
0
20
40
60
1
10

10

10

10

Phase (deg)

0
50
100
150
1

10

10
10
Frequency (rad/s)

PID: KP=32 KG=1 pu

10

No PID: KP=32 pu

Figure 7.20 Case W: Unit on-line. Bode plot of V t V ref .

PID parameters: K P = 32 pu, K I = 16 pu/s, K D = 1.5 pu-s, K G = 1 pu; T D = 0.025 s.
For No PID the AVR gain is 32 pu.

1.1

(%)

1.05

0.95
0.9
0.85
0

4
6
Time (s)

K =1
G

10

K =2
G

Figure 7.21 Case W: Unit on-line. Perturbations in terminal voltage due to a step change
in reference voltage from a steady-state value of 1.0 pu to 1.01 pu (1%). Generator is under
closed-loop control: gains KG = 1 and 2.0 pu. (Note scale range on y-axis.)

Sec. 7.7
7.7.2.5

PID compensation

345

Exploratory tuning of a PSS to improve the damping of the oscillatory mode

A PSS is typically switched into service at a lower value of real power output, say 0.1 to 0.3
pu. The procedure for tuning the PSS is the same as that outlined in Section 5.10.4. We will
assume that, for this system, Case C has also been established to be the PSS design case for
the set of generator and excitation system parameters listed in Appendix 7I.1.1.
The time-domain responses are faster for K G = 2.0 than for K G = 1.0 pu, however, for
K G = 2.0 pu the oscillatory response is less well damped. Because it may be necessary to select a value of K G such that 1 K G 2 pu, the P-Vr characteristic is derived for the design
case for the two values of K G . These characteristics, together with the synthesized characteristics for Case C, are shown in Figure 7.22.

Magnitude (dB)

20
10
0
10
20
30
1
10

10

10

10

Phase (deg)

0
50
100
150
200
1
10

10
10
Frequency (rad/s)

K =1: PVr
G
K =2: PVr
G

10

Synthesized PVr
Synthesized PVr

Figure 7.22 P-Vr characteristic for Case C for the set of PID parameters:
KP = 32 pu, KI = 16 pu/s, KD = 1.5 pu-s, TD = 0.025 s and KG = 1.0 or 2.0 pu.

The synthesized P-Vr agrees closely with actual P-Vr plots.

Including the low-pass and washout filters, the resulting PSS transfer functions based on
Case C are:
k
s5
1 + s0.357 1 + s0.0752
G PSS1 = ------------- --------------- ------------------------------------------------------------------- for KG = 1.0 pu, and
3.725 1 + s5 1 + s0.0067 1 + s0.0067

346

Introduction to the tuning of AVRs

k
s5
1 + s0.143 1 + s0.0794
G PSS2 = ------------- --------------- ------------------------------------------------------------------- for KG = 2.0 pu,
3.738 1 + s5 1 + s0.0067 1 + s0.0067

Ch. 7

(7.25)

respectively, with the damping gain k set to 20 pu on machine MVA rating 1.

With the PSS1 and PSS2 in service, the closed-loop step responses V t V ref are replotted in

1.2

1.2

1.1

1.1
Vt (%)

Vt (%)

Figure 7.23 for Cases C and W.

1
0.9

0.9

0.8

0.8
0

2
3
Time (s)

Case C

(a) KG=1.0 pu

Case W

2
3
Time (s)

Case C

Case W

(b) KG=2.0 pu

Figure 7.23 Cases C and W: Unit on-line. Perturbations in terminal voltage due to a step
change in reference voltage from a steady-state value of 1.0 pu to 1.01 pu (1%). Generator
is under closed-loop control with (a) PSS1, gain KG = 1 pu in service and (b) PSS2 for
KG = 2. PSS damping gain is 20 pu on machine MVA base. (Note scale on y axis.)
The time-domain performance and modal characteristics of the closed-loop AVR, excitation
system and generator on-line are summarized in Table 7.5 when KG is set to 1 and 2 pu.
7.7.2.6

Concluding remarks: PID Type 2A Compensation based on Cases C and W

Based solely on the limited set of operating conditions considered in the tuning of the PIDbased AVR it may be concluded that with a PSS in service the setting of the PID gain KG
should lie between 1 and 2 pu. The former and latter settings yield on-line terminal voltages
responses to a step change, shown in Figure 7.23(a) and (b), which may considered sluggish
and over-responsive, respectively. Depending on the actual performance specifications, further investigations should concentrate on values of KG, which lies in the range
1.3 K G 1.6 , say 2. The rotor mode is well damped with the PSS in service and satisfies

the damping criterion for rotor modes. Clearly there are a number of issues that should be
1. The low-pass filter time constants (6.7 ms) are very short. Such time constants should
typically be 3 or more times the cycle time of the PSS processor to reduce phase errors at
higher frequencies.

Sec. 7.8

Type 2B PID Compensation

347

reviewed in the analysis of both the on- and off-line cases before a final set of parameters
for the PID is chosen. Such issues are:
Table 7.5

Characteristics of closed-loop, on-line, step responses for KG = 1, 2 pu.

Time domain characteristics

Rotor modes

Case K G
No. pu

90%
Rise
Time
(s)

Peak
Time 2% SettOverto
ling Time PSS in service
shoot
Peak (s)
(s)
(%)

1.0

0.63 s

5.2%

1.45 s

5.3 s

2.0

0.41 s

13.4%

0.61 s

1.0

0.63 s

1.0%

2.0

0.40 s

9.2%

C
W

PSS out of
service

Mode shift

1.94 j8.90

0.05 j 9.08

1.88
j 0.18

3.5 s

1.48 j9.13

0.33 j 9.15

1.82
j 0.02

1.48 s

1.24 s

1.56 j8.72 0.24 j8.87 1.32

j 0.15

0.65 s

1.49 s

1.39 j8.78 0.11 j8.93 1.28

j 0.15

PID Parameters are K P =32 pu, K I =16 pu/s, K D =1.5 pu-s, K G ; T D =0.025 s
PSS damping gain is 20 pu on machine MVA rating.

In practice, to ensure that the specifications are satisfied it is necessary to conduct a

set of on-line studies for an encompassing range of normal and N-1 operating conditions.

What is the maximum acceptable value for the high-frequency gain (KP+KD/TD)KG
(e.g. see Table 7.2 for Type 2A PID)?

The choice of an acceptable range of transient gains ( K P K G ).

The example also reveals a basis for coordinating the tuning of the PID controls with the
tuning the PSS when the on-line responses do not meet the specifications for the rotor
modes.

7.8 Type 2B PID Compensation: Theory and Application to AVR

tuning
The theoretical basis and calculation of the parameters of the Type 2B PID are presented in
this section. However, because an application to the tuning of AVRs in a remote three-generator power station with brushless excitation systems is more detailed and complex it is
covered at the end of this chapter in Section 7.11.

2. KG should be less than 1.56 to satisfy the specified upper limit on the transient gain of
K P K G = 50 pu.

348

Introduction to the tuning of AVRs

Ch. 7

7.8.1 Tuning of Type 2B PID compensation

In contrast to the Type 2A PID, in the Type 2B PID the range of modal frequencies lies
above the corner 2 rad/s; this is illustrated in Figure 7.24.
Range of modal
frequencies

1
Magnitude
(dB)

1
-------

K
-----1

TD

Frequency f
(rad/s)

Figure 7.24 Straight-line approximation of the frequency response of the Type 2B PID
compensator transfer function G c2B s ; range of modal frequencies > 2 .
The use of the Type 2B PID compensator may occur in cases such as when the PID is required to contribute phase lead at low frequencies. In such a case the generator and exciter
time constants, T do and T ex , may be relatively long. Phase lead in the compensator is provided at the lower corner frequencies 1 and 2 ; integration ensures the steady-state error
between the reference and terminal voltages is integrated out.
Equations (7.15) to (7.22) are applied to Type 2B PID. The transient gain K TB , which ideally
applies over - or above - the range of modal frequencies is, from (7.22),
K TB = K G K D T D + K P , as s .

(7.26)

Note that K TB K P K G . Furthermore, the corner frequencies 1 and 2 in Figure 7.24 are
given by (7.19), i.e. 1 2 = K I K D + K P T D . Assuming 1 K I K P , the upper corner is
then
2 1 TD + KD KP ,

(7.27)

or K D = K P 1 2 T D .

(7.28)

Substituting (7.28) in (7.26) we find

K TB K P K G = D 2

where D = 1 T D .

Equations (7.28) and (7.29) form the basis for the Type 2B PID design.

(7.29)

Sec. 7.8

Type 2B PID Compensation

349

7.8.2 Example: Evaluation of Type 2B PID parameters.

This example demonstrates the application of the above results to the tuning of the AVRs
for a more difficult case of a remote, base-load, three-machine power station supplying energy via 132 kV lines to a high voltage grid. The purpose of this example - and its continuation in Section 7.11 - is

to illustrate the determination of the PID parameters which satisfy certain performance specifications over a wide range of N and N-1 operating conditions;

to examine a systematic and structured method for the selection of PID parameters
which are robust over the range of operating conditions;

to linearize the non-linear model of the brushless exciter and account for the variation
of its small-signal parameters with the steady-state operating conditions;

to establish the requirements for software for automating and expediting the calculations in the design process for application in practical cases.

In this application the tuning of the AVRs is more complex because the time constants T d0
and TE of the generator and brushless exciter are relatively long and the only tunable parameters in the AVR are those of the PID; the tuning is covered in some detail in Section 7.11.
However, for this application the calculation of the characteristics of a relevant set of candidate PID parameters are required and are therefore examined in the following section.
7.8.2.1

Frequency response characteristics of Type 2B PIDs

The range of modal frequencies is known to be 4 to 7 rad/s. Assume that over this frequency
range an effective value of K T 32 pu is required when KG = 1 pu. According to (7.26) and
Figure 7.24 the desired transient gain KTB must be somewhat higher, say, 40 - 70 pu; the values of KTB, 1, 2 and D are subject to the condition: 1 2 D .
The PID parameters are the calculated based on (7.28) and (7.29) using the following relationships:
T D = 1 D ; K P = 2 K TB D ; K D = K TB K P D ; K I = K P 1 .

(7.30)

For several sets of values for KTB, 1 , 2 and D the PID Type 2B parameters are derived using the above algorithm and are listed in Table 7.6. The associated frequency responses, which are shown in Figure 7.25, demonstrate the effect of modifying the
parameters in the vicinity of 1 rad/s.
It is noted from Figure 7.25 or Table 7.6:

Over the frequency range 4 to 7 rad/s the gain is close to the desired value of transient
gain, K T 32 pu.

350

Introduction to the tuning of AVRs

Ch. 7

Magnitude (dB)

60
50
40
30
20
2
10

10

10

10

10

40
Phase (deg)

20
0
20
40
60
80
2

10

10

10
10
Frequency (rad/s)

PID #1
PID #3

10

PID #2
PID #4

Figure 7.25 Frequency response plots of four Type 2B PIDs

with the parameters in Table 7.6.
Table 7.6

Calculated parameters for PID Type 2B

Selected quantities

Calculated PID parameters

Parameter
Set No.

KTB

1
(rad/s)

2
(rad/s)

D
(rad/s)

KP
pu

KI
(pu/s)

KD
(pu s)

1
2
3
4

50
70
70
70

0.30
0.50
0.65
1.0

1.2
1.4
2.2
1.9

5
7
11
9.5

12
14
14
14

3.6
7.0
9.1
14

7.60
8.00
5.09
5.89

TD
(s)
0.200
0.143
0.0909
0.1053

The minimum value of gain is about 23 dB (i.e. KP = 14 pu) at frequencies less than
2 .

With the higher values of the corner frequency 2 , Figure 7.25 shows that PID Sets
2, 3 and 4 provide additional phase lead in the range of modal frequencies 3-8 rad/s.
For PID type 2 compensation the transfer function 1 + s 2 1 + sT D in (7.21) is

Sec. 7.9

Proportional plus Integral Compensation

351

a lead-lag element, thus by increasing D with respect to 2 the peak phase-lead is

increased.
Although KTB, 1 , 2 and D are selected to calculate the PID parameters, equations
(7.30) can be rearranged to calculate the parameters based on some other choice, e.g. KP,
KTB, 1 and 2 , where KP<KTB.
The example of the detailed tuning of the AVRs in the remote three-generator power plant
is described in Section 7.11. Other methods of AVR tuning are considered in the following
Sections 7.9 and 7.10.

7.9 Proportional plus Integral Compensation

7.9.1 Simple PI Compensation
The structure of the Proportional plus Integral Compensation (PI) in the AVR is shown in
Figure 7.26. We shall refer to this as Simple PI Compensation.

KIP/s
Vref

KPP

Figure 7.26

KG

Gex

Ggen

Vt

Simple PI Compensation

The aim of the integral block is to integrate out any steady-state voltage error to zero by
providing, in effect, an infinite gain under steady-state conditions. The voltage regulation is
therefore zero or, in the steady state, the pu terminal voltage is equal to the pu reference voltage (see Section 2.10.1). As in Transient Gain Reduction a second aim for the compensator
is to provide a relatively low transient gain, K PP = K P , say 25 to 50 pu, in the forward path
over the frequencies of the rotor modes of oscillation. Let KG = 1.
The form of the simple PI compensator transfer function is:
G c s = K IP s + K PP = K IP 1 + T 1 s s , where T 1 = 1 1 = K PP K IP .

(7.31)

Let K PP = 32 pu be the desired transient gain, lets place the corner frequency
1 = K IP K PP about a decade below the lowest frequency mode, say 5 rad/s, and let
1 = 0.5 rad/s. The frequency response of the Simple PI Compensator with the resulting
integrator gain, K IP = K PP 1 = 16 pu/s, is shown in Figure 7.27. The compensator transfer function is thus:

352

Introduction to the tuning of AVRs

G PI s = 16 1 + 2s s .

Ch. 7
(7.32)

An alternative implementation of PI compensation using positive feedback is described in

Appendix 7I.3.

Magnitude (dB)

45
40
35
30
25
1
10

10

10

10

40
Phase (deg)

20
0
20
40
60
80
1

10

10
10
Frequency (rad/s)

10

Simple PI Compensation
PI with LeadLag TF

Figure 7.27 Frequency response plots of (i) Simple PI Compensation;

(ii) Simple PI Compensation with a series lead-lag transfer function block
7.9.2 Conversion to a PID Compensator with an additional lead-lag block
For a number of reasons it may be desirable to boost the gain of the AVR at higher frequencies, say, to improve the rate of response of the field voltage during the fault interval or in
the immediate post-fault period. This can be achieved by inserting a lead-lag block in series
with the PI block; the maximum phase lead of the lead-lag should occur at, or above, the
highest local-mode frequency. With this block it is also possible to improve the phase response of the compensator at the higher modal frequencies. In Section 2.12.1.4 and [8], [9]
the lead-lag compensator is described; a form of the transfer function is
G LD s = 1 + sT 2 1 + sT 2 , where the values for are less than unity.
For application with the simple PI compensator in Section 7.9.1, assume (i) the frequency
at which the maximum phase lead in the transfer function G LD j f occurs is m = 20
rad/s and (ii) the high frequency gain is to be boosted by a factor of two (6 dB). The former

Sec. 7.9

Proportional plus Integral Compensation

353

requires that m = 1 T 2 = 20 rad/s and the factor 1 = 2 ; hence the time constant of lead-lag block is T 2 = 0.07071 s. The frequency response of the Simple PI Compensator modified by the series lead-lag block is shown in Figure 7.27. This response is
similar to that shown for a Type 2A PID in Figure 7.16 for TD>0.
The PI with the series lead-lag compensator illustrated above can be converted to a PID
form. Letting T D = T 2 s, the transfer function of the PI plus lead-lag compensator is
2

K IP 1 + sT 1 1 + sT 2
K IP T 1 T 2 s + K IP T 1 + T 2 s + K IP
G c = -------------------------------- ---------------------- = ------------------------------------------------------------------------------------ .
s
1 + sT D
s 1 + sT D

(7.33)

This equation is identical in form to (7.15) for Type 2 PID with KG=1; let us equate coefficients of the powers of s in the numerators of (7.33) and (7.15). We find:
K I = K IP
K P = K I T 1 + T 2 T D for T 1 T 2 T D .

(7.34)

KD = KI T1 T2 KP TD

These equations are solved sequentially for KI, KP and KD.

In the above example the parameters for the PI of (7.31) and the series lead-lag compensator
of (7.33) are:
KIP = 16 pu-s, T1 = 2 s, T2 = 0.07071 s, TD = 0.03536 s.
By substitution of these values in (7.34) we find the equivalent PID parameters are:
KI = 16 pu-s, KP = 32.57 pu, KD = 1.111pu/s, TD = 0.03536 s.
The significance of the conversion to PID parameters is the following. It may be convenient
or simpler to determine the parameters of a PI and series lead-lag compensator rather than
to directly determine a set of PID parameters. If it is necessary to frame the compensation
in the form of a PID controller the conversion is readily calculated based on (7.34).
Note. There are further modifications possible to the frequency response of the Simple PI
Compensator to meet particular requirements. For example: The addition of a series firstorder low-pass block 1 1 + s0.01 would provide high-frequency roll-off of -20 dB/decade on magnitude at frequencies above 100 rad/s. However there would be some reduction
in the phase lead over the range of rotor modes e.g. about 5 to 6 degrees at 10 rad/s.

354

Introduction to the tuning of AVRs

Ch. 7

7.10 Rate feedback compensation

7.10.1 Method of analysis
The block diagram of the generator and excitation system with rate feedback (RFB) of the
exciter voltage are shown in Figure 7.28. For the determination of the rate feedback gain KF
and associated time constant TF of the feedback loop, both frequency response and root locus methods may be employed.
Compensation based on rate feedback employs either the output voltage of the exciter
(which is also the voltage input to the field of the generator) or a signal related to exciter
field current (if field voltage is inaccessible). As in the case of TGR tuning, the aim of the
compensation is to derive for the Excitation System (ES) a desired transient gain reduction
over the range of modal frequencies. This is achieved by appropriately determining the values of the feedback gain and time constant, KF and TF, respectively.
In the types of analyses of ESs considered, a simple low-order model of the ES is employed.
It will then be shown that the analyses can be extended to account for more complex dynamic systems.
The tuning of the ES shown in Figure 7.28 will be illustrated first. This will be followed by
an examination of the dynamic performance of the generator off-line when it is under
closed-loop voltage control with the tuned ES. Finally, the associated performance of the
generator when on-line is assessed over a range of operating conditions to ascertain if the
performance specifications are satisfied.
7.10.2 Tuning of the Excitation System (ES)
As evident from Figure 7.28 the ES comprises only the closed loop formed by the AVR, the
exciter and the rate-feedback path, the rate-feedback transfer function being:
G F s = sK F 1 + sT F .

(7.35)

Note that the block diagram in Figure 7.28 applies to the cases with the generator off-line
as well as on-line. In the latter case the block Ggen includes both the generator and the system
to which it is connected.
In the case of a brushless excitation system the output voltage of the exciter is not accessible
for measurement. In this case a signal proportional to the exciter field current, typically designated Vfe , is used as the signal for rate feedback. The exciter field current is closely related
to the field voltage of the main generator (i.e. Ef ). To cover the cases of the exciter output
voltage being available or unavailable for feedback, a generalised approach is adopted in
which a voltage V pu, not defined, is the feedback signal - as shown in Figure 7.29. For the
purposes of initially illustrating a procedure, a simple first-order system is assumed for the
forward loop where the gain K and the time constant T represent those of the AVR, or the

Sec. 7.10

Rate feedback compensation

355

AVR plus exciter, (referred to as the Plant). It will be demonstrated that this simple system
can then be modified to include additional dynamics in the forward path.

Excitation System
Vref

Verr

KA
1+sTA

AVR

Vr

Ef
Gex

Ggen

Vt

Exciter
KFs
1+sTF

Rate feedback

Figure 7.28 Generator and Excitation System with field-voltage feedback compensation.
The block Ggen accounts for the generator (and system) model when off-line (and on-line).
Excitation System
Plant
Vref

1+sT
KF s
1+sTF
Rate feedback

Figure 7.29

The Excitation System is shown as a simple closed-loop system with rate

feedback.

The closed-loop transfer function of the Excitation System shown in Figure 7.29 is
K
-------------- 1 + sT F
T
TF

V
--------- = -------------------------------------------------------------------------.
V ref
2 T + T F + KK F
1
s + ---------------------------------- s + -------------

T TF
T TF

(7.36)

This transfer function can be expressed as

1 + sT F
V
G CL s = --------- = K ------------------------------------------------------- , in which
V ref
1 + s 1 1 + s 2

(7.37)

356

Introduction to the tuning of AVRs

1 2 = 1 T T F and 1 + 2 = T + T F + KK F T T F

Ch. 7
(7.38)

The following analysis is mainly based on frequency response methods, however, the root
locus technique is used to demonstrate the fine-tuning of the rate feedback parameters
which is applicable to higher order excitation systems.
7.10.3 Rate feedback compensation using Frequency Response Methods.
The closed-loop transfer function (7.37) has one zero at s = F = 1 T F , associated
with the feedback path, and two poles at s = 1 2 . In order to obtain a more-or-less
constant transient gain reduction over the range of modal frequencies, say 1 to 10 rad/s, the
straight-line frequency response for the magnitude of the transfer function G CL j f
should have the form shown in Figure 7.30.

Magnitude (dB)

1
K

Range of modal
frequencies

2
KT

F = 1 TF

Frequency (rad/s)

Figure 7.30 Desired form of the straight-line magnitude response of the

transfer function G CL j f for the closed-loop, rate-feedback control system.
From the desired form of the transfer function shown in Figure 7.30, the corner frequency
2 should ideally be greater than, say, 10 F to cover the desired range of modal frequencies. Over that range the transient gain is KT. The lower corner 1 typically should be less
than F 10 (to minimise the effect of the corner on the modal frequency range). Assume
that the transfer function in (7.37) can be divided into two blocks in series, say,
1 + sT F
1
G c1 s = K --------------------------- and G c2 s = ---------------------- .
1 + s 2
1 + s 1

(7.39)

At frequencies greater than 1/TF, G c1 s K T = K T F 1 as s , where KT is the desired transient gain. Thus,
1 = KT K TF .

It is noted that G c2 j f 1 when f 2 10 . Moreover, from (7.38):

(7.40)

Sec. 7.10

Rate feedback compensation

1 2 = 1 T T F which, from (7.40), implies 2 = K K T T .

357
(7.41)

Note that the upper corner frequency 2 is independent of the rate feedback parameters,
KF and TF. Following substitution of (7.40) and (7.41) in (7.38), we find
KT
K KT
T F = ------ T + ---------------- K F .
K
K KT

(7.42)

If a value of the rate feedback gain, KF, is selected the associated value of the time constant
TF can be determined from (7.42), together with the lower corner frequency 1 from (7.40).
In this case the lower corner frequency cannot be specified.
However, according to (7.40), 1 = K T K T F , either 1 or TF can be selected, given values of K and KT. An approach based on the frequency response of Figure 7.30 suggests that
the corner frequency 1/TF is a more meaningful quantity to select than the rate gain KF.
Thus, given the value of TF, KF can then calculated from a rearranged form of (7.42), i.e.
KT K KT
1 K
K F = ---- ------ 1 T F 1 ------ T ---------------- T F if T is small .

K KT
K K KT

(7.43)

Note that, if T is small, (7.43) can be expressed as

KF K
TF
(7.44)
K T K 1 + --------------- = ------- , if K F K T F 1.

KF
TF
Thus, given the values of KF and TF, the ratio T F K F is an estimate of the upper limit on

the value of the transient gain KT.

An examination of the frequency response plot of (7.37) for G CL j f will reveal whether
the desired transient gain is more-or-less achieved over the range of modal frequencies.
In the above analysis a simple model has been employed for the ES. In practice, the models
of both the AVR and the exciter may be of higher order. In some cases it may be possible
to adapt the above analyses to satisfy such systems. However, in other cases it is possible to
represent the excitation control system by low-order models in order to determine initial set
of values of the rate feedback parameters, KF and TF. These parameters can then be finetuned using the more complex models and the associated frequency responses.
In (7.42) all quantities except KF and TF, the rate feedback parameters, are either known or
selectable. Using frequency response techniques an analysis of two cases will be considered:
Case 1 is based on the configuration in Figure 7.29; in Case 2 transient gain reduction or PI
compensation is included in the forward path of the AVR. Furthermore, it will been shown
that AVR and exciter transfer functions possessing additional dynamics can be incorporated

358

Introduction to the tuning of AVRs

Ch. 7

in the analyses. In all cases a check should be made that the limits imposed on the ES by
conditions such as (7.40) and (7.41) are valid
7.10.3.1

Case 1: No TGR or PI compensation in forward loop of the AVR

Three illustrative examples based on the rate feedback of the exciter output are described
for Case 1. In Example 1 below the parameters of the rate-feedback block are determined
subject to certain specifications. In Example 2 the effects of additional dynamics in the forward loop of the AVR are investigated. Finally, in Example 3 the significance of the ratefeedback parameters employed in the Sample Data for the AC2A model of the excitation
system [12] is assessed against those values calculated by the approach adopted in
Section 7.10.3.
7.10.3.1.1 Example 1, Case 1. Simple excitation system
Assuming rate feedback of the exciter output voltage E f , the parameters of the ES in Figures
7.28 and 7.29 are KA = K = 250 pu, and the exciter gain and time constant are KE = 1 and
TE = T = 0.1 s, respectively; the time constant TA is assumed negligible. The specification for
the transient gain is KT = 32 pu (30.1 3 dB) over the modal frequency range 1 to 10 rad/s.
The desired form of the straight-line frequency response of the magnitude of the closedloop excitation control system is shown in Figure 7.30. Referring to the latter figure, let us
assume the feedback time constant of is TF = 2 s. At the associated corner frequency
F = 0.5 rad/s we know, based on Section 2.12.1.3, that the transient gain KT is close to
30.1 dB +3 dB.
From (7.40), the lower corner frequency is 1 = K T K A T F = 0.064 rad/s. Similarly, based
on (7.41) the upper corner frequency is 2 = K A K T T E = 78.1 rad/s; at this frequency
the transient gain is close to 30.1 dB -3 dB. For TF = 2 s the rate-feedback gain KF = 0.0541
pu-s is calculated from (7.43). The resulting exciter frequency response is shown in
Figure 7.31 together with those for several smaller values of TF and associated values of the
gain KF. The closed-loop responses reveal that for values of TF > 1.2 s the transient gain requirements are satisfied.
The relevant characteristics of the responses are summarised in Table 7.7.

Sec. 7.10

Rate feedback compensation

Table 7.7

359

Characteristics of frequency responses for varying

rate-feedback time constant TF
F

1 (

(rad/s)

rad/s)

0.747

1.34

1.222

2.00

Ex. 1, Case 1
Study No.

TF (s)

(rad/s)

KF
(pu-s)

See Note 1.

0.171

78.13

0.020

34.4 dB 44.3

0.818

0.105

78.13

0.0329

32.3 dB 34.0

0.500

0.064

78.13

0.0541

31.1 dB 23.5

ES parameters: K = KA = 250 pu, KT = 32 pu, T = TE = 0.1 s.

Note 1. Magnitude and phase of the responses are calculated at 1.0 rad/s.

Magnitude (dB)

50
45
40
35
30
25
1
10

10

10

10

0
Phase (deg)

10
20
30
40
50
60
1
10

TF=0.75

Figure 7.31

10
10
Frequency (rad/s)
TF=1.22

10
TF=2.0

Example 1, Case 1: Closed-loop frequency responses V V ref

for the simplified rate-feedback excitation system model in Figure 7.29 for
increasing values of the time constant TF (s) in the rate-feedback block.
-------------------------------------------

360

Introduction to the tuning of AVRs

Ch. 7

7.10.3.1.2 Example 2, Case 1. Higher-order excitation system

It has been pointed out that, in practice, the models of both the AVR and the exciter may
be of higher order. Referring to Figures 7.28 and 7.29 let us assume the AVR and exciter
transfer functions, G A and G ex , possess additional dynamics and are of the forms:
1 + sT 2
KA
1
G A = ------------------ and G ex = ----------------------- ---------------------- , respectively,
1 + sT A
1 + sT E 1 + sT 1

where TA = 0.05 s, T1 = 0.02 s, T2 = 0.01 s. Assume that the rate feedback parameters are
those for Study 2 in Table 7.7, namely, TF = 1.22 s and KF = 0.033 pu-s, and K = KA = 250
pu, KT = 32 pu, and T = TE = 0.1 s. The above analysis is based on T = 0.1 s and, because
the time constants TA, T1 and T2 of the additional elements are shorter than that of the exciter, the corners associated with additional dynamics lie at higher frequencies than 1/TE.
The application of frequency response approach for Case 1 is therefore valid. The frequency
response with the additional faster time constants is shown in Figure 7.32, and is compared
with the frequency response if the faster dynamics are ignored.

Magnitude (dB)

50
40
30
20
10
1
10

10

10

10

Phase (deg)

0
50

100
150
1

10

10
10
Frequency (rad/s)

10

ECS with additional, faster dynamics

Basic Excitation Control System

Figure 7.32 Comparison of the closed-loop frequency responses of an excitation system

having additional, faster dynamics than that for the simplified system in Figure 7.29.
TF = 1.22 s and KF = 0.033 pu-s.
In this case, over the range of modal frequencies of interest, the additional faster dynamics
have little effect on the frequency response of the simplified excitation system which has

Sec. 7.10

Rate feedback compensation

361

been tuned ignoring the effects of faster dynamics. The exciter time constant of 0.1 s has
been employed in the example is long. Typically, in practice, the exciter time may be less that
0.05 s; such values cause the upper corner 2 = K A K T T E to increase in frequency. The
magnitude and phase responses, based on the frequency response for Case 1, therefore tend
to be flatter over a range beyond the 1 to 10 rad/s range of modal frequencies.
An aside: Say the exciter time constant is TE = 0.05 pu and in the forward loop there is a
first-order lag with a time constant T1 = 0.25 s (corner frequency 4 rad/s) which lies in the
modal frequency range of interest. In this event the pole at 4 Np/s can be cancelled - for
practical purposes - and the pole shifted to say, 25 Np/s.
------------------------------------------7.10.3.1.3 Example 3, Case 1: Application to excitation system model AC2A
The approach adopted in Example 1 can be adapted to excitation system (ES) models in
IEEE Standard 421.5 [12]. Type AC2A ES represents a field-controlled alternator-rectifier
exciter system (a brushless ES). A small-signal model of the AC2A is shown in Figure 7.33
in which, for present purposes, a simple model 1 K E + sT E is used for the exciter.
The following set of sample data for the Type AC2A model is provided in [12]:
KA = 400 pu, TA = 0.01 s, TB = TC = 0 s, KB = 25 pu, KH = 1 pu, KE = 1.0 pu, TE = 0.6 s,
KF = 0.03 pu-s, TF = 1.0 s.
By block diagram manipulation of Figure 7.33 with KE = 1.0 pu, it can be shown that
KB 1 + KH KB
E FD
-------------- = -------------------------------------------------------.
V A
1 + sT E 1 + K H K B

The above transfer functions reveals that the exciter time constant is reduced by a factor
1 1 + K H K B , and consequently the speed of response of the exciter is increased. The
AC2A model of the excitation system can then be expressed in the form shown in
Figure 7.29 with T = T E 1 + K H K B and K = K A K B 1 + K H K B , assuming TA = 0.
V ref

1+sTC

KA

V A

1+sTB

1+sTA

V F

KB

V R

1
sTE

KH

KFs

E FD

KE

V E

1+sTF

Figure 7.33 Small-signal model of the AC2A ES with a simplified exciter model (TA = 0)

362

Introduction to the tuning of AVRs

Ch. 7

In this example the forward loop of the ES, E FD V A , introduces a corner at 1/T = 43.3
rad/s lying above the selected range of modal frequencies 1 - 10 rad/s. The approach developed in Section 7.10.3 is thus valid since this corner lies above the upper end of the modal
frequency range.
Let us consider the following two studies:
1.

assume the desired transient gain is KT = 32 pu. Given TF = 1.0 s calculate the associated rate-feedback gains KF based on (7.43);

2.

calculate the effective transient gain KT associated with the data supplied in [12] for
the AC2A excitation system model.

In Study 2, equation (7.43) is solved for KT, i.e.

2

K T T K T KT + T F K + K K F + T F K = 0 .

(7.45)

The results are summarised in Studies 1 and 2 of Table 7.8.

Table 7.8

Comparison of the closed-loop parameters for AC2A [12] in Studies 1 & 2

Study
TF (s) KT (pu)
No.

2
KF (pu-s) Basis of calculation
(rad/s) (rad/s)

1.0

32.0

0.083

52.1

0.0286*

1.0

30.6*

0.080

54.46

0.030

Comment

Based on frequency KF calculated from

(7.43) with KT = 32
response analysis,
Case 1
pu
KT calculated from
Based on data set for
(7.45) with
exciter AC2A [12]
KF = 0.03 pu

* Calculated value, given the selected value of KT or KF.

1 , 1/TF and 2 are the corner frequencies of the closed-loop transfer function (7.37).

In Study 2 it is of interest to note that, for the AC2A ES with the rate-feedback time constant
TF = 1 s and gain KF = 0.03 pu, the calculated transient gain is KT = 30.6 pu. These values
are close to those in Study 1 when the transient gain of KT is set to 32 pu and the calculated
rate-feedback gain is KF = 0.0286 pu, i.e. the AC2A has an inherent transient gain close to
that which has been adopted in this chapter. This observation is confirmed in the frequency
response plots of Figure 7.34; moreover, the transient gain is more-or-less constant over the
selected modal frequency range, 1 to 10 rad/s.

Sec. 7.10

Rate feedback compensation

363

It is noted that the combination of KF = 0.03 pu-s and TF = 1.0 s is commonly used in the
sample data sets for a variety of ESs in IEEE Standard 421.5 [12]. According to (7.44), for
latter values the effective transient gain is K T T F K F = 33.3 pu.

Magnitude (dB)

60
50
40
30
20
10
2
10

10

10

10

10

Phase (deg)

0
20
40
60
80
2

10

10

10
10
Frequency (rad/s)

10

T =1.0 s, K =0.0286*, K =32.0 pu

F
F
T
TF=1.0 s, KF=0.0300, KT=30.6 pu*

Figure 7.34 Frequency responses V V ref of the closed-loop rate-feedback ES for Studies
1 and 2 of Table 7.8. * Calculated values when either KT or KF is the specified quantity.
------------------------------------------7.10.3.2

Case 2: Rate feedback with TGR or PI compensation in forward loop of the AVR

In Case 1, previously considered in Section 7.10.3.1, transient gain reduction or PI compensation is omitted from the forward loop.
TGR or PI compensation may be employed in conjunction with rate feedback of the exciter
voltage or AVR output (the Plant output) - as shown in general form in Figure 7.35(a).
The objective of the analysis is to derive a constant transient gain KT at frequencies in the
modal frequency range, say 1 to 10 rad/s. This implies that all corner frequencies in the
transfer function of the ES should lie outside of the latter range.

364

Introduction to the tuning of AVRs

Compensation

Vref

Plant
K
1+sT

TGR or
PI

Ch. 7

V
Vref

TGR or
PI

1+sT
TGR or
PI

KFs
1+sTF
Rate feedback

Compensation

(a)

KFs
1+sTF
Rate feedback

(b)

Figure 7.35 Compensation with rate feedback of the exciter output voltage Ef or the
equivalent exciter field current (Vfe )
(a) General form of the simple block diagram; (b) Equivalent form.
Based on the equivalent form of the block diagram in Figure 7.35(b), the following analysis
considers the behaviour of system in the lower and higher frequency ranges. The lower
range includes the corner frequencies of the TGR or PI compensation together with the corner frequency 1/TF of the rate feedback block. The higher frequencies range which exceeds
10 rad/s includes plant corner 1/T - and possibly additional higher corner frequencies We
require 1 T F 1 T - assuming 1 T F is the highest corner in the lower frequency range and 1 T 10 rad/s. Ideally, the magnitude of the transient gain KT should then be constant
over the range of modal frequencies.
Consider the limits as f , s = j f of the following transfer functions whose corners
lie in the lower frequency range:
TGR transfer function (7.1)
PI transfer function (7.31)
Rate feedback transfer function (7.35)
The plant transfer function when f 1 10T is

1 + sT C 1 + sT B T C T B ;
K PP + K IP s K PP ;
sK F 1 + sT F K F T F ; (7.46)
K 1 + j f T K .

The upper corner frequency of the TGR, 1 T C , and the corner frequency of the PI,
1 T PI = K IP K PP , are such that they are less than 1 T F .
Let KC = TC/TB or let KC = KPP (these are the high frequency gains of the TGR or PI transfer
functions in (7.46), respectively).
Under the condition that,

Sec. 7.10

Rate feedback compensation

365

f 1 T F 1 T C or f 1 T F 1 T PI ,

(7.47)

the transfer function for the systems shown in Figure 7.35 is

V
K
1
W j f = --------- = K C -------------------------------------------- ------------------------------------------------------------------------- .
V ref
1 + KK C K F T F 1 + j f T 1 + KK C K F T F

(7.48)

The gain of the transfer function (7.48) represents the transient gain KT over the range of
frequencies of interest:
K
1 + KK C K F T F

i.e., K T = K C -------------------------------------------- ,

or

KF
1
1
------- = ------ ----------- .
K T KK C
TF

(7.49)

Note the corner frequency of the transfer function W j f in (7.48) lies at a value greater
than 1/T. The application of these results is considered in Example 4.
7.10.3.2.1 Example 4, Case 2: Transient gain reduction or PI compensation with rate
feedback
The application of rate feedback with either TGR or PI compensation in a closed-loop control system is shown in Figure 7.35(a). The following parameters are provided for the compensation and the plant, i.e.:

TGR:

TC = 5 s, TB = 12.5 s, upper corner of TGR is 1/TC = 0.2 rad/s;

PI:

KPP = 0.4 pu, KIP = 0.08 pu/s, corner of PI is KIP / KPP = 0.2 rad/s.

Plant:

K = 250 pu, T = 0.05 s.

For both forms of compensation: transient gain KT = 32.0 pu. Assume TF = 2.0 s.

Condition (7.47) is valid for both types of compensator. Calculate the values of the feedback
gain KF for each compensation.
Based on (7.49), the rate feedback gains for the respective compensator types and for the
selected value of the feedback time constant are:
1

TGR: K F = ------ ---------------------- 2 = 0.0425 pu with K C = T C T B = 0.4 pu;

32 250 0.4
PI:

1
1
K F = ------ ----------------------- 2 = 0.0425 pu with K C = K PP = 0.4 pu.
32 250 0.4

For the system of Figure 7.35 the frequency responses for the cases of TGR and PI compensation are shown in Figure 7.36. For comparison, the response of the high-frequency

366

Introduction to the tuning of AVRs

Ch. 7

transfer-function model of the system, W j f (7.48), f 1 T F = 0.5 rad/s, is also displayed. All three magnitude responses satisfy the transient gain requirement of 30 dB (32 pu)
over the selected range of modal frequencies. Because the (upper) corners of the TGR and
PI compensation and the feedback time constant are the same in both cases, their phase responses are almost identical for frequencies greater than 0.5 rad/s ( = 1/TF). The upper corner frequencies of all three responses lie at 1 + KK C K F T F T = 63 rad/s, a value
greater than 1/T = 20 rad/s. For the purposes of comparison note that the parameters of
the TGR and PI compensation have been chosen such that the product K.KC in (7.49) is the
same in each case.
This example demonstrates that the relationships in (7.49) provide a basis for determining
the parameters for rate feedback analysis when coupled with other compensation functions.

Magnitude (dB)

60
50
40
30
20
2
10

10

10

10

10

Phase (deg)

0
20
40
60
80
2

10

10

10
10
Frequency (rad/s)

10

TGR
PI
High Frequency TF model (> 1/T )
F

Figure 7.36 Frequency responses for the system of Figure 7.35(a): TGR or PI compensation in the forward loop and, for comparison, the high-frequency transfer-function model,
W j f (7.48), valid for f 1 T F = 0.5 rad/s .
------------------------------------------7.10.3.3

Case 3. Effect of other or additional dynamics in the forward loop

The analysis of rate feedback compensation has been based on the excitation system (ES)
configurations of Figures 7.29 and 7.35, (i) without and with TGR or PI compensation and

Sec. 7.10

Rate feedback compensation

367

(ii) when the corner frequency of the first-order plant transfer function, K 1 + sT in the
ES, lies above the range of modal frequencies. It may be that the plant transfer function is
of higher order than one, and/or that the additional pole(s) lie in the range of modal frequencies, perhaps associated with higher-order models of linearized excitation systems.
Such additional dynamics in the forward loop can be accommodated in the analysis which
has been developed in Example 5, below.
7.10.3.3.1 Example 5, Case 3. TGR or PI compensation with rate feedback and additional
dynamics in the forward loop
Let us consider Example 4 which includes transient gain reduction or PI compensation with
rate feedback. Let us assume that there are additional blocks which introduce poles at 2.5 ,
20 and 50 Np/s in the plant. The pole at 2.5 is associated with a corner frequency of
2.5 rad/s. Unfortunately, this corner lies in the modal frequency range of 1 - 10 rad/s over
which a transient gain of 30 dB is required. Moreover, this pole being associated with a
plant parameter may vary somewhat with the plant loading between -2.3 and -2.6 Np/s.
An approach which is adopted for this scenario is to cancel the pole at 2.5 Np/s and
shift it to a higher frequency beyond the modal frequency range using the lead-lag transfer
function, say 1 + s0.42 1 + s0.042 , with corners at 2.38 and 23.8 rad/s 1. The corners
of the modified dynamics all lie above or at that of the plant corner frequency of 1/T = 20
rad/s and thus the identical design used in Case 2 above is employed, i.e. the parameters of
the TGR, PI and rate feedback parameters are the same as in Example 4. The transfer function of the block associated with the plant in Figure 7.35(a) therefore takes the form:
250
1
1 + 0.42s
1
--------------------------- ------------------------ ------------------------------ ------------------------------------------------------- .
1 + 0.05s 1 + 0.4s 1 + 0.042s 1 + 0.05s 1 + 0.02s

The frequency response of the closed-loop plant with TGR or PI compensation and rate
feedback, with and without the additional dynamics, is shown in Figure 7.37 on page 368.
Figure 7.37 reveals that the magnitude plots with and without additional dynamics agree
closely over the range 1-10 rad/s. However, the phase plots start to diverge only at 2 - 3 rad/
s and at 10 rad/s there is an additional phase lag of about 20 due to the additional dynamics. The performance of the closed-loop plant can be improved with further fine-tuning.

1. Note: there is not complete cancellation of the pole at -2.5 with the zero at -2.38. For a
disturbance to the system the magnitude of the response associated with the almost cancelled pole at -2.5 should be small.

Introduction to the tuning of AVRs

Magnitude (dB)

368

Ch. 7

60
50
40
30
20
10
0
10
2

10

10

10

10

10

0
Phase (deg)

50
100
150
200
250
2

10

TGR
TGR

10

10
10
Frequency (rad/s)

10

PI; with added dynamics

PI; without added dynamics

Figure 7.37 Frequency responses for the closed-loop plant of Figure 7.35 for TGR or
PI compensation in the forward loop and rate feedback. With additional dynamics
the poles lie at 2.5 , 20 , 23.8 , 50 , the zero at 2.38 Np/s.
------------------------------------------7.10.4 Rate feedback compensation using the Root Locus Method
What is the significance of the root locus method for the purpose of determining the ratefeedback parameters in addition to frequency response techniques? Firstly, if the AVR and/
or the exciter models are of higher order or differ from the form assumed in (7.36) / (7.37),
it may be possible to derive initial estimates only of the rate-feedback parameters KF and TF
from the analysis of Section 7.10.3. Secondly, the robustness or sensitivity of the damping
of the poorly-damped closed-loop poles to changes in a parameter value can be assessed
The use of a combination of the root locus method and frequency response techniques to
fine-tune the estimated parameters may then yield an acceptable set of parameter values.
The following is an unconventional application of the well-known root locus method described in [8] or [9]. The basis for this application of the method to determine how the
closed-loop poles of the off-line generating unit under closed-loop voltage control vary as a
parameter such as KF in the feedback path is varied from zero to infinity.
The block diagram of Figure 7.28 is manipulated in several steps into a form that is amenable to determination of the gain KF using the root locus method, i.e. the gain KF appears in

Sec. 7.10

Rate feedback compensation

369

the forward path of the open-loop system. The desired form of the open/closed loop system is that shown in Figure 7.38. In this form the generator and its voltage control loop become a feedback loop about the AVR and exciter. Note that the closed-loop poles of the
transfer functions V t V ref and E f V dum are identical, however the zeros in the two transfer
functions differ. As the terminal voltage reference V ref , shown in Figure 7.28 on page 355,
is not relevant to this scenario it is ignored in the root-locus analysis. This is now in the classical form of a closed-loop system for root locus analysis using Matlab, the gain k being
varied over the range 0 . (Note that for the unit on-line a root locus analysis can be conducted using a power system small-signal software package. A succession of eigen-analyses
is performed as the gain k in the block diagram of Figure 7.38 is varied over an appropriate
range).

Vdum
k

Figure 7.38
7.10.4.1

Ggen

Vt

Excitation System
KF s

1+sTC

KA

1+sTB

1+sTA

Rate feedback

TGR

1+sTF

AVR

Vr

Ef
Gex
Exciter

Block diagram of a generating unit off-line. An equivalent form of

Figure 7.28 for root-locus analysis with a variable gain k.

Example 6. Root Locus Method. Application to the generating unit off-line

For the analysis of the off-line performance of the generating unit the same parameters are
adopted as in Section 7.10.3.2.1 for TGR compensation with rate feedback of the exciter
voltage.
Generator and exciter parameters: T d0 = 5 s, KE = 1.0 pu, TE = 0.10 s;
AVR parameters: KA = 250.0 pu, TA = 0.05 s;
Transient gain reduction parameters: TC = 5.0 s, TB = 12.5 s;
Tuning of the feedback block parameters yielded values of KF = 0.0425 pu-s and TF = 2.0 s.
With reference to Figure 7.35, the exciter transfer function Gex(s) is included in the plant
transfer function; the feedback signal is the field-voltage perturbation. Because the condition (7.47), 1 T F f 1 T E 1 T A , applies to this scenario, the same values of KF and TF
apply to this example. Of interest is the effect on the damping of the closed-loop system of
changes in KF. This can be determined from the plot of the root loci shown in Figure 7.39
as the gain k in Figure 7.38 is varied 0 with KF set to 0.0425 pu-s. The associated
closed-loop poles for k = 1, i.e. k.KF = 0.0425, are marked on the plot and are all well
damped.

370

Introduction to the tuning of AVRs

Ch. 7

-13.0+j27.9, k=2
-10.6+j18.0, k=1
k=0, -1.0+j11.9
-8.8, k=1

k=0, -28.0

-0.74+j0.30, k=7
-3.9, k=2
-0.88, k=5.7
k=0,-0.50

Not to scale
open-loop zero
open-loop pole

k=0
-0.52, k=1
-0.54, k=2

closed-loop pole

Figure 7.39 Root locus plot for the exciter-voltage feedback gain k.KF, KF = 0.0425 pu-s
and TF = 2.0 s. Loci of the closed-loop poles start at the open-loop poles and terminate at
the finite or infinite zeros as gain k is varied 0 .
The closed-loop poles for an increase in the rate-feedback gain of 6 dB (k = 2) are also
marked on the locus plot; the dominant closed-loop pole at 0.52 Np/s (k = 1) is not significantly affected. Furthermore, it is observed that the damping of the dominant pole improves for an increase in gain of 5.7 times ( 15 dB) before the damping commences to
decrease (the damping constant changes from 0.52 Np/s to 0.88 Np/s). For a decrease
in gain, say to k = 1 5.7 = 0.18, the closed-loop pole lies just to the left of the open-loop
pole at 0.50 Np/s; the dominant pole of the off-line generator under closed-loop voltage
control is therefore robust to gain variations of 15 dB.
It is noted that the corner frequency 1/TE = 10 rad/s is a value at the upper end of the modal frequency range. Nevertheless, a frequency response plot of the ES shows that transient
gain is held constant at 30 dB from 1.5 to 20 rad/s.
The performance of the generator on-line under closed-loop voltage control with rate feedback compensation is very similar to those studied for TGR and PID compensation. The
similarity is a result of selecting the transient gain to be the same (KT = 32 pu) over the selected modal frequency, 1-10 rad/s, in all cases.
-------------------------------------------

Sec. 7.11

Tuning AVRs with type 2B PID compensation

371

7.11 Tuning of AVRs with Type 2B PID compensation in a threegenerator system

This section is a continuation of the example described in Section 7.8. Its purpose is

to demonstrate, in some detail, the more complex tuning of the AVRs;

to satisfy the dynamic and steady-state performance specifications over a wide range
of normal and line-outage operating conditions;

to analyse the performance

of a generator operating off-line at rated speed and under closed-loop voltage

control;

of one, two or three machines on-line at part and at rated real power output
for a range of reactive power generation;

to include models of the non-linear and the linearized brushless excitation system and
to determine the variation of parameters of the linearized model with operating conditions.

7.11.1 The three-generator, 132 kV power system

A power station containing three identical generators, each rated 50 MW 0.85 power factor,
is connected by double-circuit 132 kV lines to a high voltage system, represented by an infinite bus, as shown in Figure 7.40.
4

2
3

4 km

50 km

7
8

130 km

Infinite Bus

Figure 7.40 The three-generator, 132 kV power system

The parameters of the 5th order, salient-pole generator rated 58.8 MVA and its exciter are
listed in Appendix 7I.1.2. The exciter is an AC generator with a rotating rectifier and is represented by an AC8B Excitation System Model [12] shown in Appendix 7I.2, Figure 7.49.
Note that the following applies only to the analysis associated with the system shown in
Figure 7.40 and the associated AVRs with Type 2B PID compensation.
The PID parameters are to be determined assuming KG = 1, TG = 0.

372

Introduction to the tuning of AVRs

Ch. 7

Transmission line parameters in per unit /100 km on system base (100 MVA) are:
Z = 0.0632+j0.2347, b = 0.0484.
The load at bus 5 is 50 MW, 10 MVAr lag when it is on, zero when off. The most onerous
system contingency is the outage of the 130 km line a.
The transformer parameters are:
for each generator: Z = j0.20 pu on 100 MVA, tap range 10 %.
for the transformer at the Infinite Bus: Z = j0.05 pu on 100 MVA, tap range 10 %.
The range of operating conditions is summarised in Table 7.9.
For each set of study cases C*1 to C*5 and C*6 to C*10, the real power output of each generator is maintained constant for the five reactive power outputs between 25 Mvar lagging
to 20 Mvar leading. That is:

Output of each generator: 50 MW at: 25, 12.5 Mvar lag, 0 Mvar, and 10, 20 Mvar lead;

Output of each generator: 25 MW at: 25, 12.5 Mvar lag, 0 Mvar, and 10, 20 Mvar lead;

Number of generators on-line: one, two or three. Units are equally loaded; unequal
loadings are not considered in these studies.
Table 7.9
No.of
Units

Power
(MW)
*

Power system operating conditions

Cases

Load#

Line a

Cases

Load#

Line a

Mvar: 25 lag to
20 lead

in or
out

in or
out

Mvar: 25 lag to
20 lead

in or
out

in or
out

in
out
in
out
in
out
-

out
out
out
out
out
out
-

50
C01-C05
in
in
One
50
C11-C15
out
in
25
C21-C25
out
in
50
C41-C45
in
in
Two
50
C51-C55
out
in
25
C61-C65
out
in
50
C71-C75
in
in
50
C81-C85
out
in
Three
25
C91-C95
out
in
* Power output per generator
#
Load: 50 MW 10 Mvar.
Line a in or out of service.

C06-C10
C16-C20
C46-C50
C56-C60
C76-C80
C86-C90
-

The features of this generator-brushless-exciter and power system are: (i) the lines are long
with a surge impedance loading (SIL) of 45 MW; (ii) at rated output of the station the lines
are heavily loaded (about 1.7xSIL); (iii) with the outage of a line the loading on the second

Sec. 7.11

Tuning AVRs with type 2B PID compensation

373

circuit is about 3.4xSIL; (iv) the open-circuit time constants of the generator and exciter are
relatively long; (v) the only adjustable parameters in the AVR are those of the PID and the
gain KG; (vi) for planning purposes the halving time 1 of any rotor modes should be less than
5 s. An implication of items (iv) and (v) is that the PID must introduce adequate phase lead
at lower frequencies, i.e. about 1 to 4 rad/s. This not only ensures stability but also satisfies
a requirement that, for a small step-change in reference voltage, the settling time of the terminal voltage response to lie within a band of 10 % of its final value in less than 5 s when
the generator is on-line; when off-line the 10% settling time is 2.5 s.
To ensure that the tuning of the PID covers a range of operating conditions, the 75 generating/operating conditions shown in Table 7.9 are examined. However, certain system conditions are not credible because 132 kV bus voltages are outside the range of 95-108%, or
taps are at their limiting positions; several cases - such as C71 and C90 at maximum lagging
or leading reactive power output - are therefore discarded. The terminal voltage of each generator is maintained at 1 pu.
7.11.2 The frequency response characteristics of the brushless exciter and generator
The closed-loop terminal voltage control system of each generator is shown in the block diagram in Figure 7.41; note that the generator model accounts for the effects of the external
system when the unit is on-line. Because the models of both the generator and the exciter
are non-linear, the parameters of the linearized model will change with conditions at the generator terminals. In order to establish suitable parameters for the excitation control system
it is necessary to determine the variation of the generator-exciter characteristics with terminal conditions.
Voltage transducer
Vtrn
Vref

Gtrn
Vpid

Verr
Gpid

Vexf
KG

AVR

KAE Gex
Exciter

Vt

Ef

Ggen
Generator/
external system

Figure 7.41 Terminal voltage control system. The gain KAE accounts for the per unit system of the excitation control system which includes a brushless exciter.
Over a range of terminal conditions such a characteristic is the frequency response of the
generator-exciter transfer function, as measured between the exciter field voltage as input
1. See definition in Section 10.2.2.

374

Introduction to the tuning of AVRs

Ch. 7

and the transducer voltage output, V trn j f V exf j f . The use of the latter transfer function is particularly pertinent to brushless excitation systems in which the exciter output voltage is not accessible for measurement.
A 5th order model of the generator and a non-linear exciter model are available in a smallsignal power system dynamic performance package. Such models are automatically linearized at each operating condition by the software. The non-linear and linearized exciter
model are shown in Figure 7.49 and 7.50 of Appendix 7I.2.
Based on (i) the power system of Figure 7.40, (ii) the system and device parameters given in
Section 7.11.1, the frequency responses of relevant blocks in the voltage control loop are
calculated for selected operating conditions. The set of frequency responses for the generator and exciter, Vtrn/Vexf , are shown in Table 7.42 when either one, two or three generators
are on-line; the output of a generator is 25 or 50 MW at 1 pu terminal voltage.
For normal and N-1 operation of this system, operation at lagging power factors is more
likely to occur. The selection of the PID parameters may be influenced accordingly.
It is noted from Figure 7.42 that, for feasible cases C01 to C95 the gain in the generator/
exciter frequency responses in the region of 1.0 rad/s varies within 6 dB, and the phase
varies by about 25 . The variations in the responses over the frequency range are due not
only to the range of steady-state conditions at the generator terminals but also to the associated parameter values in the small-signal model of the exciter. An example of the exciter
parameters and the steady-state field voltage is illustrated in Table 7.10 of Appendix 7I.2
for operating conditions C16 to C20 in which a single machine is on-line.
The significance of the phase variation is the following. Let us assume that when the PID is
added to the forward loop the gain-cross-over frequency of the Bode plot of
V trn j f V ref j f occurs at 1 rad/s. The gain variation in the generator/exciter frequency
responses at 1 rad/s is small but the phase variation remains at about 25 . This will result
in a similar variation in the phase margin over the range of operating conditions with implications for both stability and transient response to a step change in reference voltage. The
Bode plots suggest that, when the units are under closed-loop voltage control, the greater
phase lags in the leading power factor cases (i) are not conducive to stability, and (ii) result
in the terminal voltage response to step changes in reference voltage being less-well or poorly damped.
To determine an appropriate set of PID parameters for the range of operating conditions
let us base the analysis on a condition in the middle of the band of phase variations, say Case
C17, in which the output of a single generator is 50MW, 12.5 Mvar lagging; the line a is out
of service and the load is disconnected.

Tuning AVRs with type 2B PID compensation

10

10

Magnitude (dB)

Magnitude (dB)

Sec. 7.11

20
30
40
50
0

20
30
40
50
60
1
10

10

10

50

50

Phase (deg)

Phase (deg)

60
1
10

100
150
200
1
10
C01
C11
C21

C05
C15
C25

10

C06
C16
C17

C42
C51
C61

C10
C20

10
Frequency (rad/s)
C44
C55
C65

10

C46
C56
C17

C50
C60

(b)

Magnitude (dB)

10

150

(a)

(a) One unit, (b) Two units, (c) Three units.

10

Solid lines: Maximum lagging reactive power

Dashed lines: Maximum leading reactive power

20
30
40

Real power output is 50 MW in all cases except for

C21-25, C61-65 and C92-95 when it is 25 MW.

50
60
1
10

10

10

0
50

Phase (deg)

10

100

200
1
10

10
Frequency (rad/s)

375

100
150
200
250
1
10
C72
C82
C92

10
Frequency (rad/s)
C75
C84
C95

C77
C86
C17

10
C79
C88

Cases C76, C80, C81, and others are omitted because

operating constraints are infringed and are infeasible.
For some other cases the reactive power output per
generator is reduced, e.g. from 10 to 5 Mvar for
Case 79.
Case C17 is adopted as the Base Case and is shown in
all three sets of plots.

(c)

Figure 7.42 Envelopes of frequency responses between the generator terminal voltage
transducer and the exciter field voltage, Vtrn/Vexf , for the feasible range of operating
conditions shown in Table 7.9
7.11.2.1

Calculation of the PID Type 2B parameters

In order to establish a basis for the compensation to be provided by the PID, let us consider
for Case C17 the frequency responses of the generator and exciter, V trn j f V exf j f , and

376

Introduction to the tuning of AVRs

Ch. 7

the AVR, V exf V err = PID j f . The PID parameters selected for trial are those for set
No. 2 in Table 7.6 on page 350. These two responses are shown in Figure 7.43 together with
the phase response of the open-loop transfer function V trn V ref .

Magnitude (dB)

40
30
20
10
0
10
20
2

10

10

10

10

50
Phase (deg)

0
50
100
150
200
2
10

10
10
Frequency (rad/s)

10

ExciterGen: Vtrn/Vexf
PID #2:
Vexf/Verr
Openloop TF: V /V , PM = 91 deg
trn

ref

Figure 7.43 Case C17 (one unit): Frequency responses of the component transfer
functions in the open-loop system including the PID parameter Set No. 2
(see Table 7.6 on page 350),
i.e. KP = 14 pu, KI = 7.0 pu/s, KD = 8.0 pu-s, TD = 0.143 s, KG = 1.0.
Gain cross-over frequency of the open-loop transfer function is 1.51 rad/s.
From the open-loop transfer function in Figure 7.43 it is noted that (i) the gain-cross-over
frequency occurs in the range 0.7 - 2.5 rad/s for which the possible variations in the loop
gain lie in the range 4.5 dB; (ii) the phase margin varies from 103 to 75 over the same
frequency range.
In determining appropriate PID parameters the phase margin should be more-or-less constant about the gain-cross-over frequency for robustness to gain variations. Selecting a phase
margin of 65 , say, ensures the closed-loop response of terminal voltage to a step change in
reference voltage is not significantly over-damped (for large values of the phase margin) or
under-damped (for small values of the phase margin). In the case of higher values of loop
gain associated with the gain-cross-over frequency exceeding 3 rad/s we note that the

Sec. 7.11

Tuning AVRs with type 2B PID compensation

377

closed-loop step response of terminal voltage is likely to contain a damped oscillatory component due to the electro-mechanical modal resonance at 5 to 6 rad/s.
We will therefore examine an approach to derive a more of less constant phase margin of
65 over an appropriate frequency range for Case 17. The phase matching method which
achieves this objective is described in Appendix 7I.5 in which it is shown in Figure 7.55
that the parameter set No. 4 for PID Type 2B in Table 7.6 provides the required phase margin.
The significance of the analysis of the phase margin for the Base Case 17 is revealed in its
effect on the terminal voltage response of the closed-loop system due to a step change in
reference voltage for the full set of operating conditions. As shown in Figure 7.44 the response of the system incorporating PID Set No. 2 (phase margin 96 ) is well damped. However, with PID set 4 (phase margin 66 ) a satisfactory, suitably-damped response results.
Moreover, the settling-time requirement that the response lies within 10% of its final value
within 5 s is satisfied with both PID Sets 2 and 4.

Vt (%)

1
0.8
0.6
0.4
0.2
0
0

4
6
Time (s)
PID #2

10

PID #4

Figure 7.44 Case C17. Single unit only on-line. Perturbations in terminal voltage (Vt) due
to a step change in reference voltage from a steady-state value of 1.0 pu to 1.01 pu (1%).
PID parameter Sets 2 and 4, Table 7.6. The 10 % band about the final value is also shown.
The dynamic performance of the single generator with PID Set No. 4, parameter values
KP = 14 pu, KI = 14 pu/s, KD = 5.89 pu-s and TD = 0.1053 s, appears satisfactory. The ap-

plication of this PID set to all the feasible operating cases and conditions for a generator offline and one, two and three units on-line is now examined.
7.11.2.2

Dynamic performance of a unit off-line under closed-loop terminal voltage control

When the generator is operating off-line at rated speed and under closed-loop voltage control it is required to satisfy the relevant performance specifications. For example, such specifications may require that the measured terminal voltage settles within 10% of the final
value in less than 5 s for a step change of 1% in the terminal voltage (see Section 7.4). in the

378

Introduction to the tuning of AVRs

Ch. 7

analysis.When the generator is off-line in the following analysis the 10% settling time in terminal voltage for a 1% change in reference voltage is 2.5 s.
In previous examples simple models of the exciter and the generator have been employed,
i.e. 1 K E + sT E and 1 1 + sT d0 . However, (i) at 1 pu voltage the small-signal gain of
the generator is determined by the slope of the saturation curve and is less than unity; (ii)
the perturbations in generator field current modulates the generator field voltage by two
mechanisms represented in the model of the exciter in Appendix 7I.2, Figure 7.50. The
mechanisms are (i) the effect of the demagnetization term, KDE, and (ii) the non-linear reduction in rectifier average output voltage with increase in the rectifier load, i.e the generator
field current. The latter mechanism is represented by the value of the gain KCE and the associated mode of operation of the rectifier.
The linearized model of the off-line, fifth-order salient-pole generator and the exciter are
formed automatically. The off-line unit operates a rated voltage and speed. The other elements in the voltage control loop are PID Set No. 4 (see Appendix 7I.5), the per unitizing
gain KAE and the terminal voltage transducer, time constant Ttrn . The generator and exciter
parameters are given in Appendix 7I.1.2. The Bode Plot of the open voltage-control loop,
V trn j f V ref j f , and the associated closed-loop response in generator terminal voltage
due to a +1% step in the reference voltage, are shown respectively in (a) and (b) of
Figure 7.45.
The closed-loop step response is adequately damped, as predicted by the Bode plot, and satisfies the performance specification.
7.11.2.3 Dynamic performance over a range of operating conditions; one, two and three units
on-line based on PID parameter Set No. 4.

The open-loop frequency responses for one, two and three units on-line are examined to
derive information on both the damping of the voltage control loop and the stability of the
power system under closed-loop conditions. The margins of rotor angle stability under
closed-loop conditions are also examined, assuming for planning purposes a 5 s halving time
for the dominant mode. Finally, the closed-loop responses of the generator terminal voltage
to step changes in its reference voltage are assessed to determine if the requirement that the
response lies within 10% of its final value within 5 s is satisfied over the range of operating
conditions.

Sec. 7.11

Tuning AVRs with type 2B PID compensation

379

20
0
20
40
60
80
1
10

10

10

10

90

Terminal Voltage (%)

Magnitude (dB)

40
1.2
1
0.8
0.6
0.4
0.2
0
Phase (deg)

0
135
180

4
6
Time (s)

10

(b) Perturbation in generator terminal voltage due to a

step change in reference voltage from a steady-state
value of 1.0 pu to 1.01 pu (1%).
10% Settling time < 2.5 s.

225
270
1
10

10
10
Frequency (rad/s)

10

(a) Bode Plot: Phase Margin 60 deg at 2.2 rad/s.

Figure 7.45 Generator off-line, operating at rated speed under terminal voltage control
with PID Set No. 4. (a) Open-loop Bode Plot. (b) Perturbation in closed-loop terminal voltage step-response.
7.11.2.3.1 Open-loop frequency responses; one, two and three units on-line.
In the following Bode plots for generator #1 the terminal voltage feedback path is open on
that generator, but is closed on the other generators when more than one unit is on-line. The
Bode plots are shown in Figure 7.46 for the cases when one, two or three generators are online. When all machines are under closed-loop voltage control these open-loop plots should
reveal the nature of (i) the damping in the voltage control loop on generator #1, (ii) the stability of the system, and (iii) the terminal voltage response of generator #1 to a step in its
reference voltage. One can equally well apply the above analysis to unit #2 or #3 instead.
Because the Phase Margins derived from the Bode plots in Figure 7.46 are all positive the
system is stable over the range of operating conditions. However, the Phase Margins are
much less than the desired value of 65 at higher values of leading reactive power output,
e.g. for C20 the PM is 44 at 1.26 rad/s. Thus under leading power factor operation and
closed-loop voltage control the system damping is degraded. However, in the cases of one,
two or three generators on-line at rated real power output it should be noted that the higher
values of leading reactive power output are unlikely to arise in practice. In such cases the reactive power import to the system at the infinite bus is somewhat greater than that absorbed
by the generator. For example, in case C55 the output of two units is 100 MW -40 Mvar and
the reactive import from the infinite bus is 64 Mvar. Similarly in case C15 for one generator,
output 50 MW -20 Mvar, 15 Mvar is imported from the system. Such reactive flows from
the real power sink to the real power source are unwarranted and uneconomic - especially

380

Introduction to the tuning of AVRs

Ch. 7

60

60

40

40

Magnitude (dB)

Magnitude (dB)

for the outage of line a. Thermal limits of transmission lines and transformers, which may
be relevant under outage conditions, have been ignored. Clearly, under leading power factor
operation the reactive power absorbed by the generators must be limited. Such limits would
need to be determined by further studies.

20
0
20
40
60
1

10

10

60

10

80

80
100

120
140
160
180
1

10

10
10
Frequency (rad/s)

C01
C11
C21

C05
C15
C25

C06
C16
C17

10

(a)

10

10

10

10

120
140
160
180
200
2
10

10
C42
C51
C61

C10
C20

10
10
Frequency (rad/s)
C44
C55
C65

C46
C56
C17

10
C50
C60

(b)

(a) One unit, (b) Two units, (c) Three units.

PID Set No.4 installed on all generators

40
0

Solid lines: Maximum lagging reactive power

output.
Dashed lines: Maximum leading reactive power
output.

40
80
2
10

10

10

10

10

Unit real power output is 50 MW in all cases except for C21-25, C61-65 and C92-95 when it is 25
MW.

80

Phase (deg)

40
80
2
10

10

0
20

100

200
2
10

Magnitude (dB)

Phase (deg)

Phase (deg)

80
2
10

20

120

160

200
2
10

10
C72
C82
C92

10
10
Frequency (rad/s)
C75
C84
C95

C77
C86
C17

10

(c)

C79
C88

Cases C76, C80, C81, and others are omitted because operating constraints are infringed and are
infeasible. For some other cases the reactive power output per generator is reduced, e.g. from 10
to 5 Mvar for Case 79.
Case C17 is adopted as the Base Case and is
shown in all three sets of plots.

Figure 7.46 All feasible Cases, C01 to C95: Bode plots V trn j f V ref j f of the
terminal voltage when the feedback path is open on generator #1 with PID Set No. 4.

Sec. 7.11

Tuning AVRs with type 2B PID compensation

381

Imaginary Part (rad/s)

7.11.2.3.2 System eigenvalues when generators are under closed-loop voltage control
While the Phase Margins derived from the Bode plots show that the system is stable, an examination of the eigenvalues for the rotor modes reveals the degree of stability of these
modes. The most onerous conditions most likely to yield rotor angle instability are the cases
for which line a in Figure 7.40 is out of service and the load at bus 5 is off, i.e. the rated
output of the station is carried over line b. The local and inter-machine modes for one or
more units on-line are seen in Figure 7.47.
8
6
4
2

DCC
Local Mode
A - Maximum lagging
B - Maximum leading
Inter-machine Mode
C - Maximum lagging
D - Maximum leading

0
1.5

One unit

BA
BA
BA
Halving time
=5 s

1
0.5
Real Part (Np/s)

Two units

Three units

Figure 7.47 Eigenvalues for one, two and three units on-line. Line a is out-of-service and
the load at bus 5 is off. Conditions are shown for feasible maximum lagging and leading
reactive power outputs at rated real power. PID Set No. 4 installed on all generators.
As shown in Figure 7.47, when three machines are on-line at rated real power output the 5
s halving time is breached, or nearly breached (cases C86-C88). To provide an adequate margin of stability for the most onerous condition it is therefore necessary to install power system stabilizers on the generators. (This is not considered here.)
7.11.2.3.3 Step responses for the range of feasible operating conditions; one, two or three
units on-line.
Based on the PID parameter Set 4 in Table 7.6, let us determine the terminal voltage response of the closed-loop system to a +1% step change in reference voltage of generator
#1 over the range of operating conditions C01 to C95 considered in Figure 7.42.
The perturbations in the generator # 1 terminal voltage from its initial steady-state value are
shown in Figure 7.48. For each of the Case sets in Table 7.9, e.g. C01 - C05, C50 - C60, only
the maximum feasible lagging and the maximum leading reactive power cases are plotted.
We observe the following.
1.

All Cases C01 to C95 satisfy the terminal voltage settling-time criterion. As intended,
the choice of Case C17 as the base case results in a satisfactory set of responses. The

382

Introduction to the tuning of AVRs

Ch. 7

overshoot of the terminal voltage response at leading power factors may be of concern when one unit is on-line.
2.

In Figure 7.46(a) the phase of the extreme Case C21 (25 MW 25 Mvar lag) is some
20 less than that of Case C17 at a gain cross-over frequency of ~1 rad/s. Consequently the phase margin for Case C21 is likely to be 65+20 = 85 , a value which
results in an over-damped response - as is evident in Figure 7.48(a). The converse
argument applies to Case C20 (50 MW 20 Mvar lead), i.e. the resulting step response
is lightly damped.

1.2

Vt (%)

Vt (%)

1.2
0.8
0.6

0.6

0.4

0.4

0.2

0.2
0

0
0

C01
C11
C21

4
6
Time (s)
C05
C15
C25

C06
C16
C17

10

C10
C20

(a)

C42
C51
C61

4
6
Time (s)
C44
C55
C65

C46
C56
C17

10

C50
C60

(b)

(a) One unit, (b) Two units, (c) Three units.

Solid lines: Maximum lagging reactive power output.
Dashed lines: Maximum leading reactive power output.

1.2

Vt (%)

1
0.8

1
0.8

Unit real power output is 50 MW in all cases except for C2125, C61-65 and C92-95 when it is 25 MW.

0.6
0.4

Cases C76, C80, C81, and others are omitted because operating constraints are infringed and are infeasible. For some
10
other cases the reactive power output per generator is reduced, e.g. from -10 to -5 Mvar for Case 79.

0.2
0
0

C72
C82
C92

4
6
Time (s)
C75
C84
C95

C77
C86
C17

C79
C88

(c)

Case C17 is adopted as the Base Case and is shown in all

three sets of plots.

Figure 7.48 Closed-loop operation. The perturbations are shown in the terminal voltage
of generator #1 due to a step change in reference voltage from a steady-state value of 1.0 pu
to 1.01 pu (1%) for the feasible operating conditions. Responses lie within the 10% of the
final value of the step amplitude in less than 5 s.
The lack of damping at leading power factors, highlighted in Section 7.11.2.3.1, results in the
excessive over-shoot of the terminal voltage responses when one or two generators are online. The otherwise satisfactory small-signal performance of the three-generator power system based on parameter set No. 4 for PID Type 2 compensation (Table 7.6) is demonstrated
in Figure 7.48 for N and N-1 conditions. Studies examining the provision of PSSs for the
generators, the limiting of reactive power absorption by the generators, and the performance

Sec. 7.12

Summary, Chapter 7

383

of the system for major disturbances would be undertaken in practice, but are beyond the
scope of this chapter.

7.12

Summary, Chapter 7

Suppliers of AVRs and excitation systems may have developed tuning techniques which are
peculiar to the structure and characteristics of their equipment. The aim of this chapter,
however, is to introduce some of the basic concepts in the tuning of AVRs which may help
others understand some of the approaches which could be used - as well as the relevant control system theory which underpins the analysis. The analysis of the various approaches to
tuning are complemented by examples to demonstrate the design procedure and the performance of the type of compensation employed.
The concept of transient gain, which is the effective gain of the excitation system over a selected range of modal frequencies, forms the basis for the various types of compensation.
Compensation such as Transient Gain Reduction (TGR) fulfils this objective when high
gain excitation systems are required, however the steady-state difference between the desired and the actual terminal voltage following a disturbance may be greater than that specified. The use of PI compensation provides infinite gain at zero frequency and ensures zero
error in the terminal voltage in the steady state.
The application of rate feedback of exciter voltage - or AVR output if exciter voltage is not
accessible for measurement - is also studied.
To boost the speed of response of the generator field voltage following the occurrence of a
disturbance, for example, PID compensation is employed. Assuming a general form of PID
compensation given by
sK D
KI
KG
G c s = ----- + K + ------------------- ------------------- , ((7.2) repeated)
s
1 + sT G
P 1 + sT D

three types are analysed. Type 1: TD = 0, TG = 0, KG = 1,

Type 2A or B: TD > 0, TG = 0, KG > 0, and Type 3: TD > 0, TG > 0, KG > 0.
To determine the parameters in (7.2), the user typically specifies the desired corner frequencies and the transient and high frequency gains required. Except for Type 2B PIDs, the modal frequencies and transient gain lie in a modal frequency range 1 to 2 .Type 2B PID
compensation can be applied to cases in which the PID is required to contribute phase lead
at low frequencies. Such circumstances can occur, for example, when the generator and exciter time constants, T do and T E , are relatively long.
A detailed example is given of the application of PID type 2B compensation to a remote,
three generator system over a range of operating conditions when one, two and three units
are in operation. A comprehensive set of studies of the frequency response characteristics
of the brushless-exciter-generator system for one or more machines on-line reveals that the

384

Introduction to the tuning of AVRs

Ch. 7

closed-loop step responses of terminal voltage are likely to range from poorly to heavily
damped. An emphasis in the studies has been to determine a basis for the evaluation of a
suitable set of PID 2B parameters to satisfy the dynamic performance specifications over a
wide range of operating conditions. The analysis is based on an operating condition which
is chosen that best represents those frequency response characteristics over the set of operating conditions.
The aim of new technique called the phase matching, explained in Appendix 7I.5, is to
improve the robustness of the generator controls to variations in the gain of the voltage control loop. The studies employing this technique have illustrated the importance of obtaining
good models and parameters for both exciter and generator, preferably validated by test.

7.13 References
[1]

IEEE Power & Energy Society, IEEE Tutorial Course: Power System Stabilization Via
Excitation Control, 09TP250, 2009.

[2]

IEEE Standard 421.1, IEEE Standard Definitions for Excitation Systems for Synchronous Machines, 2007.

[3]

IEEE Standard 421.2, IEEE Guide for Identification, Testing, and Evaluation of the Dynamic Performance of Excitation Control Systems, 1990.

[4]

The Mathworks, Inc. (2010, 23 August 2010). MATLAB. Available: www.mathworks.com.

[5]

D.J. Vowles and M.J. Gibbard, Mudpack User Manual: Version 10S-03, School of
Electrical and Electronic Engineering, The University of Adelaide, July 2014.

[6]

P. Kundur, M. Klein, G. J. Rogers, and M. S. Zywno, Application of power system

stabilizers for enhancement of overall system stability, Power Systems, IEEE Transactions on, vol. 4, pp. 614-626, 1989.

[7]

P. Kundur, Power System Stability and Control, McGraw-Hill, Inc., 1994.

[8]

Gene Franklin, J.D. Powell and Abbas Emami-Naeini, Feedback Control of Dynamic
Systems, 6th edition, Pearson Higher Education Inc., 2009.

[9]

R.C. Dorf and R.H. Bishop, Modern Control Systems, 11th edition, Prentice Hall,
2008.

[10]

F.P. de Mello and C. Concordia, Concepts of Synchronous Machine Stability as Affected by Excitation Control, IEEE Trans., vol. PAS-88, pp. 316-329, April 1969.

[11]

Australian Energy Market Commission, National Electricity Rules, Version 55,

March 2013, www.aemc.gov.au

Sec. 7.13

References

385

[12] IEEE Recommended Practice for Excitation System Models for Power System Stability Studies, IEEE Std. 421.5-2005 (Revision of IEEE Std. 421.5-1992), pp. 0_185, 2006.
[13]

R.J. Koessler, Techniques for Tuning Excitation System Parameters, IEEE Trans.
on Energy Conversion, Vol. 3, No. 4, December 1988.

[14]

F.G. Shinskey, Process-Control Systems, Applications, Design and Tuning. 4th Edition
1996, McGraw-Hill, New York.

[15]

K. Astrom and T. Hagglund, PID Controllers: Theory, Design and Tuning. Second
Edition 1995, Instrument Society of America, Research Triangle Park, North Carolina.

386

Appendices

Introduction to the tuning of AVRs

Ch. 7

Appendix 7I
App. 7I.1 Generator and exciter parameters
App. 7I.1.1 Parameters for the 6th order generator and a simple exciter
Note. These models are only used in the application of all types of compensation except that
in Section 7.8 and 7.11 for Type 2B PIDs.
Generator model: 6th order, classical. Saturation ignored. Values in pu on machine MVA
base (the MVA base is stated in the application). (These parameters are the same as those
used for generator TPS_4 in Table 10.23 on page 527). System frequency is 50 Hz.
D = 0,

H = 2.6 s,

ra = 0,

x d = 2.3,

T d0 = 5.0 s,

xl = 0.2,

x q = 0.40,

T q0 = 2.0 s,

x d = 0.25,

T d0 = 0.03 s.

x q = 1.7,
x q = 0.25,

x d = 0.30,
T q0 = 0.25 s,

Exciter: Simple linear first-order lag model: KE = 1.0 pu, TE = 0.1 s.

App. 7I.1.2 Parameters for the 5th order salient-pole generator and a brushless AC
exciter
Note. These models are only used in Section 7.8 and 7.11 for the application of the analysis
of Type 2B PIDs.
Generator model: 5th order, salient-pole generator; saturation is included. All values are in
per unit on machine rating (58.8 MVA) unless otherwise stated. System frequency is 50 Hz.
D = 0,

H = 5.5 s,

ra = 0,

x d = 1.5,

x q = 0.7,

T d0 = 8.0 s,

x l = 0.10,

x q = 0.16,

T q0 = 0.12 s,

x d = 0.16, T d0 = 0.04 s,

S d 1.0 = 0.15,

x d = 0.22,

S d 1.2 = 0.45.

The exciter is an AC generator with a rotating rectifier and is represented by an AC8B Excitation System Model [12] and is shown in Figure 7.49. Its parameters are:
KE = 1.0
TE = 0.7 s,
KCE = 0.1,
KDE = 1.25,
KAE = 1.75,
TA = 0.
The terminal voltage transducer is represented by a first-order lag block, Ttrn = 20 ms.

App. 7I.2 Models of the brushless AC exciter

The model of the brushless AC exciter is shown in Figure 7.49.

App. 7I.2 Models of the brushless AC exciter

VC
Vref

VS

KP

Vexf

KI
s

KG
1+sTG

387

Vr
KAE

KD
1+sTD

VE

1
sTE

EFD
FEX

FEX=fn(IN)

KE

+
+

KCE IFD
VE

VESE(VE)
KDE

IN

IFD
Exciter

AVR

Figure 7.49 The brushless AC exciter is based on the AC8B Excitation System Model [12]
The rectifier regulation modes are expressed by the three equations:
1.0 0.577I
N

F EX = f I N = 0.75 I 2
N

1.732 1.0 I N

I N 0.433
0.433 I N 0.75 .
0.75 I N 1.0

The AVR comprises the Type 2B PID (TG = 0) in which the gain KG may represent power
amplification. The gain KAE, included in the exciter model, is a factor which accounts for the
per unitization of exciter and generator quantities.
V exf

KAE

V r

1
sTE

+
+

V E

FEX0

VE0

F EX
KEX

KS

I N

KIV

+
KIF

KDE

Figure 7.50

KE

E FD

I FD

Small-signal model of the AC8B Excitation System Model

388

Introduction to the tuning of AVRs

Ch. 7

The parameters in the model are defined as follows:

0.577
I N0 0.433

I N0

K CE I FD0
K CE

0.433 I N0 0.75 , K = -------------------------- , and K IF = ---------.
K EX = ---------------------------IV
2
2
V E0

0.75 I N0
V E0

1.732
0.75 I N0 1.0

The gain KS is related to the saturation function of the exciter and is dependent on the initial
steady-state value of the field voltage V E0 . It is given by the following expression:

S E V E
K S = S E V E0 + V E0 -------------------.
V E VE = V E0

Saturation in the exciter is assumed to be negligible under steady-state operating conditions

and thus KS = 0.
As an example, the values of the parameters of the linearized exciter model for cases C16 C20 are provided in Table 7.10. The parameters of the exciter model are listed in
Appendix 7I.1.2.
Table 7.10

Parameters of the small-signal model of the exciter

Case

EF0

KIF

KIV

VE0

C16
C17
C18
C19
C20

2.16
1.92
1.61
1.37
1.20

0.044
0.049
0.059
0.069
0.079

0.041
0.046
0.056
0.065
0.075

2.26
2.03
1.70
1.45
1.27

KEX = -0.577, FEX0 = 0.945, KS = 0

App. 7I.3 PI Compensation using positive feedback

A simple positive feedback implementation can be employed for PI Compensation. The diagram of the associated control blocks is shown in Figure 7.51.

App. 7I.3 PI Compensation using positive feedback

x +

1
-----------------1 + sT 1

389

KG

G1(s)

1
-----------------1 + sT 2

Figure 7.51 PI Compensator

The transfer function of the system from x to y, i.e. excluding the gain K G , is:
1 + sT 2
G 1 s = ----------------------------------------------s sT 1 T 2 + T 1 + T 2
1 + sT 2
------------------------------- ,
T 2 s 1 + sT 1

(7.50)

for T 2 T 1

where, for the approximation T 2 T 1 , the low and high frequency corners are 1 T 2 and
1 T 1 rad/s, respectively. These corners should be respectively about a decade or more below and above the extremes of the range of frequencies of rotor oscillations.
Equation (7.50) can be rearranged into the following form representative of the PI structure:
2
T2
1
1
1
G 1 s = ------------------ --- + ------------------ ------------------------------------------------------------T 1 + T 2 s T 1 + T 2 1 + s T 1 T 2 T 1 + T 2

1 1
1
----- --- + ------------------ ,
T 2 s 1 + sT 1

for T 2 T 1

For the compensator G 1 s the effective integrator gain is 1 T 2 and the proportional gain
is 1 over the range of rotor frequencies. The respective gains of the compensator
K G G 1 s become K G T 2 and K G if K G 1 .
A plot of frequency responses of G 1 s is shown in Figure 7.52 for a range of values of T 2 .
As explained below, the value of T 1 is such that the phase angle approaches zero degrees in
the mid-range of rotor frequencies, e.g. 4 rad/s.

390

Introduction to the tuning of AVRs

Ch. 7

Magnitude (dB)

20
10
0
10
20
1
10

10

10

10

Phase (deg)

0
20
40
60
80
1

10

10
10
Frequency (rad/s)
T2=1
T2=4

T2=2
T2=5

10
T2=3

Figure 7.52 Frequency responses of the PI compensation using positive feedback for values of T2 from 1 to 5 s; associated values of T1 are such that the maximum phase angle is at
4 rad/s.
For design of the parameters of the PI compensator it may be desirable to place the phase
angle characteristic such that phase is close to zero degrees over the range of rotor modal
frequencies. From the figure we note that we can choose a frequency at which the phase angle is a maximum. This frequency, mx , occurs at the geometric mean of the corner frequencies of the exact transfer function (7.50), i.e.
1 T1 + T2
mx = ----- -----------------.
T2
T1

(7.51)

Thus, given T2 and mx , the value of T1 can be derived from (7.51) to yield:
2 2

T 1 = T 2 T 2 mx 1 1 T 2 mx for T 2 T 1 .

(7.52)

Based on (7.50) and (7.51), the value of the phase characteristic G 1 j f when f = mx is
G 1

mx

= atan mx T 2 90 atan mx T 1 T 2 T 1 + T 2
= 2 atan T 1 T 1 + T 2

(7.53)

App. 7I.4 Integrator Wind-up Limiting

391

Figure 7.52 is based on the selection mx = 4 rad/s with T2 being varied from 1 to 5 s; the
corresponding values of T1 and G 1 j f at mx are calculated from (7.52) and (7.53), respectively.

App. 7I.4

Integrator Wind-up Limiting

Two types of limiter, anti-windup 1 and windup, are encountered in excitation system models. Examples of these types are shown in Figure 7.53 in the case of a simple integrator. The
upper and lower limits are UL and LL, respectively.
UL
1
s

UL
y

1
s

LL

LL

y
u

UL
y

UL
w

time

If UL y LL then dy dt = u
If y UL and u 0 then set dy dt = 0
If y LL and u 0 then set dy dt = 0
(a) Integrator with anti-windup limiting

time

If UL y LL then w = y
If y UL then w = UL
If y LL then w = LL
(b) Integrator with windup limiting

Figure 7.53 Integrator with (a) anti-windup limiting, and (b) windup limiting.
The operation of the two types of limiters are illustrated in principle in Figure 7.53, (a) and
(b). In illustration (b), with windup limiting, the output of the integrator y(t) continues to
increase once the limit UL is reached but starts to decrease only when the input u(t) changes
sign. Limiting ceases only when the output y falls below UL. With anti-windup limiting,
however, it ceases limiting as soon as the input changes sign. The advantage of anti-windup
limiting is that it eliminates the time delay between sign reversal and wind-down to UL
that occurs in windup limiting.
Anti-windup and windup limiting occur in other types of transfer function blocks incorporating lead-lag and PI compensation for example (see [7], [12]).
1. Anti-windup limiting is also known as non-windup limiting ([7], [12]).

392

Introduction to the tuning of AVRs

Ch. 7

App. 7I.5 A phase-matching method for constant phase margin

over an appropriate frequency range
Consider the phase responses of V trn j f V exf j f for Case C17 and the parameter Set
No. 2 for PID Type 2B in Table 7.6; the responses are shown in Figure 7.54.
60
Phase (deg)

60
xxxx

xx

xx

120
x x
180
2
10

xx

+: Phase OLTF, Vtrn/Vref

1

10
10
Frequency (rad/s)

10

, ExciterGen: V /V
trn
exf
, AVR:
Vexf/Vref = PID #2

Figure 7.54 Case C17: Frequency responses of the phase of the component transfer functions of the open-loop system. The phase response of the open-loop transfer function with
PID Set No. 2 is shown by x-x-x.
Let j f and j f (deg.) be the phase responses of the generator-exciter and the PID
with parameter Set No. 2, respectively, as shown in Figure 7.54. The phase of the open-loop
transfer function is + shown by x-x-x in the figure. Depending on the location of the
gain-crossover in the range 0.7 - 2.5 rad/s, the phase margin is the difference between the
open-loop phase response ( + ) and 180 . At the gain-crossover-frequency, f = c ,
the phase margin is
PM = c + c 180 = c + c + 180 .

(7.54)

At low frequencies = 0 and, for the PID, = 90 . Let

= + 90

(7.55)

so that at low frequencies both = 0 and = 0 .

Assume the desired phase margin is PMdes, e.g. 65 . The required values of , based on
(7.54) and (7.55), are
des = PM des 90

(7.56)

(If the actual value of is greater than the desired value of des then the phase margin is
greater (i.e more stable) than the desired phase margin PM des ; and vice-a-versa.)

App. 7I.5 A phase-matching method

393

Thus, in order to match the phase margin with the desired phase margin it is necessary to
find the PID frequency response, , that closely matches the line des over the potential
range of gain-crossover-frequencies. Let
= , thus

(7.57)

des = PM des + 90 .

(7.58)

In order to illustrate a design procedure based on (7.58) let us consider the following steps.
1.

Given a selected system operating condition, choose (i) a set of parameters for a
Type 2B PID as in Table 7.6, (ii) the design case C17 for the generator-exciter transfer function V trn j f V exf j f (see Figure 7.42), and (iii) set PMdes to 65 , say.

2.

Plot (i) , the negated phase angle of the transfer function V trn j f V exf j f for
the selected operating condition, (ii) , the phase angle of the Type 2B PID
advanced by 90 , and (iii) the line showing where the response of must lie with
respect to the plot of to satisfy the Phase Margin requirement,
des = PM des + 90 = 25 . The plots of and des = 25 is
shown by x x x in Figure 7.55

3.

Based on the plot in Step 1 adjust the PID parameters systematically so that the
desired phase margin is satisfied, i.e. plots of and des match closely - or overlap
-over the desired frequency range.

4.

Check that the resulting PID satisfies the system performance criteria over the range
of operating conditions in which one or more units are on-line.

Let us consider the determination of the PID parameters based on the above steps.
For Step 1 the system operating condition Case C17 and a set of parameters have already
been selected for the analysis associated with Figure 7.54. The parameters are those in Set
No. 2, Table 7.6, KP = 14 pu, KI = 7.0 pu/s, KD = 8.0 pu-s, TD = 0.143 s, KG = 1.0. Let us
base our analysis in this step on this set of PID parameters and Case C17.
The plots of and associated with the transfer function V trn j f V exf j f for Case
C17 and the PID parameter set, respectively, are shown in Figure 7.55. Also shown is a plot
(x x x) along which the angle of the desired PID must lie in order for the open-loop transfer function V trn V ref to have the desired phase margin (assuming for this study that the
gain-cross-over frequency for the resulting open-loop transfer function (OLTF) lies in the
range 0.7 to 2.5 rad/s).

394

Introduction to the tuning of AVRs

Ch. 7

200
180

Phase Response (deg)

160
x x

Desired PM=65 deg

140

x
x

120

x
x

100
x

80
60

x
x

40
x

20

x
x

0
1
10

x
0

10
Frequency (rad/s)

10

, ExciterGen TF Vtrn/Vexf, Phase Negated

, PID #2 with phase advanced by 90 deg.
, PID #4 with phase advanced by 90 deg.

Figure 7.55 Plots of for exciter-generator transfer function (TF) for Case C17, and
of the phase-advanced PID TF together with the plot of des = 25 (x x x) which represents the desired location of the phase plot of the PID TF.
It is clear from Figure 7.55 that PID parameter Set No. 2 produces excessive phase lead and
therefore the phase margin of the OLTF is greater than the desired value PMdes = 65 . Referring to Table 7.6 or Figure 7.25 it is seen that, by increasing the values of the corner frequencies 1, 2 and D for the PID sets, the plot of in Figure 7.55 approaches the
desired phase margin plot. For PID Set No. 4 with parameters KP = 14 pu, KI = 14 pu/s,
KD = 5.89 pu-s, TD = 0.105 s, KG = 1.0, the plot of coincides with desired phase margin
plot for gain-cross-over frequencies in the range 0.9 to 2.5 rad/s.
Based on the PID parameter Sets 2 and 4, the composite OLTF for Case C17 is plotted in
Figure 7.56. From this Bode plot it is observed that (i) the gain cross-over frequencies for
the two sets are 1.5 and 1.3 rad/s, respectively, (ii) with parameter Set 4 the phase margin is
close to the desired value of 65 . The phase margin variations for PID Sets 2 and 4 for a
loop-gain variation of 6 dB are shown in Table 7.11.

App. 7I.5 A phase-matching method

395

40
Magnitude (dB)

30
20
10
0
10
20
1
10

10

60
Phase (deg)

80
100
120
140
160
180
1
10

10
Frequency (rad/s)

10

Openloop TF with PID #2, PM=91 deg.

Openloop TF with PID #4, PM=66 deg.

Figure 7.56 Case C17. Bode Plots of the OLTF comprising V trn j f V ref j f
and PID Sets 2 or 4.
Table 7.11

Set

Gain
change (dB)

Case C17. Robustness: Phase Margin variation for

loop-gain change of 6 dB
Phase Margin at
frequency of ...
PM (deg)

Phase Margin at
frequency of ...

Gain
change
(dB)
Frequency
PM (deg)
(rad/s)

Frequency
(rad/s)

102

0.54

+6

69

3.0

70

0.80

+6

64

2.4

Desired phase margin is 65 .

From Figure 7.56 and Table 7.11, it is evident that the phase margin variation of about 6
associated with PID Set No. 4 implies it is robust to variation in the loop gain; this is revealed
by the relatively small changes in its phase in the figure. In Set 2, however, not only is the
phase margin variation considerably more but the phase margin exceeds the desired value of
65 over the gain variation of 6 dB.

Chapter 8
Types of Power System Stabilizers

8.1 Introduction
In Chapter 5 a speed-PSS based on the P-Vr design approach is described. The purpose of
this chapter is to describe in detail the theoretical basis for some of the widely deployed types
of PSSs and the associated practical implications. For some other types of PSSs, including
the multi-path, multi-band PSS developed by Hydro-Qubec, only a brief overview is provided. Furthermore, the details of a number of other types of PSSs and their development
are omitted from this book, for example: delta-omega stabilizers (without and with torsional
filters) [1]; the use of notch filters to attenuate the first torsional mode [2]; the application
of the coordinated AVR/PSS, called the Desensitized Four Loops Regulator [3].
The input to the PSS in Chapter 5 is assumed to be the true rotor speed as measured directly by a high-fidelity tacho-generator, a toothed wheel, or some other device mounted on
the shaft of the turbine-generator unit. In practice there may be physical difficulties in positioning any such device on the shaft as well as locating it to minimize the introduction of the
torsional modes of the shaft into the speed signal. Moreover, other difficulties such as noise,
lateral shaft movement (runout or wobble [4]) in vertical units, may present themselves. In
this chapter, however, synthesized speed perturbations, which are assumed to accurately
represent the true rotor speed perturbations, are used as the input to the PSS. This means
that the same basis and procedure as that outlined in Chapter 5 can be employed for the design and tuning of the PSS.

397

398

Types of Power System Stabilizers

Ch. 8

The major factor in the selection of a stabilizing signal for input to the PSS is the requirement that the modes of concern, which may be the local-, inter-area, and possibly the intrastation modes, must be observable by the signal over a wide range of operating conditions.
Typically, perturbations in rotor speed, the electric power output, and the frequency at the
generator terminals are the commonly-used local signals.
Various types of pre-filters are in use which convert one or more signals derived from variable(s) other than speed into a synthesized speed signal. Such variables are electric power,
bus-voltage angle, frequency, terminal voltage and current; some manufacturers develop a
speed signal from such variables using various techniques. Lack of fidelity and resolution
of the synthesized speed signal in representing the true rotor speed are factors that result
in degradation in the performance of the PSS when the design is implemented in practice.
This chapter considers the design of the pre-filters which synthesize a speed signal, and
highlights some issues which may be detrimental to the performance of the resulting prefilter and speed-PSS.
The pre-filters which are discussed in the following sections employ as input signals:

the electric power output of the generator,

frequency (or the deviation of the frequency from its nominal value) at the generator
terminals [5], and

electric power and a speed deviation signal in the widely-used integral-of-accelerating-power pre-filter [7], [8], [9].

Some of the practical issues concerning different types of PSSs, field testing and other aspects are covered in [10], [11]. In practice, the engineer who is responsible for tuning the
PSS does not often have the ability to influence the selection of the type of PSS. This chapter
is intended to provide the reader with approaches to tuning PSSs in circumstances where
ideal performance is not possible because the most appropriate PSS may not have been
specified or provided for the application.
PSS analysis and design procedures are based on linearized models for which the inputs are
the perturbations of the above signals from their initial steady-state values.
Frequency is also derived by some manufacturers from voltage and current measurements
at the generator terminals. The analysis in this chapter concerns only that derived from the
rate of change of a bus voltage-angle.
In the design of PSSs attention must be paid to reducing the effects of the torsional modes
of the turbine-generator unit on its dynamic performance [5].

Sec. 8.2

Dynamic characteristics of washout filters

399

Though the following concerns the small-signal analysis of PSS types, it should be borne in
mind that it is necessary to limit the input to the PSS such that limiting occurs ahead of limiting at its output. This concept also applies to controllers other than PSSs.
Because the characteristics of washout filters may affect the performance of the pre-filter
and PSS significantly, the time- and frequency-domain responses of a single washout filter
and of two identical washouts in series are next examined, but in more detail than earlier in
Chapter 5.

8.2 Dynamic characteristics of washout filters

8.2.1 Time-domain responses
In Section 5.8.6.1 the washout filter is introduced with the purpose of eliminating any
steady-state offset, or DC level, in the input signal to a PSS. In this section, in addition to
those of the single washout filter, the dynamic characteristics of two identical washout filters
in series are examined and a comparison made with the dynamic performance of a single
washout.
The transfer functions of one and two washout filters having a washout time constant of
T W (seconds) are, respectively:
sT W
G 1W s = ------------------- , and
1 + sT W

(8.1)

2 2

s TW
G 2W s = --------------------------- .
2
1 + sT W

(8.2)

In analog terms, the analysis assumes a low impedance source drives the filters which then
feed into a high impedance sink. Expressions for the time-domain responses of each of the
filters to a step input of A 0 units and a ramp input of R 0 units/s are shown in Table 8.1.
Based on the definition of settling times in Section 2.8, the time-domain response of the single washout filter to a step input decays to zero with a 2% settling time of 4T W s. However,
for a ramp input the response of the single filter tends to a finite value R 0 T W - also with a
settling time 4T W s. Consequently, for a PSS having electrical power as the stabilizing signal,
and with the input being a slow ramp in electrical power, the single washout filter produces
a potentially undesirable offset in the terminal voltage of the generator. For the single washout filter the forms of the step and ramp responses are illustrated in Figure 8.1 for washout
time constants of 4 and 8 s.

400

Types of Power System Stabilizers

Table 8.1

Analytical expressions for responses of washout filters to

step and ramp inputs.
Output responses

Input signal at
time t 0 +

One washout filter

Step, A 0 units

Y 1WS = A 0 e

Ramp, R0 units/s

Ch. 8

Two washout filters in series

t TW

t TW

Y 2WS = A 0 e
1 t TW
Y 2WS 0 as t

Y 1WS 0 as t
Y 1WR = R 0 T W 1 e

t TW

Y 2WR = R 0 te

t TW

Y 2WR 0 as t

Y 1WR R 0 T W as t

10
Washout Responses

Washout Responses

1
0.8
0.6
0.4
0.2

8
6
4
2
0

0
0

10

Step in

20
30
Time (s)
Tw=4

40

50

10

Ramp in

Tw=8

20
30
Time (s)
Tw=4

40

50

Tw=8

Figure 8.1 Responses of a single washout filter to step and ramp inputs of 1 unit and
1 unit/s, respectively, for washout time constants of 4 and 8 s.
For the case of two washout filters in series the following time-domain characteristics are of
interest.
1.

The responses of two identical washout filters in series to a step input of 1 unit are
shown in Figure 8.2 for values of the washout time constant of 4 and 8 s. For a positive step input the response decays from the initial value A 0 , passes through zero at
2

time T W s and under-shoots by a value A 0 e = 0.135A 0 at 2T W s. It then decays

to within 0.02 A 0 of zero after approximately 5.5T W s.

Sec. 8.2

Dynamic characteristics of washout filters

401

Washout Responses

1
0.8
0.6
0.4
0.2
0
0.2
0

10

Step in

20
30
Time (s)
Tw=4

40

50

Tw=8

Figure 8.2 Responses of two identical washout filters in series to a step input of 1 unit for
washout time constant values of 4 and 8 s.
2.

For a positive ramp input the time domain response reaches a maximum value of
R0 TW e

= 0.368R 0 T W at time T W s. It then decays to zero. Notice that the maxi-

mum value of the response depends on the ramp rate and the value of the washout
time constant.
The responses to a ramp input of two identical washout filters in series is of particular interest in the discussion of the integral-of-accelerating-power PSS considered in Section 8.5.
Accordingly, the responses of two such filters to a ramp of 1 unit/s are shown in Figure 8.3
for a range of values of the washout time constant from 1 to 10 s.
In Figures 8.1 to 8.3 the time-domain characteristics listed in Table 8.1 are clearly illustrated.
8.2.2 Frequency-domain responses
The nature of the frequency response of a single washout filter, and its role in the dynamic
performance of speed-PSSs, are discussed in Section 5.8.6.1. Because the application of two
washouts is of interest in this chapter the frequency response of two identical washouts in
series, time constant TW, is shown in Figure 8.4. The response is normalised to a corner frequency of 1 rad/s (i.e. TW = 1 s). For example, if TW = 5 s the associated corner frequency
is 0.2 rad/s, the magnitude and phase of the response at say 0.02 rad/s (as read off Figure 8.4
at 0.02 0.2 = 0.1 rad/s) are then 40 dB and 169 , respectively.

402

Types of Power System Stabilizers

(i)

(ii)

3.5
Washout Responses

3.5
Washout Responses

Ch. 8

3
2.5
2
1.5
1

3
2.5
2
1.5
1
0.5

0.5

0
0

10
Time (s)

Input ramp
Tw=1
Tw=2
Tw=4
Tw=5

15

20

40
Time (s)

Input ramp
Tw=6
Tw=7
Tw=9
Tw=10

Tw=3

60

80

Tw=8

Figure 8.3 Responses of two identical washout filters in series to a ramp input of
R0 = 1 unit/s as the washout time constant TW is varied from 1 to 10 s.
Time-frames: (i) 0-15 s, (ii) 0-80 s. Solid lines TW 1-5 s; dashed lines TW 6-10 s.
Peak occurs at TW s.

Magnitude (dB)

0
20
40
60
80
2
10

10

10

10

10

Phase (deg)

180
150
120
90
60
30
0
2
10

10
10
10
Frequency (rad/s)

10

Figure 8.4 Frequency response for two identical washouts filters in series normalised to
a corner frequency of 1 rad/s. (For a single washout filter, halve all vertical-axis quantities.)

Sec. 8.2

Dynamic characteristics of washout filters

403

8.2.3 Comparison of dynamic performance between a single and two washout filters.
Let us compare the features of a single washout filter with two identical washouts in series.
Consider a power system in which the lowest inter-area modal frequency is 2 rad/s. For the
purpose of the design of the associated PSS, let us assume that if a single washout filter is
employed the corner frequency of the washout would be 0.2 rad/s, say, a decade below the
modal frequency. The time constant of the single filter is T1W = 5 s; the phase lead introduced by the filter at the modal frequency is 5.7 . To introduce the same phase lead at the
modal frequency for two identical washouts the corner frequency of each should be 0.1 rad/
s, i.e. T2W = 10 s. Based on these assumptions a comparison of dynamic performance is
summarized in Table 8.2.
Table 8.2

Characteristics of a single washout and two identical

washout filters in series

Corner frequency,
TW
Phase lead introduced at 2
rad/s
Step response of 1 unit:
Settling time
Under-shoot
Final value
Ramp response of 1 unit/s:
Settling time
Peak value
Final value

One washout filter

Two washout filters in series

0.2 rad/s,
5s

0.1 rad/s,
10 s

5.7

5.7

4TW = 20 s
0

~5.4TW = 54 s
-0.135A0 = -0.135 at 2TW s
0

4TW = 20 s
Peak is the final value
TWR0 = 5

By calculation
0.368R 0 T W = 3.68
0

Some observations on the characteristics of the washout filters of Table 8.2 are listed below.
1.

A reduction in time constants for both a single washout and two washouts in series
improves their dynamic performance through lower settling times.

2.

The performance of the single washout filter in Table 8.2 is superior to that of two
washouts, except that the ramp response of the single washout filter tends to a finite
value. As mentioned in Section 8.2.1, in the case of an electrical power PSS this characteristic can produce an offset in generator terminal voltage and reactive power
output when a ramp in electrical power output occurs.

3.

As noted, the time-domain performance of two identical washouts in series can be

improved by reducing the time constant. However, such a reduction is a compro-

404

Types of Power System Stabilizers

Ch. 8

mise with the increase in phase lead at the lower modal frequencies. In the case of a
reduction in the time constant, say from 10 s to 5 s, the phase lead at the modal frequency of 2 rad/s is increased from 5.7 to 11.4 . If desired, the increased phase
lead so introduced by the two washouts can be compensated for in the tuning of the
PSS main compensation blocks.
4.

If the inter-area modes are not of concern, the washout filter time constants can
likewise be determined based on the relatively higher frequency of the local-area
mode(s).

8.3 Performance of a PSS with electric power as the stabilizing signal.

8.3.1 Transfer function and parameters of the electric power pre-filter.
It has been common practice to use electrical power perturbations as a stabilizing signal. The
analysis and implementation of the associated PSS is simplified if the electrical power perturbations are converted to speed perturbations by means of a pre-filter. The transfer function of this pre-filter is now discussed.
The equation of motion of the rotor of a synchronous generator for small-signal disturbances is given in the Laplace domain by,
P m P e = 2Hs + D ,

all quantities in per unit.

(8.3)

This equation has been a basis for analysis in Chapters 4 and 5. In (8.3) P m and P e are
the perturbations in mechanical and electrical torques (or powers), respectively, acting on
the shaft; is the perturbation in rotor speed.
If we assume that perturbations in mechanical power and damping torques D acting on
the rotor are negligible, then (8.3) reduces to:
1
S = ---------- P e ,
2Hs

(8.4)

where S is a speed signal synthesized from electrical power, and therefore can be employed as a stabilizing signal as long as the assumptions stated above are justified. The structure of the PSS becomes that shown in Figure 8.5. (The negative sign at the summing
junction for V S reflects the inherent negation in (8.4)).
As shown in the figure the pseudo-integrator, or pre-filter, used in a practical PSS to replace
the ideal integrator in (8.4) is given by the transfer function:
T H 2H
s = ----------------------- P e .
1 + sT H

(8.5)

Sec. 8.3

Pe

PSS with electrical power as the stabilizing signal

Vt

1+sTH

sTW
1+sTW

Prefilter

Washout
Filter

TH/2H

405

Compensation
transfer function

Damping
Gain

LP Filter

AVR

VS
Vr

PSS

Figure 8.5 Structure of the PSS with electric power as the stabilizing signal. Note the
negative sign of the PSS output signal V S at the summing junction to the AVR.
So that transfer function of the low-pass filter T H 1 + sT H acts as an integrator 1 s over
the range of modal frequencies it is required that its corner frequency at fc = 1 T H rad/
s should be a decade or more below the lowest (inter-area) modal frequency. With this
choice of T H , the gain of the filter rolls off at -20 dB/decade over the range of modal frequencies and its associated phase angle is approximately -90 deg. This is the case for an ideal
integrator (see Section 2.12.1.2).
For example, assume the lowest (inter-area) modal frequency is 2 rad/s; the corner frequency should ideally be 0.2 rad/s or less. For a values of TH of 5.0 and 7.5 s the corner frequencies are respectively 0.2 and 0.133 rad/s; Table 8.3 shows that for these values of TH the
frequency response of the associated pseudo-integrator agrees well with that of the ideal integrator at and above 2 rad/s. While it is common to set TH = TW a higher value of TH (say
TH = 7.5 s when TW = 5 s) is sometimes used in practice.
Table 8.3

Responses of the ideal integrator and the pseudo-integrator

at lower frequencies.

1
f rad/s Ideal: s = ------j f

T H 1 + sT H
TH = 5 s

TH = 7.5 s

1.0 90

0.9806 78.69

0.9912 82.41

0.50 90

0.4975 84.29

0.4989 86.19

0.25 90

0.2497 87.14

0.2499 88.09

Because a synthesized speed signal is derived from the electrical power output using the prefilter of (8.5), the design of the compensating transfer function of the PSS follows the procedure based on a speed-stabilizing signal as outlined in Section 5.10.6. The compensating
transfer function is the same as for the speed PSS given in (5.49).
The rapid attenuation of the electric power signal with frequency f is noted in Table 8.3.
A feature of the use of electric power perturbations as a stabilizing signal is that the torsional

406

Types of Power System Stabilizers

Ch. 8

oscillations which occur on the shafts of generating units are heavily attenuated. This topic
is discussed later in Section 8.5.3.
Several cautionary comments follow.
1.

The electrical power output of the generator will closely follow any ramping of the
mechanical power output of the turbine. Depending on the ramp rate and the time
constant of the washout filter(s), there may be a significant deviation in the associated PSS output. As a result of this signal being injected into the excitation system,
there could be unacceptable variations in terminal voltage and hence in the reactive
power output of the generator [13]. This problem is ameliorated by providing an
appropriate pre-filter, such as in the Delta-P-omega stabilizer [14], or employing an
integral-of-accelerating-power PSS, to be discussed in Section 8.5.

2.

Care should be taken to ensure that negative feedback of the PSS output signal is
applied at the AVR summing junction.

3.

Prior to purchase due care should be taken to ensure that the power input PSS provides for the synthesis of a rotor-speed signal from the electrical-power input.

8.3.2 Dynamic performance of a speed-PSS with an electric power pre-filter.

A PSS designed for a speed-stabilizing signal used with an electric power pre-filter forms the
basis for the assessment of the dynamic performance of the integrated stabilizer. Five operating conditions for the sixth-order generator-SMIB system, shown in Table 5.6, are used to
illustrate the performance of the PSS. Its performance is compared to that of the PSS which
uses true rotor speed for the same fives cases.
A single washout filter with time constant of 5 s is selected in Section 5.10.6.2 for the five
Cases. The associated corner frequency of 0.2 rad/s is more than a decade below the single
rotor mode of oscillation (~ 9 rad/s) 1. The inertia constant of the unit is 3 MWs/MVA. As
revealed in Table 8.3 a suitable time constant for the pseudo-integrator is 7.5 s. The transfer
function for the electric power pre-filter is thus
T H 2H
1.25
---------------------- = ------------------- .
1 + 7.5s
1 + sT H

(8.6)

A comparison of the modes resulting from the use of an electric power pre-filter that synthesizes a rotor speed signal with those produced by a true rotor speed PSS is shown in
Table 8.4. Because there is close agreement in the values of the modes, it is concluded that
the pre-filter accurately synthesizes rotor speed perturbation with the caveat that slow variations in mechanical power may cause variations in the reactive output of the generator.

1. The corner frequency of 0.2 rad/s is a decade below any potential inter-area modal frequencies of 2 rad/s if the SMIB system represents a generator in a multi-machine system.

Sec. 8.4

PSS with bus-frequency as the stabilizing signal

Table 8.4

Case

407

Comparison of modes for Speed and Electric Power PSSs*

Generator
Output,
P, Q pu

Rotor mode
with PSS out of
service

true-speed PSS
in service*

Speed PSS with

Electric Power
pre-filter

0.9, -0.1

0.773 j9.16

1.156 j9.51

1.163 j9.48

0.9, 0

0.552 j9.12

1.271 j9.31

1.275 j9.29

0.9, 0.2

0.261 j9.02

1.338 j9.03

1.339 j9.01

0.9, 0.4

0.113 j8.98

1.305 j8.93

1.305 j8.91

0.9, -0.07
0.927 j7.98
0.409 j8.04
0.409 j8.02
G
* Results for the true-speed PSS are given in Table 5.4
# Two lines are out of service in Case G. All lines are in service in Cases
A - D.

8.4 Performance of a PSS with bus-frequency as the stabilizing signal.

The frequency of the generator terminal voltage is used by some manufacturers as a PSS stabilizing signal on the basis that bus frequency closely represents rotor speed perturbations
in magnitude and phase. The following analysis applies only to frequency signal derived
from the voltage angle. The frequency freq is the rate of change of the terminal-voltage
angle, (rad), thus
freq = 1 0 d dt pu of system frequency;

(8.7)

the associated transfer function is

freq s = 1 0 s ;

(8.8)

where 0 = 2f 0 , f 0 being system frequency (Hz). Once again, the angular perturbations
are converted to a pseudo-speed signal by means of pseudo-differentiation, pseudo-differentiation being employed to limit the gain and noise amplification at high frequencies associated with pure differentiation in (8.7). Moreover the torsional modes, if present in the terminal
voltage, are amplified. Equation (8.8) yields the transfer function of the bus-frequency prefilter:
freq s
s
------------------------ = 1 0 ----------------------- .
1 + sT F
s

(8.9)

The time-constant TF incorporates the phase-lag inherent in the measurement of the busvoltage angle or bus-frequency. It is thus a property of the measurement transducer rather
than being a tunable or selectable parameter.

408

Types of Power System Stabilizers

Ch. 8

In order that the transfer function s 1 + sT F acts as a differentiator over the range of
modal frequencies,
1.

its low-frequency response should (ideally) pass through the 0 dB axis at 1 rad/s and
roll-up at 20 dB/decade with an associated phase angle of 90 ;

2.

its corner frequency at fc = 1 T F rad/s should ideally be a decade or more above

the highest modal frequency (typically a local-area mode). However the gain introduced by the transfer function at higher frequencies may be destabilizing.

Note that the frequency response of the pre-filter of (8.9) can be deduced from that of the
single washout filter in Figure 8.4 by rearranging the pre-filter transfer function into the
form

sT F
1 0
- .
---------------------- ----------------TF
1 + sT F

Although the pre-filter transfer function of (8.9) synthesizes a speed signal from the derivative of bus-angular perturbations (frequency), the question arises how well does bus frequency represent the actual rotor speed perturbations in magnitude and phase? Let us
examine the performance of a PSS equipped with a bus-frequency stabilizing signal.
8.4.1 Dynamic performance of a speed-PSS with a bus-frequency pre-filter
The bus-frequency pre-filter delivers a synthesized speed signal to a PSS whose design is
based on a true rotor-speed stabilizing signal. This so-called bus-frequency PSS forms the
basis for the assessment of the dynamic performance of the integrated PSS. Once again, the
five Cases for the sixth-order generator-SMIB system, listed in Table 5.5, are used to investigate the performance of the pre-filter.
The pre-filter is assumed to be of the form given in (8.9); its parameters are determined as
follows.
Assuming the upper modal frequency is 10 rad/s, ideally the corner frequency of the prefilter should be set a decade higher, at 100 rad/s. However, due to the higher gains introduced at higher frequencies the choice of a corner frequency of 75 rad/s may be considered
to be a suitable compromise; thus TF = 0.0133 s. Nominal system frequency is 50 Hz,
0 = 100 rad/s. The combined transfer function of the pre-filter and the speed-PSS of
(5.49), is thus
1 2f 0 s
V s s
s5
1
1 + s0.0895 + s 2 0.00277
H s = ----------------- = --------------------------------- 20 --------------- ---------- ------------------------------------------------------------- .
s
1 + s0.0133
1 + s5 3.84 1 + s0.005 1 + s0.005

(8.10)

The mode shift associated with both the original speed-PSS and the bus-frequency PSS of
(8.10) are shown in Table 8.5; the damping gain is 20 pu for both PSSs.

Sec. 8.4

PSS with bus-frequency as the stabilizing signal

409

Table 8.5 Comparison of real parts of mode shifts for speed

and bus-frequency PSSs, damping gains 20 pu.

Gen.
OutCase put.
P, Q
pu

Rotor mode
---------------------------------------## Excitation system mode for
bus-frequency PSS ##

Rotor mode shifts, PSSs

in service

Ratio*

PSSs off

true-speed
PSS
in service

bus-frequency
PSS in service
-------------##

true-speed
PSS

bus-frequency
PSS

0.9,

0.77 j9.2

1.16 j9.5

0.46 j9.5

1.93 j0.35

1.24 j0.30

0.1

2.73 j80.0

0.9,
0

0.55 j9.1

1.27 j9.3

0.60 j9.3

1.82 j0.20

1.15 j0.17

0.97 j80.9

0.9,
0.26 j9.0
0.2
-

1.34 j9.0

0.71 j9.1

1.60 j0.02

0.97 j0.03

9.2 j81.9

0.9,
0.11 j9.0
0.4
-

1.31 j8.9

0.73 j9.0

17.7 j82.0

1.42
j0.05 0.85
j0.01 1.68 0.60
-

0.9,
0.93 j8.0
G# 0.07
-

0.41 j8.0

0.24 j8.1

1.34 j0.06

1.17 j0.14

16.3 j76.4

1/c

1.56 0.64
1.59 0.63
1.64 0.61
-

1.14 0.88
-

* Ratio: c= e (true-speed PSS mode-shift) / e (bus-frequency PSS mode-shift)

Note: Results for speed-PSS for a SMIB system are given in Table 5.5
# Two lines are out of service in Case G. All lines are in service in Cases A - D
## In column 5 the upper and lower quantities are the rotor mode and an excitation system
mode, respectively

From the table the following is observed:

1.

In Cases A and G a high-frequency mode associated with the excitation system and
q-axis variables is unstable as the generator power factor becomes leading.

2.

The real parts of the mode shifts for the bus-frequency PSS are degraded significantly (by a factor of 1/c) with respect to the speed-PSS.

3.

Therefore, because the differences in the imaginary parts of the two sets of mode
shifts are negligible, there appears to be a reduction in the loop gain of the PSSSMIB system when employing rate of change of angle of the generator terminal
voltage as the stabilizing-signal source.

Let us consider these observations commencing with no.1 above.

410
8.4.1.1

Types of Power System Stabilizers

Ch. 8

Stability of the closed-loop system for Case A

Let us examine the open-loop frequency response for the system of Case A, a leading power
factor condition, listed in Table 8.5. The open-loop transfer function is V S V ref , where
V S and V ref are the output of the PSS and the AVR reference voltage, respectively; it

includes the combined transfer function of the pre-filter and the speed-PSS given by (8.10).
A block diagram of the transfer function of the open-loop system G s H s is shown in
Figure 8.6(a) and the associated frequency responses are given in Figure 8.6 (b).
As the open-loop system possesses one unstable pole-pair at 2.73 j80.0 , the stability of the
closed-loop system can be determined from the Nyquist Criterion based on the open-loop
system G s H s 1. In the case of Figure 8.6(b) (i) it can be shown that for closed-loop stability the gain at high frequencies must be less than unity (0 dB), and thus must be attenuated. This is achieved by changing the two time-constants of the low-pass filter of the PSS in
(8.10) from 0.005 to 0.01 s. The associated response of the open-loop transfer function
shown in Figure 8.6 (b)-(ii) results in a stable closed-loop system with poles at
0.433 j9.75 and 12.0 j40.1 for the rotor and exciter modes, respectively. Further
studies are required to mitigate against instability for higher gains at high frequencies over
the range of operating conditions.
8.4.2 Degradation in damping with the bus-frequency pre-filter
Based on Table 8.5 the improvement in the damping-constant of the rotor mode due to the
frequency-PSS is substantially less than for the speed-PSS, although there is negligible
change in the modal frequency for both PSSs. This suggests that the use of bus-frequency,
derived from bus voltage-angle, results in a reduction in the loop-gain in the path through
the machine and PSS. Consider the simple system shown in Figure 8.7. The voltages and angles are E and internal to the generator and V t and at its terminals; the voltage at the
infinite bus is V b 0 . The generator internal reactance is x and that of the equivalent external circuit is xe. (One might speculate that, for a simple system such as this, the perturbations
in are roughly related to those in by a factor x e x e + x - if the angles are not large.)

1. Because the PSS output is not negated at the summing junction of the AVR, the conventional open-loop transfer G s H s must be negated for application of the Nyquist Criterion.

Sec. 8.4

PSS with bus-frequency as the stabilizing signal

411

Magnitude (dB)

10

AVR, excitation system,

generator, & power system

Vref

10
20
30
40
0
10

G(s)

10

10

10

100

VS

Phase (deg)

50

H(s)
Pre-filter &
speed PSS

0
50
100
150
0

10

(a) Open-loop SMIB system; bus-voltageangle, bus-frequency pre-filter and PSS

in feedback path.

10
10
Frequency (rad/s)
(i)

10

(ii)

(b) Open-loop frequency response

Figure 8.6 (a) Open-loop system V S V ref . (b) Open-loop frequency response for
Case A: (i) pre-filter and PSS transfer function given by (8.10) (the unstable mode in the
open-loop system is 2.73 j80.0 ). (ii) the closed-loop system is stable with modification of
the parameters of the low-pass filter.

V t
jx

Generator

V b 0
jx e

Infinite bus

Figure 8.7 A simple SMIB system

It can be shown that for small perturbations
cos 0
xe
E
= 1 + ------------------------------------- where k 0 = ----- ------ .

x V b
k 0 cos 0 0

(8.11)

Since perturbations in both rotor speed and bus frequency freq are related to
and by equations of the form of (8.7), then

412

Types of Power System Stabilizers

cos 0
= 1 + ------------------------------------- freq = c freq .

k 0 cos 0 0

Ch. 8

(8.12)

The coefficient c is greater than unity for transmission angles 0 90 . Because

freq the PSS loop-gain is, in effect, reduced by a factor of 1/c.
As an example, assume for a pseudo steady-state condition that the internal voltage E is the
voltage proportional to rotor flux linkages, E , and the reactance x is the transient reactance
x d . If x d = 0.3 x e = 0.375 , then for V t = 1.0 ,
E = 1.036 V b = 1.055 0 = 33.8 0 = 18.7 , and
k 0 = 1.23 and c = 1.80 .

Thus, using bus frequency as the stabilizing signal rather than the true rotor speed, the PSS
loop-gain is inherently reduced, in this case by a factor of 1 c = 0.56 . (Note, as surmised
above, the ratio x e x e + x is about 0.55 .)
The above somewhat simplistic example reveals the order of the magnitude of the reduction
in the PSS loop-gain for a pseudo steady-state condition. This example raises the question:
what is the effect of the closed-loop dynamics and a more accurate generator model on the
gain reduction?
Consider Case B of Table 8.5 on page 409 the rotor mode is 0.60 j9.30 when the busfrequency PSS is in service. Let us evaluate the frequency responses at 9.3 rad/s of both the
true rotor speed and the bus frequency freq for perturbations in reference voltage.
The ratio of the true rotor speed to the bus frequency at the modal frequency is 1.57; this
ratio agrees well with the value of c = 1.59 in the table. The phase difference between the
true and synthesized speeds is approximately 0 when the phase lag introduced by the corner 1/TF in the pre-filter (8.9) is accounted for. Thus, for practical purposes, the true speed
and the synthesized speeds are essentially in phase. We conclude that for the cases analysed
the use of the bus-angle perturbations as the input signal to the PSS results in a gain reduction in the machine - PSS loop. Moreover, there is a significant reduction in the mode shift
for the single rotor mode. Thus in the multi-machine context at the lower inter-area frequencies, in which the generator may participate, are there marked reductions in loop gain - and
therefore reductions in the associated mode shifts due to the use of this type of PSS?
An analysis of the performance of the PSS over the range of normal and contingency conditions, such as that in Table 8.5, suggests that the effective attenuation in gain associated
with the bus-frequency PSS is roughly x e x e + x . As in Case G (Table 8.5) when the external impedance (jxe) is increased the attenuation is significantly reduced. Accordingly, a judicious increase in PSS gain is required in order to provide a performance similar to that of
a speed-input PSS over the encompassing range of operating conditions.

Sec. 8.5

Integral-of-accelerating-power PSS

413

Note: In the event of significant transients that lead to sudden changes in bus-voltage angle,
e.g. a line fault followed by the tripping of the circuit, the synthesized rotor speed derived
from the bus-voltage angle will not necessarily be representative of the true rotor speed until
the resulting large-amplitude oscillations have markedly decayed.

8.5 Performance of the Integral-of-accelerating-power PSS

8.5.1 Introduction
A third category of the PSS models listed in the IEEE Standard [15] is the integral-of-accelerating-power PSS (IAP PSS) and is referred to in the Standard as PSS2B. As shown in
Figure 8.8, the IAP PSS consists of two main components, a pre-filter which develops a synthesized speed signal S and a conventional PSS the design of which is based on the rotorspeed stabilizing signal discussed in Chapter 5.
in
P e

PreFilter

PSS

V pss

Figure 8.8 Components of the integral-of-accelerating-power PSS

The inputs to the pre-filter are speed and electric power signals, in and P e . The
speed signal may be derived in a number of ways, for example from

frequency of the generator terminal voltage;

speed of the rotor measured by tacho-generator, a toothed wheel mounted on the

shaft. etc.;

filtered values of the instantaneous three-phase voltages and currents, and processing
of these signals.

These speed signals contain not only the inter-area and other rotor modes but also the torsional modes of the turbine-generator-exciter unit. In order that the latter modes are not excited by the PSS, the torsional modes must be significantly attenuated; this is one of the roles
of the IAP pre-filter.
If the mechanical power output of the turbine - be it hydro, gas or steam - is changing, the
electric power output of the generator will follow it closely, particularly as the mechanical
power changes occur relatively slowly under normal operating conditions. As observed in
Section 8.3, changing the mechanical - and hence electrical - power input to the PSS can perturb the terminal voltage of the generator, possibly causing undesirable swings in its reactive
power output. Assuming the variation in mechanical power is a ramp, the pre-filter incor-

414

Types of Power System Stabilizers

Ch. 8

porates a ramp tracking filter which tracks a ramp ideally with zero tracking error - and
thereby offsets the ramp in the electrical power input; this is a second role of the IAP prefilter.
Let us consider the influence of the torsional modes, the ramping of mechanical power, and
the characteristics of the ramp tracking filter.
8.5.2 Torsional modes introduced by the speed stabilizing signal
A generating unit, in the case of a steam turbine, may consist of high pressure, intermediate
and low-pressure stages, the generator and an exciter. The lumped masses are connected by
shafts whose torsional stiffness is finite. As is illustrated in Chapter 9 for a linear spring-mass
system, the rotating masses similarly exhibit modal frequencies and damping dependent on
the inertia of the masses and the stiffness of the interconnecting shafts [5], [12].
Since the mechanical stiffness of the shaft components is at least an order of magnitude
higher than the effective electro-mechanical coupling between the generator and the power
system, the entire rotating mass of the mechanical shaft of a large turbo-generator is moreor-less uniformly subject to the power systems inter- and local-area modes of frequency 1.5
to 15 rad/s. The first torsional mode for large steam turbine units can be as low as 8 Hz (50
rad/s) [5], [6]. Depending on the mode shape of the particular torsional mode, a shaft-speed
transducer that is located in a region of the shaft that closely corresponds to a peak of the
torsional oscillations (an anti-node of the mode shape) can result in a significant component
of the torsional mode in the speed signal. One way to avoid this problem is to locate the
speed transducer at a node of the modal shape [5]; this, however, is not always practical since
in some cases the node may lie inside a turbine stage.
It will be assumed in the analysis that the input speed signal to the pre-filter, C t , comprises the true rotor speed component, in t , corrupted by the first and higher torsional modes (as well as noise), t t , i.e.
C t = in t + t t .

(8.13)

8.5.3 The electric power signal supplied to the pre-filter

The electric power signal input to the pre-filter is a filtered representation of the instantaneous electric power. The filtering process typically introduces a very small phase shift over
the range of electro-mechanical modal frequencies and consequently the input power signal
closely follows the low frequency perturbations in power associated with the local- and inter-area modes. Furthermore, as mentioned earlier, if the mechanical power output of the
turbine is ramped, say, in the relatively slow process of generation despatch, i.e. changing
power output from one level to another, the electrical power output of the generator will
closely follow the mechanical power.

Sec. 8.5

Integral-of-accelerating-power PSS

415

If there is any component of the torsional modes in the electric power signal, depending on
how it is calculated, the component - being of significantly higher frequency than the rotor
mode - will be significantly attenuated by the integration in the pre-filter.
8.5.4 The Ramp Tacking Filter (RTF)
The RTF is a low-pass filter of the form,
1 + sT 8 N
F s = --------------------------- , where T 8 = M T 9 .
M
1 + sT 9

(8.14)

The RTF serves a number of purposes. Firstly, it tracks a ramp signal at its input with zero
tracking error. Secondly, it significantly attenuates signals at frequencies above the corner
frequency 1 T 9 . Thirdly, as will be demonstrated, it passes the low frequency perturbations
associated with mechanical power changes with negligible attenuation. The frequency responses for two typical sets of parameter values for the RTF are shown in Figure 8.9. (It
should be noted that the tracking feature of the RTF is defeated if T 8 deviates markedly
from T 8 = M T 9 , for example if T 8 = 0 ).

Magnitude (dB)

50

100

Phase (deg)

150
0
10

10

10

10

0
50
100
150
200
250
300
350
0

10

10
10
Frequency (rad/s)
M=5

10

M=4

Figure 8.9 Frequency responses of the Ramp Tracking Filter for

N = 1, T9 = 0.1, T 8 = M T 9 , M = 5 or M = 4.

416

Types of Power System Stabilizers

Ch. 8

It is clear from the plots that a torsional mode, say the first at 15 Hz (~95 rad/s), is attenuated by 50 dB or more. The parameters commonly used for the RTF are N = 1, M = 5 and
T9 = 0.1. The value of the time constant T9 more-or-less determines the corner frequency of
the RTF. If for example, the variations in mechanical power output are slow, and the torsional modes possess low frequency components, it may be desirable to reduce the value of
T9, and/or set N = 2.
While the RTF tracks a ramp input signal U R t = R 0 t with zero tracking error, it tracks a
2

signal which is the integral of a ramp, i.e. a parabola U P t = 0.5R 0 t , with a constant tracking error. The ramp-tracking characteristics of the filter are analysed in Appendix 8I.2.
from which it can be shown that the steady-state tracking error to a parabolic input is finite,
2

i.e. e fss = 10R 0 T 9 = 0.1R 0 when T 9 = 0.1 and N = 1, M = 5.

8.6 Conceptual explanation of the action of the pre-filter in the IAP

PSS
We will consider the action of the pre-filter in two steps, firstly without washout filters and
then considering their effects.
8.6.1 Action of the pre-filter, no washout filters
As explained in Section 8.5.2 the input speed signal to the pre-filter, C t , is assumed to
comprise the true rotor speed component, in t , corrupted by a torsional component
and high frequency noise, t t , i.e.
C t = in t + t t
(8.13) repeated.
Similarly, an input to the pre-filter is the perturbation in the electric power output of the generator P e t , which closely follows the ramping of mechanical power but also contains
perturbations in the associated local and inter-area modes.
For the purpose of explaining the conceptual basis of the pre-filter, let us assume that the
basic structure of the pre-filter is that shown in Figure 8.10 (omitting the washout filters).
C = in + t
+

P e

RTF

out

1-----2H

Figure 8.10 The basic structure of the pre-filter

In general, the relationship between the accelerating power (or torque) acting on the shaft
and instantaneous speed is given by the shaft equation (4.64) on page 115.

Sec. 8.6

Action of the pre-filter in the IAP PSS

d
P m t P e t = 2H in t pu,
dt

417

(8.15)

where P m t is the perturbation in mechanical power output of the turbine and includes a
ramp change in turbine power. Rearranging (8.15) and integrating the resulting expression,
we can express the integral of the electrical power signal, IPE, as:
1
1
IPE = ------- P e t dt = ------- P m t dt in t .
2H
2H

(8.16)

Each term in (8.16) has the dimensions of speed (pu). Consequently, on the basis of (8.16),
1
the output of the integration of the electrical power signal, IPE = ------- P e t dt , contains
2H
information not only on the mechanical power ramp and perturbations but also the true
rotor speed in t . As mentioned, if any torsional modes, which typically exceed 8 Hz (50
rad/s), are present in the electrical power signal they are heavily attenuated through the integrator transfer function P e s 2Hs , i.e. 50 dB at 50 rad/s for H=3 MWs/MVA.

Let us combine the signal IPE with the input speed signal C t of (8.13), as shown diagrammatically in Figure 8.11(i). A signal IPM results:
1

IPM = IPE + C t = ------- P m t dt in t + in t + t t ,

2H

(8.17)

1
i.e. IPM = ------- P m t dt + t t .
2H

(8.18)

Note that IPM contains only the perturbations in mechanical power and the torsional
modes, the true rotor speed signals in t in (8.17) having been cancelled out; this cancellation is an essential feature of the IAP pre-filter.
As shown in Figure 8.11(ii), the signal IPM is passed through the RTF. By judicious selection
of the parameters of the RTF it will attenuate significantly the higher-frequency torsional
modes and track the integral of the mechanical power ramp-changes with negligible tracking
error 1. An analysis of these features of the RTF are given in Appendix 8I. The output of
1
the RTF therefore contains the integral of mechanical power, V rtf = ------- P m t dt 2, the
2H

levels of the torsional modes having been attenuated significantly.

1. Strictly-speaking, because of the ideal integrator in the basic pre-filter structure shown in
Figures 8.10 and 8.11, the tracking error of the RTF to a ramp in mechanical power is
non-zero. As explained in Section 8.6.2.2 this error is very small, and is zero when there
are one or more washout filters ahead of the integrator.
2. Note that the slow changes in the integral of mechanical power are not attenuated by the
RTF (see its frequency response in Figure 8.9).

418

Types of Power System Stabilizers

1
IPM = ------- P m dt + t
2H

C = in + t
+

P e

Ch. 8

1
V rtf = ------- P m dt
2H
RTF

1
-------
2H

1
1
IPE = ------- P e dt = ------- P m dt in
2H
2H

(i)

(ii)

out
= in

(iii)

Figure 8.11 The action of the pre-filter. (i) The implementation of (8.13) and (8.16).
(ii) The ramp tracking filter attenuates the torsional modes t t in the speed input and
tracks the integral of the mechanical changes with negligible steady-state error.
(iii) The signals containing the integral of the mechanical changes are cancelled out at the
summing junction and the true speed signal out = in is synthesized.
Finally, as shown in Figure 8.11(iii), the negated signal IPE is combined with the output of
1
the RTF at the summing junction. The component ------- P m t dt present in each signal is
2H
cancelled out resulting in the output of the pre-filter being the required true rotor speed,
in s .

To compensate for a difference in the levels of the speed signal in the speed-signal path from
that derived from electric power 1, the gain k s is provided as shown in Figure 8.12. (For example, this adjustment may be required if an attenuated speed signal is derived from bus frequency, see Section 8.4.2). Furthermore, to eliminate any steady-state levels in the electrical
power and speed inputs, P e t and C t , two washout filters are added to each input;
this completes the block diagram of the IAP pre-filter. (The effect on the synthesized speed
signal out , say, of having two washouts in the speed input and one in the electric power
input path is discussed briefly in the later Section 8.6.4.2.)
We know that the RTF follows a ramp input at its terminals with zero steady-state error e fss
between its input and output. In practice there are washout filters and an integrator between
the mechanical ramp input and the input to the RTF. The input to the RTF may no longer
be a ramp, how does this affect the steady-state error?

1. The degradation in performance of the PSS in such a case is illustrated in Figure 8.20(i).

Sec. 8.6

Action of the pre-filter in the IAP PSS

sT w1
-------------------1 + sT w3

sT w2
-------------------1 + sT w2

1 + sT 8 N
-------------------------M
1 + sT 9

ks
P e

sT w3
-------------------1 + sT w3

sT w4
-------------------1 + sT w4

419

out

Ramp-tracking filter

1
---------2Hs
Ideal integrator

Figure 8.12 Block diagram of the prefilter for the IAP PSS. The gain k s is set to unity in
the following analysis.
8.6.2 Effect of the washout filters and integrators on the performance of the prefilter
In the previous section the conceptual action of the pre-filter without washout filters was
analysed; let us now consider their effect on the tracking of the RTF and the dynamic performance of the pre-filter.
In Section 8.2 the dynamic characteristics of one or two washout filters are analysed in their
own right. However, as a diversion, let us (i) assume the speed and torsional signals are negligible and (ii) examine the steady-state and dynamic performance only of the path associated
with the electric power input, namely the washout filters, the integrator and the RTF. This
path is shown in Figure 8.13. Note that a fictitious test input signal U s is used for the purposes
of this analysis and is a step, ramp, parabolic or cubic function of time only. We will also consider two cases when the integrator in the pre-filter is represented as an ideal or as a pseudointegrator; the latter is referred to as the practical integrator. In essence, in this analysis the
performance of the RTF to a particular set of characteristics of the mechanical power output
is being studied.
It has been emphasized that the component 1 2H P m t dt in the signal IPM should
pass through the RTF with zero following error so that it cancels (ideally) the same component in the signal IPE when the mechanical power is ramped. Several questions arise. Due
to the action of the washouts and the integrator, does the output of the RTF still follow its
input with zero steady-state error when that input is no longer a ramp? For example, consider the output of the washout filters in Figure 8.13. Does the RTF track with zero error
n
other mechanical power inputs, e.g. a n t , n > 1?

420

Types of Power System Stabilizers

Us

sT w
-----------------1 + sT w

sT w
-----------------1 + sT w

Ideal /
Pseudo
Integrator

U rtf

1 + sT 8 N
-------------------------M
1 + sT 9

Ch. 8

V rtf

Ramp-tracking filter

Figure 8.13 Path between the electric power input and the RTF output for a test input
U s which replaces the electric power signal.
Note that, because the blocks in the path of Figure 8.13 represent linear elements, the principle of superposition permits the performance of this sub-system to be analysed independently of the rest of the pre-filter. The behaviour of this sub-system also reflects its
behaviour when it is incorporated in the complete prefilter.
8.6.2.1

Dynamic response of the isolated path of Figure 8.13 to a ramp input

Useful insight is provided by examining both the dynamic and steady-state responses at the
input and output of the RTF as well the tracking errors for a ramp in mechanical power. We
are concerned only with the path of Figure 8.13.
Let us now demonstrate the nature of the response of the RTF for a ramp of rate
R 0 = 0.0075 pu/s in mechanical power output. For the current and later applications the
parameters of the complete pre-filter of Figure 8.12 are given below:

Washout filters: T w1 = T w2 = T w3 = T w4 = 7.5 s, assuming the lowest (inter-area)

modal frequency is 1.5 to 2 rad/s (only Tw3 and Tw4 in Figure 8.12 are relevant to the
signal path under study);

Integrator: H=3 MWs/MVA; Pseudo-integrator (as derived in Section 8.3.1):

TH = 7.5 s;

RTF: N = 1, M = 5, T 9 = 0.1 s, T 8 = MT 9 = 0.5 s. (The selection of T 9 = 0.1 s is

mentioned in Section 8.5.4).

It is shown in Figure 8.14 (a) it is noted that, for the ideal integrator, the output of the RTF
does not track the ramp in mechanical power but tends to a constant value
2

T W R 0 2H = 0.0703 in the steady state. Furthermore, the output of the RTF V rtf t
tracks its input U rtf t with negligible error which, as shown in Figure 8.14 (b), tends to zero

in the steady state. For the pseudo-integrator, however, it is observed in Figure 8.14(a) that
the output of the RTF follows the ramp in mechanical power with zero following error in
the steady-state (i.e. after some 50 s). This is because the pseudo-integrator ceases to act as
an integrator and becomes a low pass filter at low frequencies.

Sec. 8.6

Action of the pre-filter in the IAP PSS

421

For a ramp in mechanical power it is also noted in Figure 8.14 (a) that zero tracking error
between the input and output of the RTF is achieved for both the ideal and the pseudo- integrator.
7

10

Units of speed

Units of speed (%)

x 10

5
4
3
2

6
4
2
0
2

0
0

10

20
30
Time (s)

40

50

RTF In
RTF Out Ideal Int.
RTF In
RTF Out PseudoInt.
Mechanical Power Ramp

10

20
30
Time (s)

Ideal Int.

(a)

40

50

PseudoInt.

(b)

Figure 8.14 Responses to a ramp in mechanical power for ideal and pseudo-integrators
with two washout filters in the isolated path of Figure 8.13. The plots show (a) the input and
output responses of the RTF U rtf , V rtf , and (b) that the error across the RTF,
U rtf t V rtf t , in the responses is very small and tends to zero in the steady state.

(To avoid a discontinuity at time zero in Figure 8.14 (a) and (b), the initial slope of the mechanical power output is varied in parabolic fashion from zero to the ramp rate of 0.0075
pu/s at 1 s.)
8.6.2.2

The steady-state tracking - and tracking errors - of the RTF

For the RTF with the parameters given in Figure 8.9 it is known that its output V rtf tracks
a ramp change at its input U rtf with zero steady state error. However, for a parabolic input
to the RTF its output tracks the input with a constant following error after any initial transients have decayed away.
Let us examine the behaviour of the isolated path of Figure 8.13 in more detail. Firstly, for
the sake of completeness, it is of interest to ascertain the performance of the RTF not only
for the four types of mechanical power change U t , but also the effects of none, one and
two washout filters on the tracking errors. Secondly, consideration is given to the effects of
the ideal and pseudo-integrators, the transfer function of the latter being
T H 2H 1 + sT H , (8.5). Of interest are not only the steady-state values of the input to the

422

Types of Power System Stabilizers

Ch. 8

RTF but also how closely the output of the RTF tracks the input to the RTF. Consequently,
in Appendix 8I.2 expressions are derived which analyse the nature of the tracking error for
n

power changes of a general form U t = a n t .1. The results are summarised in Table 8.6.
The upper value in each row of the table is the steady-state input to the RTF, (not the power
changes at the input, U s ). The steady-state input is
U rtf

ss

= lim U rtf t = lim sU rtf s .2 .

t
s0

The lower value is the steady-state tracking error of the RTF, i.e. the difference between the
steady-state input to the RTF and its output, i.e. e fss = U rtf ss V rtf ss . Note that:

x means the quantity increases indefinitely with time.

When the both the mechanical power and the input to the RTF are increasing indefinitely with time the tracking error may be zero or finite (e.g. columns 5 to 8, parabolic
input).

Although the tracking error is zero for a ramp applied directly to the RTF (column 1), when
a ramp is applied to an ideal integrator in the path the tracking error is non-zero (Table 8.6,
ramp, col. 4). The conceptual discussion in Section 8.6 surrounding Figure 8.11, in which
there is an ideal integrator in the path, is based on the assumption that the tracking error is
zero. However, it can be shown that this error is small even for fast ramps. In practice of
course, there are one or more washout filters in the power-signal path in which case the
tracking error of the RTF is zero.
In summary, the practical case is the replacement of the ideal integrator by the pseudo-integrator of (8.5) with one or two washout filters in the electric power input path. As noted in
Table 8.6 - and analysed in Appendix 8I.2 - the steady-state tracking errors of the RTF are
zero if a pseudo-integrator is employed when the mechanical power input is a step, ramp, or
parabola.

1. The expressions are for the input to the RTF and the tracking error between RTF input
and output. For the ideal integrator these are (8.29) and (8.32), respectively; for the
pseudo-integrator they are (8.33) and(8.34).
2. Final Value Theorem. See Section 2.10.

Steady-state input to the RTF U rtf

ss

and the tracking errors following changes in mechanical power

Type of
mechanical
power change
Ut , t 0

Input to
PseudoIdeal
RTF
RTF integrator integrator
-----------only.( & RTF (no & RTF (no
Tracking
washouts) washouts)
error

One
washout,
ideal
integrator
& RTF

One
washout,
pseudointegrator
& RTF

Notes

Two
Two
washouts, washouts,
pseudoideal
integrator integrator
& RTF
& RTF

10

Step,
n=0

R0

U rtf
ss
efss

R0

K 0P R 0
0

K 1I R 0
0

0
0

0
0

0
0

Ramp,
n=1

R0 t

U rtf

K
0
0I K G R 0

K 1P R 0
0

K 2I R 0
0

0
0

K 1I K G R 0

K 2P R 0
0

K 1P = T W T H 2H

K 1P K G R 0

K 2I K G R 0

K 2P = T W T H 2H

Parabola,
2
n = 2 R0 t 2
Cubic,
n=3

R0 t 6

ss

efss
U rtf

ss

efss
U rtf

ss

efss

R0 KG

K 0P K G R 0

K G = 10T 9 ,
K 0I = 1 2H
K 0P = T H 2H ,
K 1I = T W 2H ,

Action of the pre-filter in the IAP PSS

Input to RTF, U rtf , output V rtf and tracking error of the RTF, e fss = U rtf V rtf as
ss
ss
ss
ss
t , for mechanical power changes U t applied to the path in Figure 8.13:

Sec. 8.6

Table 8.6

K 2I = T W 2H
2

RTF: N=1, M=5; Inertia constant H (MWs/MVA); Integrators: Ideal (I) 1 2Hs , Pseudo (P) T H 2H 1 + sT H

423

424

Types of Power System Stabilizers

Ch. 8

8.6.3 Dynamic performance of the complete pre-filter

The SMIB system, Case C, described in Sections 5.10 and 5.11 will be used to investigate the
performance of the pre-filter using the parameters provided in Section 8.6.2.1. In Table 5.5
the input to the PSS is rotor speed; the rotor mode of oscillation is 1.34 j9.03 . Using an
IAP PSS with the pre-filter parameters of Section 8.6.2.1 together with the SMIB PSS parameters (derived in Section 5.10.6), the value of the rotor mode is virtually unchanged at
1.32 j8.97 .
For illustrative purposes the following three disturbances are applied to the generating unit:

A ramp increase in mechanical power input is 0.45 pu per minute, or 0.0075 pu/s,
over a period of 20 s. (This rate is exaggerated to highlight certain features in the
responses.) To avoid a discontinuity at time zero, the initial slope of the mechanical
power output is varied in parabolic fashion from zero to the ramp rate of 0.0075 pu/s
at 1 s.

A relatively small step increase of 5% in terminal voltage reference at 4 s, followed by

a step decrease of 5% at 12 s.

An exaggerated, sustained torsional mode of 12 Hz (75.4 rad/s) and peak amplitude

0.25%, commencing at 12 s.

The simultaneous application of an increasing ramp, and the step change in voltage, should
reveal how the pre-filter discriminates between the changes in mechanical power input and
disassociated electrical power perturbations, oscillatory in nature, resulting from the change
in reference voltage. While responses to small changes in mechanical power at the ramp rate
specified are amenable to analysis using a small-signal model of the SMIB, the change in mechanical power of 0.45 pu per minute over a period of 20 s is not small. Although it is inconsistent to mix small- and large-signal analyses, the important issue here is the assessment
the performance of the pre-filter which is a linear element. Moreover, using the small-signal
model of the SMIB system provides to the pre-filter the electric power and rotor speed signals
inputs of the correct relative amplitudes and phase. Again, for the purposes of illustration, the
amplitude of the sustained torsional mode is exaggerated and is large, being of the same order of amplitude as the speed perturbations resulting from the step in reference voltage.
In Figure 8.15 the variable names and their locations in the pre-filter are defined for use in
subsequent figures. Variable names IPE and IPM are defined earlier in (8.16) and (8.18) respectively. The true rotor speed at the input is in ; t represents the torsional modes
present; out is the speed output of the pre-filter (and ideally is equal to the true speed
input in ). The output of the second speed washout filter is W2 and that of the second
electric-power washout filter is P W2 ; V rtf is the output signal of the RTF.

Sec. 8.6

Action of the pre-filter in the IAP PSS

t
C

in

WO2

P W2
P e

WO1

V rtf

W2 IPM
WO1

425

WO2

Washout filters, WO#

RTF

IPE
INT

Pseudo-Integrator

out
= in

IPE
RTF: Ramp-tracking filter

Figure 8.15 The variable names and their locations in the pre-filter are defined for use in
Figures 8.16 to 8.18.
For the three disturbances the responses of the variables in the pre-filter are shown in Figures 8.16 to 8.18. The left- and right-hand plots in each figure show the relevant responses
when torsional modes are absent or present, respectively. So that the responses to the
changes in reference voltage are clearly discernible, the damping gain of the PSS in Case C,
Section 5.10.6, is reduced from 20 to 10 pu.
From Figure 8.16 the following are noted:

In (a)-(i) the nature and timing of two of the input disturbances are shown.

In (a)-(ii) the decaying oscillatory responses in true speed in due to the step
changes in reference voltage are observed; the output of the second speed washout filter W2 (not shown) is identical for practical purposes.

In (a)-(ii), as predicted by (8.16), the output of the pseudo-integrator (IPE) contains

both the oscillatory rotor speed component and a component associated with the
ramp in mechanical power. Importantly, it is observed that the true speed component is
eliminated from the signal IPM which is input to the RTF.

However, in (b)-(i) the signal IPM at the input to the RTF contains a component associated with the ramp in mechanical power as well as the torsional mode, t . As mentioned, the true speed component seen in IPE is absent from IPM.

In (b)-(ii) is shown C , the torsional mode modulated by the true speed component.

Types of Power System Stabilizers

0.16

2.2

0.14

2.1
Units of Speed (%)

Mech.Power, Ref.Voltage (pu)

426

0.12
0.1

Pm

0.08
V

0.06

ref

0.04

2
1.9
1.8
1.7
1.6

0.02

IPM

(i)
0

10

15

(i)

IPE

Ch. 8

1.5

20

12

13

14

15

2.5

0.5
0.4
Units of speed (%)

Units of speed (%)

2
IPM
1.5
1
IPE

0.5
0

0.3
0.2
0.1
0
0.1
0.2
0.3
0.4

in

(ii)

in

0.5

0.5
0

Pm
IPE
in

10
Time (s)

15

Vref
IPM
C

20

12

(a)

13
14
Time (s)
IPE

in

IPM

Figure 8.16
(a) Torsional mode absent
(i) Disturbances: Ramp in mechanical power
P m and step changes in reference voltage,
V ref
(ii) Inputs: in & C . Responses of
internal variables IPE & IPM

(b) Torsional mode present

(i) Responses of internal variables
IPE & IPM
(ii) Inputs: in & C

(ii)
15

(b)

Sec. 8.6

Action of the pre-filter in the IAP PSS

427

2.5

2.2

2.1
Units of speed (%)

Units of speed (%)

Consider Figure 8.17 in which are shown the responses of internal and external variables.

1.5
V

rtf

IPM

0.5
0

1.7

15

20

12
18

16

16

14
12
Q

8
6
4

Vt

IPM
in

Vrtf
Vt

15

15

13
14
Time (s)

15

10
8
6
4
2
0

10
Time (s)

14

12

(ii)

(ii)

2
5

13

14

0
0

(i)

1.5

18

10

IPM

Vrtf

Change in Vt and Q (%)

Change in Vt and Q (%)

10

1.8

(i)

0.5
5

1.9

1.6

out in
0

20

2
12

out
Q

Vrtf
Q

IPM
Vt

(a)

(b)

Figure 8.17
(a) Torsional mode absent
(i) Responses of internal variables IPM &
V rtf ; input in & output out

(b) Torsional mode present

(i) Responses of internal variables IPM &
V rtf

(ii) Responses: terminal voltage V t and reactive power Q

428

Types of Power System Stabilizers

Ch. 8

In Figure 8.17 it is observed that:

in (a)-(i) the output of the RTF, V rtf , follows the mechanical power component
related to the input signal to the RTF, IPM, with zero tracking error;

in (a)-(i) the speed output signal out from the pre-filter is identical to the true speed
input signal in ; the associated rotor mode is clearly evident in the terminal voltage
and reactive power responses in (a)-(ii).

in (b)-(i) the torsional mode present at the input to the RTF, IPM, is not evident in the
heavily attenuated output of the RTF, V rtf .

In considering Figure 8.17(a)-(ii), it should be remembered that, as the electrical power output increases while following the mechanical power ramp, the reactive output of the generator will also ramp in order to supply the additional I2X losses. Moreover, from the figure it
is noted that there is also a step increase/decrease in reactive power output associated with
the step changes in terminal voltage; this is superimposed on the reactive power ramp. In
Figure 8.17(b)-(ii) there is no evidence of the heavily attenuated torsional mode in terminal
voltage and reactive power responses.
The output of the second washout filter in the electrical power signal path P W2 is displayed in Figure 8.18(a)-(ii), together with the output of the pseudo-integrator (IPE). The effect of the mechanical ramp change can be observed in both signals.
The responses of the speed output signal from the pre-filter out and associated response
of the PSS V pss are seen in Figure 8.18(a)-(ii). Note that there is negligible off-set in both
these signals from their zero values. Therefore, as a consequence, the offset in the output of
the pseudo-integrator (IPE) due the ramping of mechanical power will not be manifested as an
offset either in the PSS output, the terminal voltage, nor in the reactive power output of the unit.
When the torsional mode is present, due to amplification by the PSS, there is evidence of
the attenuated torsional mode in the PSS output in the expanded display of Figure 8.18(b)(ii). Bear in mind, however, the amplitude of the torsional mode, seen in Figure 8.18(b)-(i),
and the ramp rate of mechanical power have been exaggerated for illustrative purposes.
Figures 8.16 to 8.18 confirm that, due to the action of a properly designed pre-filter, the effects of neither the ramping of the mechanical power output of the turbine, nor of torsional
oscillations, are manifested in the output of the PSS. Furthermore, the swinging of terminal
voltage and reactive power output due to ramping of power is not observed.

Sec. 8.6

Action of the pre-filter in the IAP PSS

0.6
& (%)

15

0.4

in

10
5

IPE
5

0.2
0.4

out

PW2

10

0.2

PW2 & IPE (%)

429

(i)

out

15
0

10

15

20

12

13

in
14

(i)
15

1
(%)
&

out

0.5

pss

0.5
out

out & Vpss (%)

0.8

Vpss

0.6
0.4
0.2
0
0.2
0.4

out

0.6

(ii)

(ii)

pss

0.8
0

10
Time (s)
PW2
out

15

20

12

out
Vpss

IPE
Vpss

(a)

13
14
Time (s)
C
out

15

in

(b)

Figure 8.18
(a) Torsional mode absent
(i) Responses: Washout P W2 and IPE

(b) Torsional mode present

(i) Inputs: in & C
Output: out

(ii) Outputs: out & V pss

8.6.4 Potential causes of degradation in performance of the pre-filter of the IAP

PSS
Degradation in the performance of the pre-filter may be attributable to a number of causes.
Several of these are now examined.
8.6.4.1

Effects of non-ideal pre-processing the speed input signal to the pre-filter.

Any pre-processing of the speed input signal may result in incomplete cancellation of the
speed signal at the input to the RTF; complete cancellation is seen as an essential feature of
the pre-filter.

430

Types of Power System Stabilizers

Ch. 8

In Section 8.5.1 it is pointed out that the speed signal may be derived from a number of
sources, including the true rotor speed which itself may be subject to some form of processing prior to injection to the pre-filter of the PSS. In the case of a speed signal derived from
bus-frequency the signal may be subject to attenuation as established in Section 8.4.2 For
illustrative purposes it will now be assumed that the true rotor speed signal in is processed through a first-order pre-processing filter prior to input to the PSS pre-filter.
Let the transfer function G A s of the speed pre-processing filter of the true rotor speed
signal be
G A s = A in = A 1 + sT A .

(8.19)

With this transfer function the effects of attenuation - or gain - and phase shift on the output
speed signal of the PSS pre-filter, S = out are to be analysed. The output of the
speed pre-processing filter is A , A and TA are the gain and time constant. The relevant
elements of the PSS pre-filter which includes the speed pre-processing filter are shown in
Figure 8.19. Perturbations in mechanical power output and the torsional mode are assumed
to be absent; according to (8.4) the true rotor speed is
1
in = ---------- P e .
2Hs
in

P e s

GA s

1--------2Hs

A
+

A in
+

(8.20)

RTF

out

in

in

Figure 8.19 Signals in the IAP pre-filter assuming non-ideal pre-processing of the
speed input signal through a transfer function G A s ( P m = 0).
Based on (8.19) and Figure 8.19 it can be shown that the output of the pre-filter is:
A 1 sT A
out = -------------------------------- RTF s + 1 in .
1 + sT A

(8.21)

Clearly, at low frequencies out A in and at high frequencies out in . Over

the frequency range typically of interest the responses, or distortion factors out in
in the true speed, are shown in Figure 8.20 for a range of values of A and time constants TA.
The parameters of the RTF of (8.14) are N = 1, M = 5 and T9 = 0.1 s.

Sec. 8.6

Action of the pre-filter in the IAP PSS

1.4
Distortion Factor

Distortion Factor

1.4
1.2
1
0.8
0.6
0.4
0
10

1.2
1
0.8
0.6
0.4
0
10

10

30

30

20

20
Phase (deg)

Phase (deg)

431

10
0
10
20

10

10
0
10
20

30

30
0

10

10
Frequency (rad/s)
A=1.2
A=0.9

A=1.1
A=0.8

(i) A, TA=0 s

10
A=1.0

10
Frequency (rad/s)
A=1.2
A=0.9

A=1.1
A=0.8

A=1.0

(ii) A, TA=0.05 s

Figure 8.20 Distortion factors, out in , due non-ideal pre-processing of the true
speed input signal to the pre-filter. Values A: 0.8 to 1.2; (i) TA = 0, (ii) TA = 0.05 s.
Of concern in the figure are the effects of the amplitude and the phase shift on the pre-processed speed signal A over the range of frequencies of the rotor modes, 1.5 to 15 rad/s,
and their deviation from the ideal response of 1 0 . Although the range of values of A and
TA employed in Figure 8.20 may be considered somewhat extreme, the results imply that appropriate care is required in the pre-processing of the speed input signal to the pre-filter.
These results show that depending on how the speed-input signal to the pre-filter is derived
in practice, significant distortion in both gain and phase of the synthesised speed signal can
occur.
Various methods can be employed for calculating the electric power. Any pre-processing filters which are employed in the electric power input signals paths may also result in incomplete cancellation of the speed signal at the input to the RTF. This would likewise result in
distortion of the speed output of the PSS pre-filter. The effects of any pre-processing of input signals to the PSS pre-filter should therefore be examined to assess if they degrade the
performance of the PSS.

432

Types of Power System Stabilizers

Ch. 8

Consider the case of a bus-frequency stabilizing input, the associated pseudo-speed signal
freq being derived from the rate of change of terminal voltage angle as in (8.9). The degradation in the amplitude of this signal is discussed in Section 8.4.2. The effect of such degradation on the output of prefilter out is illustrated in Figure 8.20(i). Not only is the
amplitude of out modified but also is its phase- which could introduce an additional
phase lag in the PSS over the modal frequency range of interest.
8.6.4.2

One or two washout filters in the electrical-power and speed paths?

Recall that a washout filter is introduced with the purpose of eliminating any steady-state
offsets, or DC levels, in the input signal, as well as blocking very slow changes in the input.
It is thus necessary to include at least one washout filter in each path of the pre-filter.
The effect on the response of the RTF of one or two washout filters in the electrical-power
path has been examined in Section 8.6.2. The performance requirements for the pre-filter
may thus determine the number of washout filters in this path.
What are the effects of choosing a different number of washout filters in the speed and electric power paths? The following requirements must be satisfied:

When considering the presence of the local- and inter-area modes in each of the two
signal paths, the frequency response of both one or two washout filters should be ideally, or close to, 1 0 over the range of modal frequencies. This requirement dictates
the value of the washout time constant, Tw.

The frequency response of a pseudo-integrator in the electrical-power path should be

ideally, or close to, 1 2H f 90 over the range of modal frequencies. This
requirement determines the time constant of the pseudo-integrator.

If there are different numbers of washout filters in the speed and power paths an imprecise
cancellation of the true rotor-speed at the input to the RTF occurs under perturbed conditions. It is therefore desirable that the same number of washouts be employed in both input
paths.
8.6.4.3

Effect on the synthesized speed signal of setting the RTF time constant T8 to zero.

As in earlier sections, the SMIB system Case C, described in Section 5.10 will be used to investigate the performance of the pre-filter when the time constant T 8 is set to zero.
In order for the RTF to follow a ramp with zero steady-state error a requirement is that
T 8 = MT 9 in the RTF transfer function of (8.14). Setting T 8 to zero turns the RTF into a
simple low-pass filter of order M if N = 1. With this setting and for a ramp in mechanical
power the output of the RTF follows the input signal IPM with non-zero error. In
Figure 8.21, and comparing it with Figure 8.18(a)-(ii), this error is seen to manifest itself not

Sec. 8.7

The Multi-Band Power System Stabilizer

433

in

out

&

pss

(%)

only in the synthesized speed signal at the output of the pre-filter, but also in the PSS output.
Consequently there is an associated undesirable swing of the generator terminal voltage and
reactive power output during ramping of the mechanical power.
1.5
1
0.5
0
0.5
1
1.5
2
0

in

10
Time (s)

out

15

20

pss

Figure 8.21 Deviation in the synthesized speed output ( out ) from the speed input
( in ), and the consequent effect on the PSS output signal ( V pss ), due to setting T 8 = 0
in the RTF. (Compare these responses with those in Figure 8.18(a)-(ii).)

8.7 The Multi-Band Power System Stabilizer

A fourth category of the PSS models listed in the IEEE Standard [15] is called the MultiBand PSS (MB-PSS), PSS4B, which was first developed by Hydro-Qubec and is in operation on the Hydro-Qubec system [16], [17].
A block diagram of the MB-PSS structure is shown in Figure 8.22. It is noted that this PSS
has three separate tunable paths, unlike the integral-of-accelerating-power PSS, PSS2B [15],
which only has a single such path. The objective of the MB-PSS structure is to isolate and
focus the PSS tuning in three frequency bands which account for three phenomena:

0.05 - 0.2 Hz (~0.3 - 1.2 rad/s): very slow oscillations associated with the common or
global modes on a system 1;

0.2 - 1 Hz (~1.2 - 6 rad/s): low frequency, inter-area modes of rotor oscillation;

1 - 4 Hz (~6 - 25 rad/s): higher frequency, local-area and intra-plant modes of rotor

oscillation.

1. Note that it is important not to confuse the global mode with low-frequency modes
sometimes observed with hydro-turbines, for example. The latter modes may be associated with governor - water column interactions and are localized phenomena [19].

434

Types of Power System Stabilizers

SD
v
L-I

Pe
v

KL1

KL11+sTL1
1+sTL2

1+sTL/R
1+sTL

1+sTL5
1+sTL6

KL2

KL11+sTL7
1+sTL8

1+sTL
1+sTL*R

1+sTL11
1+sTL12

KI1

sTwI
1+sTwI

1+sTI/R
1+sTI

1+sTI5
1+sTI6

KI2

sTwI
1+sTwI

1+sTI
1+sTI*R

1+sTI11
1+sTI12

KH1

sTwH
1+sTwH

1+sTH/R
1+sTH

1+sTH5
1+sTH6

KH2

sTwH
1+sTwH

1+sTH
1+sTH*R

1+sTH11
1+sTH12

SD

Ch. 8

+
_

KL

+
_

+
KI

+
_

KH

Figure 8.22 Multi-Band PSS. SD: Speed Transducer. (See [16], [17] for details).
In particular, the low frequency band is introduced to provide damping for very low frequency phenomena encountered on isolated systems 1, particularly the so-called global
mode in such a system. It is stated in [17] that the MB-PSS and the integral-of-acceleratingpower PSS ... can be tuned to achieve quite similar performance in the local, intra-unit and
torsional modes ... since they both use an electric power signal to capture the high frequency
dynamics. However, having many more degrees of freedom available to modulate its phase
lead over a wide frequency range allows the MB-PSS to better balance its performance in
inter-area modes from 0.1 to 0.8 Hz (0.6 to 5 rad/s).
Low frequency oscillations have been observed, for example in hydro-systems: 0.63 rad/s
between the Northwest and Southwest power systems in the US [18]; 0.31 to 0.50 rad/s on
the Colombian system [19]. Oscillations lying in the intermediate range, associated with vortex instability in hydro machines, are reported to be less than 0.5 Hz (3 rad/s) [20], and about
1 Hz (6 rad/s) [21].
The speed signal H , input to the high frequency band, is derived from the measured generator electrical power output. A separate internal frequency transducer supplies a speed signal L I to the low and intermediate frequency bands. Washout filters are provided in the
intermediate and high frequency bands; torsional (notch) filters may be incorporated in the
PSS structure. In each of the three bands is a differential filter arrangement; it is of interest
to understand the characteristics of such a filter. An analysis of a simplified form of the filter,
shown in Figure 8.23, is conducted in Appendix 8I.3.

1. Systems may be isolated because there are no synchronous links to neighbouring systems.

Sec. 8.7

The Multi-Band Power System Stabilizer

K1

1+sT/R
1+sT

K1

1+sT
1+sT*R

435

+
_

K2

Differential filter Gdf(s)

Figure 8.23 Filter G(s)

The analysis reveals the filter takes the form of the well-known Q-filter,
K 2 2 m s
G s = --------------------------------------------------------------- ,
1 + 2 m s + s m 2

(8.22)

where m rad/s is the frequency at which the frequency response is at its maximum value
K2. The frequency response of (8.22) with variation in damping ratio is given Figure 2.21.
It is of interest to examine the nature of the frequency response of the MB-PSS omitting
washout filters, speed transducers, and torsional (notch) filters. Let the gains and centre frequencies of the three bands, evaluated in Figure 5 of [16] be KL = 5.0 pu, FL = 0.04 Hz;
KI = 25.0 pu, FI = 0.70 Hz; KH = 120 pu, FH = 8.0 Hz; respectively. The frequency responses of three bands and the output of the MB-PSS are shown in Figure 8.24; they agree
closely with Figures 5 and 6 in [16].
In [17], a detailed comparison is provided on a test system between the designs of the MBPSS (PSS4B) and the integral-of-accelerating-power PSS (PSS2B). For the MB-PSS it is
found that, by separating out the low frequency and the higher frequency bands (each of
which have their own limits and wash-out filters), the lower-frequency band limits and washout can be adjusted independently of the higher frequency bands to account for islanding
and large frequency deviations.
Figure 8.24 reveals that, for the selected parameter values, the phase response varies between 35 and 60 degrees leading. That is, the phase response is relatively level over the range
of 0.1 to 25 rad/s (0.02 to 4 Hz) in this case. However, the MB-PSS gain varies over a wide
range. Interestingly, this approach contrasts with that of the P-Vr method (Section 5.8.1) in
which the PSS transfer function attempts to account for the inherent gain and phase characteristic of the particular generator - on which the PSS is installed - over a relevant range
of modal frequencies (e.g. see Figure 5.16) and an encompassing set of operating conditions.

436

Types of Power System Stabilizers

Ch. 8

Magnitude (pu/pu)

140
120
100
80
60
40
20
0
3
10

10

10

10

10

10

Phase (deg)

100
50
0
50

10

10

10
10
Frequency (Hz)

Sum
Intermediate

10

10

Low
High

Figure 8.24 Frequency responses of the MB-PSS assuming a common speed signal input
to the three differential filters in Figure 8.22
A number of methods for the tuning the MB-PSS has been offered. For example, the parameters of the MB-PSS are selected by adjusting the centre frequency and gain of each band
so as to achieve the nearly flat phase response between 30 and 50 degrees over the range of
frequencies, say, 0.05 Hz and 3 Hz (0.3 to 20 rad/s) in order to cover the global and intrastation modes. Other approaches, including optimization techniques, are proposed in [22],
[23], [24] and [25].

8.8

Concluding remarks

In Chapter 5 a PSS based on the P-Vr design approach is described; it assumes a true speed
stabilizing signal is available. By true speed is implied that the signal faithfully represents
the generator speed in magnitude and phase, torsional oscillations being negligible. In this
chapter electric-power and bus-frequency based pre-filters are employed to yield a synthesized speed signal for input to a P-Vr based speed-PSS. A similar objective applies to the prefilter for the integral-of-accelerating-power PSS but overcomes some of the disadvantages
of the previous two pre-filters.

Sec. 8.8

Concluding Remarks

437

The frequency of the generator terminal voltage is used as a PSS stabilizing signal on the basis that bus frequency closely represents rotor speed perturbations in magnitude and phase.
Although it may be synthesized from terminal voltages, bus frequency is assumed to be derived from the rate of change of bus voltage-angle , i.e. freq = 1 0 d dt pu of
system frequency. Using the latter signal as the stabilizing signal for a true speed-PSS is
shown to reduce the effective damping gain of the PSS (by as much as 40% in the cases studied). However, the damping gain of the PSS can be increased to compensate for the gain
reduction. Because differentiation of a signal occurs, care should be taken to provide adequate attenuation at high frequencies (i) to reduce noise, and (ii) to eliminate a possible
source of instability - as is demonstrated in an example. In the signal processing for this and
other forms of bus-frequency transducers, care should be taken to avoid the introduction of
phase shifts which may degrade the design of the PSS unless they can be accounted for. The
effect on bus frequency of large, sudden disturbances at the generator terminals should be
examined.
The performance of the electric power PSS is shown to be close to that of designed for the
conventional true speed-PSS. However, in comparison with a true speed stabilizing signal
which in practice may contain torsional modes, the advantage of this pre-filter is that it significantly attenuates these modes in its output speed signal. However, the conventional PSS
has the disadvantage that ramping of the mechanical power output of the prime mover causes variations in the terminal voltage and reactive power output of the generator. This problem can be ameliorated by use of an integral-of-accelerating-power PSS.
The integral-of-accelerating-power (IAP) pre-filter generates the speed signal for a PSS designed for a true speed-stabilizing signal based on the P-Vr approach. A detailed analysis
of the pre-filter for the IAP PSS is conducted and demonstrates the role and effects of the
ramp tracking filter (RTF), and of the washout filters and the integrator in the power input
path. It is shown that the RTF itself consists of a unity feedback system with two integrations in its forward path and therefore it tracks a ramp input at its input with zero error in
the steady state. However, depending on the number of washout filters, the type of integrator, and the characteristics of the mechanical power output, the steady-state tracking errors
may be finite but are small. Because the effective operation of the pre-filter relies on the cancellation of the speed signal at the output of the integrator by the input speed signal, care
must be taken to ensure the fidelity - in amplitude and phase - of the speed input signal to
the pre-filter. If the latter signal lacks fidelity with respect to the true speed, the performance of the PSS may be markedly degraded.
PSS2B or PSS4B?
In considering the application of the multi-band and integral-of-accelerating-power PSSs
the following few items may be pertinent.
In comparison to the integral-of-accelerating-power PSS (PSS2B) the feature of the MultiBand PSS (PSS4B) is its ability to damp low-frequency and common-mode oscillations [17].

438

Types of Power System Stabilizers

Ch. 8

The sensitivity of the output of the latter PSS to ramping of mechanical power and slow system frequency drift is likely to be low because
(i) the gain in the low frequency band is relatively low (20-25% of the high frequency gain),
(ii) the corner frequencies of the washout filters in the intermediate band are 1 rad/s,
(iii) the high frequency speed signal transducer, in effect, has a washout corner frequency of
about 1.2 rad/s (0.2 Hz)
Consequently, variations in reactive power are likely to be small.
For large steam units with the first torsional mode being about 8-10 Hz, notch filters may
be required for the PSS4B. However, in the case of an integral-of-accelerating-power PSS
with a ramp tracking filter having the characteristics shown in Figure 8.9, the attenuation of
torsional frequencies at 8-10 Hz (50-60 rad/s) is 50 dB or more; notch filters may not be
needed.
Thus, in generalizing, it is necessary to consider carefully - among other factors - the system
characteristics as well of those of the generating units in order to specify the system damping
performance requirements over the low to high range of modal frequencies. Following such
an investigation it may then be possible to select the required PSS structure.

8.9

References

[1]

P. Kundur, D. C. Lee, and H. M. Zein El-Din, Power System Stabilizers for Thermal
Units: Analytical Techniques and On-Site Validation, Power Apparatus and Systems,
IEEE Transactions on, vol. PAS-100, pp. 81-95, 1981.

[2]

W. Watson and M. E. Coultes, Static Exciter Stabilizing Signals on Large Generators

- Mechanical Problems, Power Apparatus and Systems, IEEE Transactions on, vol.
PAS-92, pp. 204-211, 1973.

[3]

H. Bourles, S. Peres, T. Margotin, and M. P. Houry, Analysis and design of a robust

coordinated AVR/PSS, Power Systems, IEEE Transactions on, vol. 13, pp. 568-575,
1998.

[4]

Keay, F.W. and South, W.H., Design of a Power System Stabilizer Sensing Frequency Deviation, Power Apparatus and Systems, IEEE Transactions on, vol. PAS-90, Issue: 2, March 1971, pp. 707 - 713.

[5]

P. Kundur, Power system stability and control. New York: McGraw-Hill, 1994.

[6]

P. Pourbeik, D. G. Ramey, N. Abi-Samra, D. Brooks, and A. Gaikwad, Vulnerability

of Large Steam Turbine Generators to Torsional Interactions During Electrical Grid
Disturbances, Power Systems, IEEE Transactions on, vol. 22, pp. 1250-1258, 2007.

[7]

A. Murdoch, S. Venkataraman, R. A. Lawson, and W. R. Pearson, Integral of accelerating power type PSS. I. Theory, design, and tuning methodology, Energy Conversion, IEEE Transactions on, vol. 14, pp. 1658-1663, 1999.

Sec. 8.9

References

439

[8]

A. Murdoch, S. Venkataraman, and R. A. Lawson, Integral of accelerating power

type PSS. II. Field testing and performance verification, Energy Conversion, IEEE
Transactions on, vol. 14, pp. 1664-1672, 1999.

[9]

P. M. Paiva, J. M. Soares, N. Zeni, Jr., and F. H. Pons, Extensive PSS use in large
systems: the Argentinian case, in Power Engineering Society Summer Meeting, 1999.
IEEE, 1999, pp. 68-75 vol.1.

[10]

IEEE Tutorial Course, Power System Stabilization via Excitation Control, IEEE
Special Publication 09TP250,2009.

[11]

G. R. Berube, L. M. Hajagos, and R. Beaulieu, Practical utility experience with application of power system stabilizers, in Power Engineering Society Summer Meeting,
1999. IEEE, 1999, pp. 104-109 vol.1.

[12]

P. M. Anderson, B. L. Agrawal and J. E. Van Ness, Subsynchronous Resonance in Power Systems, New York, IEEE Press.

[13]

J.C.R. Ferraz, N. Martins, N. Zeni Jr, J.M.C. Soares, G.N. Taranto, Adverse Increase
in Generator Terminal Voltage and Reactive Power Transients Caused by Power System Stabilizers, Proceedings of IEEE Power Engineering Society Winter Meeting, 2002.

[14]

D. C. Lee, R. E. Beaulieu, and J. R. R. Service, A Power System Stabilizer Using

Speed and Electrical Power Inputs-Design and Field Experience, Power Apparatus
and Systems, IEEE Transactions on, vol. PAS-100, pp. 4151-4157, 1981.

[15]

IEEE Recommended Practice for Excitation System Models for Power System Stability Studies, IEEE Std 421.5-2005 (Revision of IEEE Std 421.5-1992), pp. 0_1-85,
2006.

[16]

R. Grondin, I. Kamwa, G. Trudel, L. Gerin-Lajoie, and J. Taborda, Modeling and

closed-loop validation of a new PSS concept, the multi-band PSS, in Power Engineering Society General Meeting, 2003, IEEE, 2003, p. 1809 Vol. 3.

[17]

I. Kamwa, R. Grondin, and G. Trudel, IEEE PSS2B versus PSS4B: the limits of performance of modern power system stabilizers, Power Systems, IEEE Transactions on,
vol. 20, pp. 903-915, 2005.

[18]

F. R. Schleif, G. E. Martin, and R. R. Angell, Damping of System Oscillations with

a Hydrogenerating Unit, Power Apparatus and Systems, IEEE Transactions on, vol.
PAS-86, pp. 438-442, 1967.

[19]

H. V. Pico, J. D. McCalley, A. Angel, R. Leon, and N. J. Castrillon, Analysis of Very

Low Frequency Oscillations in Hydro-Dominant Power Systems Using Multi-Unit
Modeling, Power Systems, IEEE Transactions on, vol. 27, pp. 1906-1915, 2012.

[20]

N. Martins, A. A. Barbosa, J. C. R. Ferraz, M. G. Dos Santos, A. L. B. Bergamo, C.

S. Yung, V. R. Oliveira, and N. J. P. Macedo, Retuning stabilizers for the north-

440

Types of Power System Stabilizers

Ch. 8

south Brazilian interconnection, in Power Engineering Society Summer Meeting, 1999.

IEEE, 1999, pp. 58-67 vol.1.
[21]

K.A. Lance, R. L.Bolden, G. Sheard, Performance of generators with static excitation on a large network. Proceedings of the Annual Engineering Conference. Hobart,
1975, pp. 226-35.

[22]

J. B. Simo, I. Kamwa, G. Trudel, and S. A. Tahan, Validation of a new modal performance measure for flexible controllers design, Power Systems, IEEE Transactions
on, vol. 11, pp. 819-826, 1996.

[23]

I. Kamwa, G. Trudel, and L. Gerin-Lajoie, Robust design and coordination of multiple damping controllers using nonlinear constrained optimization, in Power Industry Computer Applications, 1999. PICA '99. Proceedings of the 21st 1999 IEEE
International Conference, 1999, pp. 87-94.

[24]

L. Gerin-Lajoie, D. Lefebvre, M. Racine, L. Soulieres, and I. Kamwa, HydroQubec experience with PSS tuning, in Power Engineering Society Summer Meeting,
1999. IEEE, 1999, pp. 88-95 vol.1.

[25]

A. Khodabakhshian, R. Hemmati and M. Moazzami, Multi-band power system stabilizer design by using CPCE algorithm for multi-machine power system, Electric
Power Systems Research, 101 (2013) 36 48.

Appendices

441

Appendix 8I
App. 8I.1

Action of the Ramp Tracking Filter (RTF)

It is noted in Section 2.10.2.2 that, if there are two integrations in the forward path of the
unity feedback system shown in Figure 8.25, the tracking error R s C s for a ramp input is zero.
Rs

E s

Gs

Cs

Cs
Gs
W s = ----------- = --------------------Rs
1 + Gs

Figure 8.25

Closed-loop transfer function:

Structure of a closed-loop control system

Let us assume that the 5th order forward-loop transfer function of this unity feedback system is:
2 2

2 2

3 3

G s = 1 + sT 8 s T 9 10 + 10sT 9 + 5s T 9 + s T 9 .

(8.23)

(Note the double integration in G s .) The closed-loop transfer function is:

2 2

2 2

3 3

RTF s = W s = 1 + sT 8 1 + sT 8 + s T 9 10 + 10sT 9 + 5s T 9 + s T 9 .

If T 8 = 5T 9 the closed-loop transfer function becomes:

5

RTF s = 1 + s5T 9 1 + sT 9 .

(8.24)

The transfer function of the ramp-tracking filter postulated in (8.14) is of the form:
M N

RTF s = W s = 1 + sT 8 1 + sT 9 .

(8.25)

Comparing the last two equations, we note that they are identical if N = 1 , M = 5 and
T 8 = MT 9 = 5T 9 .
Based on a formal method of analysis a general result for the open-loop transfer function of
(8.23) with N = 1 and M 2 is derived:
1 + s MT 9
-.
G s = -----------------------------------------------------------------M
M
2
k

2
sT 9 sT 9
k

(8.26)

k=2

Thus, provided T 8 = MT 9 , the associated RTF has two integrations in the forward path and
consequently the RTF will track a ramp input with zero following error in the steady-state.

442

Types of Power System Stabilizers

Ch. 8

For the case when N 1 there are, in effect, N RTFs in cascade each satisfying the requirement T 8 = MT 9 . An analysis similar to that for N = 1 confirms the validity of the latter result.
Note that the RTF will track, with a finite steady-state error, a signal which is the integral of
2
1
a ramp, i.e. ---------- R 0 s . However, by extending the analysis of (8.26) it is a simple matter

2Hs

to derive the following transfer-function of a Parabolic Tracking Filter (PTF):

2

1 + MT 9 s + M M 1 2 T 9 s
PTF s = -------------------------------------------------------------------------------------- ,
M
1 + sT 9

(8.27)

which will track a parabolic input P 0 t , as well as a ramp input, with zero steady-state error.

App. 8I.2 Steady-state conditions at the input and output of the

RTF and associated tracking errors for mechanical power input
P m t = a n t

App. 8I.2.1 With and without an Ideal Integrator

The mechanical power changes of interest in the following analysis are: (i) a step of magni2

tude a 0 = R 0 (for n = 0); (ii) a ramp R 0 t , a 1 = R 0 , (n = 1); (iii) a parabola 0.5R 0 t , a 2 = R 0 2 ,

3

(n = 2); and (iv) a cubic 1 6 R 0 t , a 3 = R 0 6 , (n = 3). Note that each of the last three
n

functions is an integral of the previous input function. Since the Laplace transform of t is
n+1

, the Laplace transform of each of the input functions is simply R 0 s n + 1. Let us

consider the alternatives of either an ideal integrator or of a pseudo-integrator being employed in the integration of the mechanical power signal.
n! s

Let us assume that for the path in the pre-filter shown in Figure 8.13 consists of k ideal integrators, k = 0, 1 and m washout filters, m = 0, 1, 2; assume for the mechanical power input
n
R 0 t , n = 0, 1, 2, 3. The output of the ideal integrator, i.e. the input to the RTF, is then
m
k
m
T w R 0 2H
sT w
1 k R0
U rtf s = ----------------------- ---------- ------------ = ----------------------------------------------------------- .
n+1
m
2Hs
1 + sT w
n + k + 1 m
s
1 + sT w
s

(8.28)

Applying the Final Value Theorem of (2.27) to (8.28), the general form of the expression for
the steady-state input to the RTF, after the initial transients have decayed away, is found to
be
m

U rtf

T w R 0 2H
= lim s U rtf s = ------------------------------------- as s 0 .
n + k m
ss
s0
s

(8.29)

App. 8I.2 Tracking errors under steady-state conditions

443

The associated output of the RTF, V rtf s , is

1 + sT 8 N
V rtf s = U rtf s RTF s , where RTF s = --------------------------- . ((8.14) repeated)
M
1 + sT 9

To ascertain how well the output of the RTF tracks its input, let N = 1 and let us calculate
the tracking error E f s between the RTFs input and output:
E f s = U rtf s V rtf s
= U rtf s 1 RTF s
Gs
= U rtf s 1 --------------------- ,
1 + Gs
1
= U rtf s --------------------1 + Gs

where G s is given by (8.26). As t , The tracking error in the steady state becomes
e fss =

G s
lim sE s = lim s U s 1 --------------------
rtf
f
1
+
G s
s0
s 0
=

lim
s0

1
s U rtf s ---------------------
1
+
G

(8.30)

However, from (8.26) we can deduce

1 + sT 8
1
lim G s = lim --------------------------------------------------------------------------------- = ----------------.
2 2
2 2
2 2
3 3
s0
s 0 s T 9 10 + 10sT 9 + 5s T 9 + s T 9
10T 9 s
s0

(8.31)

Following substitution of (8.29) and (8.31) in (8.30), the latter reduces to a general expression for the tracking error between RTF input and output.
m

T w R 0 2H
2
e fss = lim ------------------------------------- 10T 9 .
n + k m 2
s0 s

(8.32)

For example, for the case of an ideal integrator, two washout filters (column 8 of Table 8.6
on page 423), and a cubic mechanical power input, i.e. n = 3, k = 1 and m = 2, (8.32) becomes:
m

T w R 0 2H
2
2
e fss = ------------------------------------- 10T 9 = K 2I K G R 0 , where K 2I = T W 2H
0
s

and

K G = 10T 9 .

444

Types of Power System Stabilizers

Ch. 8

Note, for the other inputs in column 8 when n < 3, efss = 0.

App. 8I.2.2 With a Pseudo Integrator
The same analysis as in the previous section is conducted with the transfer function of a single pseudo integrator replacing that of the ideal integrator. It can be shown that
m

T w T H R 0 2H
U rtf
= lim s U rtf s = ------------------------------------------ as s 0 , and
n m
ss
s0
s

(8.33)

T w T H R 0 2H
2
e fss = lim ------------------------------------------ 10T 9 .
n m 2
s0

(8.34)

App. 8I.3 Multi-Band PSS transfer function

Consider a differential filter of form shown in Figure 3 of [16]; it is also illustrated in
Figure 8.26 in which R is a constant ratio.
K1

1+sT/R
1+sT

K1

1+sT
1+sT*R

K2

Differential filter Gdf(s)

Figure 8.26

Differential filter Gdf

The input-output transfer function is

1 + s T R
1 + sT
G s = K 2 G df s = K 1 K 2 --------------------------- ------------------------
1 + sT
1 + s TR
2

s1 R
= K 1 K 2 T R ------------------------------------------- 1 + sT 1 + sTR

(8.35)

Equation (8.35) reveals that G s is a band-pass filter. Let the frequency m = 1 T R

and s = j m ; (8.35) then becomes:
2

1 R
G j m = K 1 K 2 --------------------- .
R1 + R

(8.36)

However, at frequency m rad/s it is required that the differential filter have unit gain; i.e.
G df j m = 1 . Hence, from (8.36):
2

1 R
G df j m = K 1 --------------------- = 1 ; thus
R 1 + R

App. 8I.3 Multi-band PSS

445
2

K 1 = R 1 + R 1 R , or

(8.37)

K 1 1 R 2 2K 1 + 1 R + K 1 = 0 .

(8.38)

Solving (8.38) for R we find:

R = 2K 1 + 1 8K 1 + 1 2K 1 1 , with K 1 1 .

Following substitution of (8.37) in (8.35) the transfer function G s becomes:

K 2 2 m s
G s = --------------------------------------------------------------- ,
1 + 2 m s + s m 2

(8.39)

where the damping ratio is = R + 1 2 R . Equation (8.39) is that of a Q-filter with

maximum gain K 2 at the centre frequency m rad/s. Its normalized frequency response is
shown in Figure 2.21 for K2 = 1 and a range of damping ratios from 0.1 to 10.

Chapter 9
Basic Concepts in the Tuning of
PSSs in Multi-Machine Applications

9.1 Introduction
The objective of the application of stabilizers in multi-machine power systems is to stabilize
the system by providing adequate damping for the critical rotor modes of oscillation. These
modes typically involve several power stations and their machines. In the case of inter-area
modes many power stations, geographically widely separated, may participate in both the local and inter-area modes. It is therefore necessary that the stabilizer which, when fitted to a
generator, contributes with stabilizers on other machines to the damping of the relevant
modes. Furthermore, because operating conditions on the system continuously change, the
performance of a fixed-parameter stabilizer should be robust to any such changes.
By employing the P-Vr method in the tuning of the PSS, as demonstrated in Chapter 5, the
inherent magnitude and phase characteristics of the generator and power system are being
utilized; for practical purposes these characteristics consistently lie in a relatively narrow
band. Not only can the method account for variations over a wide range of loading conditions on the system, line outages, etc., but the resulting PSS is most effective and beneficial
at the higher generator real power outputs as revealed in Table 5.6, and discussed in the associated text.
Prior to considering the application of the P-Vr method to the tuning of PSSs in multi-machine power systems, the use and significance of two valuable tools in the small-signal anal-

447

448

Tuning of Multi-Machine PSSs

Ch. 9

ysis of the dynamic performance of such systems are discussed. These tools concern the socalled Mode Shape and Participation Factor analyses of the system for a selected operating condition. Such analyses reveal the nature and significance of the various modes (both
rotor or other modes), the involvement - and extent of involvement - of generators in the
modes, and other insights such as the nature of the dynamic behaviour of other devices in
the system (e.g. FACTS devices and their controls).
The application of other PSS tuning methods, namely the GEP Method and the Method of
Residues, is discussed in Chapter 6. While these approaches can be adapted to the multi-machine system, for the reasons explained in the latter chapter the P-Vr method is considered
to possess some significant advantages.
9.1.1 Eigenvalues and Modes of the system
th

It has been pointed out in Section 3.5, that the h eigenvalue of the real, n n system matrix A of the state equations is the real or complex scalar quantity, h ; it is the non-trivial
solution of the equation
Av h = h v h .

(9.1)

The n -element column vector, v h , is the right eigenvector of the matrix A corresponding
to the eigenvalue h .
For low-order dynamic systems, typically with less than 2500 states, the eigenvalues are calculated using an algorithm that employs QR factorisation [1]. As the number of states approach 2500 the computation tends to become much slower. However, if fast computation
is required to determine only those eigenvalues in a selected region of the complex s-plane,
or if the system order is greater than 2500, methods such as Modified Arnoldi [2], Subspace
Iteration [3] and Multiple-Shift-Point Sparse-Eigenanalysis [3] are available. Such facilities
are normally included in software packages for the analysis of the small-signal dynamic performance and control of large power systems [4].
As has been discussed earlier, eigen-analysis is an extremely valuable tool because the n eigenvalues of the system characterize the nature of its dynamic behaviour in the following
ways:
1.

The time-domain responses of the system states and outputs to a disturbance are
weighted sums of terms of the forms a i e

i t

and b k e

k t

sin k t + k , where i = i is

a real eigenvalue and k = k j k is a complex-conjugate pair of eigenvalues. The

real and/or the real plus imaginary parts of the eigenvalues therefore clearly define the
form of its responses.
2.

The system is stable if the real parts, , of all n eigenvalues are negative.

Sec. 9.2

3.

Mode Shape Analysis

The monotonic and oscillatory terms a i e

449
i t

and b k e

k t

sin k t + k are the ith and kth

modes 1 of the system, respectively.

In the case of the oscillatory modes, assuming that the mode is unique, the right and left
eigenvectors of the complex conjugate eigenvalues are also complex conjugates. Therefore,
the mode shape and participation factors (as described below) of the mode can be identified
by considering only the eigenvectors of one of the complex conjugate pair of eigenvalues.
In addition, in later sections we will present the concept of the response of the system to a
complex frequency, namely the modal frequency of a decaying oscillatory mode. Again, it
can be shown that it suffices to evaluate the transfer function at one of the two complex
conjugate eigenvalues. For these reasons, throughout the text we have sometimes referred
to an oscillatory mode as, say, mode h where h is the index or number of the first of two
complex conjugate eigenvalues which together constitute the mode.
The above items are valuable pieces of information but they do not answer the following
questions concerning the modes of the system:

What type of mode it is? (For example, is it primarily associated with the controller of
a FACTS device?)

What states participate in this mode, in what manner and to what extent? (Do the
rotor speed states of generators i and j both participate significantly in the oscillatory
mode h ?)

Can analysis reveal the behaviour of one group of generators with respect to other
groups in the case of the electro-mechanical modes?

Which generators participate significantly in the lightly-damped or potentially unstable

modes? (In practice it may be necessary to identify some or all of the rotor modes,
particularly those that fall into the categories of being unstable or lightly damped.)

We shall therefore, in the following sections, examine two methods which are used to identify the modes by resolving the above issues, namely, Mode Shape and Participation Factor
Analyses [5].

9.2 Mode Shape Analysis

In the analysis of the dynamic performance of multi-machine power systems, the concept
of mode shapes provides a practical and meaningful tool. In essence, mode shapes assist
one to identify the mode type such as inter-area, local-area, inter-machine / intra-station.
1. See Section 3.5.2 concerning the distinction between eigenvalues and modes.

450

Tuning of Multi-Machine PSSs

Ch. 9

The theoretical basis for mode shapes is outlined in Section 3.9. It was shown that if the state
equations of the dynamic system are excited by the right eigenvector v i of a selected mode
of rotor oscillation, i , only that mode appears in the time-domain responses of the states the responses for all other modes are zero; this is succinctly summarised by (3.45), namely
xi t = vi e

i t

(9.2)

The electro-mechanical or rotor modes of oscillation are usually identified with the perturbations of rotor speed about synchronous speed. The mode shape is therefore identified
mainly from the phase of the elements of the right speed-eigenvector of the selected mode.
Rather than considering a complex multi-machine system, the significance and application
of mode shapes are illustrated more simply - and in some detail - initially using a two-mass
spring system.
9.2.1 Example 1: Two-mass spring system
A two-mass spring system which is constrained to move freely in the positive x-direction
from a reference position is shown in Figure 9.1(a). The instantaneous position and speed
of the centre of mass j is x j t (m) and v j t (m/s), respectively, are highlighted in
Figure 9.1(b). M j is the mass (kg), B j is the viscous damping coefficient (N/m/s) between
the mass and the ground plane, K jk is the spring stiffness coefficient (N/m), and f j t is an
externally applied force (N).
v2

v1

x2

x1

K20

Figure 9.1

xj

K01

K12
M2

M1

B2

B1
(a)

vj

Mj

f1(t)
Reference plane,
node 0

Kij

Kjk

fj(t)

Bj
(b)

(a) A two-mass system free to move in the x-direction on a flat surface,

(b) the general form of the parameters and variables for the jth mass.

Based on Figure 9.1(b), a general form of the equation of motion for mass M j can be expressed as [6], [7], [8]:
dv j
f j = K ij x i + M j
+ B j v j + K ij + K jk x j K jk x k .
dt

This equation can be rewritten in state equation form as follows:

(9.3)

Sec. 9.2

Mode Shape Analysis

451

K ij
K jk
Bj
K ij + K jk
1
v j = ------ v j + ------- x i --------------------- x j + -------- x k + ------ f j ,
Mj
Mj
Mj
Mj
Mj

and

x j = v j .

(9.4)

Applying the above relationships to the two masses in turn, a fourth-order set of state equations is formed in the state variables v 1 v 2 x 1 x 2 ; the derivation of the set of equations is
left as an exercise to the reader.
Consider the following parameters for the four-mass spring system:
M 1 = 2 M 2 = 4 B 1 = 1 B 2 = 0.5 , and K 01 = 10 K 12 = 8 K 20 = 10.
For these values of the system parameters the eigenvalues of the system are given in
Table 9.1.
Table 9.1

Eigenvalues of the two-mass spring system

Eigenvalue number and value

1
-0.214+j3.21

2
-0.214-j3.21

3
-0.098+j1.77

4
-0.098-j1.77

We note that there are two stable oscillatory modes having damping ratios of 0.067 for mode
A (which is associated with the complex conjugate eigenvalue pair 1,2) and 0.055 for mode
B (eigenvalue pair 3,4). However, there is no information that reveals the nature of the system performance; for example, what is the relative characteristic behaviour of the masses
for mode A?
The right speed-eigenvectors for the two oscillatory modes are shown in Table 9.2. It is noted for mode A, when it alone is excited, that the speed states v 1 and v 2 of masses 1 and 2
are essentially in anti-phase. The mass M 1 is said to swing against mass M 2 . The displacement states x 1 and x 2 , which are almost in anti-phase, lag their respective speed states by
nearly 90 . When the right eigenvectors are normalised to 1 0 for the state with the largest
magnitude (the speed state v 1 for mode A, v 2 for mode B), the modal behaviour of the
states is interpreted more easily using the polar plots for the relevant modes as shown in
Figure 9.2.
Often in mode-shape analysis only the speed elements in the right eigenvector are plotted.
In this event the plot is the same as that in Figure 9.2 except all other states are omitted.

452

Tuning of Multi-Machine PSSs

Table 9.2

Ch. 9

Right eigenvectors for the oscillatory modes

Right eigenvectors

Mode A: 0.214 j3.21

Mode B: 0.098 j1.77

Magnitude

Angle

Magnitude

v1

0.904

180

0.492

-5.3

v2

0.309

-9.6

0.719

x1

0.281

86.2

0.277

-98.4

x2

0.096

-103.4

0.405

-93.2

State

Mode A: -0.214+/-j3.21

0.5

Mode B: -0.098+/-j1.77

0.5

v2

x2

v1

-0.5

v2

x1

-0.5

-1
-1

Angle

-0.5

0.5

-1
-1

v1

x1
x2

-0.5

0.5

Figure 9.2 Normalised right eigenvectors of speed (v) and displacement (x)
for the oscillatory modes
Let us now consider the time-domain responses of the states when the mass-spring system
is excited by the right eigenvector consisting of the real parts of its elements for each of the
T

modes in Table 9.2, e.g. by the initial condition 0.904 0.305 0.0188 0.0223 for mode
A. The transient response to this initial condition is shown in Figure 9.3.
Note in Figure 9.3 the instantaneous phase relationship between the states is consistent with
Figure 9.2 and/or Table 9.2. From the figure it is seen that

the time constant and the period of the response are consistent with the single mode,
0.214 j3.21 ;

Sec. 9.2

Mode Shape Analysis

453

0.8
0.6
0.4

Amplitude

0.2
0
-0.2
-0.4

v1
v2

-0.6

x1
-0.8
-1

x2
0

5
6
Time (s)

10

Figure 9.3 First ten seconds of the transient response speed (v) and displacement states
(x) for the two-mass spring system to an initial condition which excites only
mode A, 0.214 j3.21.

the speed states v 1 and v 2 as well as the displacement states x 1 and x 2 are, respectively, nearly in anti-phase;

x 1 and x 2 , respectively, lag v 1 and v 2 by nearly 90 ;

as might be expected for mode A, 0.214 j3.21 , in which the masses swing against
each other as shown in Figure 9.3, the amplitude of the oscillation of the smaller mass
is larger.

Likewise, as seen in Figure 9.4 if the system is excited by an initial condition

T

0.490 0.719 0.0407 0.0224 on the four states in Table 9.2, only mode B is excited. In
this case the speed states v 1 and v 2 as well as the displacement states x 1 and x 2 are, respec-

tively, nearly in-phase, i.e. the two masses swing together with respect to the reference
frame. Again, the form of the responses is consistent with the results in Figure 9.2 and/or
Table 9.2.

454

Tuning of Multi-Machine PSSs

Ch. 9

0.8
v1
v2
x1
x2

0.6

Amplitudes

0.4
0.2
0
-0.2
-0.4
-0.6
-0.8

5
Time (s)

10

Figure 9.4 First ten seconds of the transient response of the two-mass spring system to
an initial condition which excites only mode B, 0.098 j1.77 .
It is noted above for mode A, in which the masses swing in anti-phase, the amplitude of the
oscillation of the smaller mass is larger. This suggests that the nature of the oscillations observed in the responses of Figure 9.3 and Figure 9.4 are associated with the interchange of
energy between the energy storage elements. Let us calculate the instantaneous stored ener2

gies in the masses and the spring. The instantaneous stored energy in a mass is M j v j 2 and
2

that in a spring is K jk x j x k 2 . For mode A the time responses of the stored energy in
each of the five elements for the relevant initial conditions are plotted in Figure 9.5.
As is to be expected, the envelope of the decay of the stored energies decays with a time
constant of one-half of that of mode A 1. Further we note:

The stored energy in each of the two masses peak more-or-less simultaneously; at that
time the stored energy in each of the three springs is zero;

1. Assume that the response of a speed state of a mass is v t = V 0 e

2

2 2t

energy will decay as v t = V 0 e

. The stored

Sec. 9.2

Mode Shape Analysis

455

A quarter cycle later of the modal frequency (3.21 rad/s, period approximately 2 s),
the latter condition is reversed, i.e. the stored energies in the springs peak more-or-less
simultaneously; at that time the stored energy in each of the masses is zero.

If the losses during the interchange were zero (i.e. no viscous damping, B = 0 ), the
system would oscillate indefinitely with constant amplitude and the peaks and troughs
in the responses would coincide exactly.
Stored energy in each of 2 masses and 3 springs (joule)

0.8

Spring 0-1

0.7

Spring 1-2

0.6

Spring 2-0
Mass 1
Mass 2

0.5
0.4
0.3
0.2
0.1
0
0

10

Time (s)

Figure 9.5 Stored energy response in each of the masses and springs for an initial condition which excites only mode A, 0.214 j3.21 .
A plot of the stored energy responses, similar to Figure 9.5, for mode B ( 0.098 j1.77 )
can be predicted from the mode shape shown in Figure 9.2 or the amplitude responses of
Figure 9.4. This is left as an exercise to the reader.
The interchange of energy between energy storage elements every quarter of a cycle of the
oscillatory behaviour is explained in any text book on the fundamentals in physics or engineering. The significance of mode shapes in the analysis of dynamic performance is that it
reveals the nature of the behaviour of the masses (or inertias) in selected modes - normally
the electro-mechanical modes in power system dynamic performance.

456

Tuning of Multi-Machine PSSs

Ch. 9

This example illustrates that, for the two-mass-spring system, there is one oscillatory mode
representing the relative dynamic behaviour between the two masses. The second mode
portrays the behaviour of the masses with respect to the reference frame.
9.2.2 Example 2: Four-mass spring system
To highlight some further relevant issues a somewhat more complex mass-spring system
than that in Example 1 (Section 9.2.1) is analysed; the system is shown in Figure 9.6.
Unlike the previous example there are no springs restraining movement between the masses
and the reference plane. The previous example of the two-mass-spring system is simple
enough to demonstrate not only the concepts of mode shapes, but also the associated transient responses and the responses of the stored energy. However, the purpose of this example is to demonstrate for higher-order systems the types of interactions between elements
that are revealed through mode-shape analysis. Moreover, the more complex system provides additional insight into the use of participation factor analysis described in Section 9.3.
v4

v3

x4

v2

x3

v1

x2

K34

x1

K23

K12

f1(t)

M4

M3

M2

M1

B4

B3

B2

B1

Figure 9.6

A four mass, three-spring system

The following are the parameters for the four-mass spring system:
M 1 = 1 M 2 = 2 M 3 = 4 M 4 = 2 ; B 1 = 0.3 B 2 = 0.8 , B 3 = 2.4, B 4 = 0.6 , and
K 12 = 10 K 23 = 8 K 34 = 15. The units of these parameters are supplied in

Section 9.2.1.

The eight eigenvalues and five modes of system are given in Table 9.3.
Table 9.3

Eigenvalues and modes, A to E, of the four-mass system

Eigenvalue number and value

Mode A (Oscillatory)
1

Mode B Mode C
(Mono.) (Mono.)
3

Mode D (Oscillatory)
5

Mode E (Oscillatory)
7

-0.18+j4.12 -0.18-j4.12
-0.46
0
-0.19+j1.82 -0.19-j1.82 -0.21+j3.43 -0.21-j3.43
Mono.: a monotonically increasing or decaying mode

Sec. 9.2

Mode Shape Analysis

457

We note that there is a pole at the origin in the case of mode C (eigenvalue 4), the remaining
eigenvalues constitute modes that are stable, and that the oscillatory modes have a damping
ratio between 0.04 for mode A (eigen-pair 1,2) and 0.10 for mode D (eigen-pair 5,6).
In the analysis of multi-machine power system dynamics it is common practice to employ
the normalised right speed-eigenvectors in assessing mode shapes for the electro-mechanical modes. These eigenvectors of the oscillatory modes for the four-mass system are shown
in both Table 9.4 and the plots of Figure 9.7. It is observed for oscillatory mode A, the only
mode excited, that

the speed states of masses M 1 and M 3 move together essentially in anti-phase with
those of masses M 2 and M 4 ;

the lighter masses M 1 and M 2 have the larger amplitudes;

the frequency of oscillation of this mode is the highest of all the modes.
Table 9.4

Four mass system: Normalised right speed-eigenvectors for

the three oscillatory modes
Normalised right speed-eigenvectors for oscillatory modes

State Mode A: 0.18 j4.12 Mode D: 0.19 j1.82 Mode E: 0.21 j3.43
Magnitude

Angle

Magnitude

v1

1.00

v2

0.703

-177.9

0.668

v3

0.167

8.2

v4

0.131

-173.3

Angle

1.00

Magnitude

Angle

0.516

6.5

-1.1

0.096

-161.6

0.306

-178.3

0.579

-174.9

0.549

-176.6

1.00

The speed-eigenvector plot for the monotonically decaying real mode (B in Table 9.3) reveals that all masses move in-phase with respect to the reference when this mode is
excited 1. In the case of the oscillatory mode D (eigen pair 5,6), the lighter masses M 1 and
M 2 swing together against M 4 and the heaviest mass M 3 ; the frequency of oscillation is the
lowest of all the modes. Similarly, for mode E (eigen pair 7,8) the lighter masses M 4 and M 1
1. In Table 9.3 the elements of the speed eigenvector of mode C (eigenvalue 4), which represents a pole at the origin of the s-plane, are all zero. If mass 4 were attached to the reference plane through a spring with non-zero coefficient K40, one is likely to find that
modes B and C represent a fourth complex conjugate pair which would constitute a
common oscillatory mode in which all four masses oscillate in-phase against the reference.

458

Tuning of Multi-Machine PSSs

Ch. 9

swing predominantly against the heaviest mass M 3 . It is these types of phenomena that
mode shapes are particularly useful in revealing when this analysis is applied to the electromechanical modes of a multi-machine power system. For example, when two large groups
of generators swing against each other, the frequency of oscillation is typically low (e.g. 2 to
5 rad/s), but if a single generator swings against the rest of the machines the frequency tends
to be relatively much higher (e.g. 7-10 rad/s).
Mode A: -0.18+/-j4.12

1
0.5
0

0.5
x4

v2

x2

v4 x3
x1

-0.5
-1
-1

v1

v3

-0.5

0.5

Mode D:-0.19+/-j1.82

0.5
v4

Figure 9.7

v4 v3

x2 x
1

v1 v2

-0.5

v1
v2

x2

0.5

Mode E: -0.21+/-j3.43

x3
x2
v3

-0.5

0.5

x1
-0.5

-1
-1
1

x4
x3

v3

-0.5
-1
-1

x3 x4

-0.5

Mode B, Mono.: -0.46

0.5

-1
-1

-0.5

v2 x
1
x4

v1
1

0.5

v4

Normalised right speed-eigenvectors for the three oscillatory modes (A, D, E)

and the monotonically decaying real mode B.

In this example of a four mass-spring system, there are three oscillatory modes representing
the relative dynamic behaviour between the four masses. A fourth real or complex mode
typically portrays the behaviour of all four masses with respect to a reference. Typically, if
there are N masses, there are N-1 modes representing the dynamic characteristics of interactions between the masses. Instead of an analysis of masses which are subject to linear displacement an analysis of rotating masses can be conducted using a similar approach. Thus
for N generators connected to a multi-machine system, there are N-1 modes of rotor - or
electro-mechanical - oscillation representing the relative dynamic interactions / behaviour between the N rotating masses. The N modes are associated with N pairs of complex conjugate
eigenvalues - that is, a total of 2N eigenvalues.

Sec. 9.3

Participation Factors

459

However, a note of warning is appropriate here. The mode shapes do not reveal the relative
extent in which the set of states participate in a selected mode, or the relative extent for
which a selected state participates in the set of system modes. As seen in Figures 9.3 and 9.4
the elements of the right eigenvector v i in (9.2) represent the relative amplitude of the states
at time zero and thereafter. These elements are not dimensionless; the first four elements
have the dimensions of speed in m/s, the second four, displacement in m. The set of states
in a model of a multi-machine power system is comprised of states of very different types,
e.g. fluxes, control and excitation system variables, as well as rotor speed and angle. Again,
the elements of the right eigenvector are not dimensionless and would alter if the per-unit
system employed were changed. It is therefore not possible to measure the relative participation, say, of each of the system states in a selected mode based on the right eigenvector,
unless the measure is expressed in a dimensionless form. This is achieved using the concept
of participation factors.

9.3 Participation Factors

In the case of a multi-machine power system we may suspect from the mode shapes that the
speed state or a field voltage state of a generator is involved in a particular rotor mode of
oscillation, but how extensively are they involved, and how much relative to the involvement
of other generators? Although an analysis of modes shapes reveals the phase relation between speed states in a given rotor mode of oscillation, the participation in that mode by
some other state may be greater than that of the rotor speed state, e.g. field voltage (if the
PSS damping gain is set to a high value).
Firstly, if v h and w k are the right and left eigenvectors, respectively, it is shown in
Section 3.10, based on [5], that the dimensionless participation factor p hk = v kh w hk provides a measure of the relative extent to which each of the n states participates in the h th
eigenvalue at time t = 0 . Recall also that the sum of the participation factors for eigenvalue
h is unity.
Secondly, it is also shown in Section 3.10.1 that the participation factor also provides a measure of the relative extent to which eigenvalue h participates in state k at time t = 0 . Bear
in mind, however, that at some later time t = t 1 some modes will have decayed away and
only the more dominant modes prevail; consequently, the participation of the h th eigenvalue in state k will have changed from the values at t = 0 . The participation factor is thus a
meaningful measure only at t = 0 - which is the basis of its definition.
The first of the alternative forms of applying the participation factor concept is employed in
the following chapters.

460

Tuning of Multi-Machine PSSs

Ch. 9

9.3.1 Example 4.3

In Example 4.2 the mode shapes for a four-mass spring system are derived. Due to the few
number of states involved, let us now consider the associated participation factors for the
states in the relevant modes for this simple system.
The participation matrix can be calculated as the element by element product of the left eigenvector matrix and the transpose of the right eigenvector matrix. The hth row then represents the participation factors of the states in the hth eigenvalue, as shown in Tables 9.5 and
9.6. Likewise, the kth column represents the participation factors of the modes in the kth
state.
Table 9.5
Mode
No.
A. (Osc)
B. (Mon)
C. (Mon)
D. (Osc)
E. (Osc)

Table 9.6

Mode
No.
A. (Osc)
B. (Mon)
C. (Mon)
D. (Osc)
E. (Osc)

Participation factors for the speed states v i in the system modes.

Eigenvalue Eigenvalue
No
1
3
4
5
7

Participation factors 10 for the eight states

v1

-0.18+j4.12
23.5-j0.2
-0.46
10.8
0
0
-0.19+j1.82 17.5+j1.4
-0.21+j3.43
3.6+j0.7
Osc.: Oscillatory mode.

v2

v3

23.2+j1.5
2.5+j0.7
21.9
45.2
0
0
15.6+j0.7
6.5+j0.9
0.2+j0.1
18.4+j2.8
Mon.: Monotonic mode

v4
0.8+j0.2
22.2
0
10.4+j2.1
27.8-j0.7

Participation factors for the displacement states x i in the system modes.

Eigenvalue Eigenvalue
No
1
3
4
5
7

-0.18+j4.12
-0.46
0
-0.19+j1.82
-0.21+j3.43

Participation factors 10 for the eight states

x1

x2

x3

x4

23.4-j1.9
3.7
7.3
17.4-j1.5
3.7+j0.4

23.2-j0.8
2.7
19.5
15.4-j2.8
0.2+j0.1

2.6+j0.3
-14.2
58.5
6.6-j1.2
18.7-j0.4

0.8+j0.1
7.6
14.6
10.5+j0.4
27.5-j3.1

It can be confirmed that the sum of the participation factors covering the eight eigenvalues
is 1 + j0 . Note that the participation factors are complex but, as is often the case for the
larger factors, they are almost real. Therefore, for ease of interpretation the magnitudes of
the participation factors are plotted in the bar-chart form illustrated in Figure 9.8. The bar

Sec. 9.3

Participation Factors

461

chart for mode C, 4 = 0 , is not shown as the participation factors for the speed states are
all zero.
For this simple case the participation of the set of modes in a selected state can be read from
Figure 9.8, e.g. the magnitude of the participation factors for the eight eigenvalues, shown
in Table 9.3, in the speed state v 2 are 0.232 0.232 0.219 0 0.156 0.156 0.002 0.002 .
Mode A: -0.18+/-j4.12

Mono. Mode B: -0.46

v1
x1
x2
v2
x3
v3
v4
x4

v3
v4
v2
x3
v1
x4
x1
x2
0

0.1

0.2

0.3

0.4

Mode D: -0.19+/-j1.82

0.2

0.4

0.6

0.8

Mode E: -0.21+/-j3.43

v4
x4
x3
v3
v1
x1
v2
x2

v1
x1
x2
v2
v4
x4
x3
v3
0

0.05

0.1

0.15

0.2

0.1

0.2

0.3

0.4

Figure 9.8 Participation factors of the set of states v 1 v 4 x 1 x 4

in selected modes.
A comparison between the plots of the participation factors and the modes shapes of
Figures 9.8 and 9.7 is instructive. For oscillatory modes A, D and E the participation factors
of the pairs of states v i x i are almost identical but the magnitudes of the mode shapes for
the same pairs differ significantly, typically by a factor of two or more to one in this example.
As demonstrated through the two examples, very pertinent information on the dynamic performance characteristics of the system is provided by the combination of modes shapes and
the participation factors. It is the relative phase information provided by modes shapes that
is particularly useful; the relative amplitude of the states depends the units of the states. As
mentioned earlier if the system of units were changed, as in the selection of the per unit sys-

462

Tuning of Multi-Machine PSSs

Ch. 9

tem, the relative amplitudes of the states could change. On the other hand, the participation
factors indicate the relative degree of involvement not only of all the states in a mode on a
dimensionless basis but also of the modes in a state; the participation factors are therefore
a characteristic of the system, invariant to change in units. In the two examples the amplitudes of the right speed (note, only speed) eigenvectors for a given mode shape appear to
correlate fairly well with the participation factors for the same mode; this may lead to the
misconception that the amplitudes in the mode shape represent the participation of the
speed states in the selected mode.
The application of these tools will be demonstrated in analysing the dynamic behaviour of
a multi-machine power system in Chapter 10.

9.4 Determination of the PSS parameters based on the P-Vr

approach with speed perturbations as the stabilizing signal
9.4.1 The P-Vr transfer function in the multi-machine environment
Earlier, in Section 5.8, the tuning of a speed-input PSS for a generator in a single-machine
infinite-bus system is outlined in detail. The P-Vr transfer function is introduced in
Section 5.8.2, and its application to the tuning of the PSS compensating transfer function
G c s is described in the subsequent subsections of Chapter 5. It is pointed out in those sections that the PSS tuning and use of the P-Vr transfer function are applicable to multi-machine systems; this is the case, but issues such as interactions between PSSs outlined in a later
chapter need to be accounted for.
The comment in Section 5.8.2 for SMIB systems also applies to the tuning of PSSs in multi-machine systems, namely: The PSS must be tuned to be robust to a full range of N and N-1 operating
conditions. For this purpose it is necessary to select a set of appropriate operating conditions
which encompass, and therefore include, the full range of conditions. By examining the border-

ing conditions this approach reduces the number of cases for which the P-Vr characteristics
must be evaluated.
In the case of the multi-machine system the P-Vr transfer function is defined as follows: The
P-Vr transfer function for generator i is the transfer function from the voltage reference V ri s
of the AVR of generator i to the electrical torque output of the generator, P ei s , with the shaft
dynamics of all generators in the power system disabled. Note that shaft dynamics on all gen-

erator are disabled for the purposes of the analysis.

Diagrammatically, this shown by a comparison of Figure 9.9(a), for the intact generators,
and Figure 9.9(b) in which the shaft dynamics of all generators are disabled. An examination
of Figure 9.9(b) reveals that there are a number of signal-flow paths associated with the PVr transfer function. For example, not only is there the direct path from the reference voltage input V ri s on generator i to its electrical torque output P ei s - but there are also

Sec. 9.4

PSS parameters based on the P-Vr approach

463

paths through both the power system and the other generators to the electrical torque output on i. A question is: Compared to the SMIB system, do these additional paths diminish
the effectiveness of the P-Vr approach to PSS tuning?

Vti

Vti

P
O
W
P
i
ei
E
/(sMi)
R
S
Di
Pdi Pmi Y
S
T
o/s
E
i M

AVRi

Pei

i
Vri

Machine i in a n
generator system

Di

AVRi

Pdi
o/s

(a)

Vri

Machine i in a n
generator system

(b)

Figure 9.9 (a) Model of a generator in a multi-machine power system; (b) conceptually,
with shaft dynamics on all machines disabled.
The terminology used here, i.e. `P-Vr transfer function', as defined above for the multi-machine context is that introduced in [9]. However, this same transfer function has been determined by different techniques elsewhere. For example, for the tuning of the PSS of a
generator in a multi-machine power system, phase information on the P-Vr transfer function has been determined by field tests [10] or is based on SMIB models with the machine
inertia constant set to a very large value on the generator of interest [11], [12]. In references
[10], [11]and [12] no attempt is made to employ the P-Vr transfer function for the formal
tuning of PSSs in a multi-machine system - including the concept and setting of the PSS
damping gain, or as a basis for the coordination of PSSs.
The method adopted here for calculating the P-Vr transfer function is that presented in [9].
The significance of this approach is that a simple direct method is provided for determining
both the magnitude and phase response of the P-Vr transfer function for each generator.
The theoretical basis for the P-Vr characteristic of generator i, P ei s = i s V ri s , in
a multi-machine system of N generators is considered in [14], [15]. With the shaft dynamics
of all generators disabled it is shown that the electrical power or torque of the N generators
is given by
P e s = A v s V r s + B s s .

(9.5)

Furthermore, it is shown that matrices Av and B are essentially diagonal or block diagonal
due to the diagonal dominance property of the reduced network admittance matrix into
which the generator dynamic admittances are embedded as network elements. For brevity,

464

Tuning of Multi-Machine PSSs

Ch. 9

let us now consider only the first term in (9.5) associated with the P-Vr-like matrix of generators in the multi-machine system, i.e.
P e s = A v s V r s , where from [15],

(9.6)

Av s = G3 s Y G1 s G5 s + G6 s
= G 3 s Z s G 5 s + G 6 s

and Z s = Y G 1 s ,

(9.7)
(9.8)

and where Y is the reduced network admittance matrix. In the first term of the summation
in (9.7), G 3 s - and in (9.8) G 1 s - are essentially functions of the steady-state conditions
and are modified by the generator operational reactances x d s and x q s . Moreover, in
(9.7) the first term is determined mainly, and diagonally dominated, by the network admittance matrix Y. The Thvenin equivalent of the network as seen from the terminals of generator i is not much affected by the dynamics of the other generators in the system. The
second term G 6i s in (9.7) depends only on the parameters of the generator i, its excitation
system and a scalar multiplier vdo_i , the d-axis steady-state terminal voltage, i.e.:
G 6i s = v do_i G gen_i s G avr_i s x di s ,

(9.9)

where Ggen_i , Gavr_i and xdi are respectively the operational transfer functions of generator
i, its AVR / exciter, and its direct-axis synchronous reactance; these functions are independent of the external system.
The phase characteristic of G 6i s is independent of operating conditions in the external
system, however, the magnitude of the low-frequency response varies only with the scalar
gain vdo. The magnitude characteristic thus retains its shape over the range of operating conditions. Consider firstly the variation of v d0 with generator reactive power output at constant real power (P) and 1 pu terminal voltage as shown in Table 9.7. The low frequency gain
of the P-Vr characteristics decreases with increasing lagging reactive power (Q); this observation is reflected in the P-Vr characteristics of Figure 5.16 for the SMIB system.
Likewise, as illustrated in Table 9.8 and manifested in the P-Vr characteristics of Figure 5.22,
v d0 decreases with decrease in real power output at unity power factor. At rated power output vdo is relatively large, but tends to zero as the real power output is reduced.
A consideration of Table 9.7 suggests that it is prudent to include a range of reactive power
outputs in the set of encompassing operating conditions.

Sec. 9.4

PSS parameters based on the P-Vr approach

Table 9.7

465

Variation of v d0 with increasingly lagging Q at constant real power outputs

Corresponding values of Q and v d0
P=0.9, Q pu

-0.2

0.2

0.4

v d0 pu

0.930

0.851

0.766

0.686

P=0.7, Q pu

-0.2

0.2

0.4

v d0 pu

0.892

0.783

0.680

0.591

Table 9.8

Variation of v d0 with reduction in P at unity power factor

Corresponding values of P and v d0

Q=0, P pu

0.9

0.7

0.5

0.3

0.1

v d0 pu

0.851

0.783

0.669

0.475

0.177

Because G 6i s is the more significant term of the two in (9.7), it determines the consistently
narrow bands of the frequency responses of the P-Vr characteristics of generator i given by
A v_ii s = P ei s V ri s . For example, for the SMIB system the relatively minor variations in the phase of the P-Vr characteristic with steady-state operating conditions observed
in Figure 5.16 are caused by the contribution of first term in (9.7), G 3 s Z s G 5 s , over
the modal frequency range.
In [15] the authors imply that the P-Vr characteristic of generator i can be calculated if the
network is represented by a SMIB system connected at the generator terminals. However,
in multi-machine cases, it may not be clear what value should be attributed to the impedance
of the Thvenin equivalent, particularly as it will change with line outages, whether electrically close-by machines are on/off line, the effect of close-by loads, etc. It is then simpler
and more efficient to calculate the P-Vr characteristics of generator i 1 for each operating
condition using the complete model of the multi-machine system. Moreover, each generator
may participate in a range of local- and inter-area modes as well as intra-station modes, not
in a single mode as is the case in the SMIB system.
These results in [15] provide a theoretical basis for the observation in [13] that the P-Vr
transfer function is relatively robust to changes in the system operating conditions in multimachine systems. That is, for higher values of generator real power outputs both the gain,
phase and the shapes of the frequency response of the P-Vr transfer functions do not vary
appreciably, for practical purposes, over a wide range of operating conditions and system
1. Or the characteristics of generating station i if there are a number of identical units in
the station.

466

Tuning of Multi-Machine PSSs

Ch. 9

configurations. Consequently, in multi-machine systems, individual PSS designs that are based
on the synthesized P-Vr transfer function using the methodology adopted in Section 5.10 are also
robust over a wide range of operating conditions. Typically, this applies for generator real power outputs exceeding 0.5 pu. An examination of Figures 5.21, 5.22 and Tables 5.5 and 5.6
reveals that the mode shifts are essentially real, an observation which supports the above
statement.
The robustness and application of the P-Vr characteristic has been demonstrated and verified for generators on very large systems [20].
For

the

multi-machine

system

the

P-Vr

characteristics

of

each

generator,

H PVri j f = P ei j f V ri j f , are calculated in a similar fashion to that for the single-

machine infinite-bus system, except that the characteristics are calculated for the entire network with the shaft dynamics of all machines disabled. The calculation is similar to that described in Section 5.10.3 in which rows and columns of the A, B and C matrices associated
with the speed states in the states equations (3.9) are eliminated; the D matrix is usually a null
matrix. The relationship between perturbations in electric power (or torque) as the output
quantity and voltage reference as the input quantity can then be formed, and the frequency
response evaluated for the set of encompassing operating conditions and over the range of
modal frequencies.
The derivation of the synthesized transfer function,
,
H PVrS s = P ei s V ri s
i
Synth

(9.10)

which is selected from the family of P-Vr frequency response characteristics as the most
suitable basis for the tuning of the PSS, has been covered in Section 5.10.6.
9.4.2 Transfer function of the PSS of generator i in a multi-machine system
The basic concepts for the determination of the parameters of a PSS in a single machine system have been outlined in Chapter 5. The approach in the case of a generator in a multimachine systems follows along similar lines in Section 5.8.1 and therefore can be summarized fairly briefly.
Consider the model of generator fitted with a speed-PSS in multi-machine system as shown
in Figure 9.10. Note that the transfer functions from V ri or V si to P ei are identical.
It has been established in Section 5.9.1 that the PSS transfer function for a generator - say
the ith - takes the form
H PSS s = k i G i s = k i G Wi s G ci s G LPi s ,
i

(9.11)

Sec. 9.4

PSS parameters based on the P-Vr approach

467

where k i is the damping gain; G ci s is the PSS compensation block; G Wi s and G LPi s
are the transfer functions of the washout and low-pass filters, respectively.

Vti
i

Pei

/(sMi)

P
O
W
E
R

AVRi

S
Di P
Y
di Pmi
S
T
o/s
E
i M

Vsi

Vri

Machine i
PSSi

Figure 9.10

Model of generator i fitted with a PSS in a multi-machine power system

As has been outlined earlier in Section 5.14:

1.

The aim of the tuning procedure is to introduce on the generator shaft a damping
torque (a torque proportional to machine speed); this causes the modes of rotor oscillation to be shifted directly to the left 1 in the complex s-plane.

2.

The compensation transfer function G ci s is tuned to achieve the desired left-shift in

the complex s-plane of the relevant modes of rotor oscillation.

3.

The damping gain k i (on machine MVA rating) of the PSS determines the extent of
the left-shift.

Based on item 1, the ideal transfer function between speed i and the electrical damping
torque perturbations P ei due to the action of the PSS i over the range of modal frequencies should ideally be:
P ei s PSSi = D ei i s ,

(9.12)

where D ei is a damping torque coefficient and is a real number (p.u. on generator MVA rating). The transfer function G ci s compensates in magnitude as well as phase for the synthe-

1. By direct left-shift is implied that the mode shift is j0 , 0 . As explained in

Chapter 13, deviations from the direct left-shift of modes are mainly due to interactions
between multi-machine PSSs and non-real generator participation factors.

468

Tuning of Multi-Machine PSSs

Ch. 9

sized P-Vr transfer function of machine i, H PVrS s , defined in (9.10). With rotor speed
i

being used as the input signal to the PSS, whose output is V si s , the expression (9.12) for
D ei can also be expressed in terms of the P-Vr and PSS transfer functions as:
P ei s V si s
D ei = ------------------- ------------------- = H PVrS s k i G ci s ;
i
V si s i s

(9.13)

hence, rearranging (9.13), we find

k i G ci s = D ei H PVrS s .
i

(9.14)

It follows from an examination of (9.14) that

k i = D ei and G ci s = 1 H PVrS s ,
i

(9.15)

Note from (9.15) that k i is a damping torque coefficient. Assuming that the synthesized
transfer function H PVrS s is of the general form
i

1 + sT b1
i
-,
H PVrS s = k ci --------------------------------------------------------------------------i
1 + c 1i s + c 2i s 2 1 + sT a1

(9.16)

then, from (9.15), the compensation block transfer function is

2
1 1 + c 1i s + c 2i s 1 + sT a1i
-,
G ci s = ------ --------------------------------------------------------------------------k ci
1 + sT b1

(9.17)

where k ci c 1i c 2i T a1 T b1 are parameters determined from the synthesized P-Vr

i

characteristic in the tuning procedure. Note that in the form of (9.16) (i) real and complex
zeros can be accommodated in the synthesized P-Vr transfer function; (ii) the coefficient of
s 0 is unity.

Substitution of (9.17) in (9.11), and incorporating the washout and low pass filters, yields the
PSS transfer function:
2
sT Wi
1 1 + c 1i s + c 2i s 1 + sT a1 i
1
H PSSi s = k i --------------------- ------ ---------------------------------------------------------------------------- ------------------------------------------------------- .
1 + sT Wi k ci
1 + sT b1
1 + sT 1i 1 + sT 2i
i

(9.18)
For generator i, T Wi is the time constant of the washout filter; T 1i T 2i are the time constants of the low-pass filter which may be added (i) to ensure H PSSi s is proper, (ii) to mitigate against excitation of the torsional modes of the rotating turbine/generator/exciter
shaft system.

Sec. 9.5

Synchronising and damping torque coefficients

469

Earlier the gain k i has been referred to as the damping gain of the PSS. The gain k ci is the
DC gain of P-Vr characteristic of generator i. If the washout filter is ignored the DC gain of
the PSS transfer function is k i k c ; conventionally this is referred to as the PSS Gain. (However, the PSS gain k i k ci has been attributed little meaning because the significance of the
gain k ci has not been recognized.)
Note, assuming that the synthesized P-Vr characteristic H PSSi s for generator i closely
matches that for the selected operating condition, H PVr s , it follows that
i

H PVr s k i G ci s k i + j0 ,
i

(9.19)

(i.e. equal to the damping gain) over the modal frequency range 1. In the next chapter, by
examining the damping torque coefficient, this result will be used to confirm that the designed damping gain ki of PSS i is, in fact, achieved (see Section 10.6).
In the multi-machine PSS tuning methodology the PSS is designed not only to swamp any
negative (destabilizing) inherent damping torque coefficients on that machine over the range
of frequencies of the rotor modes, but also to provide sufficient damping so that the associated damping criteria of the multi-machine system are satisfied [9], [13]. These issues, together with the contribution to damping by stabilizers installed on FACTS devices, are
considered in a later chapters.
Examples of the application of the P-Vr approach to the tuning of PSSs in multi-machine
power systems have been presented in several publications [9], [13], [16], [17], [18] and [19].
In Chapter 10 the tuning the PSSs of generators in an inherently unstable 14-generator, multi-machine power system is described. Based on this system, the features of the PSS tuning
technique discussed above are illustrated through an example for which the complete system
data is provided.

9.5 Synchronising and damping torque coefficients induced by PSS i

on generator i
The concepts of synchronising and damping torques over a range of frequencies s = j f
are explained in the context of the single-machine system in Sections 5.3 and 5.5. The same
concepts are employed in the multi-machine application to assess the synchronising and
damping torque coefficients developed by PSS i on generator i. However, let us first assess
the significance of Figure 9.11 which is derived from Figure 9.10.

1. Note G ci s is the compensation which applies over the range of modal frequencies.
The washout and low-pass filter time constants lie outside the latter range and are not
included in G ci s , but are included in the PSS transfer function, H PSS s .

470

Tuning of Multi-Machine PSSs

Vti
Pei

i
Di

Pdi
Sdel
o/s

P
O
W
E
R
S
Y
S
T
E
M

AVRi

Ch. 9

Vri
Vsi

Machine i
PSSi

SPSS

Figure 9.11 Model of generator i in a multi-machine system with (i) shaft dynamics on all
machines disabled and (ii) switches Sdel and SPSS in rotor angle and PSS paths, respectively.
Consider generator i. When the shaft dynamics on all generators are disabled with switch
Sdel in the rotor angle path closed and the PSS out of service, the signal flow paths can be
deduced from Figure 9.11. As in the case of the SMIB system there are signal flow paths
directly from i through i to P ei (or to P 2 in Figure 5.2). However, in the multimachine case there are paths from i through the network to perturbations in rotor angles
and terminal voltages on other generators, then to the inherent torque output P ei . The
principle of superposition in linear systems analysis says that these paths remain when the
shaft dynamics are enabled and the full system is reinstated.
The object of the following analysis is to determine in the multi-machine cases if the PSS
performance is consistent with its design basis, i.e. if H PVr j f k i G ci j f k i + j0 , over
i

the modal frequency range for an encompassing range of operating conditions. In other
words, is the per unit damping gain ki of the PSS the realised? This objective is illustrated in

Figure 9.11 when the PSS is in service with switch SPSS closed and the rotor angle path is
open by means of switch Sdel.
The synchronising and damping torque coefficients for generator i are defined in Section 5.3
and apply to generator i in Figure 9.11:
f P ei j f
P ei j f
k si = ------ ------------------------ and k di = ------------------------ ,
w 0 i j f
i j f

(9.20)

i.e. components on generator i of torques in quadrature and in phase with rotor speed on
unit i.

Sec. 9.6

References

471

The theoretical basis for concept of the inherent synchronising and damping torques for
generators in a multi-machine is derived from [14], [15] and the associated equation (9.5).
Recall that the shaft dynamics of all generators are disabled (i.e. switch Sdel is closed and SPSS
is open in Figure 9.11. The torque-like relationship of interest for generator i is extracted
from (9.5), i.e.
P e s = B s s , where

(9.21)

B _ii s G 3i s Z ii s G 2i s + G 4i s .

(9.22)

In (9.22) G 2i s , G 3i s and G 4i s are essentially functions of the steady-state conditions

and are modified by the generator operational reactances x d s and x q s . Notice that the
first term, through (9.8), is determined mainly by the reduced network admittance matrix Y.
According to [14] both terms in (9.22) are essentially functions of the steady-state conditions
at the generator terminals. However, the first term in (9.22) also involves the Thvenin
equivalent of the network as seen from the terminals of generator i, and the generator parameters. Unlike the second term G 6i s in (9.7), G 4i s does not display phase invariance
in its frequency response over a wide range of operating conditions. This implies that, unlike
the P-Vr characteristics, the inherent synchronising and damping torques coefficients for
generators in a multi-machine system do vary with changes in operating conditions. Moreover, both G 2i s and G 4i s incorporate a multiplying factor 0 s with the result that at
high and low frequencies the frequency response rolls off at 20 dB/decade and exhibits a
constant phase of 90 . This observation is illustrated in the SMIB cases by a comparison
of the inherent torque coefficients in Figures 5.5(a) and 5.18.
As implied in Figure 9.9(a), in reality with the shaft dynamics of all generators enabled, there
are also paths through the power system, the AVRs, and the other generators. A disturbance
in the speed of generator j will therefore induce a torque on the generator i, the unit of interest, at the inter-area and other modal frequencies. The components of this torque in
phase, or lagging by 90 , with the speed of generator i are damping or synchronizing torques
induced on unit i by j. The associated torques coefficients, without and with PSSs in operation and with the shaft dynamics of all generators enabled, are analysed in Chapter 12 and are
referred to as Modal Induced Torque Coefficients.

9.6 References
[1]

G. Golub and C. F. Van Loan, Matrix Computations, Third Edition, John Hopkins
University Press, 1996.

[2]

W. E. Arnoldi, The principle of minimized iterations in the solution of the matrix eigenvalue problem, Quarterly of Applied Mathematics, vol. 9, pp. 1729, 1951.

472

Tuning of Multi-Machine PSSs

Ch. 9

[3]

K. J. Bathe, The subspace iteration method Revisited, Comput Struct (2012),

http://dx.doi.org/10.1016/j.compstruc.2012.06.002.

[4]

D. J. Vowles and M. J. Gibbard, Mudpack User Manual: Version 10S-03, School of

Electrical and Electronic Engineering, The University of Adelaide, July 2014.

[5]

F. L. Pagola, I. J. Perez-Arriaga and G. Verghese, On sensitivities, residues and participations: Applications to oscillatory stability analysis and control, Power Systems,
IEEE Transactions on, vol. 4, pp. 278-285, 1989.

[6]

J.DAzzo and C.H. Houpis, Linear Control System Analysis and Design, Conventional
and Modern, McGraw-Hill International Editions, 3rd Edition, 1988.

[7]

Gene Franklin, J.D. Powell and Abbas Emami-Naeini, Feedback Control of Dynamic
Systems, 5th edition, Prentice Hall, October 2005.

[8]

R.C. Dorf and R.H. Bishop, Modern Control Systems, 10th edition, Prentice Hall, April
2004.

[9]

M. J. Gibbard, Coordinated design of multimachine power system stabilisers based

on damping torque concepts, IEE Proc. Part-C, pages 276-284, July 1988.

[10]

E. V. Larsen and D.A. Swann, Applying power system stabilizers: Part I-III, IEEE
Trans. PAS, pp. 3017-3046, 1981.

[11]

P. Kundur, M. Klein, G. J. Rogers, and M. S. Zywno, Application of power system

stabilizers for enhancement of overall system stability, Power Systems, IEEE Transactions on, vol. 4, pp. 614-626, 1989.

[12]

F.P. De Mello, J. S. Czuba, P. A. Rusche, and J. Willis, Developments in application

of stabilizing measures through excitation control, International Conference on Large
High Voltage Electric Systems, CIGRE, paper 38-05, 1986.

[13]

M. J. Gibbard, Robust design of fixed-parameter power system stabilizers over a

wide range of operating conditions, Power Systems, IEEE Transactions on, vol. 6, pp
794-800, 1991.

[14]

D. M. Lam, Eigenvalue analysis and stabilizer design for electrical power systems, PhD
Thesis, University of Sydney, 1995.

[15]

D. M. Lam and H. Yee, A study of frequency responses of generator electrical torques for power system stabilizer design, Power Systems, IEEE Transactions on, vol.
13, pp. 1136-1142, 1998.

[16]

M.J. Gibbard, D.J. Vowles and P. Pourbeik, Interactions between, and effectiveness
of, power system stabilizers and FACTS stabilizers in multimachine systems, Power
Systems, IEEE Transactions on, vol. 15,pp. 748-755, 2000.

Sec. 9.6

References

473

[17]

M.J. Gibbard, D.J. Vowles, Reconciliation of methods of compensation for PSSs in

multimachine systems, Power Systems, IEEE Transactions on, vol. 19, pp. 463-472,
2004.

[18]

CIGRE Technical Brochure no. 166 prepared by Task Force 38.02.16, Impact of Interactions among Power System Controls, published by CIGRE in August 2000.

[19]

M.J. Gibbard, N. Martins, J.J. Sanchez-Gasca, N. Uchida, V. Vittal, and L. Wang,

Recent Applications in Linear Analysis Techniques, Power Systems, IEEE Transactions on, vol. 16, pp. 154-162 2001.

[20]

P. Pourbeik, T. Cain, and R. Bottoms, Application of small-signal stability tools and

techniques to a large power system, in The Proceedings of the IEEE Power & Energy
Society General Meeting, 2009. IEEE PES '09, pp. 1-11.

Chapter 10
Application of the PSS Tuning Concepts
to a Multi-Machine Power System

10.1

Introduction

The previous chapter introduced some important concepts in the tuning of PSSs in multimachine power systems. The purpose of this chapter is to demonstrate the application of
the associated techniques for the analysis and tuning of PSSs in a fourteen-generator power
system which, without continuously acting PSSs, is inherently unstable. Each generator in
this system, in fact, represents a power station which accommodates between one and
twelve units; the number of units in-service (nu) depends on the particular operating condition. The units in a power station are assumed to be identical, therefore the rating of the
equivalent generator for a station is nu times the rating of a single unit. It is assumed that the
individual generators in each power station are fitted with identical excitation systems and
PSSs.
In a later chapter a class of stabilizers known as Power Oscillation Dampers (PODs) are discussed; these are stabilizers that can be fitted to power-electronic based transmission devices
such as FACTS (e.g. Static Var Compensators) and HVDC transmission. The analysis and
tuning of POD stabilizers are demonstrated by means of examples in Chapter 11. In the
fourteen-generator power system described in this chapter the Static Var Compensators
(SVCs) are fitted with continuously acting voltage regulators controlling bus voltage, but are
not fitted with stabilizers.

475

476

PSS Tuning in Multi-Machine Systems

Ch. 10

The steps in the tuning of PSSs of machines in a multi-machine system are explored, commencing with (i) the eigen-analysis of the system with all PSSs out of service, and (ii) the associated analysis based on Mode Shapes and Participation Factors. The PSSs are then tuned
using the P-Vr approach discussed in Section 9.4. Having completed the determination of
the PSS parameters, the effect on the shifts of eigenvalues associated with the rotor modes
are assessed as the damping gains of the PSSs are increased; ideally over the range of operating conditions such shifts are directly to the left in the complex s-plane.
In practice a new power station is built to supply energy to an existing power system in
which many of the existing generators may already be fitted with PSSs. The latter PSSs
would have been tuned and their parameters set to fixed values. The PSSs in a new power
station have to be tuned to satisfy the damping and other performance criteria of the system
operators over the range of system operating conditions and contingencies. However, in the
following example the PSSs fitted to all generators are tuned at the same time, and the effect
on damping established as the PSS damping gains are increased from zero to 30 pu on machine base. This analysis reveals a number of issues that are not found in the analysis associated with the new power station in an existing system. Nevertheless, the approach adopted
in the example is applicable to the tuning of PSSs for additional generation.
Earlier work has investigated the tuning of PSSs to adequately damp both local- and interarea modes [1] [2]. It will be demonstrated that the design of PSSs based upon the P-Vr concept inherently damps both types of modes.
Although each power station is represented by a single composite generator formed from
the nu units in service, it is often necessary to represent the individual machines in the station. Because the PSS tuning techniques do not directly determine the nature of the intrastation modes (i.e. modes of oscillation between machines in a single power plant), the effects of the PSS tuning on these modes are examined in Section 10.8 to assess their characteristics. If the damping of the intra-station modes is poor, it will be necessary to determine
what action needs to be taken to remedy the problem.
Normally the main emphasis is placed on the dynamic performance of the multi-machine
power system following large-signal disturbances. Such disturbances are major faults on the
system, switching of heavily-loaded transmission lines, the tripping of a generator, the loss
of a significant load, etc. Notwithstanding the non-linear nature of the limiting action of
controllers immediately following the fault, the dynamic performance is determined by the
non-linear nature of differential-algebraic equations. The question arises: what is the relevance and significance of small-signal analysis to the analysis and understanding of large-signal dynamic behaviour? These issues are investigated in Section 10.9 by examining the
transient response of the 14-generator system following the incidence of a major fault in a
critical location. Furthermore, interesting recent developments are establishing a bridge
between small- and large-signal analysis; this has been achieved by including the second-order terms in the Taylor series expansion about the steady-state operating condition as dis-

Sec. 10.2

Fourteen generator model of a power system

477

cussed in Section 10.9.2. To provide further understanding of the nature of the system
behaviour, concepts of modal interactions and their significance with respect to both
large- and small-signal dynamic performance are briefly discussed.

10.2

A fourteen-generator model of a longitudinal power system

The simplified system of 14 power stations 1 is shown in Figure 10.1. It represents a long,
linear system as opposed to the more tightly meshed networks found in Europe and the
USA. For convenience, the system has been divided into 5 areas in which areas 1 and 2 are
more closely coupled electrically. There are in essence 4 main areas and hence 3 inter-area
modes, as well as 10 local-area modes. Without PSSs installed on generators in this system,
many of these modes are unstable.
For the purpose of designing generator PSSs in practice a wide range of both normal operating conditions and contingencies 2 are considered. However, to simplify the procedures
for illustrative purposes in this and the following chapters, a limited number of cases encompassing a range of fairly diverse, normal conditions is employed. The encompassing range of
operating conditions 3, system loads and major inter-area flows are listed in Table 10.1.
Table 10.1

Load Condition

Six normal steady-state operating conditions

Case 1

Case 2

Case 3

Case 4

Case 5

Case 6

Heavy

Mediumheavy

Peak

Light

Medium

Light

Total generation (MW)

23030

21590

25430

15050

19060

14840

Total load (MW)

22300

21000

24800

14810

18600

14630

Inter-area flows

(North to
south)

(South to
north)

(Area 1 to N (Area 2 to N
& S)
& S)

(N & S to
Area 1)

(~Zero
transfers)

Area 4 to Area 2 (MW)

500

500

-500

-200

300

Area 2 to Area 1 (MW)

1134

1120

-1525

470

740

270

Area 1 to Area 3 (MW)

1000

1000

1000

200

-200

Area 3 to Area 5 (MW)

500

500

250

200

250

1. In the analysis a power station with n units on-line is represented as a single generator.
Consequently the station is often referred to as a generator
2. These are referred to as N and N-1 conditions respectively.
3. The term encompassing range of operating conditions in defined Section 5.1. It is
assumed that, for the subsequent analysis, a reduced set of steady-state conditions are
selected which encompass those conditions for which the stabilizers are to be tuned.

478

PSS Tuning in Multi-Machine Systems

Ch. 10

For the six cases the ratings of generators, the number of units on line, and their real and
reactive power outputs are listed in Table 10.2. The power stations are designated *PS_<area number>, e.g. HPS_1 refers a power station (PS) called H in area 1. Note that the number of units on-line in certain stations can vary considerably over the range of operating
conditions. A number of the units in the hydro station HPS_1 can operate as synchronous
compensators, or as synchronous motors driving pumps in a pump-storage mode of operation.
Table 10.2

Generation conditions for six power flow cases.

Case 6:
Case 5:
Case 4:
Case 3
Case 2
Power Station/Bus Case 1
No. units No. units No. units No. units No. units No. units
#
MW
MW
MW
MW
MW
MW
Rating
Mvar
Mvar
Mvar
Mvar
Mvar
Mvar
Rated power factor
HPS_1 / 101
12 x 333.3 MVA
0.9 power factor lag

4
75.2
77.9

3
159.6
54.4

12
248.3
21.8

2
0
-97.4
Syn.Cond

3
-200.0
-26.0
Pumping

2
0
-102.2
Syn.Cond

BPS_2 / 201
6 x 666.7 MVA
0.9 power factor lag

6
600.0
95.6

5
560.0
38.9

6
550.0
109.1

4
540.0
-30.8

5
560.0
38.7

3
560.0
-53.5

EPS_2 / 202
5 x 555.6 MVA
0.9 power factor lag

5
500.0
132.7

4
480.0
60.5

5
470.0
127.6

3
460.0
-2.5

4
480.0
67.2

3
490.0
-7.3

VPS_2 / 203
4 x 555.6 MVA
0.9 power factor lag

4
375.0
132.8

3
450.0
82.4

2
225.0
157.0

3
470.0
9.4

2
460.0
83.1

3
490.0
3.7

MPS_2 / 204
6 x 666.7 MVA
0.9 power factor lag

6
491.7
122.4

4
396.0
17.8

6
536.0
96.5

4
399.3
-43.6

4
534.4
55.2

3
488.6
-61.2

LPS_3 / 301
8 x 666.7 MVA
0.9 power factor lag

7
600.0
142.3

8
585.0
141.1

8
580.0
157.6

6
555.0
16.6

8
550.0
88.1

6
550.0
9.4

YPS_3 / 302
4 x 444.4 MVA
0.9 power factor lag

3
313.3
51.5

4
383.0
63.3

4
318.0
49.6

2
380.0
-9.3

3
342.0
43.8

2
393.0
-6.9

TPS_4 / 401
4 x 444.4 MVA
0.9 power factor lag

4
350.0
128.7

4
350.0
116.5

4
350.0
123.2

3
320.0
-21.9

4
346.0
84.9

3
350.0
-32.6

Sec. 10.2

Fourteen generator model of a power system

479

Case 6:
Case 5:
Case 4:
Case 3
Case 2
Power Station/Bus Case 1
No. units No. units No. units No. units No. units No. units
#
MW
MW
MW
MW
MW
MW
Rating
Mvar
Mvar
Mvar
Mvar
Mvar
Mvar
Rated power factor
CPS_4 / 402
3 x 333.3 MVA
0.9 power factor lag

3
279.0
59.3

3
290.0
31.4

3
290.0
32.0

2
290.0
-2.4

3
280.0
45.4

3
270.0
4.7

SPS_4 / 403
4 x 444.4 MVA
0.9 power factor lag

4
350.0
52.3

4
350.0
47.2

4
350.0
47.3

3
320.0
14.2

4
340.0
46.3

2
380.0
25.2

GPS_4 / 404
6 x 333.3 MVA
0.9 power factor lag

6
258.3
54.5

6
244.0
39.8

6
244.0
40.0

3
217.0
-3.5

5
272.0
50.4

3
245.0
3.9

NPS_5 / 501
2 x 333.3 MVA
0.9 power factor lag

2
300.0
25.3

2
300.0
-8.8

2
300.0
6.5

2
280.0
-52.5

2
280.0
-35.2

1
270.0
-42.2

TPS_5 / 502
4 x 250 MVA
0.8 power factor lag

4
200.0
40.1

4
200.0
53.0

4
180.0
48.8

3
180.0
-1.8

4
190.0
0.1

4
200.0
-9.7

PPS_5 / 503
6 x 166.7 MVA
0.9 power factor lag

4
109.0
25.2

5
138.0
36.9

6
125.0
32.6

1
150.0
2.2

2
87.0
3.5

2
120.0
-11.2

10.2.1 Power flow analysis

Data for the power flow analysis of the six normal operating conditions given in Table 10.1
are supplied in Appendix 10I.2. Included in Appendix 10I.2 are relevant results of the
analysis such as reactive outputs of generators and SVCs, together with tap positions on generator and network transformers. This information permits the power flows to be set up on
any power-flow platform and the results checked against those provided in this document.
10.2.2 Dynamic performance criterion
The dynamic performance criterion requires

that all modes are stable;

for all normal and N-1 system conditions the damping of the electro-mechanical
modes is to be such that the associated halving times are 5 s or less.

The halving time is defined as the time for the mode or its envelope to decay to half its
initial amplitude. The real parts of the electro-mechanical modes must therefore be less than
= 0.139 (since exp 5 = 0.5 ) to satisfy the latter requirement.

480

PSS Tuning in Multi-Machine Systems

Ch. 10
406

SPS_4

GPS_4
407

403

404
405

CPS_4

AREA 4

408

402

409
401
TPS_4

410
411

413
414
412

415
416

BPS_2

ASVC_2 SVC

SVC

BSVC_4
VPS_2

207

201

205

203

206
208

EPS_2
202
209
MPS_2

211

NPS_5

210
204

501

215

504

212
213

TPS_5
502

AREA 2

214
505

PSVC_5
507

506

216

508

503
PPS_5

AREA 5

SVC

217
102

509

SVC

HPS_1

SSVC_5
311

315

310

101

305

309

308
307

301

LPS_3
AREA 3
YPS_3

RSVC_3
314

Figure 10.1

303

304

306

SVC

AREA 1

313

312

302

Simplified fourteen-generator system.

Sec. 10.3

Eigen-analysis, mode shapes and participation factors

481

10.3 Eigen-analysis, mode shapes and participation factors of the 14generator system, no PSSs in service
In order to gain some insight into the dynamic performance and characteristics of the system, a series of analyses is conducted without - and later with - PSSs in service on all generators.
10.3.1 Eigenvalues of the system with no PSSs in service
The preliminary objective of the eigen-analysis is to identify the nature of the unstable and
lightly-damped modes.
Let us consider the eigen-analysis of Case 1, a heavy load condition, in the 14-generator
power system with no PSSs or SVC stabilizers in service. In this case there are 125 states and
consequently 125 modes. Because there are Ng=14 generators in service there are Ng-1=13
rotor modes that reflect the modal interplay between generation. A fourteenth real or complex mode typically portrays the behaviour common to all fourteen rotating masses with respect to a reference (see Section 9.2.2).
For this simple system the eigenvalues are calculated using an algorithm that employs QR
factorisation. The unstable modes are displayed either as a listing of the eigenvalues, or on
a plot in the complex s-plane. For Case 1 such a plot, with eigenvalue designations, is shown
in Figure 10.2 for a limited region about the positive imaginary axis.

A
D
F
J

E
G

K
M

Figure 10.2 Plot of eigenvalues for Case 1, no PSSs in service

482

PSS Tuning in Multi-Machine Systems

Ch. 10

Figure 10.2 reveals that there are five unstable oscillatory modes, one stable oscillatory mode
that does not satisfy the dynamic performance criterion and seven other lightly-damped oscillatory modes with damping ratios less than 0.1.
Valuable information on the stability of the modes is provided by the eigenvalue plot but it
does not reveal the type or nature of modes. In this case it is desirable to identify all thirteen
electro-mechanical modes, particularly those which are unstable or are lightly damped. Participation Factor and Mode Shape Analyses are employed for this purpose.
10.3.2

Application of Participation Factor and Mode Shape Analyses to Case 1

Consider in Figure 10.2 the unstable, oscillatory mode 0.088 + j2.60 (designated Mode L).
Let us view the plots not only of the magnitudes of its participation factors (PFs) but also
of its mode shape (MS); the plots are shown in Figure 10.3.

Figure 10.3 Magnitude of the participation factors (left) and the mode shape (right)
for the unstable mode L, 0.088 + j2.60 . No PSSs are in service.
(In the plot of the participation factors W and DEL are the rotor speed and angle perturbations, respectively.)
Recall that the concepts of participation factors (PFs) and mode shapes (MSs) were discussed in Chapters 3 and 9. In this case the participation factor is the participation of the
states in the selected mode arranged in decreasing values of the magnitude of the PFs. For
the selected mode the mode shape is the plot of the normalised magnitude and phase of the

Sec. 10.3

Eigen-analysis, mode shapes and participation factors

483

right speed eigenvectors and reveals, for example, that a group of generators swing with - or
against - another group of machines.
According to the plot of the PFs in Figure 10.3 the two states, rotor speed and angle, of a
number of generators dominate the involvement of the states in mode L; this mode is therefore an electro-mechanical mode. There are a total of 125 states in this system model. The
MS reveals that the generators in Areas 5 and 4 swing against those in Areas 1 and 2; the
participation factors of those in Area 3 are small. Mode L is therefore classified as an interarea mode. However, note that:

although the magnitude of the MS phasor of generator NPS_5 is the second largest,
the PF of its speed state is the twelfth largest;

as highlighted in Section 9.2.2, some care should therefore be attached to interpreting

the lengths of the MS phasors. The length of the MS phasors for some generators is
shorter than for others because their inertias on system MVA base may be significantly greater. In Figure 10.3, for example, the relative lengths of the MS phasors for
BPS_2 and PPS_5 are in the ratio 0.274:1, the ratio of their inertias is 2.56:1. The most
useful feature of the MS plot is therefore the relative phase information that it provides.

the PFs for mode L are nearly real (e.g. PF is 0.101 + j0.013 for both the speed and
rotor angle states of PPS_5). When in a later Chapter 13 we analyse the mode shift
contributed by the PSS of a given generator we shall find that the complex value of its
PF plays a major role [4].

For a second unstable mode, 0.115 + j3.97 , the PF plot in Figure 10.4 reveals that this mode
is also an electro-mechanical mode; the MS shows that generators in Areas 3 and 1 swing
against machines in Areas 2 and 5. This mode, called K, is also an inter-area mode.
For reference in later studies the PFs and MS for the third inter-area mode (M) are shown
in Figure 10.5. In this case Areas 5, 3 and 2 swing against Area 4.

484

PSS Tuning in Multi-Machine Systems

Ch. 10

Figure 10.4 Participation factors and mode shape for the unstable mode K
( 0.115 + j3.97 ), no PSSs are in service.

(a)

(b)

Figure 10.5 Case 1. Participation factors and mode shape for the inter-area mode M
( 0.016 + j2.03 ), no PSSs are in service.

Sec. 10.3

Eigen-analysis, mode shapes and participation factors

485

To determine the nature of other lightly-damped or unstable oscillatory modes with frequencies between 7 and 11 rad/s, shown in Figure 10.2, the plots of their MSs and PFs are
examined. Such plots are displayed in Figure 10.6; each plot reveals a rotor mode of oscillation. All three are found to be local-area modes:

in mode 0.17 + j10.4 VPS_2 swings against EPS_2;

in the unstable mode 0.11 + j9.58 SPS_4 swings mainly against CPS_4 and GPS_4;

in the unstable mode 0.04 + j8.96 BPS_2 swings mainly against EPS_2, VPS_2 and
TPS_4.

Figure 10.6 Case 1. Participation factors and mode shapes for three lightly-damped
modes, A, B & C, respectively 0.17 + j10.4 0.11 + j9.58 0.04 + j8.96 ; the latter
two modes are unstable.
The behaviour and type of the thirteen electro-mechanical modes in the fourteen machine
system are summarised in Table 10.3.
Though not shown in the eigen-plot of Figure 10.2 there is an oscillatory mode at about
1.4 j2.8 which could be of interest since it lies in the frequency range of the rotor modes.
In the PF plot, shown in Figure 10.7, it is observed that the states mainly participating in the
mode are associated with the direct axis of the generators at NPS_5, i.e. the field voltage and
the AVR. Thus, this mode is likely to be a controller mode associated with the AVR and

486

PSS Tuning in Multi-Machine Systems

Ch. 10

generator dynamics of NPS_5. Because such an examination of the PFs of a selected mode
quickly establishes the nature of the mode, participation factor analysis proves to be a very
useful tool.
Table 10.3

Behaviour and type of the electromechanical modes, Case 1;

no PSSs in service

Mode
No.

Real

Imag

A
B
C
D
E
F
G
H
I
J
K
L
M

-0.17
0.11
0.04
-0.56
-0.26
-0.61
-0.44
0.01
-0.19
-0.62
0.12
0.09
-0.02

10.44
9.58
8.96
8.63
8.37
8.05
7.96
7.81
7.72
7.43
3.97
2.60
2.03

0.02
-0.01
-0.01
0.06
0.03
0.08
0.06
-0.00
0.02
0.08
-0.03
-0.03
0.01

Mode Behaviour
VPS_2<-->EPS_2, BPS_2
SPS_4<-->CPS_4, GPS_4
EPS_2, VPS_2<-->BPS_2
NPS_5<-->TPS_5
CPS_4, SPS_4<-->GPS_4, TPS_4,
HPS_1, MPS_2<-->EPS_2, VPS_2, LPS_3
MPS_2, HPS_1<-->EPS_2, BPS_2, VPS_2
TPS_4<-->GPS_4, SPS_4, MPS_2
YPS_3, MPS_2<-->LPS_3, EPS_2
PPS_5<-->TPS_5, NPS_5
Area 3 <--> Area 5, Area 2
Area 5, Area 4 <--> Area 2
Area 5, Area 3 <--> Area 4

Mode Type
Local Area

Local Area
Inter-area

Inter-area

- damping ratio.
<--> means ... swings against ....
In Mode Behaviour, generators or areas are listed in descending order of their participation factors.

Figure 10.7 Participation factor plot for an oscillatory mode that participates mainly in
states associated with the direct axis of NPS_5.
The analysis of the behaviour of the electro-mechanical modes demonstrated above for
Case 1, Table 10.3, is repeated for the other cases 2 to 6 for all PSSs out of service. In Tables
10.4, 10.15 and 10.16 the modes for each case are sorted such that each row contains the

Sec. 10.4

P-Vr characteristics of the generators

487

modes of the same behaviour and type. For example,. in Table 10.4 the modes J in row 10
for Cases 3 and 4, 0.58 j7.62 and 0.19 j7.20 , respectively, are modes in which the
same generators are the main participants and both are local-area modes. This type of information will prove useful in a later chapter.
Table 10.4

Rotor modes of oscillation and damping ratios, Cases 3 and 4,

peak and light loads. No PSSs in service

Mode
No.
A
B
C
D
E
F
G
H
I
J
K
L
M

Case 3. Peak load

Case 4. Light load

Real

Imag

Real

Imag

-0.38
0.10
-0.30
-0.58
-0.18
-0.13
-0.14
-0.19
-0.08
-0.58
0.01
0.02
-0.03

11.11
9.56
9.02
8.66
8.48
6.31
8.26
7.91
7.38
7.62
4.08
2.67
2.05

0.03
-0.01
0.03
0.07
0.02
0.02
0.02
0.02
0.01
0.07
-0.00
-0.01
0.01

0.20
0.03
-0.17
-0.51
-0.18
-1.54
-0.56
-0.43
-0.21
-0.19
0.17
0.02
-0.01

10.48
9.67
9.37
8.52
8.78
8.28
8.58
8.21
8.28
7.20
4.74
3.57
2.68

-0.02
-0.00
0.02
0.06
0.02
0.18
0.07
0.05
0.03
0.03
-0.03
-0.01
0.00

is the damping ratio

10.4 The P-Vr characteristics of the generators and the associated

synthesized characteristics
The basis for the P-Vr characteristics is outlined and illustrated in Chapter 5 for single-machine infinite-bus systems. The extension for their application to multi-machine systems is
explained and illustrated in Section 9.4. This section highlights the calculation of the P-Vr
characteristics, an examination of their forms, and the synthesis of a P-Vr characteristic representing a set of P-Vrs of a particular generator.
For each of the 14 generators the P-Vr characteristics are determined (with all shaft dynamics disabled) and are shown in each of Figures 10.8 to 10.21 for the six operating conditions;
each characteristic is in per unit on the generator rating given in Table 10.23. These characteristics, determined with all shaft dynamics disabled, are calculated using a software package
for the analysis of the small-signal dynamic performance and control of large power systems
[5].

488

PSS Tuning in Multi-Machine Systems

Ch. 10

To avoid unnecessary complexity it should be noted in this analysis that the limited number
of encompassing operating conditions on which the power flows - and thus the P-Vr characteristics - are based are normal operating conditions. In practice, the P-Vr characteristics
for a relevant encompassing set of contingency conditions must be included in the determining the synthesized characteristic.
Examination of Figures 10.8 to 10.21 reveals that, over the modal frequency range of 1 to
15 rad/s, the bands of P-Vr characteristics 1 for any generator under normal operating conditions may possess the following features:

Magnitude plots: The width of the bands is typically less than 6 dB; the variation about
a characteristic lying in the centre of the band is therefore 3 dB or less.

Phase plots: The maximum width of the bands at the relevant frequency is typically
less than 15 ; the variation about a central characteristic is thus 7.5 or less.

Magnitude (dB)

Magnitude (dB)

20
0
20
40
60
1
10

10

10

10

50
100
150
200
1
10
Case 1
Case 5

0
10
20
30
1
10

10

10

10

0
Phase (deg)

Phase (deg)

10

10
10
Frequency (rad/s)

Case 2
Case 3
Synthesized PVr

Figure 10.8 P-Vr Xstics, HPS_1

10

50
100
150
200
1
10

10
10
Frequency (rad/s)

Case 1
Case 2
Case 4
Case 5
Synthesized PVr

Case 3
Case 6

Figure 10.9 P-Vr Xstics, BPS_2

1. The word characteristics is shortened to Xstics in the following figure captions.

10

Sec. 10.4

P-Vr characteristics of the generators

20
Magnitude (dB)

Magnitude (dB)

20
0
20
40
1

10

10

10

10

50
100
150
200
250
1
10

10
10
Frequency (rad/s)

Case 1
Case 2
Case 4
Case 5
Synthesized PVr

Case 3
Case 6

20
30
1
10

10

10

10

50
100
150
0

10
10
Frequency (rad/s)

Figure 10.11

10

Case 3
Case 6

P-Vr Xtics, MPS_2

10
Magnitude (dB)

Magnitude (dB)

10

Case 1
Case 2
Case 4
Case 5
Synthesized PVr

20
10
0
10
0

10

10

10

0
10
20
30
40
1
10

10

10

10

0
Phase (deg)

0
50
100
150
1
10

200
1
10

10

Figure 10.10 P-Vr Xtics, EPS_2

20
1
10

10

0
Phase (deg)

Phase (deg)

Phase (deg)

489

10
10
Frequency (rad/s)

Case 1
Case 2
Case 4
Case 5
Synthesized PVr

Case 3
Case 6

Figure 10.12 P-Vr Xtics,VPS_2

10

50
100
150
200
1
10

10
10
Frequency (rad/s)

Case 1
Case 2
Case 4
Case 5
Synthesized PVr

Case 3
Case 6

Figure 10.13 P-Vr Xtics, LPS_3

10

490

PSS Tuning in Multi-Machine Systems

Magnitude (dB)

20

0
20
40
60
1
10
0
50
100
150
200
250
300
1
10

10

10

10

10
10
Frequency (rad/s)

Case 1
Case 2
Case 4
Case 5
Synthesized PVr

Case 3
Case 6

20
30
1
10

10

10

10

50
100
150
0

10
10
Frequency (rad/s)

10

Case 3
Case 6

Figure 10.15 P-Vr Xtics, CPS_4

20
Magnitude (dB)

Magnitude (dB)

10

Case 1
Case 2
Case 4
Case 5
Synthesized PVr

20
10
0
10
20
30
1
10

10

10

10

10
0
10
20
1
10

10

10

10

0
Phase (deg)

0
Phase (deg)

200
1
10

10

Figure 10.14 P-Vr Xtics, YPS_3

50
100
150
200
1
10

10

0
Phase (deg)

Phase (deg)

Magnitude (dB)

20

Ch. 10

10
10
Frequency (rad/s)

Case 1
Case 2
Case 4
Case 5
Synthesized PVr

Figure 10.16

10

Case 3
Case 6

P-Vr Xtics, GPS_4

50
100
150
1
10

10
10
Frequency (rad/s)

Case 1
Case 2
Case 4
Case 5
Synthesized PVr

Figure 10.17

Case 3
Case 6

P-Vr Xtics, SPS_4

10

P-Vr characteristics of the generators

20
10
0
10
20
30
40
50
1
10

10

10

Magnitude (dB)

Magnitude (dB)

Sec. 10.4

10

Phase (deg)

Phase (deg)

0
50
100
150
200
1
10

10
10
Frequency (rad/s)

Case 1
Case 2
Case 4
Case 5
Synthesized PVr

20
10
0
10
20
30
40
50
1
10

Case 3
Case 6

P-Vr Xtics, TPS_4

10

10

Phase (deg)

Phase (deg)

100
150
200
0

10
10
Frequency (rad/s)

Case 1
Case 2
Case 4
Case 5
Synthesized PVr

Figure 10.20

10

10

10
10
Frequency (rad/s)

10

Case 3
Case 6

P-Vr Xtics, NPS_5

20
10
0
10
20
30
1
10

10

10

10

0
50

250
1
10

0
50
100
150
200
250
300
1
10

10

Figure 10.19

10

20
10
0
10
20
30
40
1
10

Case 1
Case 2
Case 4
Case 5
Synthesized PVr

Magnitude (dB)

Magnitude (dB)

Figure 10.18

10

491

10

Case 3
Case 6

P-Vr Xtics,TPS_5

50
100
150
200
1
10

10
10
Frequency (rad/s)

Case 1
Case 2
Case 4
Case 5
Synthesized PVr

10

Case 3
Case 6

Figure 10.21 P-Vr Xtics, PPS_5

For each set of generator P-Vr characteristics a synthesized P-Vr characteristic is derived
based on the following:

492

PSS Tuning in Multi-Machine Systems

Ch. 10

The synthesized characteristic is a best fit of a generators P-Vr characteristics for the
range of cases examined over the modal frequency range of interest, 1.5 to 15 rad/s.
As outlined in Section 5.10.6.1 the best fit characteristic for these studies is considered to be that lying in the centre of the magnitude and phase bands formed by the PVr characteristics 1.

If particular P-Vr characteristics tend to lie outside the bands formed by the majority
of the characteristics, the synthesized P-Vr may be offset towards the band formed by
the majority (e.g. see Figures 10.17 and 10.19). However, weighting of P-Vrs depends
on knowledge of the system, the contingencies and engineering judgement.

The transfer function of the synthesized P-Vr characteristic, PVR(s), for each of the 14 generators is given in Table 10.5.
In several figures, e.g. Figures 10.12 and 10.21 for generators VPS_2 and PPS_5 respectively, the bands of the low-frequency responses for the magnitude plots are much wider than
those in other figures, e.g. Figure 10.20 for TPS_5. An examination of the generation conditions for the six power flow cases in Table 10.2 reveals that the generator real power outputs vary from 45% to 98% of rated real power for VPS_2, and 60% to 100% for PPS_5;
on the other hand, the variation for TPS_5 is much smaller, 90-100%. These observations
are consistent with those in Section 5.11 and Figure 5.16, namely, that the low-frequency
magnitude response (the gain) of the P-Vr characteristic decreases as the real power output
of the generator is reduced. This phenomenon is explained in Section 9.4.1. It is shown that
the gain of the P-Vr characteristic varies only with the scalar gain vdo, the steady-state d-axis
component of the terminal voltage, but retains its shape over the range of operating conditions. At rated power output vdo is relatively large, but tends to zero as the real power output
is reduced. However, from 70% to 100% of real power output the magnitude characteristic
is, for practical purposes, lie within a band of less than 3 dB from the Design Case 2. (Similarly, at constant real power output the magnitude of the gain decreases as the reactive power output is varied from maximum leading to maximum lagging power factor. See Tables
9.7 and 9.8).
The more-or-less invariant nature of the phase responses of the P-Vr characteristics is also
explained in Section 9.4.1.

1. A least squares estimation procedure, or the MATLAB Signal Processing Toolbox routine invfreqs.m, can be employed to determine.the parameters for the synthesized transfer function.
2. Design Case is defined in Section 5.10.6.1.

Sec. 10.5

Synthesized P-Vr and PSS transfer functions

Table 10.5
Generation

10.5

493

Transfer functions of synthesized P-Vr characteristics, PVR(s)

PVR(s)
2

HPS_1

PVR s = 1.3 1 + s0.373 + s 0.0385

BPS_2
EPS_2

PVR s = 3.6 1 + s0.128 + s 0.0064

PVR s = 4.3 1 + s0.286 1 + s0. 111 1 + s0.040

MPS_2

PVR s = 3.0 1 + s0.01 1 + s0. 1 + s 0.0051

VPS_2

PVR s = 3.5 1 + s0.0292 1 + s0.0708

LPS_3

PVR s = 1.6 1 + s0.168 + s 0.0118

YPS_3

2
PVR s = 3.35 1 + s0.05 1 + s0.509 + s 0.132

CPS_4

PVR s = 4.25 1 + s0.278 1 + s0.100

GPS_4

2
PVR s = 3.3 1 + s0.115 + s 0.00592

SPS_4

PVR s = 3.16 1 + s0.0909 + s 0.00207

TPS_4

PVR s = 2.8 1 + s0.208 1 + s0.208

NPS_5

PVR s = 5.13 1 + s0.3 1 + s0.033 1 + s0.3 + s 0.111

TPS_5

PVR s = 3.4 1 + s0.500 1 + s0.0588 1 + s0.0167

PPS_5

5.62 1 + s0.350 1 + s0.0667

PVR s = ---------------------------------------------------------------------------------------------------------------- 1 + s0.02 1 + s0.167 1 + s0.187 1 + s0.2

The synthesized P-Vr and PSS transfer functions

Because the forms of the transfer functions of the synthesized P-Vr transfer functions - and
consequently those of the PSS compensation and low-pass filters - vary significantly between generators, it is instructive to list the parameters that have been evaluated for all PSSs.
The parameters of the compensation transfer function of the PSS are based on those of the
synthesized P-Vr transfer function given by (5.45). However, a more general form of the
synthesized function, which includes (say) additional poles and zeros as required by the form
of the design-case P-Vr, is
1 + sT b1
i
-.
H PVrS s = k ci --------------------------------------------------------------------------2
i
1 + c 1i s + c 2i s 1 + sT a1
i

The transfer function of the associated speed PSS, the structure of which is shown in
Figure 10.22, incorporates the compensation transfer function and the other elements as described by (10.1), i.e.

494

PSS Tuning in Multi-Machine Systems

Ch. 10

sT Wi
1 + c 1i s + c 2i s 2 1 + sT a1i
1
H PSSi s = k i --------------------- K ci ---------------------------------------------------------------------------- -----------------------------1 + sT Wi
1 + sT b1i
1 + sT 1i

(10.1)

K ci = 1 k ci .

where

sTW
1+sTW
Washout
Filter

Figure 10.22

Compensation
transfer function

LP Filter

Vs

Damping
Gain

Structure of the PSS for analysis and design purposes

The damping gain, the compensation transfer function and the low-pass filter can be included in the one structure if the relevant number and type of blocks are provided in the PSS.
Blocks which can accommodate complex poles and zeros are desirable in the PSS structure,
as will be seen from the form of PSS transfer functions in (10.4) and (10.6) below.
The damping gain for all PSSs is assumed to be k i = 20 pu on generator MVA rating, a value that is assumed to be a medium value of gain. Similarly, in all PSSs the washout time constant T W is set at 7.5 s. Its corner frequency of 0.133 rad/s is more than a decade below the
lowest inter-area modal frequency of about 2 rad/s; the phase lead it introduces at the latter
frequency is therefore small, less than 5 . Omitting the damping gain ki and the washout
filter, the compensation transfer function and the low-pass filter are combined in the following transfer function (TF), i.e.
1 + c 1i s + c 2i s 2 1 + sT a1
i
-.
H ci s = K ci ---------------------------------------------------------------------------------------- 1 + sT b1i 1 + sT 1i 1 + sT 2i

(10.2)

However, for simplicity and illustrative purposes in this example the time constants of the
low-pass filters are all selected to be 0.00667 s; the associated corner frequency is 150 rad/
s 1. The reason for the selection of the value of the corner frequency is to reduce phase lags
introduced by the filter on the phase lead provided by the compensation at the frequencies
of the local-area modes, e.g. for three low-pass filter poles the filter contributes a phase lag
of 11 at 10 rad/s. These issues have been discussed in more detail for the single machine
case in Sections 5.8.5 and 5.8.6.
Form of the third-/fourth-order compensation TF having real zeros, and a low-pass filter:

Based on (10.2) the form of this TF follows in (10.3); its parameters are listed in Table 10.6.
1. For illustrative purposes the very short time constants (6.7 ms) of the low-pass filter are
used here to minimise its influence in the range of modal frequencies. However, such
time constants should typically be 3 or more times the cycle time of the PSS processor to
reduce phase errors at higher frequencies.

Sec. 10.5

Synthesized P-Vr and PSS transfer functions

495

1 + sT a 1 + sT b 1 + sT c 1 + sT d
H c s = K c ------------------ ------------------ ------------------ ------------------ .
1 + sT e 1 + sT f 1 + sT g 1 + sT h

Table 10.6

(10.3)

Compensation and LP Filter Parameters

for PSS based on (10.3).

Generator

Kc

Ta

Tb

Tc

Td

EPS_2

0.233

0.286

0.111

0.040

PPS_5

0.178

0.200

0.187

0.167

0.020

TPS_5

0.294

0.500

0.0588

0.0167

Te

Tf

0.00667* 0.00667*
0.350

0.0667

0.00667* 0.00667*

Tg

Th

0.00667*

0.00667* 0.00667*
0.00667*

* Low-pass filter parameters

Form of the fourth-order compensation TF having real and complex zeros, and a low-pass filter:

Based on (10.2) the form of this TF is:

2
1 + sT a 1 + sT b
1 + as + bs
H c s = K c ------------------ ------------------ --------------------------------------------- .
1 + sT 1 + sT 1 + sT 1 + sT
d

(10.4)

The associated parameters are given in Table 10.7:

Table 10.7

Compensation and LP Filter Parameters

for PSS based on (10.4)

Ta

Tb

Td

Te

Tf

Tg

0.333

0.010

0.10

0.0051

0.00667*

0.00667*

0.00667*

YPS_3

0.298

0.050

0.5091

0.1322

0.00667*

0.00667*

0.00667*

NPS_5

0.195

0.033 0.033

0.30

0.1111

0.300

0.00667*

0.00667*

0.00667*

Generator

Kc

MPS_2

* Low-pass filter parameters

Form of the second-order compensation TF having real zeros, and a low-pass filter.

The form of this low-order TF is:

1 + sT a 1 + sT b
H c s = K c ------------------ ------------------ .
1 + sT e 1 + sT f

Its parameters are provided in Table 10.8.

(10.5)

496

PSS Tuning in Multi-Machine Systems

Table 10.8

Ch. 10

Compensation and LP Filter Parameters

for PSS based on (10.5)

Generator

Kc

Ta

Tb

Te*

Tf*

TPS_4

0.357

0.2083

0.2083

0.00667

0.00667

CPS_4

0.235

0.2777

0.1000

0.00667

0.00667

VPS_2

0.286

0.0708

0.0292

0.00667

0.00667

* Te and Tf are low-pass filter parameters

Form of the second-order compensation TF having complex zeros, and a low-pass filter
The TF is:
2

1 + as + bs
H c s = K c -------------------------------------------- .
1 + sT e 1 + sT f

Table 10.9

(10.6)

Compensation and LP Filter Parameters

for PSS based on (10.6)

Generator

Kc

Te*

Tf*

HPS_1

0.769

0.3725

0.03845

0.00667

0.00667

BPS_2

0.278

0.1280

0.00640

0.00667

0.00667

LPS_3

0.625

0.1684

0.01180

0.00667

0.00667

GPS_4

0.303

0.1154

0.005917

0.00667

0.00667

SPS_4

0.316

0.0909

0.002067

0.00667

0.00667

* Te and Tf are low-pass filter parameters

10.6 Synchronising and damping torque coefficients induced by PSS

i on generator i
The concepts of synchronising and damping torques coefficients are explained in the context of the single-machine system in Sections 5.3 and 5.10.6.3. The basis for the application
of the same concepts in the multi-machine case is explained in Section 9.5 to assess the synchronizing and damping torque coefficients developed by PSS i on generator i. The object
of the following analysis is to determine if the PSS transfer function k i G ci j f is consistent
with its design basis, i.e. H PVr j f k i G ci j f k i + j0 (eqn. (9.19)), over the modal frei

Sec. 10.6

Synchronising and damping torque coefficients

497

quency range for the selected operating condition. In other words, recalling that ki is also a
damping torque coefficient, is the desired per unit damping gain ki of PSS i realized?
The relevant part of Figure 9.11 is shown in Figure 10.23 in which the rotor dynamics on all
generators are disabled.

Vti
Pei

i
Di

Pdi
Sdel
o/s

P
O
W
E
R

AVRi

S
Y
S
T
E
M

Vri
Vsi

Machine i
PSSi

Figure 10.23

SPSS

Model of generator i fitted with a PSS in a multi-machine power system;

shaft dynamics on all machines are disabled.

The damping torque coefficient is defined in Section 5.3 and applies to generator i in
Figure 10.23:
P ei j f
k di = ------------------------ .
i j f

(10.7)

Firstly, with the path through Sdel in Figure 10.23 closed, and SPSS open, let us examine the
inherent frequency responses of the torque coefficients P ei i for several generators in
the fourteen-generator system when the individual machines are either heavily or lightly
loaded (see Tables 10.2 and 10.10). The responses are shown in Figure 10.24 in per unit on
generator base.
As anticipated in Section 9.5, at high and low frequencies the frequency response characteristically rolls off at 20 dB/decade and exhibits a constant phase of 90 . Over the range of
modal frequencies, however, the phase varies about 90 implying that the inherent damping torque coefficient is negative when the phase is less than 90 , and positive when greater
than 90 . Unlike the P-Vr characteristics, it appears that it is not possible to characterize
the variation of the damping torque coefficients; as foreshadowed in Section 9.5 the torque
coefficients depends mainly on the steady-state conditions of the generator.

498

PSS Tuning in Multi-Machine Systems

Table 10.10

Generator loading and study case-numbers

Heavily loaded

Units

Ch. 10

Case

Number &
output*

Lightly loaded
Case

BPS_2
1
6 @ 100%
4
4 @ 90%
VPS_2
6
3 @ 98%
3
2 @ 45%
PPS_5
4
1 @ 100%
5
2 @ 58%
* Number of equally loaded units on-line and percentage of
rated real power output

Magnitude (dB)

80
70
60
50
40
30
20

Phase Response (deg)

10
1
10

10

10

10

10

60
70
80
90

100
1
10

10
10
Frequency (rad/s)

BPS_2 Case 1 Hy
VPS_2 Case 6 Hy
PPS_5 Case 4 Hy

Case 4 Lt
Case 3 Lt
Case 5 Lt

Figure 10.24 Frequency responses of the inherent torque coefficients P ei i when

the unit is either heavily (Hy) or lightly (Lt) loaded (in pu on unit rating)
Secondly, let us examine separately the frequency responses of the inherent and the PSS-induced synchronizing and damping torque coefficients for generators whose inherent torque

Sec. 10.6

Synchronising and damping torque coefficients

499

coefficients are negative at low frequencies. For the calculations the shaft dynamics are disabled and, as above, in the case of the inherent torque coefficients the PSS path is open. For
the PSS-induced synchronizing and damping torque coefficients the rotor-angle path is
open and the PSS path closed (see Figure 10.23). The responses are shown in Figure 10.25
for two generators for the operating condition Case 1; the coefficients are in per unit on generator rating.

Damping Torque (pu)

25
20
15
10
5
0

Synch. Torque (pu)

5
2
10

10

10

10

10

10

4
2
0
2

10

10
10
10
Frequency (rad/s)

EPS_2 PSS only

LPS_3 PSS only

EPS_2 Inherent
LPS_3 Inherent

Figure 10.25 Case 1. Synchronizing and damping torque coefficients for EPS_2 (5x555.6
MVA) and LPS_3 (7x666.7 MVA); the coefficients are in per unit on generator base.
Over the range of modal frequencies, 1.5 to 15 rad/s, the following are observed from
Figure 10.25.
(i) The PSS gain is more-or-less flat at the desired damping gain setting of 20 pu on generator MVA rating; the deviations from 20 pu are typically accounted for by the factors listed
in Section 5.10.6.3. (Thus the question raised at the beginning of this section (is the damping
gain ki realized?) - is successfully answered.)
(ii) For each machine the positive damping torque coefficient induced by the PSS swamps
the inherent negative damping torque coefficient.
(iii) It is desirable to attenuate the PSS output signal at higher frequencies to avoid exciting
torsional modes at 50 rad/s or greater. This is achieved by means of an integral-of-acceler-

500

PSS Tuning in Multi-Machine Systems

Ch. 10

ating-power PSS (see Section 8.5); other types of PSSs may require the use of torsional
notch-filters as mentioned in Chapter 8.
The significance of the above analysis and observations is that they confirm - or provide a
check - that the PSS transfer function designed for each machine is being realized.
It will be shown that more meaningful information on synchronizing and damping torque
coefficients can be derived through Modal Induced Torque Coefficients - which are the
subject of Chapter 12.

10.7 Dynamic performance of the system with PSSs in service

10.7.1 Assessment of dynamic performance based on eigen-analysis
PSSs are assumed installed on all generators in each of the power stations. Two units only
are on-line at HPS_1 in Cases 4 and 6 and are operating as synchronous compensators. In
Case 5, however, three units at HPS_1 are operating as pumps with their PSSs in service.
Due to the motoring action in the latter case the sign on the PSS output for HPS_1 units is
negated.
The damping gain on each PSS is set to 20 pu on generator MVA rating. In Table 10.11 is
shown the values of the modes of rotor oscillation for a heavy and a light load condition,
Cases 1 and 4, with the PSSs out of and in service. Also shown are mode shifts and damping
ratios for each of the modes. The corresponding set of results are provided for remaining
cases in Tables 10.15 and 10.16.
It is instructive to track the shifts in the rotor modes as the PSS damping gains are jointly
increased from zero to 150% of the damping gain of 20 pu on machine MVA rating. For
Cases 1 to 6 the modes shifts for each of modes A to M are tracked in Figure 10.26 as the
damping gain on all unit is increased in 25% (5 pu) steps; the value of 100% corresponds to
20 pu on machine MVA rating. The plots in the figure are
(a) Case 1 (heavy load),
(b) Case 2 (medium-heavy load),
(c) Case 3 (peak load),
(d) Case 4 (light load),
(e) Case 5 (medium load), and
(f) Case 6 (light load).
See Table 10.3 for the details of the nature and types of modes A to M.

Rotor modes and modes shifts for heavy and light loads, Cases 1 and 4.
All PSS damping gains are 20 pu on generator MVA rating..

Case 1. Heavy load

No PSSs

A
B
C
D
E
F
G
H
I
J
K
L
M

Case 4. Light load

All PSSs in service

Mode Shift

No PSSs

All PSSs in service

Real

Imag

Real

Imag

Real

Imag

Real

-0.17
0.11
0.04
-0.56
-0.26
-0.61
-0.44
0.01
-0.19
-0.62
0.12
0.09
-0.02

10.44
9.58
8.96
8.63
8.37
8.05
7.96
7.81
7.72
7.43
3.97
2.60
2.03

0.02
-0.01
-0.01
0.06
0.03
0.08
0.06
-0.00
0.02
0.08
-0.03
-0.03
0.01

-2.19
-1.98
-1.93
-2.51
-1.95
-1.97
-1.87
-1.78
-2.06
-1.89
-1.04
-0.39
-0.52

10.39
9.74
9.29
8.86
8.26
8.49
7.76
7.64
7.87
7.59
3.64
2.40
1.80

0.21
0.20
0.20
0.27
0.23
0.23
0.23
0.23
0.25
0.24
0.28
0.16
0.28

-2.02
-2.09
-1.97
-1.95
-1.69
-1.36
-1.44
-1.79
-1.87
-1.26
-1.16
-0.47
-0.51

-0.06
0.16
0.33
0.22
-0.11
0.44
-0.21
-0.17
0.15
0.16
-0.33
-0.20
-0.23

0.20 10.48 -0.02

0.03 9.67 -0.00
-0.17 9.37 0.02
-0.51 8.52 0.06
-0.18 8.78 0.02
-1.54 8.28 0.18
-0.56 8.58 0.07
-0.43 8.21 0.05
-0.21 8.28 0.03
-0.19 7.20 0.03
0.17 4.74 -0.03
0.02 3.57 -0.01
-0.01 2.68 0.00

Mode Number.

Imag

Mode Shift

Real

Imag

Real

Imag

-2.37
-2.16
-2.27
-2.49
-2.27
-1.69
-2.50
-2.28
-2.55
-1.32
-1.08
-0.56
-0.59

10.77
9.95
9.81
8.83
8.79
8.17
9.06
8.28
8.44
7.49
4.58
3.32
2.51

0.22
0.21
0.23
0.27
0.25
0.20
0.27
0.27
0.29
0.17
0.23
0.17
0.23

-2.57
-2.19
-2.10
-1.98
-2.09
-0.15
-1.94
-1.85
-2.34
-1.13
-1.25
-0.59
-0.58

0.29
0.29
0.44
0.30
0.01
-0.11
0.48
0.07
0.17
0.29
-0.16
-0.25
-0.17

Dynamic performance with PSSs in service

No.

Sec. 10.7

Table 10.11

is the damping ratio..

In Case 4 the PSS of HPS_1 is OFF as the machine operates as a synchronous compensator.

501

502

PSS Tuning in Multi-Machine Systems

12

12

B
B
B
B
B
B
C
C
C
C
C
C
D
D
D
D
D
F
F
F
E FE F E F EF E
E
E
I H GI HG I GH I H
I
G I GH G
H GJ HJ J J J J J

=0.1

=0.2

Case 1

0
4

K K
K K

LL
LLLLL
MMM
MMMM

2
1
Real Part (Np/s)

(a)

10

G D

B C B C B
C B C BC B
G D G D G DG
DG
DG
D E
E
EF
F
FE
F
FE
EF
EF
J J H J HJ JH J H J H
H
I
I
I
I
I
I
IH
=0.1

=0.2

K K K K K
K K
LL L L
LLL
MM M
M
MM M

Case 2

0
4

2
1
Real Part (Np/s)

(b)

A
B

C
C
C D C D C D C
D
EG EDG
EG
EG
EG
EG
E
H
H
H
H
J H J HJ J J J J
I
I
I
I
I
IH
I

CD
G

=0.1

=0.2

K K K
K K

Case 3
2

LL L
LLLL
M
M M MMM

A
BC

10
Imaginary Part (rad/s)

10
Imaginary Part (rad/s)

12
A

CB

CB

CB

CB

B
C
G
G
GD E G E
G
E
D
D
E
D
D GD GD
E
E
I HI H
H I H IH
FFFFFF H
H I
J J J J J
J

A
B

C
E
I

J
=0.1

=0.2
K

K K K K K
LLL L
LLL
M
M MMMMM

Case 4
2

0
4

2
1
Real Part (Np/s)

(c)

3
2
1
Real Part (Np/s)

(d)

12

12

10

B
D G
F

E
I F
H

B
B
C
C
G
D
D
E
E
F I FH
I H
J J J

A
B

C
G
D
E
I FH
J J

B
B
C
C
C
G
G
E D E D EG
IFH IFH I H
J J
=0.1

Case 5

=0.2

4
PSS damping gain (%):
2 150 125 100 75 50 25 0

K
K K K K
K K
LL
L
L
L
LL
MMMM

MMM

AB
C

10
Imaginary Part (rad/s)

Imaginary Part (rad/s)

C B

12

Imaginary Part (rad/s)

Imaginary Part (rad/s)

A
A

A
B
C
D

10

Ch. 10

8 I

AB

D
E
H

D
E

AB
C

AB
C

AB
AB
C
C
G
C
G
G
D
D
D
D ED G
E
IH E H I EHFF IE
I
H
H
FFFFF
HI
J J J J J
J J
C
G

A
B

=0.1

=0.2
K

Case 6
2

LL L L
LL L
M MMMMM M

Local area modes: A to J;

Interarea modes: K to M.

0
4

(e)

2
1
Real Part (Np/s)

(f)

3
2
1
Real Part (Np/s)

Figure 10.26 Tracking of rotor modes for values of PSS damping gain 0 to 150% (30 pu)

Sec. 10.7

Dynamic performance with PSSs in service

503

It is noted:
1.

the modes, particularly the local-area modes, shift more-or-less horizontally to the
left in the complex s-plane;

2.

the extent of the left shift is least for the peak and heavily-loaded conditions, and
most for the light-load cases;

3.

for a given Case and a selected mode the extent of the left shift for each 25% increment in damping gain is fairly uniform; the amount of the left shift for the selected
mode varies from Case to Case;

4.

items 1 to 3 above satisfy the definition for robustness in item 3 of Section 1.2;

5.

a variation to items 1 to 3 above applies to the inter-area mode L. The increment in

the left shift progressively reduces at the higher values of the damping gains; this
phenomenon will be discussed in Chapter 14 (there is inadequate damping support
for mode L). Employing the P-Vr concept to determine the parameters for decentralized PSSs is shown in the above studies to improve the damping performance of
both the local- and inter-area modes. It will be demonstrated in Chapter 13 that the
smaller increments in the mode shift of the inter-area modes with PSS gain are due
(i) the smaller values of the participation factor of generators participating in the
mode, and (ii) the affect of interactions between their PSSs [4].

10.7.2 Assessment of dynamic performance based on participation and modeshape analysis

It is interesting to establish if the nature of the rotor modes for this system have changed
between the case when all PSSs are out of service to that when all are in service with the
damping gain set to 20 pu on machine MVA rating.
The participation factor and mode shape plots for representative modes are therefore re-examined for Case 1, the medium-heavy load condition for the system. The plots for an interarea mode and a set of three local-area modes are shown in Figures 10.27 and 10.28, respectively.
The mode shapes in both figures reveal that the nature of the four modes has not changed
from the case when the PSSs are out of service (see Figures 10.4 and 10.6). For example, for
the inter-area mode the Area 3 generators continue to swing against generators in Areas 5
and 2. However, a comparison of the participation plots with PSSs out and in service in the
four figures demonstrates that, with the PSSs in service, the following states associated with
the action of the PSSs participate noticeably:

internal states in the PSSs (e.g. x081);

field voltage (EF);

voltage behind d-axis transient reactance (Eq).

504

PSS Tuning in Multi-Machine Systems

Ch. 10

Figure 10.27

Case 1. Participation factors and mode shape for the inter-area mode K
( 1.04 + j3.64 ) when all PSS damping gains are set to 20 pu. (Compare with Figure 10.4)

Figure 10.28 Case 1. Participation factors and mode shapes for the local-area modes
A ( 2.19 + j10.4 ), B ( 1.98 + j9.74 ), and C ( 1.93 + j9.29 );
PSSs in service, damping gains are 20 pu. (Compare with Figure 10.6)

Sec. 10.7

Dynamic performance with PSSs in service

505

10.7.3 Assessment of dynamic performance based on time responses

The transient responses of a two-mass spring system to an initial condition which only excites a single mode are demonstrated in Figures 9.3 and 9.4. As an example of the same concept applied to the dynamics of a multi-machine power system let us consider the case of a
stable, lightly-damped system when the PSS damping gains for Case 1 are each set to 5 pu
on machine MVA rating. This low gain setting is chosen because the oscillatory nature of
the responses will be more pronounced than at higher gains.

Normalized rotor speed perturbations

The plot of the rotor modes for Case 1 with increasing gain is shown in Figure 10.26(a). Let
us consider the inter-area mode labelled M associated with a PSS damping gain 25% (5 pu);
the value of this mode is 0.144 j1.98 . The initial conditions for the transient response
are the real parts of all elements of the right speed eigenvector; none of the control inputs
is excited. The mode shape for this scenario and the transient response of representative machines which are the most responsive in this mode are shown in Figure 10.29.
1
0.8
Area 5
0.6
Area 3
0.4
0.2
0
-0.2
-0.4
Area 4
-0.6
-0.8
0

10

15

Time (s)
PPS_5
CPS_4

(a)

NPS_5
LPS_3

GPS_4
YPS_3

(b)

Figure 10.29 Case 1. PSS damping gains 5 pu on machine MVA rating.

(a) Mode shape for the inter-area mode M ( 0.144 j1.98 ).(Compare with Figure 10.5)
(b) Time responses of rotor speed perturbations for initial conditions comprising the real
parts of all elements of the right speed eigenvector.
The nature of the transient response reflects the relative phase and magnitude information
provided not only by the mode shape but also confirms the values of the real and imaginary
parts of the complex eigenvalue (by rate of decay and frequency of oscillations). The relative
magnitude and phase relationships apply for all t 0 . Thus, as emphasized earlier, these

506

PSS Tuning in Multi-Machine Systems

Ch. 10

tools provide valuable aids for the rapid assessment of the characteristics of the dynamic behaviour of the system.
An alternative method of exciting mainly this mode in an analysis of the transient response
is to apply small step changes in mechanical power to appropriate generating units. In this
case, guided by the mode shape, step increases in power are applied to units which swing
together in phase and step decreases in power to those that swing together in anti-phase.
The magnitudes of the steps must be adjusted to accentuate the mode of interest and to reduce the influence of other modes which might also be excited, such as some local area
modes. The sum of the positive and negative changes in mechanical power should amount
to zero.
8

Electric power (MW)

6
4
2
0
2
4
6
8
0

Time (s)
k = 0
k =15

k = 5
k =20

k =10

Figure 10.30 Case 6, light load operating condition. Simulated time responses for a step
change in reference voltage on the single generator on-line and under test at SPS_4. The PSS
damping gain (k) is varied from 0 to 20 pu on machine MVA rating; all other PSS gains are
set to 20 pu.
In commissioning a PSS the recording of the time responses of generator outputs to small
step changes in the generators reference voltage is often used to verify that the parameters
of the PSS have been correctly set. Such verification is conducted by comparing the measured response with those of the time responses predicted by simulation. For illustrative purposes, the nature of the time responses for a set of step changes are displayed in Figure 10.30
for a range of gain settings on the PSS of a generator at SPS_4 when only the unit on test at
the generating station is on-line under a light load condition. The damping gains on the latter
PSS are varied from zero to 20 pu (100%) on machine MVA rating; the damping gains of all

Sec. 10.8

Intra-station modes of rotor oscillation

507

other PSSs are set to 20 pu. For a given setting of the PSS damping gain the superimposed
step responses from the commissioning test and that derived from simulation should agree
closely.

10.8

Intra-station modes of rotor oscillation [6], [7]

In the previous studies and the associated eigen-analysis the generators in a power station
have been treated as a single generator which is assumed to represent the number of identical, equally-loaded units on-line.
In the studies all units within a power station could have been individually represented.
However, for our purposes this would added complexity to both the analysis and assessment
of results. In practice, representation of individual units may be necessary, (i) if the loadings
on individual unit differs markedly for different operating conditions, (ii) if there are machines of different rating and parameters in the station, and (iii) in order to understand the
nature of the intra-station modes and how the PSS tuning affects these modes. If there are
m machines in a station, there are m 1 modes of rotor oscillations; we will refer to these
as the intra-station or inter-machine modes.
In Table 10.12 are shown the three intra-station modes when the four unequally loaded units
at SPS_4 and PPS_5 are represented individually for the heavy load condition, Case 1. The
three intra-station modes in each station are well damped when all machine PSSs are set to
20 pu on machine MVA rating; the values of the other 13 modes, both local-area and interarea, are close to those given in Table 10.11 for Case 1. The frequencies of the intra-station
modes for the SPS_4 machines are significant higher than all other rotor modes, primarily
because the inertia constant of each unit is relatively low at 2.6 MWs/MVA. On the other
hand, the frequencies of the intra-station modes for the PPS_5 machines are relatively lower,
the inertia constant of each unit being greater at 7.5 MWs/MVA.
By means of the participation factors and mode shapes the nature of the intra-station modes
is demonstrated in Figure 10.31 for SPS_4. Unit #1 in SPS_4 predominantly swings against
the other three machines in the case of mode 105. For mode 107 SPS_4 unit #2 swings
mainly against machine #3 whilst for mode 109 unit #4 swings mainly against machine #3.
Because of the level of the damping gain of the PSSs, the PSS and d-axis states participate
more markedly in these modes.
For a light-load condition, Case 4, three units at SPS_4 are in service and one at PPS_5 (see
Table 10.2). From a comparison of light and heavy load conditions in Tables 10.13 and
10.12, respectively, it is noted that the two intra-station modes for SPS_4 are comparable.

508

PSS Tuning in Multi-Machine Systems

Table 10.12

Generator

SPS_4
SPS_4
SPS_4
SPS_4
PPS_5
PPS_5
PPS_5
PPS_5

Figure 10.31

Case 1: Heavy load. Intra-station modes for four units

at SPA_4 and PPS_5.

Generator output

no. 1
no. 2
no. 3
no. 4
no. 1
no. 2
no. 3
no. 4

Ch. 10

MW

Mvar

400
367
333
300
149
122
96
69

58.3
54.4
50.7
47.4
30.2
26.7
24.0
21.9

Inter-machine modes
PSSs off

PSSs on

Mode shift

0.08 j12.2

2.75 j13.6

2.83 j1.42

0.21 j12.2

2.78 j13.6

2.99 j1.41

0.33 j12.2

2.79 j13.5

3.12 j1.38

0.43 j7.25

1.34 j7.43

0.91 j0.18

0.81 j7.17

1.57 j7.40

0.76 j0.22

1.10 j7.05

1.81 j7.22

0.71 j0.17

All PSS damping gains set to 20 pu on machine MVA rating

Case 1. Participation factors and mode shapes for intra-station modes

2.75 + j13.6 , 2.78 + j13.6 and 2.79 + j13.5 for 4 units on-line at SPS_4.

All PSS damping gains are all set to 20 pu.

Sec. 10.9

Small- and large-signal dynamic performance

Table 10.13

Generator

SPS_4 no. 1
SPS_4 no. 2
SPS_4 no. 3

509

Case 4: Light load. Intra-station modes

for the three units at SPS_4

Generator output

Intra-station modes

MW

Mvar

PSSs off

PSSs on

Mode shift

330
320
310

4.5
3.4
2.4

0.20 j12.0

2.79 j13.3

2.99 j1.36

0.24 j12.0

2.80 j13.3 3.04 j1.35

All PSS damping gains set to 20 pu on machine MVA rating

Note that the PSS design procedure based on the P-Vr characteristic does not explicitly attempt to shift the intra-station modes directly to the left in the complex s-plane. For the
SPS_4 units, from the condition in Table 10.12 when all PSSs are off to that when all PSSs
are in service and damping gains set to 20 pu, there is a marked increase in modal frequency
in the intra-station mode shifts (i.e. by ~ 1.4 rad/s); however, such a mode shift does not
apply to the intra-station modes for the four PPS_5 units (~ 0.20 rad/s). As a matter of
course in the design process the effects of PSS tuning on the intra-station modes should be
assessed to ensure they are adequately damped, and that there are no unexpected interactions between controllers.
The design of an ancillary controller specifically to damp the intra-station modes is proposed
in [8].

10.9 Correlation between small-signal dynamic performance and that

following a major disturbance
In Section 1.10 the question: how small is small in small-signal analysis is discussed. In
practice, of particular concern is the stability and dynamic performance of the power system
following a major disturbance, i.e. a large-signal disturbance. To name a few examples,
such disturbances are faults, tripping of a large generators, the opening of transmission lines,
the loss of significant loads. It is important to realize that the nature of the responses following a major disturbance correlates with the performance predicted from small-signal
analysis, particularly after limiting action by controllers has ceased. In other words, smallsignal analysis provides significant insights into, and understanding of, the nature of the
large-signal dynamic performance - or the transient stability - of the system.
10.9.1 A transient stability study based on the fourteen-generator system
We shall examine the dynamic behaviour of the simplified fourteen-generator system of
Figure 10.1 to a three-phase fault at a major busbar on the high-voltage side of a large power
station, i.e. busbar #206 at BPS_2. Because there is no line switching or other system changes associated with this busbar fault, which is cleared in 0.120 s, the system configuration and

510

PSS Tuning in Multi-Machine Systems

Ch. 10

steady-state operating conditions in the post- and pre-fault periods are the same. The system
modes are therefore unchanged.
In order to reveal features of the dynamic responses following the clearance of the fault, the
low value of the damping gain of 5 pu on machine MVA rating is adopted for all the PSSs.
As is seen in Table 10.14 or Figure 10.26(a) for Case 1, a heavy load condition, the system
is stable and the real parts of the rotor modes lie between -0.05 and -1.00. The mode behaviour shown in the table does not differ significantly from that of Table 10.3 when all PSSs
are out of service.
Table 10.14

Behaviour and type of the rotor modes, Case 1;

damping gain of PSSs is 5 pu.

Mode
No.

Real

Imag

A
B
C
D
E
F
G
H
I
J
K
L
M

-0.68
-0.39
-0.42
-1.00
-0.68
-0.88
-0.81
-0.40
-0.64
-0.92
-0.18
-0.05
-0.14

10.47
9.65
9.06
8.73
8.38
8.27
7.80
7.82
7.83
7.48
3.93
2.57
1.98

0.065
0.041
0.046
0.114
0.081
0.106
0.103
0.052
0.082
0.123
0.046
0.021
0.073

Mode Behaviour

VPS_2<-->EPS_2
SPS_4<-->CPS_4, GPS_4
BPS_2<-->EPS_2, VPS_2
NPS_5<-->TPS_5
CPS_4, SPS_4<-->TPS_4, GPS_4,
HPS_1, EPS_2<-->MPS_2, LPS_3
HPS_1, MPS_2<-->EPS_2, BPS_2
TPS_4<-->GPS_4, SPS_4, MPS_2
YPS_3, MPS_2, HPS_1<-->LPS_3, EPS_2
PPS_5<-->TPS_5, NPS_5
Area 3 <--> Area 5, Area 2
Area 4, Area 5 <--> Area 2
Area 5, Area 3 <--> Area 4

Mode Type

Local Area

Local Area
Inter-area

Inter-area

<--> means ... swings against .... Generators or areas are listed under Mode Behaviour are in
descending order of their participation factors.

The responses of speed perturbations about synchronous speed following the incidence of
the three-phase fault are shown in Figure 10.32 for selected generators. As stated, the fault
occurs at the 330 kV bus at BPS_2 (bus 206) and is cleared in 0.120 s. The responses are
divided into three time intervals so that the various features of the modal behaviour in each
interval can be examined; the time intervals are (a) 0 to 7 s, (b) 7 to 16 s, (c) 16 to 30 s. (Note
the changes of scales on both axes.)

Sec. 10.9

Small- and large-signal dynamic performance

x10

150

511

-4

(a)
100
50
0
-50
-100

BPS_2, EPS_2 & HPS_1 respond to the local mode

0.42 j9.06 . Although remote from the fault,
PPS_5 is excited by inter-area mode 0.05 j2.57 .

-150
-200

Rotor speed perturbations (pu)

-4

x10

15

(b)
10
5
0

Mode 0.42 j9.06 decays and BPS_2, EPS_2 & HPS_1 responses merge into the inter-area mode 0.05 j2.57 .
LPS_3 response is that of inter-area mode 0.18 j3.93 .

-5
-10
7

10

11

12

13

14

15

16

x10

Except for LPS_3,the responses are

dominated by the inter-area mode
0.05 j2.57 . LPS_3 responds to the
inter-area mode 0.18 j3.93 .

-4

6
5
4
3
2
1

(c)
0
-1
16

18

20
BPS_2
LPS_3

22
24
Time (s)
EPS_2
HPS_1

26

28

30
GPS_4
PPS_5

Figure 10.32 Rotor speed perturbations of selected generators following a 3-phase fault

512

PSS Tuning in Multi-Machine Systems

Ch. 10

During the interval 0 to 4 s the responses shown in Figure 10.32(a) for selected rotor speed
perturbations is dominated by the mode C, 0.42 j9.06 , subject to the caveat discussed
later. The phase relationship between the principal participants in the response appears close
to that predicted by the mode shape in Figure 10.33. Although remote from the faulted bus,
PPS_5 is excited by the inter-area mode L, 0.05 j 2.57 , in which machines in Area #2 also
participate, as revealed in Figure 10.33. The same comment applies to LPS_3 with respect
to the inter-area mode K, 0.18 j3.93 .
During the interval 7 to 14 s shown in Figure 10.32(b) the responses principally associated
with mode C, 0.42 j9.06 , decay away and merge into the modal behaviour revealed in
the mode shape in Figure 10.33 for the inter-area mode L, 0.05 j 2.57 . After 16 s, except
for LPS_3, the machines participate in the slowly decaying mode L, with a 5% settling time
of ~56 s. LPS_3 continues to participate in the more rapidly-decaying mode K,
0.18 j 3.93 , (see Figure 10.33(c)).

(a)

(b)

(c)

*
*

*
*

Figure 10.33 Participation factors and mode shapes for the principal modes in the response, the local-area mode C ( 0.42 + j9.06 ) and the inter-area modes K ( 0.18 + j3.93 )
and L ( 0.05 + j2.57 ); all PSS damping gains in Case 1 are all set to 5 pu on machine MVA
rating. (<*PS_area> refers to all generators in the numbered area.)
10.9.1.1 Benefits of small-signal analysis of large power systems

This example demonstrates how small-signal analysis complements that based on transient
stability studies.

Sec. 10.9

Small- and large-signal dynamic performance

513

The above example reveals the important features of small-signal analysis, that is, it furnishes not only an understanding of the underlying modal structure of the power system and but
also provides insights into a system's dynamic characteristics that cannot easily be derived
from time-domain simulations for large magnitude disturbances. It is the case in
Figure 10.33 that only a few of the thirteen modes appear to be excited; the nature and location of the fault does not significantly excite the local-area modes outside the faulted area
at all. Understanding the nature of the small-signal modal behaviour therefore yields a synoptic view of the system characteristics which would require many large-signal studies of
faults in different locations to gain similar, but not exact, information [9].
Knowledge of the behaviour of certain local and inter-area modes has revealed the nature
of the responses of the speed states following a major disturbance on the system. However,
as stated earlier, the behaviour of the system is highly non-linear during the initial phase of
the response. During the first 0.6 s certain exciters reach their ceiling voltages and some
PSSs, together with most SVCs, hit limits on their outputs. In the context of the magnitude
of rotor speed oscillations, the question is asked in Section 1.10, how small is small?. The
peak amplitudes of the speed perturbations in Figure 10.32(a) are 1.5 to 2% which are not
small. The functional non-linearities come into play and therefore the small-signal analysis
is based is not strictly accurate. In the following section the applicability and validity of the
small-signal analysis that has been conducted in this section is reviewed.
10.9.2 The analysis of modal interactions [10], [11], [12]
As has been discussed earlier, small-signal analysis is based on the first-order approximation
of the non-linear power system equations, both differential and algebraic, about a steadystate operating condition. Strictly speaking, such analysis is valid as the perturbations in variables become vanishingly small. Consequently, once limiting by controllers has ceased following a large-magnitude disturbance, techniques based on linear analysis are unlikely to
provide accurate information on the dynamic behaviour of the system when the variations
in system variables is large. This is likely to be valid particularly for so-called stressed conditions in which the system is heavily loaded and/or the system performance is bordering on
instability in the period immediately following the disturbance.
In [10] the significance and application of extending first-order (linear) system analysis to
include the second-order terms is reviewed. The Taylor series expansion about the steadystate operating condition now includes both the first- and second-order terms, but no thirdor higher-order terms. Based on the second-order form of the expansion for a state equation, and employing Normal Form analysis, it is shown in [10] that the ith state equation can
be expressed as:
n

xi t =

v ij z j0 e
j=1

j t

j
v ij h2 kl z k0 z l0 e
j=1

k = 1l = 1

k + l t

(10.8)

514

PSS Tuning in Multi-Machine Systems

Ch. 10

where j , k and l are conventional eigenvalues of the state matrix A ; v ij is an element

of the right eigenvector corresponding to the eigenvalue or mode j ; z p0 is a function of
j

the initial conditions; h2 kl is a function of 1 k + l j , k + l j 0 .

Equation (10.8) reveals the relation between the state variables x i t , the first-order system
modes 1 n , and the second-order modes,
1 + 1 , 1 + 2 , , n 1 + n , n + n .

Note the following:

The terms associated with the mode pairs k + l represent modal interactions that
arise due to the inclusion of the second-order terms.

The second-order terms supplement information provided from the first-order linear
approximation of the power system equations.

If the system is stable, the second-order mode k + l lies to the left of either of its
constituent modes, k or l , in the complex s-plane; it therefore decays more rapidly
than either of the individual modes.

The interaction coefficients, h2 kl z k0 z l0 , of the exponential terms k + l in (10.8)

provide a measure of the participation of any of the mode pairs in the state variable.
Firstly, for the first- and second-order modes discussed in the following, let us assume the
linear coefficient term, v ij z j0 , and the interaction coefficients in (10.8) are not negligible.
Secondly, we will assume j = j + j j and j + 1 = j j j are the complex conjugate
pair of the dominant first-order mode, normally an inter-area mode. When k = j and l = j
in (10.8), the mode pair k + l = 2 j + j2 j ; likewise when k = j + 1 and l = j + 1 , the
mode pair k + l = 2 j j 2 j . Thus, due to modal interactions, a second-order mode of
double the frequency and double the damping constant of the first order mode is introduced
into the response; significantly, however, it decays in half the settling time of the linear mode.
Thirdly, let us assume there is some other, more heavily damped first-order mode present,
r r + 1 = r j r . When k = j and l = r in (10.8), the second-order mode pair
k + l = j + r + j j + r is introduced. Thus the resulting complex second-order

mode will be of higher frequency than the dominant first-order mode and, because r j ,
it will decay with a settling time of less than half the settling time of the linear mode.

Sec. 10.9

Small- and large-signal dynamic performance

515

200

Rotor angles (deg.)

150
100
50
0
50
100
150
0

BPS2
SPS4

5
Time (s)
EPS2
HPS1

10

MPS2
PPS5

Figure 10.34 Case 1. Responses of rotor angles of selected generators to a three-phase

fault at the hv bus at the terminals of the transformers at BPS_2 cleared in 0.250 s. Angles
are relative to that of LPS_3. All PSS gains set to 20 pu on machine MVA rating.
The transient response of the rotor angles of generators is shown in Figure 10.34 for a threephase fault at bus #206, the high-voltage bus at the terminals of the generator transformers
at BPS_2. The system conditions are the same as those in Figure 10.32 except the fault clearing time has been increased from 120 ms to 250 ms. The system, which is marginally stable
with rotor angle differences across the system reaching 260 at about 1 s, is heavily stressed
in the immediate post-fault period. The essential point of the analysis in this section is that in
this period, during or after which no limiting by controllers occurs, one should be aware that
second-order modes of some significance may arise. However, because such modes are better damped than their constituent first-order modes, they tend to decay more rapidly.
The analysis of modal interactions based on Normal Forms for a large system is computeintensive and complex, mainly because of the size of the system and the number of combinations of both the second-order modes k + l and the associated interaction coefficients.
Moreover, given the identical system conditions and type of disturbance, the latter coefficients will vary depending on the instant in the transient response at which initial conditions
are selected.
In the particular cases of Figures 10.32 and 10.34 it is clear from (10.8) that the first-order
modes exist in the responses. However, without a detailed analysis based on Normal Forms
it is unclear what modal interactions are present, and their magnitude at any instant - at least

516

PSS Tuning in Multi-Machine Systems

Ch. 10

up to one half of the settling time of the dominant mode when the responses of secondorder modes k + l have effectively decayed away. From studies in the literature it appears
that for less stressed systems the effects of modal interactions dissipate well within the latter
time. For the Study Case 1, shown in Figure 10.32, this may well be the situation. In [14],
[15] interesting comparisons are made between the transient response of a stressed system
to major disturbances and the first- and second-order responses based on the results of Normal Form analysis. For the scenarios considered the second-order responses agree closely
with those derived from the transient responses based on the step-by-step simulation.
The above summary of modal interactions and their significance is necessarily very brief.
More extensive details are provided in other papers referenced in [10], [14], [15], [16].

10.10 Summary: Tuning of PSSs based on the P-Vr approach

The case study illustrates the basis and benefits of the P-Vr method but also helps to identify
some of the limitations of the basic design technique. (These limitations motivate certain developments which follow in later chapters.)

The P-Vr method provides a systematic approach and a formal basis for the design of
PSSs. The phase of P-Vr characteristics are, for practical purposes, more-or-less invariant over the prudently-selected set of encompassing operating conditions (see
Section 9.4.1). At higher real power outputs, typically 0.5 to 1 pu of rated power, the
magnitude response of the P-Vr characteristic retains its shape and consistently lies in a
band of 3 dB from the Design Characteristic.

These encompassing conditions should not only cover normal operation but also
include various contingencies, line outages, and perhaps some potentially extreme
conditions in order to ensure that the PSS is adequately tuned.

The calculation of P-Vrs for normal operation and for contingencies is easily automated, resulting in the display of a full set of P-Vr characteristics and providing a basis
for the synthesizing of the PSS transfer function.

In tuning fixed-parameter PSSs using the P-Vr approach the concept of robustness is based
on the following considerations:

There are two important components of a fixed-parameter PSS transfer function

kG c s which are essentially decoupled for practical purposes;
(a)

(b)

the rotor modes are more-or-less directly left-shifted by the PSS compensating transfer function G c s with increase in the PSS damping
gain, k ;
the extent of the left-shift of the rotor modes is determined by the
damping gain, k ;

Sec. 10.10

Summary
(c)

517

the value of the damping gain should be such that the damping torque
contributions induced by the PSS swamp the negative inherent contributions by the generator.

Ideally, the incremental left-shifts of the rotor modes are linearly related to increments
in PSS gain for changes about selected nominal values.

Such considerations should apply over the set of encompassing operating conditions
and an appropriate range of rotor modes.

In Section 10.7.1, and from an examination of the modes shifts induced by the PSSs as
shown in Tables 10.11, 10.15 and 10.16, we observe that the above considerations for robustness are - in essence - satisfied. However, there are two factors which cause the a deviation from a direct left shift of the modes, an increase or decrease on modal frequency with
increase in damping gain.
Firstly, in Figure 10.26 it is noted that mode shifts for the selected gain increment vary with
the type of mode (e.g. a local mode) and the machines participating in the mode. In
Section 5.9.3 it is foreshadowed that the shift in the complex rotor mode h is given by
h
h = -------------- H PVr h G c h k ,
2H

((5.36) repeated)

(10.9)

where k is an increment in the damping gain of the PSS and h is the complex participation factor of the generators speed state in the mode h , evaluated with the PSS in service with a damping-gain setting, k0 . It is shown in Section 13.3 that h is essentially real
for generators participating strongly in the mode, but for those participating with a relatively
small participation factor it ( h ) may acquire a not insignificant positive or negative imaginary component. However, being small the contribution by the generator to the mode
shift h may be minor.
Secondly, in Chapter 13 it is shown that in the multi-machine environment the modes shift
in (10.9) can either be enhanced or degraded by the action of PSSs installed on other generators. This is caused by the production of a positive or negative damping torque being induced on generator i by the action of the PSS fitted to machine j [4]. Furthermore, we
observe in Figure 10.26 that the mode shifts associated with the inter-area modes are smaller
than those of the local modes. This feature is also considered in Chapter 13.
We have examined the intra-station modes and emphasized that their damping should be
examined because information on how the design methods (including the GEP and Residue
Methods) influence these modes is not readily available. Though exciter modes, which can
become lightly damped or unstable, have not been examined, the same comments are relevant [6], [7].

518

PSS Tuning in Multi-Machine Systems

Ch. 10

10.11 References
[1]

P. Kundur, M. Klein, G.J. Rogers and M.S. Zywno, Application of power system
stabilizers for enhancement of overall system stability, Power Systems, IEEE Transactions on, vol. 4, no. 2, May 1989, pp. 614-626.

[2]

M. J. Gibbard, Co-ordinated design of multimachine power system stabilisers based

on damping torque concepts, IEE Proceedings C Generation, Transmission and Distribution, vol. 135, pp. 276-284, 1988.

[3]

P. Pourbeik, Design and Coordination of Stabilisers for Generators and FACTS devices
in Multimachine Power Systems, PhD Thesis, The University of Adelaide, Australia,
1997.

[4]

M.J. Gibbard, D.J. Vowles and P. Pourbeik, Interactions between, and effectiveness
of, power system stabilizers and FACTS stabilizers in multimachine systems, Power
Systems, IEEE Transactions on, vol 15, pp. 748-755, 2000.

[5]

D.J. Vowles and M.J. Gibbard, Mudpack User Manual: Version 10S-03, School of
Electrical and Electronic Engineering, The University of Adelaide, July 2014.

[6]

G.J. Rogers, The application of power system stabilizers to a multigenerator plant,

Power Systems, IEEE Transactions on, vol 15, pp. 350-355, 2000.

[7]

M.J. Gibbard, D.J. Vowles, G.J. Rogers, Discussion of The application of power system stabilizers to a multigenerator plant, [and reply], Power Systems, IEEE Transactions on, vol 15, pp. 1462 - 1464, 2000.

[8]

N. Martins and T. H. S. Bossa, A Modal Stabilizer for the Independent Damping

Control of Aggregate Generator and Intraplant Modes in Multigenerator Power
Plants, Power Systems, IEEE Transactions on, vol. 29, pp. 2646-2661, 2014.

[9]

P. Pourbeik, T. Cain, and R. Bottoms, Application of small-signal stability tools and

techniques to a large power system, in Power & Energy Society General Meeting,
2009. PES '09. IEEE, pp. 1-11.

[10]

J.J. Sanchez-Gasca, V.Vittal, M.J. Gibbard, A.R. Messina, D.J. Vowles, B.S. Liu and
U.D. Annakkage, Inclusion of higher-order terms for small signal (modal) analysis:
Committee Report - Task Force on assessing the need to include higher-order terms
for small-signal (modal) analysis, Power Systems, IEEE Transactions on, vol. 20, pp.
1886-1904, 2005.

[11]

ibid, Analysis of Higher Order Terms for Small Signal Stability Analysis, Conference
Proceedings, IEEE Power Engineering Society General Meeting, San Francisco, California, USA, 12-16 June 2005.

[12]

V. Vittal, N. Bhatia, and A. A. Fouad, Analysis of the Inter Area Mode Phenomenon
in Power Systems Following Large Disturbances, Power Systems, IEEE Transactions
on, vol. 6, pp. 1515-1521, 1991.

Sec. 10.11

References

519

[13] IEEE Recommended Practice for Excitation System Models for Power System Stability Studies, IEEE Standard No:421.5-2005, ISBN:0-7381-4787-7.
[14]

E. Barocio and A. R. Messina, Application of perturbation methods to the analysis

of low frequency inter-area oscillations, in Power Engineering Society Summer Meeting, 2000. IEEE, pp. 1845-1850 vol. 3.

[15]

A. R. Messina, J. Arroyo, and E. Barocio, Analysis of modal interaction in power systems with FACTS controllers using normal forms, in Power Engineering Society General Meeting, 2003, IEEE, p. 2117 Vol. 4.

[16]

L. Shu, A. R. Messina, and V. Vittal, A Normal Form Analysis Approach to Siting

Power System Stabilizers (PSSs) and Assessing Power System Nonlinear Behavior,
Power Systems, IEEE Transactions on, vol. 21, pp. 1755-1762, 2006.

App. 10I.1

520

Appendix 10I
Modes of rotor oscillation for Cases 2, 3, 5 and 6

Table 10.15

Case 3. Peak Load

Case 2. Medium-heavy load

No
.

No PSSs

Real Imag
0.07
0.10
-0.25
-0.53
-0.18
-0.70
-0.92
-0.21
-0.06
-0.49
0.19
0.05
0.08

10.74
9.56
9.26
8.67
8.48
8.29
8.61
7.93
7.39
7.57
3.77
2.86
1.92

All PSSs in service

-0.01
-0.01
0.03
0.06
0.02
0.08
0.11
0.03
0.01
0.06
-0.05
-0.02
-0.04

Real Imag
-2.40
-2.04
-2.37
-2.49
-2.04
-2.44
-2.81
-2.03
-2.02
-1.81
-0.77
-0.45
-0.43

10.96
9.72
9.64
8.94
8.38
8.37
8.96
7.74
7.49
7.77
3.54
2.54
1.76

0.21
0.21
0.24
0.27
0.24
0.28
0.30
0.25
0.26
0.23
0.21
0.17
0.24

Mode Shift

No PSSs

Real Imag Real Imag

-2.47
-2.14
-2.12
-1.96
-1.86
-1.74
-1.88
-1.82
-1.96
-1.33
-0.96
-0.50
-0.51

Mode Number

0.22
0.16
0.38
0.27
-0.10
0.08
0.35
-0.19
0.11
0.20
-0.24
-0.32
-0.16

All PSSs in service

Real Imag

-0.38 11.11 0.03 -1.91 11.24

0.10 9.56 -0.01 -2.04 9.72
-0.30 9.02 0.03 -2.28 9.10
-0.58 8.66 0.07 -2.52 8.91
-0.18 8.48 0.02 -2.03 8.38
-0.13 6.31 0.02 -2.03 5.91
-0.14 8.26 0.02 -1.95 8.49
-0.19 7.91 0.02 -2.01 7.73
-0.08 7.38 0.01 -1.93 7.53
-0.58 7.62 0.07 -1.93 7.80
0.01 4.08 -0.00 -1.12 3.71
0.02 2.67 -0.01 -0.43 2.42
-0.03 2.05 0.01 -0.58 1.86

- damping ratio.

0.17
0.21
0.24
0.27
0.24
0.32
0.22
0.25
0.25
0.24
0.29
0.17
0.30

Mode Shift

Real Imag
-1.53
-2.14
-1.98
-1.94
-1.84
-1.90
-1.81
-1.82
-1.86
-1.36
-1.13
-0.45
-0.55

0.14
0.16
0.08
0.25
-0.10
-0.41
0.27
-0.18
0.15
0.18
-0.37
-0.25
-0.19

PSS Tuning in Multi-Machine Systems

A
B
C
D
E
F
G
H
I
J
K
L
M

Rotor modes and modes shifts for medium-heavy and peak loads, Cases 2 & 3.
All PSS damping gains are 20 pu on generator MVA rating.

Appendices

The following Tables 10.15 and 10.16, together with Table 10.11, show the values of the modes of rotor oscillation for Cases 1 to 6
with the PSSs out and in service. When the PSSs are in service, the damping gains are all set to 20 pu on machine MVA rating.

Ch. 10

Rotor modes and modes shifts for medium and light loads, Cases 5 & 6.
All PSS damping gains are 20 pu on generator MVA rating.

Case 5. Medium load

A
B
C
D
E
F
G
H
I
J
K
L
M

All PSS in service

No PSSs

Case 6. Light load

Mode Shift

All PSS in service

No PSSs

Mode Shift

Real

Imag

Real

Imag

Real

Imag

Real

Imag

Real

Imag

Real

Imag

0.18
0.09
-0.16
-0.50
-0.26
-0.52
-0.18
0.01
-0.16
-0.77
0.19
0.01
0.06

10.94
9.57
9.17
8.55
8.45
7.98
8.70
7.90
7.74
7.24
4.15
3.12
2.15

-0.02
-0.01
0.02
0.06
0.03
0.07
0.02
0
0.02
0.11
-0.05
-0
-0.03

-2.41
-1.99
-2.09
-2.47
-2.02
-2.38
-2.19
-1.85
-2.12
-1.86
-0.88
-0.46
-0.50

11.26
9.76
9.39
8.83
8.38
7.97
9.12
7.81
7.87
7.45
3.90
2.89
1.96

0.21
0.20
0.22
0.27
0.24
0.29
0.23
0.23
0.26
0.24
0.22
0.16
0.25

-2.59
-2.07
-1.90
-1.98
-1.76
-1.86
-2.01
-1.86
-1.96
-1.09
-1.08
-0.46
-0.56

0.32
0.19
0.22
0.27
-0.07
-0.01
0.42
-0.09
0.13
0.21
-0.25
-0.23
-0.20

0.28
0.32
-0.13
-0.46
-0.14
-1.51
-0.23
-0.30
-0.21
-0.36
0.20
0.05
0.04

10.39
10.14
9.42
8.74
8.58
8.24
8.92
8.13
8.29
7.25
4.81
3.55
2.60

-0.03
-0.03
0.01
0.05
0.02
0.18
0.03
0.04
0.03
0.05
-0.04
-0.02
-0.01

-2.22
-2.14
-2.07
-2.38
-2.32
-1.70
-2.52
-2.08
-2.58
-1.60
-1.08
-0.57
-0.52

10.71
10.65
10.02
9.02
8.51
8.17
9.54
8.24
8.45
7.55
4.64
3.30
2.45

0.20
0.20
0.20
0.26
0.26
0.20
0.26
0.24
0.29
0.21
0.23
0.17
0.21

-2.49
-2.50
-1.94
-1.93
-2.18
-0.19
-2.29
-1.78
-2.37
-1.24
-1.28
-0.62
-0.56

0.32
0.51
0.59
0.29
-0.07
-0.07
0.62
0.11
0.17
0.30
-0.17
-0.25
-0.15

Mode Number

Rotor modes, Cases 2, 3, 5 and 6

No
.

App. 10I.1

Table 10.16

- damping ratio. PSS of HPS_1 is OFF as it operates as a synchronous compensator in this case.

521

522

PSS Tuning in Multi-Machine Systems

App. 10I.2

Data for steady-state power flow analysis

Table 10.17

SVC name /
Bus No.

Ch. 10

SVC bus numbers, ratings and operating conditions for Cases 1 to 6.

Reactive
Power
Range
(Mbase)

Qmax

Qmin

Case 2
Voltage
Mvar

Case 1
Voltage
Mvar

Case 3
Voltage
Mvar

Case 4
Voltage
Mvar

Case 5
Voltage
Mvar

Case 6
Voltage
Mvar

Mvar @ 1.0 pu voltage

ASVC_2 /
205

650.0

430.0 -220.0

1.055
-68.3

1.055
41.8

1.02
-5.2

1.045
-39.3

1.045
-118.3

1.045
-29.4

RSVC_3 /
313

800.0

600.0 -200.0

1.015
71.4

1.015
129.4

1.015
158.8

1.015
86.7

1.015
54.9

1.015
54.2

1430.0 1100.0 -330.0

1.000
58.2

1.000
63.9

1.000
83.8

1.000
-52.2

1.000
22.8

1.000
-0.2

BSVC_4 /
412
PSVC_5 /
507

500.0

320.0 -180.0

1.015
22.6

1.040
36.8

1.043
18.0

1.010
-4.0

1.015
13.8

1.000
-3.7

SSVC_5 /
509

550.0

400.0 -150.0

1.030
10.6

1.027
50.2

1.050
-63.4

1.030
-109.3

1.030
-123.8

1.030
-109.3

Note: System frequency is 50 Hz.

Table 10.18

Switched Shunt Capacitor / Reactor banks (C/R)

in service, Cases 1-6 (Mvar)

Bus
Number

Case 1

Case 2

211

Case 3
100 C

Case 4

Case 5

Case 6

212

400 C

150 C

150 C

400 C

400 C

400 C

216

300 C

150 C

150 C

300 C

300 C

300 C

409

60 C

60 C

60 C

60 C

60 C

60 C

411

30 C

30 C

30 C

30 C

30 C

30 C

414

30 R

30 R

30 R

30 R

30 R

30 R

415

60 R

60 R

60 R

60 R

60 R

60 R

416

60 R

60 R

60 R

60 R

60 R

90 R

90 R

90 R

504

App. 10I.2 Data for steady-state power flow analysis

Table 10.19
From bus
/ to bus

Line
No.

102
102
102
102

217
217
309
309

1,2
3,4
1,2
3

205
205
206
206
206
207
207
208
209
210
211
211
212
214
214
215
215
216
303
303
304
305
305
306
307

206
416
207
212
215
208
209
211
212
213
212
214
217
216
217
216
217
217
304
305
305
306
307
307
308

1,2
1,2
1,2
1,2
1,2
1,2
1
1,2,3
1
1,2
1,2
1
1
1
1
1,2
1,2
1
1
1,2
1
1
1,2
1
1,2

Transmission Line Parameters; Values per circuit

Line: r+jx, b
(pu on 100MVA)
0.0084
0.0078
0.0045
0.0109

523

0.0667
0.0620
0.0356
0.0868

0.817
0.760
0.437
0.760

0.0096 0.0760 0.931

0.0037 0.0460 0.730
0.0045 0.0356 0.437
0.0066 0.0527 0.646
0.0066 0.0527 0.646
0.0018 0.0140 0.171
0.0008 0.0062 0.076
0.0031 0.0248 0.304
0.0045 0.0356 0.437
0.0010 0.0145 1.540
0.0014 0.0108 0.133
0.0019 0.0155 0.190
0.0070 0.0558 0.684
0.0010 0.0077 0.095
0.0049 0.0388 0.475
0.0051 0.0403 0.494
0.0072 0.0574 0.703
0.0051 0.0403 0.494
0.0010 0.0140 1.480
0.0011 0.0160 1.700
0.0003 0.0040 0.424
0.0002 0.0030 0.320
0.0003 0.0045 0.447
0.0001 0.0012 0.127
0.0023 0.0325 3.445
continued ...

Note: System frequency is 50 Hz.

From bus
/ to bus

Line
No.

... contd
309 310
310 311
312 313
313 314
315 509

1,2
1,2
1
1
1,2

Line: r+jx, b
(pu on 100MVA)
0.0090
0.0000
0.0020
0.0005
0.0070

0.0713
-0.0337
0.0150
0.0050
0.0500

0.874
0.000
0.900
0.520
0.190

405
405
405
406
407
408
409
410
410
410
411
414
415

406 1,2
408
1
409 1,2,3
407 1,2
408
1
410 1,2
411 1,2
411
1
412 1 to 4
413 1,2
412
1,2
415 1,2
416 1,2

0.0039
0.0054
0.0180
0.0006
0.0042
0.0110
0.0103
0.0043
0.0043
0.0040
0.0012
0.0020
0.0037

0.0475
0.0500
0.1220
0.0076
0.0513
0.1280
0.0709
0.0532
0.0532
0.0494
0.0152
0.0250
0.0460

0.381
0.189
0.790
0.062
0.412
1.010
0.460
0.427
0.427
0.400
0.122
0.390
0.730

504
504
505
505
506
506
507
507

507
508
507
508
507
508
508
509

0.0230
0.0260
0.0008
0.0025
0.0008
0.0030
0.0020
0.0300

0.1500
0.0190
0.0085
0.0280
0.0085
0.0280
0.0190
0.2200

0.560
0.870
0.060
0.170
0.060
0.140
0.090
0.900

1,2
1,2
1
1
1
1
1
1,2

524

PSS Tuning in Multi-Machine Systems

Table 10.20

Transformer Ratings and Reactances.

Buses
Number
From

To

101
201
202
203
204
209
213
301
302
304
305
305
308
401
402
403
404
413
501
502
503

102
206
209
208
215
210
214
303
312
313
311
314
315
410
408
407
405
414
504
505
506

g
g
g
g
g
4
4
g
g
2
2
2
2
g
g
g
g
3
g
g
g

Rating,
each Unit
(MVA)

333.3
666.7
555.6
555.6
666.7
625.0
625.0
666.7
444.4
500.0
500.0
700.0
370.0
444.4
333.3
444.4
333.3
750.0
333.3
250.0
166.7

Reactance per
transformer
% on
Rating

per unit on
100MVA

12.0
16.0
16.0
17.0
16.0
17.0
17.0
16.0
15.0
16.0
12.0
17.0
10.0
15.0
17.0
15.0
17.0
6.0
17.0
16.0
16.7

0.0360
0.0240
0.0288
0.0306
0.0240
0.0272
0.0272
0.0240
0.0338
0.0320
0.0240
0.0243
0.0270
0.0338
0.0510
0.0338
0.0510
0.0080
0.0510
0.0640
0.1000

g - Generator/transformer unit; in service if

associated generator is online.
Note: System frequency is 50 Hz.

Ch. 10

App. 10I.2 Data for steady-state power flow analysis

V1

t:1

V1/t

From
bus

525

Zt

V2
To
bus

Taps-ratio convention employed

Figure 10.35

Transformer Taps Convention

The transformer tap ratios listed in Table 10.21 are based upon the convention shown in
Figure 10.35.
Table 10.21

Transformer Tap Ratios for power flow Cases 1 to 6

Buses
From

To

101
201
202
203
204
209
213
301
302
304
305
305
308
401
402
403
404
413
501
502
503

102
206
209
208
215
210
214
303
312
313
311
314
315
410
408
407
405
414
504
505
506

Case 1

Case 2

Case 3

Case 4

Case 5

Case 6

0.939
0.943
0.939
0.939
0.939
0.976
1.000
0.939
0.952
0.961
1.000
1.000
1.000
0.939
0.952
0.952
0.952
1.000
0.952
0.962
0.962

0.948
0.948
0.948
0.948
0.948
0.990
1.000
0.935
0.952
0.961
1.000
1.000
0.960
0.939
0.952
0.952
0.952
1.000
0.952
0.930
0.930

0.948
0.939
0.939
0.939
0.939
0.976
1.000
0.930
0.952
0.948
1.000
1.000
1.000
0.939
0.952
0.952
0.952
1.000
0.952
0.930
0.930

1.000
1.000
1.000
1.000
1.000
0.976
1.000
1.000
1.000
0.961
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000

1.000
0.971
0.971
0.971
0.971
0.976
1.000
0.961
0.961
0.961
1.000
1.000
1.000
0.952
0.952
0.952
0.952
1.015
0.985
0.995
0.985

1.000
1.010
1.010
1.010
1.010
0.976
1.000
1.000
1.000
0.961
1.000
1.000
1.000
1.010
1.000
1.000
1.000
1.000
1.015
1.020
1.020

526

PSS Tuning in Multi-Machine Systems

Ch. 10

For simplicity, loads are assumed to behave as constant impedances in the small-signal analysis.
Table 10.22
Case 1

Busbar Loads (P MW, Q Mvar) for Cases 1 to 6

Case 2

Bus
No.

102
205
206
207
208
211
212
215
216
217
306
307
308
309
312
313
314
405
406
407
408
409
410
411
412
504
507
508
509

450
390
130
1880
210
1700
1660
480
1840
1260
1230
650
655
195
115
2405
250
990
740
0
150
260
530
575
1255
300
1000
800
200

45
39
13
188
21
170
166
48
184
126
123
65
66
20
12
240
25
99
74
0
15
26
53
58
126
60
200
160
40

Case 3

380
330
110
1600
180
1445
1410
410
1565
1070
1230
650
655
195
115
2405
250
1215
905
0
185
310
650
700
1535
200
710
520
70

38
33
11
160
18
145
140
40
155
110
123
65
66
20
12
240
25
120
90
0
20
30
65
70
155
40
140
105
15

P
475
410
140
1975
220
1785
1740
505
1930
1320
1450
770
770
230
140
2840
300
1215
905
0
185
310
650
700
1535
300
1100
800
100

Case 4

Q
50
40
15
200
25
180
180
50
200
140
150
80
80
25
15
290
30
120
90
0
20
30
65
70
155
60
220
160
20

P
270
235
80
1130
125
1060
1000
290
1105
750
900
470
620
140
92
1625
180
730
540
0
110
190
390
420
922
180
640
490
122

Case 5

Q
30
25
10
120
15
110
110
30
120
80
90
50
100
15
10
165
20
75
55
0
10
20
40
45
100
20
65
50
15

Load Characteristics: Constant Impedance

P
340
290
100
1410
160
1275
1245
360
1380
940
1085
580
580
170
105
2130
222
990
740
0
150
260
530
575
1255
225
750
600
150

Case 6

Q
35
30
10
145
20
130
125
40
140
95
110
60
60
20
15
220
25
100
75
0
15
30
55
60
130
25
75
60
15

P
270
235
80
1110
125
1035
1000
290
1105
750
900
470
620
140
92
1625
180
730
540
0
110
190
390
420
922
170
565
450
117

Q
30
25
10
120
15
110
110
30
120
80
90
50
100
15
10
165
20
75
55
0
10
20
40
45
100
20
65
50
15

App. 10I.3 Data for dynamic performance analysis

527

App. 10I.3 Data for dynamic performance analysis

The parameters of the fourteen generators are listed in Table 10.23.
Table 10.23

Generator

Bus

Order

Generator Parameters ##

H
RatNo. of MWs
ing
/
Units
MVA
MVA

Xa
pu

Xd
pu

Xq
pu

Xd'
pu

Tdo'
s

Xd"
pu

Tdo"
s

Xq'
pu

Tqo'
s

Xq"
pu

Tqo"
s

HPS_
101
1

333.3

12

3.60

0.14

1.10

0.65

0.25

8.50

0.25 0.050

0.25 0.200

BPS_
201
2

666.7

3.20

0.20

1.80

1.75

0.30

8.50

0.21 0.040

0.70

0.30

0.21 0.080

EPS_
202
2

555.6

2.80

0.17

2.20

2.10

0.30

4.50

0.20 0.040

0.50

1.50

0.21 0.060

MPS_
204
2

666.7

3.20

0.20

1.80

1.75

0.30

8.50

0.21 0.040

0.70

0.30

0.21 0.080

VPS_
2

203

555.6

2.60

0.20

2.30

1.70

0.30

5.00

0.25 0.030

0.40

2.00

0.25 0.250

LPS_
301
3

666.7

2.80

0.20

2.70

1.50

0.30

7.50

0.25 0.040

0.85

0.85

0.25 0.120

YPS_
302
3

444.4

3.50

0.15

2.00

1.80

0.25

7.50

0.20 0.040

0.20 0.250

CPS_
402
4

333.3

3.00

0.20

1.90

1.80

0.30

6.50

0.26 0.035

0.55

1.40

0.26 0.040

GPS_
404
4

333.3

4.00

0.18

2.20

1.40

0.32

9.00

0.24 0.040

0.75

1.40

0.24 0.130

SPS_
4

403

444.4

2.60

0.20

2.30

1.70

0.30

5.00

0.25 0.030

0.40

2.00

0.25 0.250

TPS_
401
4

444.4

2.60

0.20

2.30

1.70

0.30

5.00

0.25 0.030

0.40

2.00

0.25 0.250

NPS_
501
5

333.3

3.50

0.15

2.20

1.70

0.30

7.50

0.24 0.025

0.80

1.50

0.24 0.100

TPS_
502
5

250.0

4.00

0.20

2.00

1.50

0.30

7.50

0.22 0.040

0.80

3.00

0.22 0.200

PPS_
5

166.7

7.50

0.15

2.30

2.00

0.25

5.00

0.17 0.022

0.35

1.00

0.17 0.035

503

## Classically-defined operational parameters (see Section 4.2.12.2 and Section 4.2.13).

Generator reactances in per unit on machine rating as base. System frequency is 50 Hz.
For all generators the stator winding resistance (Ra) and damping torque coefficient (D) are both assumed to be zero.

528

PSS Tuning in Multi-Machine Systems

Ch. 10

App. 10I.3.1 Excitation System Parameters

Two basic types of excitation systems are employed, AC4A and AC1A [13]. The parameters
of the AVR have been tuned to ensure that the open-circuit generator under closed-loop
voltage control is stable and satisfies the performance specifications.
1
1+sTr

Vc
Vr

Vt

1+sTC 1+TC1
1+sTB 1+sTB1

Vs

KA
1+sTA

EF

Figure 10.36 Small-signal model of a type AC4A Excitation System

1
1+sTr

Vc
Vr

Vs

VF

1+sTC
1+sTB

Vt
KE
KA
1+sTA

sTE

EF

sKF
1+sTF

Figure 10.37 Small-signal model of a type AC1A Excitation System; demagnetizing effect
of field current neglected

App. 10I.3 Data for dynamic performance analysis

Table 10.24

529

Excitation System Parameters: 14-generator system

HPS_1

BPS_2

EPS_2

MPS_2

VPS_2

LPS_3

YPS_3

Type

AC4A

AC4A

AC1A

AC4A

AC4A

AC4A

AC1A

Tr (s)

KA (s)

200

400

400

400

300

400

200

TA (s)

0.10

0.02

0.02

0.02

0.01

0.05

0.05

TB (s)

13.25

1.12

1.12

0.70

6.42

TC (s)

2.50

0.50

0.50

0.35

1.14

KE

1.0

1.0

TE (s)

1.0

1.333

KF

0.029

0.020

TF (s)

1.0

0.8

Table 10.25

Type

Excitation System Parameters: 14-generator system (continued)

CPS_4

GPS_4

SPS_4

TPS_4

NPS_5

TPS_5

PPS_5

AC4A

AC4A

AC4A

AC4A

AC1A

ST5B

AC4A

Tr (s)

0.02

KA

300

250

300

300

1000

400

300

TA (s)

0.05

0.20

0.01

0.10

0.04

0.50

0.01

TB (s)

9.80

0.0232

0.70

40.0

16.0

0.8

TC (s)

1.52

0.1360

0.35

4.00

1.40

0.2

TB1 (s)

0.05

TC1 (s)

0.60

KE

1.00

TE (s)

0.87

KF

0.004

TF (s)

0.27

530

PSS Tuning in Multi-Machine Systems

Vt

Vref

Vs

Vd

KA
s

2.5
1+sTd

Kd

KS

1----KS

Q/Vt]

SVC name / bus

number
ASVC_2 / 205
RSVC_3 / 313
BSVC_4 / 412
PSVC_5 / 507
SSVC_5 / 509

Mbase
(Mvar)
650
800
1430
500
550

Ch. 10

KA

KS

500
500
500
250
250

6.5
8.0
14.3
5.0
5.5

B and Q V t are in per-unit on MBASE.

K d = 0.01 pu on SBASE and T d = 0.005 s

Figure 10.38

Small-signal model of the controller for the SVCs.

Chapter 11
Tuning of FACTS Device Stabilizers

11.1 Introduction
In the 1990s the development of high power semiconductor devices found application in
power electronic equipment in power systems. Such transmission systems and associated
devices are generally known as Flexible AC Transmission Systems (FACTS); a comprehensive description of the technology, the devices and references to the literature are given in
[1] (published in 2000).
In this chapter the tuning of stabilizers is outlined for FACTS devices such as Static Var
Compensators (SVCs), the converters at the ends of High Voltage Direct Current (HVDC)
transmission lines, Thyristor-Controlled Series Capacitor (TCSC), and other similar FACTS
devices. Such stabilizers are generally known as Power Oscillation Dampers (PODs), however, the role of PSSs is also to act as power oscillation dampers - hence we will refer to
PODs as FACTS Device Stabilizers (FDSs) to emphasize the application to FACTS devices.
Consider the studies for Cases 1 to 6 presented in the previous chapter. Referring to Tables
10.11, 10.15 and 10.16 it is noted that, for all PSSs in service with the damping gain set to
20 pu on machine MVA rating, the real parts of the mode shifts for the local-area modes
typically vary from -1.3 to -2.5 Np/s over the encompassing range of operating conditions
covered by the six cases. However, the real parts of the mode shifts for the inter-area modes,
modes K, L and M, roughly vary over a much smaller range, from -0.4 to -1.1 Np/s for the
same operating conditions. The damping of all modes in these cases is good, the lowest
damping ratio being about 15%. However, because the damping of some modes may be

531

532

Tuning FACTS Device Stabilizers

Ch. 11

poor, stabilizers installed on FACTS devices can provide a significant improvement in the
damping of targeted modes. By reducing PSS damping gains to 5 and/or 10 pu on machine
MVA ratings, cases of poorer damping are also examined in which the damping ratios of the
inter-area modes are in the range 2 to 8%.
The common configuration of the FACTS device and controllers is shown in Figure 11.1.
In the case of a Static Var Compensator (SVC), for example, the controller regulates the
voltage at its terminals or at an electrically close, high-voltage busbar where voltage support
is required [1], [2]. The location of the SVC in the network may be such that a stabilizer installed on the SVC is effective in improving the damping of certain inter-area modes. An
effective stabilizing signal may be the perturbations in frequency at its terminals, an appropriate power flow, etc. [3].
Control Signal

Controller
Reference
Signal

FACTS
Controller
Vs

FACTS
Device

FDS
Stabilizing Signal

.
Figure 11.1

Configuration of the FACTS device, its controller and stabilizer (FDS)

The objective of FDS tuning is to improve the damping of lightly damped modes, ideally
without degrading the damping of other modes, or compromising the performance of the
primary control function of the device. As foreshadowed above, an inter-area mode is typically - but not necessarily - the mode which is targeted for enhanced damping.
As background, a simplistic tuning procedure for a SVC is considered to illustrate the intent
of the FDS tuning methods. The theoretical basis of the Method of Residues, already analysed in Chapter 6, is briefly summarized and will provide the basis for the tuning of FDSs
[4], [5], [8]. However, in the multi-machine case there is a major difference with respect to
the SMIB case of Section 6.3; the FDS may be tuned to provide damping over a range of
modal frequencies.
A variety of other methods for tuning the stabilizers of a range of FACTS devices are described in the literature, [9] to [19]. Reference [20] provides a more detailed account of modelling shunt FACTS devices such as SVCs and Static Compensator. The tuning methods and
approaches investigated in this chapter are presented using simple models for the FACTS
devices; however, the methods are equally applicable for more sophisticated systems. The
models presented in [20] provide detailed descriptions of modern control features such as

Sec. 11.2

A simplistic tuning procedure for a SVC

533

coordinated control of nearby switched capacitor banks by the SVC. Such functions are not
considered any further in this book in which only small deviations about a steady-state operating condition are relevant to small-signal analysis.

11.2 A simplistic tuning procedure for a SVC

The application of a simplistic procedure to the tuning of a SVC, BSVC_4, at bus number
410 in the fourteen-generator network in Figure 10.1 is now examined. As stated in
Table 10.17, the maximum and minimum reactive power generation for BSVC_4 is 1100
and 330 Mvar, respectively, giving a reactive range (Mbase) of 1430 Mvar. The operating
condition selected is the heavy load condition of Case 1 (see Table 10.2). For all PSSs in service, with their damping gains set to 20 pu on machine MVA rating, the local and inter-area
modes are shown in Table 10.11. For illustrative purposes we will consider the more lightly
damped of the complex inter-area modes, M ( 0.52 j1.80 ).
The terminal voltage bus frequency, freq (pu of system frequency), which is the rate of
change of the terminal-voltage angle, rad, is employed as the stabilizing signal. As shown
in (8.9) the transfer function of the bus-frequency pre-filter is:
freq s
s
------------------------ = 1 0 ------------------ ,
1 + sT F
s

(11.1)

where 0 = 2f 0 (rad/s) and f 0 is the system frequency (Hz). At f 0 = 50 Hz,

1 0 = 0.003183 ; TF is normally set so that high frequency noise above the selected cor-

ner frequency (1/TF) is attenuated, say, TF = 0.005 s 1.

The block diagram of the FACTS device controller and stabilizer is shown in Figure 11.2.
Let us assume that the transfer function of the FDS is V s s F rq s = k fds , a real gain
(i.e. omitting compensation, washout and low-pass filters). Lets calculate the values of the
mode M for a range of gain values (not knowing as yet what constitute high gain values). As
noted in Figure 10.38, the value of Mbase is 1430 Mvar, Sbase = 100 Mvar.

1. The time constant TF (5 ms) is very short. Such time constants should typically be 3 or
more times the cycle time of the PSS processor to reduce phase errors at higher frequencies.

534

Tuning FACTS Device Stabilizers

Ch. 11

Vt
Vref

Vd

Vs

500
s

2.5
1+s0.005

0.01

KS

FDS
F(s)

1
KS

KS = Mbase / Sbase
B and Q V t are in

Q/Vt]

per-unit on Mbase

Frq
F-T

F-T: Frequency
Transducer

Figure 11.2 The controller and stabilizer, F(s), for SVC BSVC_4 showing terminal voltage
control, the provision of droop, and the frequency stabilizing signal F rq .
In Table 11.1 the mode shifts in mode M for Case 1 are shown as the stabilizer gain kfds is
increased from zero with the stabilizer in service. The mode shift for a gain of 30 pu is
shown in Figure 11.3. Ideally, to introduce pure damping to the mode, the mode shift should
lie at 180 . Phase lag compensation must therefore be provided for the multi-machine system in this example noting that the required lag compensation angle increases with increasing gain. Although the lag compensation which the stabilizer transfer function should
provide is as much as 16 for the selected gain range, let us derive the transfer function of
the lag compensation with a lag angle of 11 at s = 0 + j1.8 (1.8 rad/s) for the stabilizer
gain of 30 pu.
Table 11.1

Case 1. Shifts in inter-area mode M with increasing FDS

transfer function gain kfds a

kfds (pu)

Mode M

Mode Shift

Angle b

0.522 j1.797

0.649 j1.786

0.127
j0.011

5.0

20

0.778 j1.763

0.256
j0.034

7.6

30

0.905 j1.722

j0.075
0.383

11.1

40

1.024 j1.656

j0.141
0.502

15.7

10

Note: (a) FDS is a pure gain transfer function.

(b) Required lag compensation angle

Sec. 11.2

A simplistic tuning procedure for a SVC

535

Imag
0.86 j1.84

0.34 j0.05
0.52 j1.80
0.38
j0.08

= 11.1

0.91 j1.72
Real

Figure 11.3

Kfds= 0
Kfds= 30 pu
FDS transfer
function

Shift in mode M both for Kfds= 30 pu and for FDS transfer function
F s (11.2).

The calculation of the transfer function of the lag compensator is similar to that for lead
compensation in the example in Section 2.12.1.4 and is based on frequency response analysis with s = 0 + j . The simple compensator transfer function for the lag angle of 11 at
1.8 rad/s is 1 + s0.458 1 + s0.674 . When washout and low pass filters, with corner frequencies 0.17 and 30 rad/s respectively, are included the transfer function of the FDS is:
V s
6s
1 + s0.458
1
F s = ------------ = k fds --------------- ------------------------- ------------------------- , with k fds = 30 pu.
1 + 6s 1 + s0.674 1 + s0.033
F rq

(11.2)

With the FDS of BSVC_4 in service with the above transfer function the resulting value of
mode M is 0.86 j1.84 for k fds = 30 pu compared to the value of 0.91 j1.72 for the
scalar transfer function k fds = 30 in Table 11.1. While the FDS enhances the damping of
mode M relative to the case when the FDS is out of service, the mode shift 0.34 j0.05
is not quite that desired; moreover, its modal frequency is increased from that with the stabilizer off-line. There are therefore a number of observations that can be found in this simplistic procedure.

The agreement between the value of the targeted mode using the simplistic procedure to evaluate the stabilizer transfer function is not as close as desirable. (Further
iterations of the procedure could improve the result.)

The lag compensation of the stabilizer transfer function is based on the frequency
response calculation using s = j f rather than the complex value in the vicinity of the
targeted mode, s = + j f . This problem is compounded when the washout and
low-pass filters are added. A more rigorous, iterative process is required to converge
on a lag transfer function for the stabilizer - with the specified filters - in the vicinity of
the targeted mode. (With the FDS out of service, mode M varies between
0.43 j1.76 and 0.59 j2.51 over the six operating conditions, see Tables 10.11,
10.15 and 10.16.)

536

Tuning FACTS Device Stabilizers

Ch. 11

No cognizance has been given to the suitability of the stabilizer transfer function
(11.2) over an encompassing range of operating conditions (including outages, etc.) in
enhancing the damping of the targeted mode.

Although the damping of the targeted mode may be enhanced over the range of operating conditions, the damping of other modes may be degraded.

Under some operating conditions the presence of zeros or modes (other than rotor
modes), in the vicinity of the targeted mode may significantly affect the trajectory of
the mode as the stabilizer gain is increased.

From Figure 10.26 it is observed that the frequency of the inter-area mode M
decreases with increasing PSS damping gains. To improve synchronizing torques it
may be desirable to tune the FDS to enhance not only the damping of the targeted
mode but also to increase its oscillatory frequency.

It is clear that a method for tuning the stabilizers is desirable that better takes account of the
range of operating conditions, the filters and the complex value of the targeted mode.

11.3

Theoretical basis for the tuning of FACTS Device Stabilizers

Some of the relevant theoretical material, based on the Method of Residues, is described in
Section 6.2.1 and is summarized here for ease of reference.
Let the stabilizer transfer function be:
F s = k fds H s = k fds G c s G W s G LP s ,

(11.3)

where the transfer function G c s of the stabilizer in this application is tuned to provide the
appropriate phase compensation and is assumed to consist of m lead or lag blocks of the
form:
1 + Tn s m
G c s = ------------------ , s = + j .
1 + Td s

(11.4)

The FDS gain setting in (11.3) is k fds (note, this is not the damping gain value). The washout and low-pass filter transfer functions, G W s and G LP s , are given by (5.29) and (5.30),
respectively. It is assumed that the values of the time constants in the latter two transfer
functions have been appropriately selected (see Section 5.8.6). The objective of the tuning
procedure for the ith stabilizer is to determine the values of the parameters k fdsi , Tni and Tdi,
i = 1m in (11.4) that satisfy the relevant requirements on damping.

The following analysis (which repeats part of that in Section 6.2.1) assumes that (i) initially
the FDS is out of service, and then it is in service with the FDS gain set to k fds , (ii) the FDS
feedback is positive (see Figure 6.1 and 11.2). It is shown in (6.8) that the mode shift in the
targeted mode h is:

Sec. 11.3

Theoretical Basis for tuning FDSs

537

k fds r h H h
h = ------------------------------------------------- ,

1 r h k fds
H h
h

(11.5)

where r h is the residue of the transfer function V stab h V ref h (no FDS); V stab is
the stabilizing signal selected to be the input to the FDS.
If in (11.5) the gain k fds is chosen such that r h k fds

H h 1 then (11.5) reduces to

h

yield the approximate value of the mode shift, i.e.:

h k fds r h H h .

Let the value of k fds for which r h k fds

(11.6)

H h = 1 be k Rm , a quantity which provides

h

a nominal measure of an upper value of the gain. According to (11.6) for values of k fds k Rm
the mode shift increases linearly with stabilizer gain. Thus it follows from the definition of
k Rm that

k Rm = 1 r h
H h .
h

(11.7)

In order for the mode shift h in (11.6) to be 180 , i.e. a direct left-shift of h in the
complex s-plane,
arg r h H h = 180 .

(11.8)

Therefore the compensation angle provided by the FDS is

= arg H h = 180 arg r h .

(11.9)

Typically k fds is selected to be less than 0.1k Rm . However, in multi-machine cases the effect
on the actual modal trajectories of other system poles and zeros, as the FDS gain is increased
from k fds = 0 , may result in mode shifts estimated from the above analysis differing substantially from actual shifts, even at gains much less than 0.1k Rm .
Other comments in Section 6.2.1 are also applicable to FDS tuning.
Consider now the application of the above results to the tuning of a FDS in a multi-machine
system. It may be necessary to tune the FDS to improve the damping of several rotor modes
and to accommodate the associated variation in magnitude and phase of the associated residues.
The application of the Method of Residues is now illustrated by a number of studies; two
studies illustrate the tuning of a FDS for a SVC using bus frequency or real power flow as
stabilizing signals. A study on a different FACTS device concerns the tuning of a stabilizer

538

Tuning FACTS Device Stabilizers

Ch. 11

for a thyristor-controlled series capacitor (TCSC). In the latter case the stabilizer transfer
function is required to accommodate power flows in both directions through the TCSC. In
all studies the trajectories of selected inter-area modes are tracked as the stabilizer gain is increased from zero to an appropriate value. The aims of mode-tracking studies are (i) to determine the stability of the system, (ii) to investigate the characteristics of the mode shifts
with increasing stabilizer gain, (iii) to compare the estimated mode shifts calculated using
(11.5) or (11.6) with those calculated by eigen-analysis, and in some cases (iv), to account for
the nature of the deviation between estimated and calculated values.

11.4
nal

Tuning SVC stabilizers using bus frequency as a stabilizing sig-

As mentioned earlier, a SVC is primarily installed for voltage support and control, typically
in areas more remote from generation - such in the vicinity of loads or at intermediate substation buses on higher voltage transmission lines.
In this application of FDS tuning it is assumed that there are inter-area modes whose damping may be improved by a FDS installed on a SVC close to a major load centre. Conceptually, when close to a major load centre the FDS should modulate the load-area voltage such
that load real power is reduced concomitant with a fall in system frequency - thereby enhancing the damping of the mode. This suggests that frequency may be a suitable stabilizing
signal.
Because it has been the basis of a number of studies the 14-generator power system employed in Chapter 10 is used as the study system. From Tables 10.11, 10.15 and 10.16 it is
observed that the inter-area modes L and M typically have values in the vicinity of
0.5 j2.8 and 0.5 j1.9 , respectively, over the range of the normal cases 1 to 6 with all
PSS damping gains set to 20 pu. Inter-area mode K is generally well damped, but may be
enhanced by the FDSs.
With reference to the system diagram in Figure 10.1, the SVCs BSVC_4 in Area 4 and
PSVC_5 in Area 5 will be used to establish what improvements in damping of the interarea modes can be achieved using perturbations in local frequency as a stabilizing signal. It
will also be found that it is desirable to install a SVC in Area 2; this is considered in
Chapter 14.
Based on the results in Chapter 10 when all PSS damping gains are set to 20 pu it may be
considered that it is not necessary to install stabilizers on any FACTS device. On the other
hand, say, can the PSS damping gain settings be reduced with the installation of FDSs? Let
us therefore establish a whether a FDS transfer function tuned for 20 pu PSS damping gains
adequately covers a lower range of PSS damping gain settings, say 10 to 20 pu.

Sec. 11.4

Tuning a SVC stabilizer using bus frequency

539

For illustrative purposes and to avoid complexity, line outages and other conditions which
cause a degradation in the damping of the inter-area modes have not been included in the
limited analysis which follows. The tuning of a frequency-stabilized FDS for BSVC_4 is now
investigated.
11.4.1 Use of bus frequency as a stabilizing signal for the SVC, BSVC_4
The perturbations in local bus frequency is synthesized from angular perturbations (rad)
in the terminal voltage of the SVC at bus 412 in Figure 10.1. The basis for employing busfrequency perturbations F frq = freq (pu of system frequency) as a stabilizing signal is
outlined in Section 11.2; the transfer function of the frequency transducer is given by (11.1).
Initially it is of interest to learn which of the inter-area modes over the encompassing range
of normal operating conditions, Cases 1 to 6, are best damped by means of the FDS on
BSVC_4. Mbase for BSVC_4 is 1430 Mvar (see Figure 10.38).
11.4.1.1

Determination of the stabilizer transfer function for BSVC_4

Referring to (11.3) and (11.4), the aim of the analysis is to determine k fds G c s , the transfer
function of the compensation, as well as the parameters of the washout and low-pass filters
such that the damping of the mode(s) satisfies the relevant performance criteria. Furthermore, the improvement in damping of any inter-area mode should not lead to an unacceptable degradation in the damping of other inter-area modes or of local modes in the vicinity
of the SVC.
Let us assume (i) all SVCs are in service and controlling the voltage on their respective buses,
(ii) the FDS path in Figure 11.2 is open, and (iii) all PSSs are in service and their damping
gains are set to 20 pu on machine MVA rating. The residues of the transfer function
F rq s V ref s for the inter-area modes K, L and M are then calculated for the operating conditions 1 to 6. Depending on the characteristics of the residues as revealed by their
polar plots, it is of interest to ascertain if the compensation should in fact target any one of
the three inter-area modes. It is also possible, for example, that compensation based on the
residues for mode L may enhance or degrade the damping on mode M, or vice-versa.
Using the Mudpack small-signal, power system dynamic performance package [21] the polar
plot of the residues for modes L and M is shown in Figure 11.4. The residues are of the SVC
transfer function F rq s V ref s for the range of operating conditions, Cases 1 to 6.
The residues of the inter-area mode K are negligible and are omitted from the plot. The values of modes L and M are listed in Tables 10.11, 10.15 and 10.16.
As foreshadowed in the Section 11.4, it is desirable to establish whether the same FDS transfer function adequately covers the 10 and 20 pu sets of PSS damping gains. The polar plot
of the residues for the lower set of PSS damping gains is shown in Figure 11.5.

540

Tuning FACTS Device Stabilizers

Ch. 11

Figure 11.4 Polar plot of the residues for the transfer function, F rq V ref , for modes
L and M and six operating conditions. All PSS damping gains set to 20 pu on machine MVA
rating. The values of modes L and M are in the vicinity of 0.47 j2.8 and 0.5 j1.9 , respectively. Note: the magnitude scale is to be multiplied by 0.1.

Figure 11.5 Polar plot of the residues as for Figure 11.4 with all PSS damping gains set to
10 pu on machine MVA rating. Note scaling.

Sec. 11.4

Tuning a SVC stabilizer using bus frequency

541

A comparison of the magnitudes of the residues in Figures 11.4 and 11.5 reveals (i) the residues for mode M are about four times greater than those of mode L, and (ii) the band of
phase angles of the residues is narrower than that for mode L. The compensation will therefore target mode M for which, as revealed in Figure 11.4, the residues lie in a relatively narrow phase-band of approximately 13 with a spread in magnitudes between 0.013 and 0.018
units.
For the range of operating conditions it is now necessary to select representative values of
(i) the compensation angle for the calculation of the compensator transfer function, (ii) the
magnitude of the residues for determining the nominal upper gain value, and (iii) a single
mode value considered to cover the modes of interest or concern.
From Figure 11.4 for PSS damping gains set to 20 pu a representative angle for the residues
of mode M is selected to be 168 which lies in the mid-range of values. The required compensation angle is therefore 12 (or 12 lagging). For mode M the maximum value of the
residues of 0.0176 pu (on SVC base) is selected from Figure 11.4. (These decisions may depend on the application, e.g. whether to weight certain operating conditions more heavily,
or whether to abide by the system criteria which specify the minimum level for damping,
say, for the outage of a critical circuit.) For mode M, and for the range of modal values over
the encompassing operating conditions, a targeted value of complex frequency is selected to
be 0.5 j1.9 , a value which tends to favour the heavier load conditions.
An associated set of representative values can be deduced from Figure 11.5 when all PSS
damping gains are set to 10 pu; similarly a set for 5 pu is derived. The values are summarized
in Table 11.2.
Table 11.2 Representative values for evaluation of
compensation transfer function, mode M
PSS
gain (pu)

Phase spread
(deg)

Representative phase
angle (deg)

Compensation angle
(deg)

Maximum
residue

Representative modal
frequency*

20

-164 --> -176

-168

-12

0.0176

0.5 j1.9

10

-168 --> -176

-172

-8

0.0151

0.25 j2.2

-171 --> -184

-178

-2

0.0149

0.12 j2.2

* Representative modal frequencies over the range of operating conditions

In Mudpack [21] there are facilities to calculate iteratively the compensation transfer function of the stabilizer given the desired compensation angle, a representative or target complex tuning frequency, the order of the lag or lead compensator, and the required number of

542

Tuning FACTS Device Stabilizers

Ch. 11

washout and low-pass filters and their parameters (see Appendix 6I.1). Based on (11.3) and
(11.4) the form of stabilizer transfer function is given in (11.10).
TW s w 1 + Tn s m
z
1
F s = k fds H s = k fds ------------------- ------------------ --------------------- .
1 + T LP s
1 + TW s
1 + Td s

(11.10)

There are also facilities in the software to estimate, for a selected stabilizer gain, the mode
shifts for the target and other selected modes, according to (11.5), for each of the operating
conditions. The provision of estimates of local as well as inter-area modes can reveal if a local mode, say, is unduly degraded by the stabilizer and therefore may be of concern. However, it is also necessary to establish through Bode- or eigen-analysis the range of FDS gains
k fds for which the closed-loop system is stable. It may happen that a mode other than a rotor mode becomes unstable.
Two cases of FDS tuning could now be considered. In the first case the washout and lowpass filters are selected to cover a wide band of modal frequencies; in the second the filters
provide a narrow band which specifically targets mode M. Wide-band compensation only is
now considered; a practical example of narrow band compensation is analysed in [6] using
the Method of Residues. The objectives of the former are to improve the damping of the
inter-area modes as well as local-area modes, if possible. To cover the range of operating
conditions and rotor modes in Tables 10.11, 10.15 and 10.16, a first-order compensator is
specified, together with first-order washout and low-pass filters with parameters TW = 6 s
and TLP = 0.033 s. The phase shifts of the filters, which lie a decade above and below the
corner frequencies of 0.17 and 30 rad/s, respectively, are less than 5 . The FDS transfer
function thus takes the form:
sT W
F s = k fds H s = k fds ------------------1 + sT W

1 + Tn s
----------------1 + Td s

1
--------------------- .
1 + T LP s

(11.11)

The representative values for 20 pu PSS damping gain settings in Table 11.2 is used to calculate the compensation transfer function because
1.

In practice the PSS damping gain settings may tend towards the higher value of 20 pu
because normal, outage and N-1 operating conditions must all satisfy the system
damping performance specifications.

2.

The range of residue angles in Table 11.2 for 10 and 5 pu PSS damping gains are
essentially covered by that for the 20 pu gain settings.

3.

The representative residue angles differ by 10 at most, and the associated range of
compensation angles lie between 2 to 12 lagging. Over compensation in this study
is likely to increase the frequency of the inter-area modes at the lower gain settings and
thereby improve synchronizing torques. (This may help to offset the decrease in the
inter-area frequencies, observed in Figure 10.26, with increase in PSS damping gains.)

Sec. 11.4

Tuning a SVC stabilizer using bus frequency

543

Based on the representative values in Table 11.2 for the evaluation of the compensation
transfer function, the iterative procedure described in Appendix 6I.1 is used to calculate
the parameters of the lag transfer function for PSS damping gains of 20 pu. The PSS transfer
function (11.11) is found to be
s6
F s = k fds H s = k fds --------------1 + s6

1 + s0.346
------------------------1 + s0.498

1
------------------------- ,
1 + s0.033

(11.12)

the nominal upper gain value being k Rm = 398 pu.

11.4.1.2

Range of stabilizer gains for stability

Before the trajectories of the selected modes are calculated by eigen-analysis it is desirable
to ascertain for what range of gains the system with the stabilizer transfer function calculated
above is stable. It may not be clear if some other mode (e.g. a controller mode) becomes
unstable as the stabilizer gain is increased - or if instability occurs, say, for some value of gain
less than 0.1k Rm where k Rm is the nominal upper value determined by the Residues Method.
From Tables 10.11, 10.15 and 10.16 for the six operating conditions with all PSS damping
gains set to 20 pu on machine MVA rating, and for no FDSs in service, it is known (i) that
the system is stable, and (ii) that the stabilizer transfer function of (11.12) possesses left-half
plane poles. Therefore, with no right-half plane poles, we can use the open-loop Bodes plots
to determine closed-loop stability as well as the gain and phase margins for a selected stabilizer gain.
Lest us insert the stabilizer transfer function (11.12) in the feedback path in Figure 11.2. The
feedback path at the summing junction is left open in order to calculate the open-loop transfer functions V s j f V ref j f . For the Case 1, a heavy load condition, the associated
Bode plot is shown in Figure 11.6 remembering that, for stability analysis using the Bode
plot, negative feedback is assumed (positive feedback of the stabilizer output is specified in
Figure 11.2).
Note in Case 1 that the gain margin for stability (673 pu) is greater than 10% (i.e. ~40 pu)
of the upper gain value ( k Rm = 398 pu) necessary to satisfy the nominal upper value of gain
as determined by the Method of Residues. The stability limits for cases 1 to 6 are confirmed
by eigen-analysis. Within the gain range of 0 to 40 pu the selection of the gain setting k fds
is dependent on a number of factors: for example: (i) Can the desired damping of the target
mode be achieved with lower gain settings such that the reactive power output of the SVC
is not continually hitting limits for acceptable variations in frequency? (ii) Is the damping of
other modes unduly degraded? (iii) Can we be confident about the accuracy of the models
of the devices and the system? (iv) What are the effects of high controller gains on unmodelled dynamics, etc.

544

Tuning FACTS Device Stabilizers

Ch. 11

Magnitude (dB)

20
40
60
80
100
2
10

10

10

10

10

10

10

Phase (deg)

50
0
50
100
150
2

10

Figure 11.6

10

10
10
10
Frequency (rad/s)

Case 1: FDS plus SVC. Open-loop frequency response V s V ref for

k fds = -1 pu on device base. The gain margin is 56.6 dB (673 pu on device base) at

241 rad/s. All PSS damping gains are set to 20 pu.

11.4.1.3

Inter-area modal trajectories as the stabilizer gain is increased

To ascertain the effectiveness of the FDS tuning, which assumes all PSS damping gains are
set to 20 pu on machine MVA ratings, the eigenvalue trajectories are calculated as the FDS
gain is increased from zero to 100 pu on the SVC base. For the same FDS parameters the
trajectories are also evaluated for the case when all PSS damping gains are set to 10 pu. Based
respectively on Cases 1 and 4 both heavy and light load conditions are considered. The trajectories of modes L and M are shown in Figure 11.7.
From the modal trajectories, it is observed that:

For increases of stabilizer gain up to 40 pu the shift in the inter-area mode M is moreor-less directly to the left in the s-plane with small changes in modal frequency at
higher gains.

The shift in mode L is negligible. It may be necessary to investigate whether stabilizers

on other SVCs in the system enhance the damping of mode L.

The use of the FDS transfer function, whose tuning is based on a damping gain setting of 20 pu on all PSSs, is satisfactory for (i) both the heavy and light load cases
investigated, (ii) both PSS damping gain settings of 10 and 20 pu.

Sec. 11.5

Tuning a SVC stabilizer using real-power flow

545

As predicted from Figure 11.6 no evidence of closed-loop instability is found over the
FDS gain range 0 - 100 pu for the six cases investigated.

Imaginary Part (rad/s)

Modes:

l
lll
l
L lll
L
L
L
LLL

3.4
3.2

L M, Case 01,
PSS Gains 20 pu

3
m

2.8

2.6
2.4
2.2
2
1.8

m
M

M M

M
M

m
M

l m, Case 01
PSS Gains 10 pu

mZ m m m m m
m
llll
M M M MM
LL ll
M
L
L

L M, Case 04,
PSS Gains 20 pu

m Z m m m m mm
M M M MMM

l m, Case 04,
PSS Gains 10 pu

1.6
1.4
3

2
1
Real Part (Np/s)

Figure 11.7 Cases 1 and 4. Trajectories of the inter-area modes L and M as the stabilizer
gain k fds is increased in 10 pu steps from zero (shown by an arrow) to 100 pu on the SVC
base. All PSS damping gains are set to 10 or 20 pu on machine base.
Z: Estimated mode values from (11.5) for stabilizer gain k fds = 40 pu.

11.5 Use of line real-power flow as a stabilizing signal for a SVC

Consider the case when a major load is connected to two separate areas of generation
through high voltage transmission lines. Associated with these areas there is an inter-area
mode which is assumed to be lightly damped. If a SVC is located close to the load centre the
FDS may be able to improve the damping of the oscillatory power flow between the areas
by modulating the load-area voltage - and hence the real power flow into the load.
Referring to the system diagram in Figure 10.1, it is observed that a major load and the SVC
BSVC_4 are both connected to bus 412. It is also noted that the power flow on transmission
lines between buses 410 and 412 and between buses 411 and 412 supply the net real power
to the load at 412. The total net power flow perturbations, P tot , at bus 412 into the load
bus will be considered to be a potential stabilizing signal.
For the purposes of calculating the residues of the transfer function P tot h V ref h ,
the SVC is placed under closed-loop voltage control with the FDS path in Figure 11.8 open.
All PSSs are in service with their damping gains set to 20 pu on machine base. A polar plot
of the residues is shown in Figure 11.9 from which it is noted that, by comparison with that
of Figure 11.4, the spread of amplitude and phase over the range of operating conditions is

546

Tuning FACTS Device Stabilizers

Ch. 11

greater. Note: we cannot compare the residues derived from power signals with those derived from bus frequency (see Section 3.7).
Vt
Vref

Vs

Vd

500
s

2.5
1+s0.005

0.01

Q/Vt]

Ptot
FDS

Figure 11.8 The controller and stabilizer for SVC BSVC_4 showing terminal voltage and
droop controls, as well as the real power stabilizing signal P tot .

Figure 11.9 BSVC_4. Polar plot of the residues of the transfer function P tot V ref for
modes L and M and six operating conditions. PSS damping gains 20 pu.
Note: the magnitude scale is to be multiplied by 10.
As previously discussed, the objectives of the compensation is to improve the damping of
the inter-area mode M and, if feasible, mode L and the local-area modes as well. To cover
the range of rotor modes (see Tables 10.11, 10.15 and 10.16), a compensation angle of 45

Sec. 11.5

Tuning a SVC stabilizer using real-power flow

547

is selected together with a representative value for the residue of 2.0 pu on SVC base. A firstorder compensator is specified, together with parameters of the first-order washout and
low-pass filters of TW = 7 s and TLP = 0.02 s, respectively. The complex tuning frequency is
again 0.5 j1.9 , targeting mode M.
Using an iterative procedure in Mudpack the parameters of the compensator are calculated
based on (11.11). The resulting FDS transfer function is
s7
F s = k fds H s = k fds --------------1 + s7

1 + s0.188
------------------------1 + s0.799

1
---------------------- ;
1 + s0.02

the nominal value of the upper gain is k Rm = 2.02 pu.

With the FDS in service, the trajectories of the inter-area modes L and M, together with that
of a controller mode X associated with the FDS, are plotted in Figure 11.10 for Case 1. It is
seen that mode X becomes unstable when the stabilizer gain k fds = 3.47% or 0.070 pu on
SVC base. This result is confirmed from the Bode plot of the open-loop transfer function
V s j f V ref j f .
2.5

LLLLL

Imaginary Part (rad/s)

kfds=5%
k

=0

fds

M
MMM MM
MMMM

1.5

1
k
k

=3.5%

=5%

fds

fds

0.5
k

=1%

fds

0
1

X
X

0.5
0
Real Part (Np/s)

0.5

Figure 11.10 Case 1 with power flow FDS. Trajectories of the inter-area modes L and M
and stabilizer mode X as the stabilizer gain k fds is increased in 0.5% steps from zero to 5%
(0.101 pu on SVC base). All PSS damping gains set to 20 pu.
In Table 11.3 the estimated rotor mode shifts for the inter-area mode M are compared with
the eigen-analysis-based (actual) values for a range of stabilizer gains. The estimated mode

548

Tuning FACTS Device Stabilizers

Ch. 11

shifts are based on (11.5) and the actual shifts are calculated by the software package for
the full system.
Table 11.3

Estimated a and actualb mode shifts in mode M for Case 1, k fds = 1 to 4%

Rotor mode
shift

Stabilizer gain, k fds

1% (0.020 pu) 2% (0.040 pu) 3% (0.061 pu) 4% (0.081 pu)

Estimated shift

0.02
j0.02

0.05
j0.04

0.07
j0.06

0.09
j0.07

Estimated mode

0.54 j1.78

0.57 j1.76

0.59 j1.74

0.61 j1.72

Actual mode shift

0.03
j0.02

0.08
j0.02

0.12 j0.00

0.15 j0.05

Actual mode value

0.55 j1.78

0.60 j1.78

0.65 j1.80

0.68 j1.85

a: Based on (11.5). b: Calculated from eigen-analysis

Value of mode M at k fds = 0 is 0.52 j1.80 .

From the table it is observed: (i) With increasing gain the left shift in the real part of the
mode calculated from eigen-analysis is 30 to 70% greater than the corresponding estimated
shifts. (ii) The system is unstable at k fds = 3.47 % (0.070 pu), however, the Method of Residues does not indicate that the instability of a controller mode occurs.
Allowing for a 10 dB (3.1 times) margin the gain setting k fds should be 0.070/3.1=0.022 pu
(1.1%). The associated mode shift is small and therefore the operation of a power-stabilized
FDS may not be justified for this system based on this study alone.

11.6

Use of bus frequency as a stabilizing signal for the SVC, PSVC_5

It was noted in Section 11.4.1.1 that the frequency-stabilized FDS installed on the SVC,
BSVC_4, did not usefully contribute to the damping of inter-area modes K and L. For the
purposes of coordination of stabilizers in Chapter 14 it is of interest to ascertain if the SVC,
PSVC_5 at bus 507, contributes to the damping of any of the inter-area modes. The maximum and minimum reactive power generation for PSVC_5 is 320 and 180 Mvar, respectively, giving a reactive range (Mbase) of 500 Mvar. The relevant details are provided in
Table 10.17 and in the block diagram of Figure 11.2; Sbase = 100 Mvar.
A similar procedure to that outlined for BSVC_4 is followed: (i) to determine the parameters
of the FDS, and (ii) to evaluate the damping performance of PSVC_5 by means of the modal
trajectories for increasing FDS gain. For this purpose the FDS at BSVC_4 is out of service.
The local bus frequency Frq is used as the stabilizing signal and is synthesized from angular
perturbations (rad) in the terminal voltage of the SVC, bus 507 in Figure 10.1. For Cases

Sec. 11.6

Bus frequency stabilizing signal for SVC, PSVC_5

549

1 to 6 the polar plot of the residues of the inter-area modes K, L and M are shown in
Figure 11.11; the damping gain of the PSSs is set to 20 pu.

Figure 11.11 PSVC_5. Polar plot of the residues for the transfer function F rq Vref
for modes K, L and M and six operating conditions. All PSS damping gains are set to 20 pu
on machine MVA rating. The values of modes K, L and M are in the vicinity of
1.0 j4.0 , 0.45 j2.6 and 0.5 j1.9 , respectively.
Note: the magnitude scale is to be multiplied by 0.1.
Because the magnitudes of the residues for mode L in Figure 11.11 are two to three times
greater than those for modes M and K, the FDS tuning targets mode L. It is evident from
Table 11.4 that the representative values for PSS damping gains of 20 pu are likely to lead
to a satisfactory FDS design for the case when all PSSs are set to the lower value of 10 pu.
Table 11.4 Representative values for evaluation of
compensation transfer function, mode L
PSS
gain (pu)

Phase spread
(deg)

Representative phase
angle (deg)

Compensation angle
(deg)

Maximum
residue

Representative modal
frequency

20

-111 --> -151

-131

-49

0.0294

0.45 j2.6

10

-123 --> -148

-136

-44

0.0251

0.29 j2.7

550

Tuning FACTS Device Stabilizers

Ch. 11

The FDS transfer function, based on the PSS damping gains of 20 pu is found to be:
6s
F s = k fds --------------1 + 6s

1 + 0.140s
------------------------1 + 0.685s

1
------------------------- , k Rm = 312 pu;
1 + 0.033s

(11.13)

the same washout and low-pass filter time constants as for the FDS of BSVC_4 are employed to cover the ranges of the inter-area modal frequencies.

4.8

4.8

4.4

KKKKKK
4.4 KKK

4
3.6

Imaginary Part (rad/s)

Imaginary Part (rad/s)

To evaluate the effectiveness in the damping introduced by the FDS on PSVC_5 a similar
set of modal trajectories to those in Figure 11.7 are plotted in Figure 11.12.

kkkkkkk
KKKKKKKK

3.2
2.8
2.4
2
1.6
1.25

ll l
Ll
L LlLll
LLl
LLLlLlLl
m
m
m
m
m
mmm
MMM
mm m
M
MMMMM
1 0.75 0.5 0.25
Real Part (Np/s)

kkkk
kk
k

4
3.6

l l
ll
L lL
L
l
L
lllll
LLLL
L

3.2
2.8

mm m m
MmMmMmmm
Mm
MMMMMMMm

2.4
2
1.6
1.25

1 0.75 0.5 0.25

Real Part (Np/s)

Modes:
K L M, Case4, PSS Gains 20 pu
K L M, Case1, PSS Gains 20 pu
k l m, Case1, PSS Gains 10 pu

k l m, Case4, PSS Gains 10 pu

Figure 11.12 Mode trajectories for Cases 1 and 4 as the FDS gain on PSVC_5 varies from
zero (shown by an arrow) to 100 pu on device base in 10 pu steps. In each case all PSS damping gains are set to 10 or 20 pu on generator MVA rating.
For the range of FDS gains 0 to 100 pu this system is stable.
From the trajectories of the inter-area modes the following are observed.

Mode L is left-shifted in the s-plane with a slight decrease in frequency when all PSSs
are set to 20 pu. However, its improvement is limited in the light-load condition, Case
4, when the FDS gain exceeds 30 pu.

Improvement in the damping of mode M is also limited for Case 1, the heavy load
condition, when the stabilizer gain exceeds 30 to 40 pu.

There are marginal improvements in the damping of mode K, but are limited for FDS
gains exceeding 30 pu.

Sec. 11.7

Tuning a FDS for a TCSC using power flow signal

551

Because the FDSs on BSVC_4 and PSVC_5 induce marked shifts in modes M and L respectively, coordination between the FDSs is desirable to achieve the best improvement in the
damping of these modes within the limitations which have been observed. The coordination
of stabilizers, PSSs and FDSs, will be considered in Chapter 14.

11.7 Tuning a FDS for a TCSC using a power flow stabilizing signal
A series capacitor is primarily installed to reduce the series inductive reactance of transmission lines thereby improving both the voltage and rotor-angle stability of the interconnected
system [2], [22] and [23]. It also reduces the voltage drop between buses straddling the line
and series capacitor - as well as reducing the I2X losses in the circuit. The proportion of the
lines series inductive reactance which the series capacitor cancels out depends on a number
of factors which are determined by the characteristics of the system, [23]; such factors are
beyond the scope of this discussion.
For present purposes it is assumed that in a Thyristor-Controlled Series Capacitor (TCSC)
the series reactance is effectively perturbed by an amount X through the action of the stabilizer [24]. Conceptually, for perturbations in real power flow in the line the action of the
FDS is to reduce the effective series reactance of the line when the power flow tends to increase, and vice-versa. Damping of both the relevant modes and the line flow perturbations
is thereby improved. Due to the action of its washout filter the FDS does not respond to
relatively slow changes in the lines real power flow associated with changes in load or in
generation dispatch.
The Method of Residues is again employed for the tuning of the stabilizer for the TCSC;
this technique is applied in [4], [5], [25], [26], and Appendix A of [27]. Other techniques are
covered in [28], [30] and [31].
It is assumed that equivalent single series capacitance, located between buses 310 and 311
in the simplified 14-generator system (see Figure 10.1), is thyristor controlled [29]. The
MVA base (Mbase) for the TCSC is selected to be 300 Mvar, The relevant section of the
network and the format of the stabilizing controls are shown in Figure 11.13 (a) and (b), respectively.
In Figure 11.13(b), for the purposes of analysis, (i) a dummy reference is inserted in the controller, and (ii) the gain in the forward path is KA = 1 pu on Mbase. Based on the transmission line data in Table 10.19, the effective series reactance of the two capacitors in parallel
is j0.01685 pu on system base (Sbase = 100 MVA).

552

Tuning FACTS Device Stabilizers

VP_ref
Pt

309

Vp
311

310
Area 3

(a)

P-T: Power Transducer

FDS

Area 1

(b)

KA

102

Ch. 11

KS

P-T

Pt

Pt pu on Sbase; KA, X pu on Mbase

Figure 11.13
(a) TCSC in the simplified 14-generator system (see Figure 10.1 for bus numbering).
(b) Model of FDS using the perturbation in the real power flow P t (pu) through the TCSC
as the stabilizing signal. (KS = Sbase/Mbase=100/300)
X (pu) is the perturbation in the series reactance of the equivalent single capacitor.
For the purposes of testing the variation of the residues with PSS damping gains it is assumed that the inter-area modes may be heavily or lightly damped for all the operating conditions. Such damping is implemented by setting all PSS damping gains to 20 pu or by
reducing all such gains to 5 pu on generator MVA rating. The effect of the gain reduction
on all the rotor modes in Cases 1 to 6 can be seen in Figure 10.26; for the inter-area modes
the associated eigenvalues are listed in Table 11.5. It is of interest to learn if an improvement
in the damping of the inter-area modes such that their damping ratios exceed 0.1 is achievable with the FDS installed on the TCSC.
In order to derive a transfer function for the FDS we follow the procedure outlined in the
previous studies.
With no stabilizers in service on the SVCs and at the TCSC, the values of the lightly- and
heavily damped inter-area modes are listed in Table 11.5 for PSS damping gains of 5 and 20
pu; the residues are calculated for these PSS damping gains and modes.
According to Table 11.5 for the operating conditions in Cases 1, 3 and 4 the power flow
through the TCSC is from Area 1 to Area 3, and from Area 3 to Area 1 for Cases 2, 5 and
6. It is therefore proposed to use the modulus of the total power Pmod = P t through
the TCSC as the stabilizing signal with the object of deriving a single FDS transfer function
covering flows in both directions. For the purpose of calculating the residues of the transfer
function Pmod h V ref h the FDS path in Figure 11.13(b) is open. Polar plots of the
residues are shown in Figures 11.14 and 11.15 for PSS damping gain settings of 20 and 5 pu,
respectively.

Sec. 11.7

Tuning a FDS for a TCSC using power flow signal

553

Table 11.5 Modes K, L, M. No FDSs on SVCs and at TCSC.

PSS gains 5 and 20 pu on generator MVA rating.
Case

TCSC/ PSS
Line* gains
(MW) (pu)

763/
984

-1291/
-1023

730/
984

58 /
199

-379 /
-201

-141 /
0

Mode K

Mode L

Mode M

Value

Value

Value

0.18 j3.93

0.05

0.05 j2.57

0.02

0.14 j1.98

0.07

20

1.04 j3.64

0.28

0.39 j2.40

0.16

0.52 j1.80

0.28

0.05 j3.75

0.04

0.11 j2.81

0.03

0.050 j1.88

0.03

20

0.77 j3.54

0.21

0.45 j2.54

0.17

0.43 j1.76

0.24

0.28 j4.02

0.07

0.12 j2.63

0.05

0.16 j2.01

0.08

20

1.12 j3.71

0.29

0.43 j2.42

0.17

0.58 j1.86

0.30

0.14 j4.74

0.03

0.13 j3.53

0.04

0.16 j2.64

0.06

20

1.08 j4.58

0.23

0.56 j3.32

0.17

0.59 j2.51

0.23

0.08 j4.13

0.02

0.12 j3.08

0.04

0.09 j2.12

0.04

20

0.88 j3.90

0.22

0.46 j2.89

0.16

0.50 1.96

0.25

0.12 j4.80

0.03

0.11 j3.51

0.03

0.11 j2.57

0.04

20

1.08 j4.64

0.23

0.57 j3.30

0.17

0.52 j2.45

0.21

* Total power flow through (i) TCSC at and from bus 310 (upper value); (ii) Line, from bus 102 to 309 at
- Damping ratio
309 (lower value)

Figure 11.14 reveals that a lag compensation angle exceeding 60 would cause a degradation
in the damping of Mode L for Case 2. Four assumptions are therefore made in determining
the representative values for calculating the compensation transfer function. (i) The degradation in mode L is ignored unless it becomes excessive (i.e. other stabilizers are capable of
providing additional damping for this mode). (ii) The target mode for improvement in
damping is mode K. (iii) The FDS on the TCSC is switched off-line only when the steadystate power flow in the TCSC is less than 200 MW, i.e. in Cases 4 and 6. (iv) It is anticipated
that PSS damping gains are normally in the vicinity of 20 pu on machine base.
For the FDSs designed for SVCs it is noted in mode trajectories, such in Figure 11.12, that
the imaginary parts of the modes tend to decrease with increasing gain when the compensation shifts the residue such that the imaginary part of the residue is negative. For example,
in Figure 11.14 the residue for mode K, Case 5, is 6.1 70.7 ; if the compensation angle
were 90 lagging, say, the residue is shifted to 161 . It is therefore decided to provide overcompensation with a compensation angle of 110 in order to increase the frequency of oscillation of the inter-area modes; this applies to Cases 1, 2, 3, and 5.

554

Tuning FACTS Device Stabilizers

Ch. 11

Figure 11.14 TCSC: Polar plot of the residues for the transfer function
Pmod h V ref h for modes K, L and M and six operating conditions.

All PSS damping gains set to 20 pu. Note: magnitude scale is to be multiplied by 100.

Figure 11.15 TCSC: Polar plot of the residues for the transfer function
Pmod h V ref h for modes K, L and M and six operating conditions.
All PSS damping gains set to 5 pu. Note: magnitude scale is to be multiplied by 100.

Sec. 11.7

Tuning a FDS for a TCSC using power flow signal

555

For the FDSs designed for SVCs it is noted in mode trajectories, such in Figure 11.12, that
the imaginary parts of the modes tend to decrease with increasing gain when the compensation shifts the residue such that its imaginary part is negative. For example, in Figure 11.14
the residue for mode K, Case 5, is 6.1 70.7 ; if the compensation angle were 90 lagging,
say, the residue is shifted to 161 . It is therefore decided to provide over-compensation
with a compensation angle of 110 in order to increase the frequency of oscillation of the
inter-area modes; this applies to Cases 1, 2, 3, and 5.
It is decided to base the calculation of the compensation transfer function on PSS damping
gains of 20 pu because:

the nature of the residues in Figures 11.14 and 11.15 are comparable for PSS damping
gains set to 20 and 5 pu, respectively;

the representative values in Table 11.6 for 5 pu PSS damping gains are essentially covered by those for the 20 pu gain settings.
Table 11.6 Representative values for evaluation of the compensation
transfer function, mode K (excluding Cases 4 and 6)
PSS
RepresentPhase spread
damping
ative phase
for K (deg)
gain (pu)
angle (deg)

CompensaRevised
Representtion angle Comp. angle
ative residue
(deg)
(deg)

Representative modal
value

20

-71 --> -93

-82

-98

-110

20.3

1 j4

-86 --> -100

-93

-87

-110

20.1

0.14 j4.2

The transfer function is therefore based on the representative values for PSS gains of 20 pu.
Because the required phase lag is greater than 60 and less than 120 a second-order transfer function is selected (see Appendix 6I.1). The transfer function is therefore:
Vref
6s
1 + 0.078s 2
1
------------------- = k fds H s = k fds --------------- ------------------------- ------------------------- k Rm = 0.404 pu. (11.14)
Pmod
1 + 6s 1 + 0.429s
1 + 0.025s

11.7.1 Gain range for the stability of TCSC with the FDS in service
Maximum power flow through the TCSC from buses 310 to311 occurs in Cases 1 and 2 (763
and -1291 MW, respectively). Analysis to establish the range of gains for which the system
is stable is based on the Bode plot of the open-loop transfer function V p j f V ref j f
and assumes (i) negative feedback at the open-loop summing junction 1, (ii) there are no
open-loop poles in the right-half of the s-plane. For Cases 1 and 2 and with all PSS damping
gains set to 20 pu the Bode plot of the open-loop transfer function V p j f V ref j f is
shown in Figure 11.16. With the FDS in closed-loop operation the plot establishes for
1. Positive feedback of the FDS transfer function is assumed in Figure 11.13.

556

Tuning FACTS Device Stabilizers

Ch. 11

Magnitude (dB)

Case 1 that the system is stable - theoretically - over the gain range zero to infinity. In practice due to unmodelled dynamics the gain range for system stability may be limited. However, for Case 2 the Bode Plot reveals that the system is unstable for stabilizer gains exceeding
0.21 pu on the TCSC base (this limit corresponds to 51.5% of the nominal upper gain of
0.404 pu). Eigen-analysis of the closed-loop system reveals that a mode associated with a
stabilizer state migrates into the right-half of the s-plane at the limiting value of gain.

20
10
0
10
20
2

10

10
10
10
Frequency (rad/s)

Phase (deg)

10

10

10

10

10

0
180
360
10

Case 1

Case 2

Figure 11.16

TCSC. Cases 1 and 2: Open-loop frequency responses of the FDS and SVC,
V p V ref , k fds = -1. Case 1 is stable over the gain range; for Case 2 the gain margin is
13.6 dB at 0.35 rad/s. All PSS damping gains 20 pu on generator MVA ratings.

11.7.2 Inter-area mode trajectories with increasing stabilizer gain

The stabilizer transfer function is given by (11.14). With the stabilizer loop closed, it is desirable (i) to assess the nature of the variation of the inter-area modes K, L and M as the
value of the gain k fds is increased, and (ii) to compare modal values with those estimated
based on (11.5) of the Residue Method. For the heavier load Cases 1 and 2, and for stabilizer
gain settings between zero and 0.135 pu, the trajectories of the inter-area modes are plotted
in Figure 11.17. The modal trajectories are the eigen-value plots calculated by the Mudpack
software package.
In Figure 11.17 for Case 1 the estimated values based on (11.5) of the Residue Method agree
closely with those from eigen-analysis. However, for Case 2 in Figure 11.17 the modal values
calculated from eigen-analysis diverge from those estimated for FDS gains greater than

Sec. 11.8

Concluding comments

557

0.045 pu. As predicted from the plot of residues for Case 2, the damping of mode L degrades
slightly with increasing FDS gain.

3.5

4
K K KKK
KKKKK

3
2.5

LLLLLLLLLL

MMMMM
MMMM

1.5
1.6

Imaginary Part (rad/s)

Imaginary Part (rad/s)

1.2
0.8
0.4
Real Part (Np/s)

kk

3.5

k k k kZk kZ k k

3
llll

2.5
2
1.5

2.5

mm
mm
m
m
m
Zm
m
Zm

2 1.5 1 0.5
Real Part (Np/s)

Case 2, Modes k l m

Case 1, Modes: K L M

Figure 11.17 TCSC, Cases 1 and 2. Trajectories of the inter-area modes K, L and M as the
stabilizer gain is increased from zero (shown by an arrow) to 0.135 pu in 0.015 pu steps.
All PSS damping gains are set to 20 pu on machine MVA rating.
Z: Estimated mode values from (11.5) for stabilizer gain k fds of 0.03, 0.06 and 0.09 pu on
TCSC base.
For this study there are only a few feasible operating conditions on which to base the tuning
of the FDS with confidence. In addition to line and other outages conditions it would be
desirable to include operating conditions in which the flow through the TCSC from Area 1
to Area 3 is of a comparable magnitude to that in Case 2, i.e. about 1300 MW. Clearly the
range of encompassing conditions needs to be widened - and the benefits established - in
order to justify an expensive FACTS device such as a TCSC with stabilizing controls.

11.8 Concluding comments

11.8.1 Improving the damping of inter-area modes using FACTS devices
A number of papers on the tuning of a stabilizer for a FACTS device consider a SMIB system or simple four machine system with a single inter-area mode. However, the examples
in this chapter illustrate the potential difficulties of developing such a stabilizer at a particular
location in a multi-modal, multi-machine system, particularly as the location of the device is
chosen primarily for reasons other than damping rotor modes. Consequently, the stabilizer
may not be able to provide the desired damping for some or all the lightly-damped modes
of concern. The examples do suggest that a number of locations at which FACTS devices
are situated could be examined to determine their suitability for improving the damping particular modes. While the frequency-stabilized FDS might be feasible for BSVC_4 for improving the damping of the inter-area mode M (in Section 11.4.1), it has a much smaller

558

Tuning FACTS Device Stabilizers

Ch. 11

effect on mode L. A FDS in some other part of the network may, because of its location,
provide superior damping for mode L but may be less effective for mode M. It might therefore be useful to screen the appropriate FACTS devices to ascertain if a stabilizer installed
on the device might be effective in damping particular modes.
In considering the FACTS devices in the 14-generator system analysed in Chapter 10 there
are a number of sites at which SVCs are located. However, an examination - for example of the residues of the SVC in Area 3, RSVC_3, reveals that over the range of operating conditions a FDS installed on this SVC produces relatively much greater mode shifts for the
inter-area mode K than for modes L and M. Furthermore, other analyses show that incrementing the PSS damping gain of the generators at LPS_3 in Area 3 is very effective in enhancing the damping of mode K. A stabilizer installed on this SVC would have been of more
interest if it were relatively more effective in damping mode L, say.
11.8.2 Robustness of FDSs
In tuning fixed-parameter FDSs the concept of robustness is based on the following considerations:

there are two important components of a fixed-parameter FDS transfer function

k fds G c s which should be decoupled for practical purposes;
(a) the rotor modes are more-or-less directly left-shifted by the FDS compensating transfer function G c s with increase in the FDS gain, k fds 1;
(b) the extent of the left-shift of the rotor modes is determined by the gain,
k fds ;

ideally, the incremental left-shifts of the rotor modes are linearly related to increments
in FDS gain for changes about selected nominal values.

such considerations should apply over the set of encompassing operating conditions
and an appropriate range of rotor modes.

The securing of a predominately left-shift of the relevant modes with increasing stabilizer
gain is a requirement for the simultaneous coordination of PSSs and FDSs in Chapter 14,
[32].
From the various studies presented in this chapter it is clear that the task of ensuring robustness is complex and time-consuming. In particular the FDS typically enables the damping
of certain modes only, such damping being found to be dependent upon the location of the
FACTS device in the system and the type of stabilizing signal employed.

1. This is not a damping gain which is associated with P-Vr based PSS tuning.

Sec. 11.9

References

559

Unlike the P-Vr characteristics which contributes robustness in the analysis of generator
PSS parameters, there appears to be no corresponding system characteristics for the determination of FDS parameters. The issue of robustness of FDSs is a deficiency associated with
the Method of Residues.
11.8.3 Estimated versus calculated mode shifts
From the mode trajectories for increasing stabilizer gains in Figures 11.7 and 11.17 it is observed that the value of the estimated mode, derived from the mode shift calculated using
(11.5), diverge significantly at higher gains from those calculated using eigen-analysis.
The estimated mode shifts are based on the simple relations in (11.5) or (11.6). These do not
account of the characteristics of a multi-machine system. In particular, as is well-known in
the root-locus analysis of transfer functions, with increasing gain the modes migrate from
the open-loop poles to the finite system zeros or those zeros at infinity. As the gain of the
FDS is increased from zero, the influence of system zeros arises and the mode trajectories
deviate from the ideal direct left-shift, or approach a close-by zero. Moreover, these modes
may diverge from the estimated left-shift at relatively low or high gains. If there are openloop zeros in the right-half s-plane, poles may migrate towards them resulting in instability,
possibly at a relatively low value of stabilizer gain.
11.8.4 The notion of a nominal upper gain for FDSs
The value of the stabilizer gain can be expressed as a fraction or a percentage of the so-called
nominal upper gain, k Rm (per unit). Note that (11.7) shows that k Rm is inversely proportional to the representative value selected for the magnitude of the residue. If the largest
magnitude is chosen then the value of k fdsR = 0.1 k Rm may provide a notional warning
that the mode shift may no longer increase linearly with increase in stabilizer gain for certain
modes or operating conditions. As stated above, it has been shown that in multi-machine
systems a number of other factors may result in linear changes in mode shift ceasing at lower
values of k fdsR . Depending on the application, the latter values may provide a convenient
or meaningful gain limit - or a warning for the user.

11.9 References
[1]

Narian G. Hingorani and Laszlo Gyugyi, Understanding FACTS: concepts and technology of flexible AC transmission systems, IEEE Press, Piscataway, NJ, 2000.

[2]

P. Kundur, Power System Stability and Control, McGraw-Hill, Inc., 1994.

[3]

E. V. Larsen, J. J. Sanchez-Gasca, and J. H. Chow, Concepts for design of FACTS

controllers to damp power swings, Power Systems, IEEE Transactions on, vol. 10, pp.
948-956, 1995.

560

Tuning FACTS Device Stabilizers

Ch. 11

[4]

P. Pourbeik, Design and Coordination of Stabilisers for Generators and FACTS devices
in Multimachine Power Systems, PhD Thesis, The University of Adelaide, Australia,
1997.

[5]

P. Pourbeik and M.J. Gibbard, Tuning of SVC stabilisers for the damping of interarea modes of rotor oscillation, in Proceedings of Australasian Universities Engineering Conference, Perth, Sept. 1995, pp. 265-270.

[6]

S. Hiley, Redesign of the Blackwell Static Var Compensator Power Oscillation Damper
Controller. Undergraduate Thesis Report, School of Information Technology and
Electrical Engineering, The University of Queensland, Australia, 2007.

[7]

T. Smed and G. Andersson, Utilizing HVDC to damp power oscillations, Power

Systems, IEEE Transactions on, vol. 8, pp. 620-627, 1993.

[8]

P. Pourbeik and M.J. Gibbard, Damping and synchronising torques induced on generators by FACTS stabilizers in multimachine power systems, Power Systems, IEEE
Transactions on, vol. 11, pp. 1920-1925, 1996.

[9]

H. F. Wang and F. J. Swift, Capability of the static VAr compensator in damping

power system oscillations, Generation, Transmission and Distribution, IEE Proceedings-, vol. 143, pp. 353-358, 1996.

[10]

H. F. Wang and F. J. Swift, FACTS-based stabilizer designed by the phase compensation method. Part II: Multi-machine power systems, Proc. 4th International Conference on Advances in Power System Operation and Management, APSCOM-97, pp.644649, Hong Kong, 1997.

[11]

H. F. Wang, Phillips-Heffron model of power systems installed with STATCOM

and applications, Generation, Transmission and Distribution, IEE Proceedings-, vol.
146, pp. 521-527, 1999.

[12]

H. F. Wang, Selection of operating conditions for the coordinated setting of robust

fixed-parameter stabilisers, Generation, Transmission and Distribution, IEE Proceedings-, vol. 145, pp. 111-116, 1998.

[13]

J. J. Sanchez-Gasca, Coordinated control of two FACTS devices for damping interarea oscillations, Power Systems, IEEE Transactions on, vol. 13, pp. 428-434, 1998.

[14]

H. F. Wang and F. J. Swift, A unified model for the analysis of FACTS devices in
damping power system oscillations. I. Single-machine infinite-bus power systems,
Power Delivery, IEEE Transactions on, vol. 12, pp. 941-946, 1997.

[15]

H. F. Wang, F. J. Swift, and M. Li, A unified model for the analysis of FACTS devices in damping power system oscillations. II. Multi-machine power systems, Power
Delivery, IEEE Transactions on, vol. 13, pp. 1355-1362, 1998.

Sec. 11.9

References

561

[16]

H. F. Wang, Selection of robust installing locations and feedback signals of FACTSbased stabilizers in multi-machine power systems, Power Systems, IEEE Transactions
on, vol. 14, pp. 569-574, 1999.

[17]

N. Mithulananthan, C. A. Canizares, J. Reeve, and G. J. Rogers, Comparison of PSS,

SVC, and STATCOM controllers for damping power system oscillations, Power
Systems, IEEE Transactions on, vol. 18, pp. 786-792, 2003.

[18]

N. Tambey and M. L. Kothari, Damping of power system oscillations with unified

power flow controller (UPFC), Generation, Transmission and Distribution, IEE Proceedings-, vol. 150, pp. 129-140, 2003.

[19]

J. Guo, M. L. Crow, and J. Sarangapani, An Improved UPFC Control for Oscillation

Damping, Power Systems, IEEE Transactions on, vol. 24, pp. 288-296, 2009.

[20]

P. Pourbeik, D. J. Sullivan, A. Bostrom, J. Sanchez-Gasca, Y. Kazachkov, J. Kowalski, A. Salazar, A. Meyer, R. Lau, D. Davies, and E. Allen, Generic Model Structures
for Simulating Static Var Systems in Power System Studies - A WECC Task Force
Effort, Power Systems, IEEE Transactions on, vol. 27, pp. 1618-1627, 2012.

[21]

D.J. Vowles and M.J. Gibbard, Mudpack User Manual: Version 10S-03, School of
Electrical and Electronic Engineering, The University of Adelaide, July 2014.

[22]

G. D. Breuer, H. M. Rustebakke, R. A. Gibley, and H. O. Simmons, The Use of Series Capacitors to Obtain Maximum EHV Transmission Capability, Power Systems,
IEEE Transactions on, vol. 83, pp. 1090-1102, 1964.

[23]

P. M. Anderson and R. G. Farmer, Series Compensation of Power Systems, Encinitas,

Calif.: PBLSH Inc.,1996.

[24]

Yong Hua Song and A. T. Johns (Editors), Flexible ac transmission systems (FACTS),
The Institution of Electrical Engineers, UK, 1999.

[25]

N. Yang, Q. Liu and J.D. McCalley, TCSC controller design for damping interarea
oscillations, Power Systems, IEEE Transactions on, Vol. 13, no. 4, 1998, pp. 1304 1310.

[26]

L. Rouco and F. L. Pagola, An eigenvalue sensitivity approach to location and controller design of controllable series capacitors for damping power system oscillations, Power Systems, IEEE Transactions on, vol. 12, pp. 1660-1666, 1997.

[27]

Impact of the Interactions Among Power System Controls, CIGRE TF 38.02.16,

CIGRE, Paris, France, Technical Report 166, 2000.

[28]

J. J. Paserba, N. W. Miller, E. V. Larsen, and R. J. Piwko, A thyristor controlled series

compensation model for power system stability analysis, Power Delivery, IEEE
Transactions on, vol. 10, pp. 1471-1478, 1995.

562

Tuning FACTS Device Stabilizers

Ch. 11

[29]

E. V. Larsen, K. Clark, S. A. Miske, and J. Urbanek, Characteristics and rating considerations of thyristor controlled series compensation, Power Delivery, IEEE Transactions on, vol. 9, pp. 992-1000, 1994.

[30]

A. M. Simoes, D. C. Savelli, P. C. Pellanda, N. Martins, and P. Apkarian, Robust Design of a TCSC Oscillation Damping Controller in a Weak 500-kV Interconnection
Considering Multiple Power Flow Scenarios and External Disturbances, Power Systems, IEEE Transactions on, vol. 24, pp. 226-236, 2009.

[31]

A. B. Leirbukt, J. H. Chow, J. J. Sanchez-Gasca, and E. V. Larsen, Damping control

design based on time-domain identified models, Power Systems, IEEE Transactions
on, vol. 14, pp. 172-178, 1999.

[32]

P. Pourbeik and M. J. Gibbard, Simultaneous coordination of power system stabilizers and FACTS device stabilizers in a multimachine power system for enhancing
dynamic performance, Power Systems, IEEE Transactions on, vol. 13, pp. 473-479,
1998.

Chapter 12
The Concept, Theory, and Calculation of
Modal Induced Torque Coefficients

12.1

Introduction

In this chapter the concept, the theory, and calculation of modal induced torque coefficients
(MITCs) in multi-machine power systems are introduced. The concept of a modal induced
torque coefficient is new [1], [2]. It forms the basis for calculation of the shifts in rotor
modes when the stabilizer gains of one or more PSSs and/or FDSs are incremented by k
(pu) on device base. Based on the concept of MITCs, the background theory of the rotor
modes shifts, together with analysis of the effectiveness of, and interactions between, PSSs
and FDSs in multi-machine systems are described in Chapter 13.
The theoretical development of MITCs in this chapter is fairly detailed and can be omitted
if the practical applications of the analysis of rotor modes shifts are of primary interest.
Where relevant, references are made in Chapter 13 to the results and equations that are developed in this chapter. A case study in the latter chapter demonstrates the significance of
the MITCs and the insights that they provide into the dynamic performance of a multi-machine power system.
In essence, the concept of a modal induced torque coefficient is a further development of
the concepts of damping and synchronising torque coefficients based on frequency response analysis (i.e. s = j f ) [1]. In this chapter the torque coefficients are evaluated at the

563

564

Modal Induced Torque Coefficients

Ch. 12

complex rotor modes ( h = j). The frequency-analysis-based torque coefficients are

introduced in Section 5.3 for a SMIB system and in Sections 9.5 and 10.6 for the multi-mah

chine case. With shaft dynamics enabled, the modal induced torque coefficient, T ij , for the
complex rotor mode h is defined as a complex torque coefficient which is induced on generator i due to a perturbation in the stabilizing signal of stabilizer j. The stabilizer in question
may be a power system stabilizer (PSS) or a FACT device stabilizer (FDS) installed on a
FACTS device. Such FACTS devices are static var compensators (SVCs), high voltage DC
links, thyristor controlled series capacitors (TCSCs) among others. It will be shown that the
effect of any such stabilizer on the damping of any mode of rotor oscillation can be quantified. The concept can be extended to other devices such as wind turbine generators, photovoltaics and any other power-converter based transmission or generation equipment. The
calculation of the MITC is from any controller to the effective induced torque coefficient
on a specific synchronous generator. The controller may be installed on another synchronous generator on or any other dynamic device.
The analysis of the torque coefficients for generator i in the earlier chapters is based on the
frequency response of the transfer function P 0i s i s s = j f , all machine dynamics
being disabled. However, it is possible that due to perturbations in the speed of machine j a

torque coefficient is induced on generator i. However, because we cannot relate speed perturbations on machine j to those on generator i when the shaft dynamics are disabled, the
component of electro-magnetic torque induced by perturbations on generator i, P 0ij j f .
in phase with speed perturbations on generator i, i j f , cannot be calculated, i.e. the
damping torques induced by other machines on generator i are not available. However, the
concept of modal induced torque coefficients (MITCs) overcomes this problem and facilitates, among other outcomes, the calculation of synchronizing and damping torque coefficients at modal frequencies. Essentially the analysis is divided in parts, (i) analysis with rotor
dynamics enabled; (ii) analysis based on part (i) to derive the MITCs; (iii) analysis based on
the MITCs to derive mode shifts due to stabilizer gain increments (this analysis is conducted
in Chapter 13). The advantage of this approach is that it facilitates the study of (i) the effects of
controls on individual rotor modes, and (ii) the relative effects of controls on a set of selected
modes. It should be emphasized that the analysis of a large system with shaft dynamics enabled

is complex, and the effects of controls on selected modes may be difficult to separate out.
For the purposes of generality in the initial analysis of MITCs, it is assumed that the controller to which all n speed stabilizing signals are fed is a centralized PSS, a full n n matrix transfer function. Following the derivation of a set of general results, decentralized
PSSs are employed in which the PSS matrix transfer function is diagonal. The decentralized
stabilizer is, of course, the practical form of the PSS. Likewise for FACTS devices, the z local stabilizing signals are transmitted to a centralized FDS which is represented by a full z z
matrix transfer function. Each output of the centralized FDS is then fed to the summing

Sec. 12.2

Concept of Modal Induced Torque Coefficients

565

junction of the controller on each FACTS device. In the practical form of the decentralized
FDS the matrix transfer function is diagonal.
This chapter is structured as follows. In Sections 12.2.2 to 12.5 the concept of the modal
induced torque coefficient (MITC) is introduced and its physical significance is explained.
In order to apply the concepts to a multi-machine system a transfer function model of the
system and its controllers is derived. in Section 12.3. Furthermore, a method is outlined for
calculating MITCs when either a centralized PSS or FDS is in service. The application to
decentralized controllers follows in Section 12.6. Using parameter-perturbation analysis in
the remaining parts of the chapter, the relationship between MITCs and stabilizer gains is
established. On this relationship is based the calculations in Chapter 13 of the shifts in the
rotor modes of oscillation caused by an increment in the gain of any or all stabilizers.

12.2

The Concept of Modal Induced Torque Coefficients (MITCs)

12.2.1 Conventional frequency response techniques versus modal analysis

In previous chapters the analysis of decentralised, fixed-parameter PSSs and FDSs has been
based mainly on frequency response techniques, i.e. with s = j f . However, in the analysis
that follows we are interested in the components of torques of electromagnetic origin induced on the shaft of a generator at a modal frequency of rotor oscillation, s = = j ,
through the action of a generator or FACTS device stabilizer. Because certain relationships
[(12.67) and (12.69)] apply only at a modal frequency of rotor oscillation s = , modal analysis must be employed in the associated analysis. [It is easy to show that for a complex modal
frequency s = = j , which is a damped-sinusoid, the response of the transfer function G s at that modal frequency (and only at that modal frequency), is G . The derivation of this result is similar to that in Section 2.11 for G s ]. For the sake of generality,
however, in parts of Sections 12.2.4 to 12.6 which are applicable in both the domains
s = j f and = j the relevant expressions are expressed as functions of the form
Fs .

12.2.2 Modal torque coefficients induced by the action of a power system stabilizer
Two generators in a linearized representation of a multi-machine power system are shown
in Figure 12.1(a). The speed-input PSS on machine j is assumed to be in service; machine i
is not fitted with a stabilizer. A small system disturbance is assumed to occur which results
in (i) only the hth complex mode of rotor oscillation, h , being excited; (ii) the rotors of the
generators being perturbed from synchronous speed. For machines i and j, the relative magnitude and phase of perturbations in their speeds, i and j , respectively, are related by
(12.67) in Appendix 12I.1 (repeated here):

566

Modal Induced Torque Coefficients

v jh
j h
-,
-------------------- = -----v ih
i h

Ch. 12

(12.1)

where, for the eigenvalue h , v ih and v jh are the ith and jth elements of the right-eigenvector
corresponding to the respective speed states of the two generators.

j
P Tij

P
O
Pei W
E
R

TPij
/(sMi)

S
Di
Y
Pdi Pm S
T
o/s
E
i M

P Tjj

TPjj
j

Pej

/(sMj)

P
O
W
E
R

S
Dj P
Y
P
dj
mj S
T
o/s
E
j M

AVRi

Vri

P Tij

TFij

Vti
i

/(sMi)
Di

Machine i

S
Y
Pdi Pmi S
T
o/s
E
i M

P
O
W
E
R

Vtj
AVRj
Vsj

Vrj

Vti
AVRi

FACTS

Machine j

Vri

Machine i

Vtj
Cont.j

S
Y
S
T
E
M

Device j

Vrj

FDSj

PSSj

(b) Action of FDS j

(a) Action of PSS j

Figure 12.1

P
O
Pei W
E
R

(a) Conceptually, at complex modal frequency s = h , due to speed perturh

bation on machine j acting through its PSS, the MITCs T Pij and T Pjj for mode h are induced on machines i and j respectively.
h

(b) Likewise, the electrical torque coefficient T Fij is induced on machine i due to the perturbation in the local stabilizing signal j acting through the FDS on FACTS device j.
Consider in Figure 12.1(a) the signal path from the speed perturbation j through PSS j
to the torque of electro-magnetic origin acting on the rotor of generator j. As has been discussed in Section 10.6, there will be in the latter torque a component P Tjj induced on the
shaft of generator j by its own PSS. However, there is also a signal path from j through
PSS j to the electrical torque component P Tij on the rotor of generator i. The component

Sec. 12.2

Concept of Modal Induced Torque Coefficients

567

of P Tij on generator i in phase with the speed perturbation i on generator i is a damping

torque induced on that generator; this torque arises from the perturbation in j and the action
of PSS j [3]. Consequently, for the complex modal frequency h , a complex modal induced

torque coefficient, T hPij , can be defined which relates the change in a component of the
torque of electro-magnetic origin P Tij on generator i to the change in speed i on the
same machine, the electrical torque being a result of the perturbation, say, of the input signal
on stabilizer j, i.e.
h

P Tij h = T Pij i h .

(12.2)

Using (12.1) the torque coefficient, T hPij , may be expressed as follows:

h
T Pij

pss j

pss j

P Tij
j
v jh
P Tij
P Tij
--------- h = -------------s
------- .
= -------------- h = -------------- s
i
j
i
j
v ih
s = h
s = h

The term in (12.3), P Tij j

pss j
s = h

(12.3)

, is the transfer function from the speed of machine

j, j , through PSS j to the electrical torque component P Tij . The effect of the term,
v jh v ih , is to relate the perturbation in P Tij to a perturbation in speed i on machine i,

rather than machine j. Because, in (12.3), there are components of P Tij on machine i in
phase and quadrature with i , the concept of damping and synchronising torques can be
employed [4]. Conceptually, the complex modal induced torque coefficient, T hPij , can be
considered to be embedded in the linearized model of machine i, as shown in the dashed
blocks of Figure 12.1(a).
An examination of the linearized model of Figure 12.1(a) reveals that a modal torque coefficient is also induced on machine i due to a speed perturbation on machine j acting through
the feedback path of the rotor angle j of machine j; this path is parallel to that of PSS j.
Though not shown in Figure 12.1(a), this is an inherent torque coefficient. Since the inherent
torque coefficient is associated with a path which is independent of the PSS feedback path,
it is induced on the generators both in the absence and presence of the PSS. However, the
emphasis in this and Chapter 13 is on the role of stabilizers in enhancing the damping performance of the multi-machine power system. Nevertheless, it is important to ensure that
negative inherent modal torque coefficients are not significant enough to swamp out the
positive torque coefficients induced by the PSS.
12.2.3 Modal torque coefficients induced by the action of a FACTS device stabilizer
Consider a FACTS device in service in a multi-machine power system. In Figure 12.1(b) are
shown representative elements of a linearized model of the power system consisting of n

568

Modal Induced Torque Coefficients

Ch. 12

generators and z FACTS devices and z FDSs. An examination of the figure reveals there is
a signal path from the local stabilizing signal, j , through FDS j to the electric torque component P Tij on generator i.
In the analysis of the matrix transfer function representation of the multi-machine system
in Section 12.3 a system output vector is defined in (12.10) as:
T

Y = P e1 P en 1 z = y 1 y n y n + 1 y n + z ,

(12.4)

where P ei is the electrical power output of generator i; the vector of the z FDS stabilizing
T

signals is = 1 z . As a result of a perturbation in the stabilizing signal j

a complex modal torque of electromagnetic origin is induced on the shaft of the generator
i. The electric torque component P Tij induced on the shaft of machine i can be expressed
as:
P Tij
P Tij h = -------------- h
j

j h ,

j = 1 z .

(12.5)

fds j

Again, of particular interest for generator i is the component of P Tij in phase with the
speed on machine i. To determine the induced damping and synchronising torques, (12.5)
must be modified such that the induced electrical torque component on machine i is related
to its own speed perturbation, i . If a single mode of rotor oscillation, h , is excited then
the relative magnitude and phase of the two signals, i and the FDS stabilizing signal, j ,
can be calculated using (12.69) in Appendix 12I.1,
cq* v h
j
y
*--------- h = --------q- h = --------------, where
i
i
v ih

(12.6)

q = (n+j) since j is both the (n+j)th element of the output-vector Y and the jth
of ;

c q is the qth row vector of the output state matrix C;

v h is the right-eigenvector of the eigenvalue h ; and

v ih is the ith element of v h corresponding to the speed-state of generator i.

For the purposes of analysis, however, it is again more convenient to define a complex induced torque coefficient rather than consider the induced torque itself. The modal torque coh

efficient induced on machine i due to the action of FDS j is T Fij therefore defined as
h

P Tij h = T Fij i h .

(12.7)

Sec. 12.3

Transfer function matrix representation of a linearized system

569

Using (12.6) an expression for the torque coefficient T Fij is derived:

cq* v h
P Tij
j
P Tij
P Tij
*--------- h = ------------- h
--------------T hFij = -------------- h = -------------- h
.
i
j
i
j
v ih
fds j
fds j

(12.8)

The term, P Tij h j h , in (12.8) is the transfer function from the stabilizing signal
j through FDS j to the torque component P Tij on machine i, evaluated at the mode of

interest. This gives the component of electrical torque induced on machine i due to perturbation in j . The second term, c q * v h v ih , relates the perturbation in torque P Tij to
*

the perturbation in speed i on machine i. Because in (12.7) there are components of

P Tij on machine i in phase and quadrature with i , the concept of damping and synchronizing torques can again be applied. As was the case in Figure 12.1(a), the complex
h

MITC T ij can also be considered, conceptually, to be embedded in the linearized model of

machine i, as shown in Figure 12.1(b).
12.2.4 Modal torque coefficients induced by centralized stabilizers
In the following analysis it is assumed that a single centralized PSS accepts n speed input
signals and supplies n output signals to the n generators. Similarly, a centralized FDS accepts
z stabilizing signals and delivers z output signals at the z summing junctions of the FDSs.
The assumption that the stabilizing signals of the centralized PSS are shaft speeds is made
for convenience, but is not a necessary condition for the analysis; the theory of modal induced torque coefficients can be developed for an appropriate stabilizing signal.
In order to analyse the contributions to the MITCs on generators by the PSSs and FDSs in
a multi-machine system a transfer function matrix model of a power system is first derived.

12.3 Transfer function matrix representation of a linearized multimachine power system and its controllers
To calculate the modal induced torque coefficients it is necessary to identify the signal paths
which cause torques of electromagnetic origin to be developed on the rotors of generators,
and to be able to evaluate the associated components of torque. Consequently, in this section a transfer function matrix (TFM) model of a power system is derived which facilitates
the calculation of the shaft torques that result from the action of PSSs and FACTS Device
Stabilizers (FDSs) [1], [2]. Moreover, the associated state equations not only form the basis
for the calculation of the MITCs, but also the design of stabilizers for FACTS devices, the
coordination of FDSs with the PSSs, the interactions between stabilizers, and the development of the so-called stabilizer damping contribution diagrams.
Let the number of states be N, the number of generators be n and the number of FACTS
devices be z. The input vectors V r and U r represent voltage reference inputs to the gen-

570

Modal Induced Torque Coefficients

Ch. 12

erator AVR and the FACTS controllers, respectively. Similarly the vectors P e and
represent, respectively, the generator electrical power outputs and the local signals which act
as inputs to FACTS device stabilizers. The vector of system states can be divided into three
groups, , and q which represent the generator speeds, the generator rotor-angles
and the set of all other system states, respectively.
Consider the state-space model:

x = Ax + Bu ,

y = Cx + Du ,

(12.9)

where
x = q

N 1 ; y = P e

u = V r U r
A A A q
A = o I 0
0
0

n + z 1 ,

n + z 1 ;

N N ; A = diag K di M i

(12.10)

n n ;

(12.11)

A q A qq
0 0 B qVr
B =
0 0 B qUr
C =

0 C P C Pq
0 C C q

D = 0

N n + z ;

(12.12)

n + z N ;

(12.13)

n + z n + z .

(12.14)

Note that it is assumed that the system matrix, D, is a zero matrix; the physical interpretation
of which is that there is no instantaneous relationship between system outputs and system
inputs. This is a valid assumption in the context of power system models. Furthermore, assume that in the above state-space model:
(i) there are no PSSs fitted to the generators,
(ii) governors and turbines are not modelled, and
(iii) there are no stabilizers fitted to the FACTS devices in the system.
In order to disable the dynamics of all generator shafts, let us (i) temporarily remove the state
equations describing the shaft dynamics from (12.9), and (ii) treat as an input vector.
As explained in Sections 5.10.2 and 5.10.3 this is similarly achieved by eliminating the rows
associated with in matrices A, B and C, and transferring the columns associated with
the speed states in these matrices to expanded B and D matrices to form a new set of state
equations. Therefore, (12.9) reduces to

Sec. 12.3

Transfer function matrix representation of a linearized system

571

V r
q = A qq q + B qVr B qUr A q U , and
r

(12.15)

V r
0 0 C P
y =
q +
U r .
C q
0 0 C

(12.16)

C Pq

Using (12.15) and (12.16) we can write a TFM equation which relates the vector of system
T

outputs, y in (12.10), to the vector of system inputs, V r U r , with shaft dynamics

disabled. That is,
V r
H PVr s H PUr s H P s
=
U r ,
H Vr s H Ur s H s
s

P e s

(12.17)

where
1

(12.18)

(12.19)

H PVr s = C Pq sI A qq B qVr ;
H PUr s = C Pq sI A qq B qUr ;
1

H P s = C Pq sI A qq A q + C P ;

(12.20)

(12.21)

(12.22)

H Vr s = C q sI A qq B qVr ;
H Ur s = C q sI A qq B qUr ;
1

H s = C q sI A qq A q + C .

(12.23)

The above equations represent the TFMs from perturbations (i) in generator reference voltages V r , (ii) in the FACTS device reference inputs U r , and (iii) in generator rotor-angles
, to perturbations both in generator electric power outputs P e and in stabilizing signals
of the local FACTS devices. Details of the calculation of the matrices C P and C Pq

are supplied in Appendix 12I.3.

The shaft dynamics equations, which were temporarily eliminated from (12.9) in forming
(12.15), can be expressed as
s = J s P m s P e s ,

s = N s s ,

(12.24)

572

Modal Induced Torque Coefficients

Ch. 12

where
J s = diag 1 M i s + K di i 1 n
.
N s = diag 0 s n n

(12.25)

The generator shaft dynamics represented by (12.25), and the rest of the system dynamics
represented by equations (12.18) to (12.23), are combined with the TFM representing the
PSSs, FDSs and governors to form a TFM model of the power system. This model is shown
in Figure 12.2.

PSSs

Generating Units

Hpss(s)
Vs
Vr

1
PVr

HPVr(s)

Governors

Pm

Hgov(s)

PUr
HVr(s)

Transmission System

J(s)

Pe

HPUr(s)

HP(s)

HUr(s)
Ur

Us

FDSs

N(s)

H(s)

Hfds(s)
FACTS Devices

Figure 12.2 Transfer function matrix model showing three paths through which components of electrical modal torques are induced on the generator rotors.
The significance of each TFM block in the figure is examined below.
H pss s : is initially a TFM of a centralized speed PSS and is a full n n matrix. In later sec-

tions, because in practice each generator is fitted with a single, decentralized PSS, H pss s
becomes a diagonal matrix.

Sec. 12.3

Transfer function matrix representation of a linearized system

573

H fds s : is initially a TFM of a centralized FACTS device stabilizer and is assumed to be a

full z z matrix. The inputs are the local stabilizing signals given by the vector , z 1 .
The outputs of the centralized FDS, U s , are inputs to the summing junctions of the
FACTS device controllers. Again in later sections, H fds s becomes a diagonal matrix representing the transfer functions of decentralized FDSs.
H gov s is a TFM of the governors; it is a n 1 matrix.
H PVr s : is a full n n TFM which relates the perturbations in torque contributions of elec-

tromagnetic origin on generators, P Vr s , n 1 , due to perturbations in the reference inputs, V r , n 1 , on the AVRs when V s = 0 . Of particular significance is that the diagonal
element, H PVr_ii s , is the P-Vr transfer function of machine i.
H PUr s : is a full n z TFM which relates the perturbations in torque contributions of elec-

tromagnetic origin on generators, P Ur n 1 , due to perturbations in the reference signals

on the FACTS device controllers, U r , z 1 , when U s = 0 .
J s : is a n n diagonal TFM which represents the shaft dynamics of the units.
N s : is a n n TFM of diagonal elements 0 s ; it relates the perturbations in rotor angle

due to speed to perturbations on the units.

H P s : is a full n n TFM which relates the perturbations in torque contributions of elec-

tromagnetic origin, P , due to perturbations in the rotor angles of machines. It is through

this TFM that the inherent torques are produced.
H s , H Vr s and H Ur s : are full TFMs which relate the perturbations in the stabilizing signals, , due to perturbations in the rotor angle ( ), the machine reference volt-

age, ( V r ), and the FACTS controller reference signals, ( U r ), respectively, when V s and
U s = 0 .
In Figure 12.2 three distinct paths are shown through which components of electrical torque
are induced on the shafts of generators. The first path, #1, is through the speed-PSS feedback path, the second, #2, is through the FDS feedback path and the third, #3, is via the
rotor-angle feedback path. From Figure 12.2 it is revealed that the torque induced by the
third path is given by
P s = H P s N s s .

(12.26)

574

Modal Induced Torque Coefficients

Ch. 12

Substitution of (12.20) and (12.25) for H P s and N s , respectively, into (12.26) yields
1

P s = C P sI A A + C P diag 0 s s
1

= C P sI A A + C P s

(12.27)

The term C P is a matrix of real, constant, inherent synchronising torque coefficients (compare this with the coefficient K1 in the SMIB case of Figure 5.1). The term
1

C P sI A A + C P represents the inherent feedback paths from through the

generator electromagnetic circuits, the network and the AVR/exciter to the electrical torque
induced on the rotors of the generators. Because the term is complex at any frequency, its
real and imaginary components represent the inherent damping and synchronising torque
coefficients on each machine.
Following a large-magnitude disturbance, high gain AVRs tend to increase synchronising
power flows and thus enhance first swing stability [6]; however, high-gain AVRs have a tendency to reduce damping torques [7]. As has been discussed in earlier chapters, the objective
of PSS and FDS design in small-signal analysis is to induce positive damping torques on generator rotors for all modal frequencies of rotor oscillation. In order to achieve a constant
damping torque coefficient over the range of rotor modes of oscillation, the frequency response (for s = j f ) of the coefficient induced by the PSSs should ideally be flat with negligible phase shift and must swamp the negative inherent damping torque coefficients (see
Section 10.6). Furthermore, the PSS or FDS should not significantly reduce the inherent
synchronising torque coefficients. Paths #1 and #2 in Figure 12.2 will therefore be analysed
in more detail to determine quantitatively the modal torque coefficients induced on each
generator by the PSSs and FDSs.

12.4 Modal torque coefficients induced by a centralized speed PSS

The analysis in the previous section is formulated on a centralized stabilizers. The modal induced torque coefficients for the centralized PSS are derived below. The results for the decentralized PSS are then developed in Section 12.6.
Based on Figure 12.2 it is seen that the total electrical torques induced on the rotors of generators are the sum of the torque components associated with paths #1 to #3 and are expressed by the vector
P e s = P Vr s + P Ur s + P s , where

(12.28)

P Vr s - is the vector of contributions to the electrical torques resulting from rotor speed

perturbations being fed back through the centralized PSS, path #1;
P Ur s - is the vector of contributions resulting from the perturbations in the local stabilizing signals for FACTS devices being fed back through the centralized FDS, path #2;
P s - is the vector of contributions to the inherent electrical torques, path #3.

Sec. 12.4

MITCs induced by a centralized PSS

575

Note the effects of speed-governors are ignored here, but they can be included in the TFM
G m s , as shown in Appendix 12I.2, without affecting the following analysis.
Let us consider the contribution made by the TFM of the centralized speed-PSS to electrical
torques on the generators. It is shown in Appendix 12I.2 that the block diagram in
Figure 12.2 can be reduced to that illustrated in Figure 12.3; the latter shows only the loop
associated with the centralized PSS feedback path. The rest of the system dynamics, including that of the centralized FDSs, etc., has been incorporated in the TFM G P s .
Hpss(s)
Vs

PVr
Gp(s)

HPVr(s)

Vr

Figure 12.3 TFM representation of a linearized multi-machine system showing the loop
associated with PSS feedback paths. All other dynamics are included in Gp(s).
From Figure 12.3, the vector of contributions of electrical torque, P Vr , resulting from generator speed perturbations, , fed to the centralized speed PSS H pss s is
P Vr s = H PVr s H pss s s = P s s ,

(12.29)

where P s = H PVr s H pss s is a n n TFM which relates speed perturbations on any

machine to the PSS-induced torque perturbations on any machine. Based on (12.29), the total PSS-induced torque on the rotor of generator i due to the speed perturbations on all n
generators is given by
P Vr_i s =

P Vr_ij s =

j=1

Pij s j s ,

(12.30)

j=1

where the elements of the speed-torque TFM P s are given by

Pij s =

H PVr_il s H pss_lj s .

(12.31)

l=1

The jth element of the summation in (12.30), P Vr_ij , which is the component of the PSSinduced torque on machine i due to speed perturbation on machine j, is
P Vr_ij s = Pij s j s .

(12.32)

The above equation relates to torques induced on generator i by speed perturbations on i

and all other machines. However, to calculate the synchronizing and damping torques induced on generator i we need to explicitly determine torques in-quadrature and in-phase
with the speed perturbations on generator i, caused by speed perturbations on all generators.

576

Modal Induced Torque Coefficients

Ch. 12

Bearing in mind the comments in Section 12.2.1, consider now a complex mode of rotor oscillation, s = h . The torque P Vr_ij can therefore be defined in terms of the modal inh

duced torque coefficient T Pij in (12.2), i.e.

P Vr_ij h = T hPij i h ,

(12.33)

where the MITC is given in (12.3) by

h
T Pij

P Vr_ij
P Vr_ij
= ------------------- h = ------------------- s
i
j

pss j
s = h

v jh
------- .
v ih

(12.34)

Substitution in (12.34) of Pij s for P Vr_ij s j s from (12.32) thus yields

v jh
T hPij = Pij h ------- .
v ih

(12.35)

This equation provides an expression for calculating the MITC on machine i due to speed
perturbations on machine j, at the single modal frequency h . Equation (12.35) is substituted into (12.30) to give,
P Vr_i h =

v jh

n h
-----

Pij h v
i h = T Pij i h .

j=1
ih
j = 1

(12.36)

The term nj = 1 T hPij is the total MITC on machine i due to perturbations in the speed of all
n generators being fed back through the PSS TFM, H pss s . The total MITC on machine i

is defined as
hPi =

h
T Pij .

(12.37)

j=1

Using this definition, (12.36) becomes

P Vr_i h = hPi i h .

(12.38)

Hence, considering all n machines in the system, (12.29) can be written TFM form as
h

P Vr h = P h

(12.39)

at the modal frequency h , where the matrix P = diag hPi i [1,n] .

At this point it is important to reiterate the physical significance of the MITCs. Because the
h

real part of T Pij represents a damping torque coefficient, the total damping torque coefficient induced on generator i by the action of all stabilizers is hi .

Sec. 12.5

MITCs induced by a centralized FDS

577

In summary, for a centralized speed-PSS whose elements H pss_lj s are known, the total
induced torque coefficient hPi for generator i and mode h can be calculated by successively evaluating (12.31), (12.35) and (12.37).

12.5

Modal torque coefficients induced by a centralized FDS

PUr

Ur

Us

HPUr(s)

GF(s)

HUr(s)

L(s)

FDSs

Hfds(s)

Figure 12.4 TFM representation of a linearized multi-machine system showing the loop
associated with FDS feedback paths; other dynamics are included in TFMs GF(s) and L(s).
Let us consider now the contribution made by the TFM of a centralized FDS to the modal
electrical torque induced on the generators, P Ur . It is shown in Appendix 12I.2 that the
transfer function block diagram in Figure 12.2 can be simplified to that illustrated in
Figure 12.4; the latter shows only the loops associated with the centralized FDS feedback
path. All other system dynamics and feedback paths, such as the PSSs, have been absorbed
into the TFMs G F s and L s . This figure shows that there is a closed path from , the
vector of generator speeds, to P Ur , through the TFM of the FDSs. Therefore, associated
with speed perturbations there will be components of torque in phase and quadrature with
speed induced on the shaft of each machine through the feedback path of the FDSs. Based
on Figure 12.4 the stabilizing signals supplied to the FDSs are given by the vector
s = L s s + H Ur s U r s + H fds s s ,

(12.40)

where H fds s is the centralized FDS TFM defined in Section 12.2.4. The components of
torque induced through the feedback path of the FDSs are
P Ur s = H PUr s H fds s s + H PUr s U r s .

(12.41)

The vector U r is the vector of reference set points at the summing junction of the FACTS
device controllers (see Figure 12.2). Let us assume that there is no perturbation in these signals, i.e. U r = 0 . Therefore, (12.40) and (12.41) become
s = L s s + H Ur s H fds s s ,

(12.42)

578

Modal Induced Torque Coefficients

P Ur s = H PUr s H fds s s
= fds s

Ch. 12

(12.43)

Solving (12.42) for s and substituting this into (12.43) we find

1

H PUr s H fds s I H Ur s H fds s L s

P Ur s =

n n

(12.44)

= F s s

-----------------------AN ASIDE: As in the case of the centralized speed-PSS (12.30), the effective action of the transfer function matrix of the centralized FDS, H fds s in Figure 12.4, can be replaced by the equivalent n n speed-torque TFM F s of (12.44).
If the power system model used here represented a SMIB system, with a single FACTS device in
the system, evaluating F s would be relatively easy; F s would represent the transfer function between the generators speed and electrical torque, due to the action of the FDS. Therefore,
for this case, a FDS could be designed based on damping torque concepts as for PSSs. That is,
based on (12.44), if the FDS transfer function were designed such that
P Ur
1
------------- = H PUr s H fds s 1 H Ur s H fds s L s = k fds ,
(12.45)

then ideally a real damping torque coefficient, equal to kfds, would be induced on the

shaft of the generator. By solving (12.45) for H fds s we find the FDS transfer function
to be
k fds
H fds s = ---------------------------------------------------------------------- .
H PUr s L s + k fds H Ur s

(12.46)

This is equivalent to the design procedure, based on a frequency response approach, derived in
[8].

-----------------------In the case a multi-machine system evaluating the TFM F s in (12.44) is not a trivial task,
especially since H Ur s is not a diagonal matrix [9]. Therefore, no attempt will be made
here to evaluate F s ; moreover (12.46) cannot be extended to the multi-machine case in
an approach analogous to that for PSSs in Section 12.4.
In contrast to the tuning of PSSs, which is described in Chapters 5 and 10 and which is based
on frequency response methods, a different approach to the tuning of FDSs in a multi-machine system is adopted here. First an expression will be derived which relates the perturbation in the electrical torque induced on the shaft of generator i to the perturbations in the

Sec. 12.5

MITCs induced by a centralized FDS

579

signal j . Then, as in section Section 12.4, an expression will be derived for this component of torque at a selected mode of rotor oscillation, h .
Based on (12.43), the torque induced on generator i due to the action of the z FDSs is
P Ur_i s =

j=1

j=1

P Ur_ij s = fds_ij s j s ,

(12.47)

where, from (12.43), the elements of the speed-torque TFM fds s are given by
fds_ij =

H PUr_il s H fds_lj s .

(12.48)

l=1

Let us consider the jth element of the summation in (12.47), i.e.

P Ur_ij s = fds_ij s j s .

(12.49)

Equation (12.49) relates to torques induced on generator i by stabilizing signal perturbations

on FDS j. However, to calculate the synchronizing and damping torques induced on generator i we need to explicitly determine torques on generator i in quadrature and in-phase with
the speed perturbations on generator i, caused by stabilizing signal perturbations on all
FDSs.
At a selected rotor mode of oscillation, s = h , the torque of electromagnetic origin
P Ur_ij can also be defined in terms of the MITC T hFij , as in (12.33),
P Ur_ij h = T hFij i h .

(12.50)

Referring to (12.69) in Appendix 12I.1, the MITC is given by (12.8) becomes

T hFij

P Ur_ij
= ------------------- s
i

fds j
s = h

P Ur_ij
= ------------------- s
j

fds j
s = h

P Ur_ij
j
--------- h = ------------------s
i
j

fds j
s = h

cq* u h
*---------------(12.51)
u ih

where q = (n+j) and j is the jth element of the vector and the (n+j)th element of the
system output vector Y (see (12.4)).
Following substitution of the transfer function fds_ij h = P Ur_ij h j h from
(12.49) into (12.51) the MITC on machine i associated with the signal path through FDS j
becomes
T hFij = fds_ij h c q * u h u ih .
*

(12.52)

We can also define the total MITC on machine i, T hFij , associated with signal paths through
all z FDSs, i.e.

580

Modal Induced Torque Coefficients

hFi =

h
T Fij .

Ch. 12

(12.53)

j=1

Consequently, at the single modal frequency h , (12.49) can be written as

P Ur_i h = hFi i h .

(12.54)

Once again, an examination of (12.50) and (12.54) reveals that the real parts of both T hFij
and hFi represent damping torque coefficients on the ith generator.
In summary, for a centralized FDS whose elements H fds_lj h are known, the total MITC
for generator hFi can be calculated by successively evaluating (12.48), (12.52) and (12.53)
for the mode h .

12.6 General expressions for the torque coefficients induced by conventional, decentralized PSSs & FDSs
In Sections 12.4 and 12.5 centralized speed-PSSs and FDSs are assumed. However, for a decentralized PSS, the stabilizing signal on machine i is derived only from the speed signal of
machine i. Consequently, the TFM H pss s is a diagonal matrix, i.e. H pss_lj = 0, l j , and
thus (12.31) becomes
Pij s =

H PVr_il s H pss_lj s = H PVr_ij s H pss_jj s .

(12.55)

l=1

An expression for the modal torque coefficient induced on machine i due to speed perturbation, j , on machine j acting through PSS j is derived by substituting (12.55) into
(12.35), i.e.
v jh
T hPij = H PVr_ij h H pss_j h ------- ,
v ih

(12.56)

where H pss_jj h is simply written as H pss_j h . The total MITC for generator i, hPi , is
given by (12.37).
Similarly, decentralized, practical FDSs are designed for the various FACTS devices. That
is, for each FACTS device a local stabilizing signal is selected as an input to its FDS and thus
the TFM H fds s becomes a diagonal matrix. Consequently, (12.48) becomes
fds_ij s =

H PUr_il s H fds_lj s = H PUr_ij s H fds_jj s .

(12.57)

l=1

Substituting (12.57) into (12.52) we find the modal torque coefficient induced on machine i
due to perturbations in the local stabilizing signal of FDS j is

Sec. 12.6

MITCs induced by decentralized stabilizers

581

T hFij = H PUr_ij h H fds_j h c q * v h v ih ,

*

(12.58)

where H fds_j h is the jth element of the diagonal TFM H fds h .

Let us assume that the transfer function of any decentralized stabilizer is k j G j s . This is the
transfer function between the qth system output (which represents a local stabilizing signal)
and the jth system input (the summing junction of either a generator or FACTS device).
Based on (12.56) and (12.58), the following generalised form which applies to both PSSs and
FDSs is proposed for the MITC:
T hij = H ij h G j h c q * v h v ih k j .
*

(12.59)

Note that for PSSs H ij h = H PVr_lj h ; for FACTS devices H ij h = H PUr_ij h .

The transfer function H ij h is evaluated at the modal frequency h with all machine shaft
dynamics disabled. To calculate the TFMs H PVr h and H PUr h at a number of modal

frequencies, (12.18) and (12.19) are employed, e.g.

1

H PVr s = C Pq sI A qq B qVr ,

(12.18) repeated.

Following the calculation of the eigenvalues and eigenvectors of A qq the TFM is decomposed into a form which allows it to be evaluated easily at the required modal frequencies.
12.6.0.1

Relationship between: MITC T ii and PSS damping gain ki

Any appropriate stabilizing signal can be used for a PSS. Note that a speed-input PSS is a
special case because the speed of a machine is both a state and an output of the system.
Therefore if a speed-PSS is installed on the l th generator then the l th row of the output mac v

l h
- in (12.59)
trix of the state equations (12.9) is c l = 0 010 0 ; hence the term -------------

v ih
v ih
reduces to ------- = 1 . Furthermore, it is of interest to note from (12.59) that the MITC inv ih

duced on generator i due to its own PSS is

v ih
T hi = H PVr_ii h G i h ------- k i k i ,
v ih

(12.60)

since for a well-tuned PSS G i s 1 H PVr_ii s 1. That the MITC is equal to the nominal
damping gain of PSS i is consistent with the design objectives for PSSs discussed in
Chapter 5 and summarized in Section 5.14.

1. As in Chapter 5 this assumes that for the mode of interest s = h j f .

582

Modal Induced Torque Coefficients

Ch. 12

12.6.1 The total modal induced torque coefficients for systems with both PSSs and
FDSs
The total induced torque coefficient on generator i due to PSSs and FDSs is given in (12.37)
h

and (12.53), respectively. Because in both these equations the MITC T ij can be replaced by
the generalised expression of (12.59), the total induced modal torque coefficient for mode
h can be evaluated from the contribution of all n PSSs and z FDSs, i.e.
h

i =

n+z

h
T ij .

(12.61)

j=1

12.6.2 A relationship between modal induced torque coefficients and incremental

stabilizer gains
Let the transfer function of the jth stabilizer (PSS or FDS) be H stab_j s = k j G j s ; the
torque coefficient induced by this stabilizer on machine i is then given by (12.59). For a selected operating condition and stabilizer gain setting k j , the calculation of the eigenvalues
h and eigenvectors in this equation is based on the model of the system dynamics de-

scribed by (12.9). However, the relationship between T hij and the stabilizer transfer function
H stab_j s , given by (12.59), is based on this system with its shaft dynamics and rotor angle
feedback paths disabled. Into this disabled and therefore different system are injected the
mode h and right-eigenvector components v ih and v jh selected from eigenanalysis of the

original system. Therefore, in differentiating (12.59) with respect to k j neither h nor its eigenvector components are functions of k j in the disabled system. Hence, for small changes
in the stabilizer transfer functions,
T hij

dT hij
c q * v * h
------------------------------- H stab_j h = H ij h --------------- H stab_j h .
=
dH stab_j h
v ih

(12.62)

The increment H stab_j h is given by

H stab_j h
H stab_j h
H stab_j h = -------------------------------- h + -------------------------------- k j .
h
k j
H

(12.63)

stab_j h
stab_j h
- h ------------------------------- k j for PSSs. For FDSs the
As is shown in [1], the term -------------------------------

k j

stabilizer gain is chosen to satisfy a gain criterion which ensures that the term
H stab_j h
-------------------------------- h is negligible. Therefore ignoring the latter term, equation (12.63) reduces
h

to:
H stab_j h
H stab_j h = -------------------------------- k j ,
k j

Sec. 12.7

References

583

and, by substituting it into (12.62), we find

c q * v * h H stab_j h
c q * v * h
- -------------------------------- k j = H ij h --------------- G j h k j .
T hij = H ij h --------------
k j
v ih
v ih

(12.64)

This is a first-order Taylor series approximation, and is based on parameter perturbation

analysis, i.e. T hij and k j represent increments in the parameters T hij and k j , respectively.
h

An array of incremental MITCs, T = T hij , n n + z , is now defined. Its elements are

complex numbers. The n rows of the array represent the n generators on which the incremental MITCs are induced, while the n + z columns represent the stabilizers (the n PSSs and
z FDSs) on which the gain increments are made. The significance of the array can be explained as follows:

The element, T hii , is the incremental MITC on machine i due to the increment in the
nominal gain of its own PSS i, k i .

The jth element of row i, T hij , is the incremental MITC on machine i due to the increment k j on stabilizer j.

All elements T hij for which j n are due to PSSs and all those for which j n are
due to the FDSs.

For mode number h the summation of all elements in row i of the array yields the total inh

cremental MITC, i , on machine i due to the gain increments on all PSSs and FDSs in
the system It is derived from (12.61), i.e.
h

i =

n+z

h
T ij .

(12.65)

j=1

The practical application and significance of the arrays of incremental and total MITCs,
which are given by (12.64) and (12.65), are addressed in Chapter 13 and are illustrated with
several case studies.

12.7

References

[1]

P. Pourbeik, Design and Coordination of Stabilisers for Generators and FACTS devices
in Multimachine Power Systems, PhD Thesis, The University of Adelaide, Australia,
1997.

[2]

P. Pourbeik and M. J. Gibbard, Damping and synchronizing torques induced on

generators by FACTS stabilizers in multimachine power systems, Power Systems,
IEEE Transactions on, vol. 11, pp. 1920-1925, 1996.

584

Modal Induced Torque Coefficients

Ch. 12

[3]

M. J. Gibbard, Coordinated design of multimachine power system stabilisers based

on damping torque concepts, IEE Proceedings C Generation, Transmission and Distribution, vol. 135, pp. 276-284, 1988.

[4]

F. P. Demello and C. Concordia, Concepts of Synchronous Machine Stability as Affected by Excitation Control, Power Apparatus and Systems, IEEE Transactions on,
vol. PAS-88, pp. 316-329, 1969.

[5]

H. F. Wang, Selection of operating conditions for the coordinated setting of robust

fixed-parameter stabilisers, Generation, Transmission and Distribution, IEE Proceedings-, vol. 145, pp. 111-116, 1998.

[6]

P. M. Anderson and A. A. Fouad, Power System Control and Stability, IEEE Press,
1993.

[7]

R. L. Bolden, P. J. Wallace, and A. W. Grainger, Considerations in the improvement

of damping in the South East Australian interconnected system, CIGRE, paper 3105, 1982.

[8]

K. R. Padiyar and R. K. Varma, Damping torque analysis of static VAR system controllers, Power Systems, IEEE Transactions on, vol. 6, pp. 458-465, 1991.

[9]

J. M. Maciejowski, Multivariable feedback design, Electronic systems engineering series. Addison Wesley, 1989.

[10]

P. Pourbeik, M. J. Gibbard, and D. J. Vowles, Proof of the Equivalence of Residues

and Induced Torque Coefficients for Use in the Calculation of Eigenvalue Shifts,
Power Engineering Review, IEEE, vol. 22, pp. 58-60, 2002.

Appendices

585

Appendix 12I
App. 12I.1

Appendix: System response at a single modal frequency

The natural response of a linearized system is given by the differential equation

x t = Ax t . For a given vector of initial conditions x 0 the solution of this equation

for the vector of states is

N t
1 t 2 t
x t = V diag e e e W x 0 ,

(12.66)

where V , W are the right and left modal matrices of A , respectively. The condition for
exciting only one mode, e.g. mode h = h + j h , is x 0 = v h , where v * h is the hth
*

column vector of V , and is an arbitrary real constant, e.g. unity. When only mode h is
excited, (12.66) reduces to
x t = v h e
*

h t

.
h t

v qh
v qh e
x q t
- = -------------------= -------- .
Then the ratio of the q and k responses is --------------h t
v kh
x k t
v kh e
th

th

Thus in the case of the linearized power system model (12.9), the speed perturbation of machine i is related to the speed perturbation of machine j, when only one mode is excited, by
v jh
j h
-.
-------------------- = -----v ih
i h

(12.67)

Furthermore, since the vector of system outputs is given by y t = Cx t , then the ratio
of the qth to the kth outputs is
cq v h
y q h
* *------------------- = --------------, and
ck v h
y k h
* *

(12.68)

cq v h
y q h
* *-------------------,
= --------------v ih
i h

(12.69)

where c q * is the qth row vector of the C matrix. Thus, if only the hth mode, h , is excited
by setting x 0 = v h , at any instant of time the ratios of any two system output respons*
es is related by constant magnitude and phase. By mathematical induction it may be shown
that if x 0 is set to some linear combination of right-eigenvectors then all of the corresponding modes will be excited.

586

Modal Induced Torque Coefficients

App. 12I.2
and 12.4

Ch. 12

Reducing the TFM model in Figure 12.2 to those in 12.3

The multi-machine TFM power system model in Figure 12.2 can be reduced to that in
Figure 12.3 by closing the mechanical torque loop, i.e.
G m s = J s I H gov s J s

n n ,

where H gov s is a diagonal TFM representation of the governor/turbine dynamics and

J s is the diagonal TFM of machine shaft dynamics given by (12.25), i.e.
J s = diag 1 M i s + K di

N s = diag 0 s

n n .

Based on Figure 12.2, the following expressions may be written (Note: for convenience the
Laplace operator s is omitted):
P = H P N ,

P Ur = H PUr U r + U s ,

U s = H fds , and

= H N + H Vr H pss + V r + H Ur U r + U s .

The expressions for the TFM H P H , etc. are given by (12.18) to (12.23). Therefore
1

P + P Ur = H P N + H PUr U r + I H Ur H fds
.
H fds H N + H Vr H pss + V r + H Ur U r

(12.70)

Let the perturbation in the reference-set-point inputs at the summing junctions on generators and FACTS devices, V r and U r , respectively, be zero-vectors. Hence, (12.70) becomes

1
P + P Ur = H P N + H PUr I H Ur H fds H fds H N + H Vr H pss
.

= Z

Close loops #2 and #3 in Figure 12.2, around the shaft and governor /turbine dynamics,
G m s , by letting G P = G m I + ZG m

. Figure 12.2 thus reduces to Figure 12.3.

A similar analysis can be adopted to reduce Figure 12.2 to Figure 12.4, i.e.
P + P Vr = H P N + H PVr H pss
.
= Z F

and by letting

GF = Gm I + ZF Gm

, and L = H N + H Vr H pss .

App. 12I.3 Elements of the output matrix, C

587

App. 12I.3 Elements of the output matrix, C

The following is an important aside. In the above formulation of the TFM power system
model, the matrix partitions A and A q may appear to have been ignored, because neither of the two matrix partitions appear in equations (12.18) to (12.23). It is important to
show that these terms are not ignored because by doing so it will become apparent that certain partitions of the system matrices A and C are related. These relationships are then used
in a proof in [10]. To show that they have not been ignored, consider the following argument. From (12.20) and (12.21):
= A + A + A q =

K d1 M 1 1

+ A + A q q .

K d n M n 1n

(12.71)

Comparing this with (4.64) on page 115, and noting that governor/turbines have not been
modelled in (12.9) (i.e. P m = 0 ), it is clear that
d iag 1 M i P a = A + A q q

(12.72)

where P a is the accelerating power acting on the generator shaft, which is equal to the machine electric power output, P e , plus the stator copper losses. Neglecting copper losses
and since, based on (12.13) and (12.9), P e = C P + C Pq q it can be shown that
C P = d iag M i A , and

(12.73)

C Pq = d iag M i A q ;

(12.74)

the latter two equations can be substituted in (12.18) to (12.20) as required.

Chapter 13
Interactions between, and effectiveness of,
PSSs and FDSs in a multi-machine power system

13.1

Introduction

In this chapter the theoretical basis and a case study are used to illustrate the concepts of
interactions between, and effectiveness of, PSSs and FDSs in a multi-machine power system. The theoretical relationships between the incremental modal induced torque coefficients (MITCs), the associated mode shifts, and increments in stabilizer gains are outlined.
The case study will illustrate how the method developed for estimating rotor mode (eigenvalue) shifts can be used to assess the relative effectiveness of stabilizers and, thereby, gain
some important insights which form a basis for the coordination of stabilizers [1], [2], [3],
[4].
Techniques have been described in the literature for determining shifts in the modes of rotor
oscillation due to changes in stabilizer parameters [5], [6], [7], [8]. These techniques have
been used not only for determining optimal locations for PSSs and FACTS devices [6], [7],
[8] but also for tuning PSS parameters [9], [10].
In this chapter the theory and analysis is used to:

develop a new method, based on incremental MITCs, for estimating the mode or
eigenvalue shifts;

589

590

Interactions, effectiveness of PSSs &FDSs

Ch. 13

develop, for a given rotor mode, a method for estimating the contributions to damping of selected stabilizers for a selected increment in stabilizer gain, be they PSSs or
FDSs;

deduce the relative effectiveness of selected stabilizers in contributing to damping, say,

of an inter-area mode;

assess the effect of interactions

modes;

provide a basis for the systematic coordination of both PSSs and FDSs in multimachine systems [4].

between PSSs, particularly for inter-area and local

13.2 Relationship between rotor mode shifts and stabilizer gain

increments
Recall from Section 12.6 that the MTIC T hij is the modal (complex) torque coefficient for
the hth mode, h , this torque coefficient being induced on the shaft of the ith generator by
the jth stabilizer. It is established in (12.59) for both PSSs and FDSs that T hij is dependent
on the gain setting k j (in pu on device base) of stabilizer j. In Section 12.6.2 a relationship
is developed between the incremental MITCs and increments in stabilizer gains; this is then
employed in the following analysis to determine the eigenvalue shifts due to increments in
any or all of the gains of the n PSSs and z FDSs. This allows us to calculate for generator i
the change in T hij , T hij , caused by a change in stabilizer gain k j . An expression will now
h

be derived relating the shift in the hthmode of rotor oscillation, ij , due to the change T hij .
The gain increments k j may differ in magnitude and sign, i.e. they may be positive, negative or zero.
h

The proposed technique provides a direct relationship between the eigenvalue shift, ij ,
h

and an increment in gain k j on any stabilizer, be it a PSS or FDS. Note that ii is the
contribution to the mode shift by generator i due to an increment in gain on its own PSS;
this is, in essence, the objective of the PSS design methods discussed in Chapters 5 and 9.
h

The shift ij also represents the contribution to the shift of the mode h by generator i
due to a gain increment, k j , on some other stabilizer j. Depending on its sign, the shift
h

ij may enhance or degrade the damping of the mode h ; the mode shift ij therefore
1. Such interactions have been observed earlier [11]

Sec. 13.2

Relationship between mode shifts and stabilizer gains

591

represents the contribution of the interaction between stabilizer j and generator i to the shift
in the hth mode. The proposed method provides a basis for quantifying the effects of such
interactions for selected modes of rotor oscillation; this analysis is discussed in
h

Section 13.2.2. Moreover, by appropriately summing the mode shifts, ij , the contribution
made either by each generator or by each stabilizer to the total shift in a rotor mode can be
calculated for a set of stabilizer gain increments. This provides a basis for the coordination
of all stabilizer gains, both PSSs and FDSs [4].

Ps
Pm +
_

Ppss

K1
_

Pa
P3

/(sM)
k

o/s

Generator
model

Kd
Ideal PSS

Figure 13.1 A simple model of the generator with an ideal speed-PSS

Consider the simple generator model with an ideal PSS in Figure 13.1. The PSS damping
gain K d shown in this figure is a real damping torque coefficient because it induces on the
shaft of machine i a component of torque, P pss = K d i , which is in phase with the machines speed perturbation. Using a first-order Taylor series approximation, the change in
the eigenvalue for an incremental change in the torque coefficient K d , is
h
h
= --------- K d .
K d

(13.1)

Assume that n generators and z FACTS devices are fitted with stabilizers in a multi-machine
system; consider an ideal PSS on generator i. Though the torque coefficient, K di , by definition is real and the modal induced torque coefficient T hij on generator i by stabilizer j is
complex, conceptually they both have the similar effect on machine i. Thus, by using the
n+z

h
artifice of replacing K di in (13.1) by the total incremental MITC i = T hij due to
j=1

all n + z stabilizers, the variation in h due to the increments in gains of all stabilizer gains
is given by the total differential,

592

Interactions, effectiveness of PSSs &FDSs

n+z

h =

h
h
- T h1j + + ----------- T hnj
----------K
K

j = 1

d1

dn

h
h
h
h
= ------------ 1 + + ------------ n
K d1
K dn
n+z

Alternatively h =

j=1

Ch. 13

(13.2)

h
h
h
- T hij .
ij , where ij = ---------
K
i=1
di

(13.3)

If h is an eigenvalue of the system matrix A, and if w h and v h are respectively the associated left and right eigenvectors, it is shown in Section 3.11 that
h
A
---------- = w h ----------- v h .
K

K di
di

((3.51) repeated)

(13.4)

The system matrix A is given by (12.11). Differentiation of (12.11) with respect to K di yields
0

A
----------- = 1 M ,
i
K di

and thus (13.4) becomes

ih
w hi v ih
h
- = ------,
---------- = -------------K di
Mi
Mi

(13.5)

where ih = w hi v ih is the complex participation factor of the ith system state - namely the
rotor speed perturbation of the ith generator i - in the rotor mode h ; M i is twice the
inertia constant (H) of generator i. Substitution of (13.5) into (13.3) yields the expression for
estimating the contribution to the eigenvalue or mode shift by generator i due to an incremental change in the torque coefficient, T hij , on machine i, i.e.
hij = ih M i T hij .

(13.6)

Recall for mode h , that the incremental MITCs for generator i are related to the incremental gains on the n PSSs and the z FDSs by
((12.64) repeated)).
T hij = H ij h c j * v h v ih G j h k j ,
*
where j = 1 n n + 1 z , the gains being in pu on the base of the device.

(13.7)

Sec. 13.2

Relationship between mode shifts and stabilizer gains

593

Finally, by substitution of (13.7) in (13.6), a general expression is derived for the contribution to the mode shift by generator i due to an increment in gain on stabilizer j, namely
h

ij = ih M i H ij h c j * v h v ih G j h k j ,
*

(13.8)

where stabilizer j may be a PSS or a FDS. Note that for PSSs H ij s = H PVr_ij s ,
j = 1 n ; for FACTS device stabilizers, H ij s = H PUr_ij s , j = 1 + n z . The
transfer function H ij h is evaluated at the modal frequency h with all machine shaft dynamics disabled.

In addition to (13.8) three additional expressions will be employed in later sections. The first
is the contribution to the mode shift by all n generators as a result of an increment in gain
on any PSS j,
n

h stab_j = ih M i H ij h c j * v h v ih G j h k j .
*
i = 1

(13.9)

The second expression is the contribution by generator i to the mode shift caused by increments in the gains in some or all of the n PSSs and z FDSs:
h

gen_i = ih M i

n+z

H ij h c j * v * h v ih G j h k j .

(13.10)

j=1

Thirdly, the total contribution to the mode shift by all n generators as a result of increments
in the gain on some or all n+z stabilizers is:
h

n+z n

ih M i H ij h c j * v * h v ih G j h k j .

(13.11)

j =1 i = 1

13.2.1 Relationship between residues and MITCs in calculation of mode shifts

It is shown in Chapters 6 and 11 that, for mode h and generator i, the contribution to the
mode shift by generator i is related to an increment in gain on stabilizer j by:
h

ij = r ij G j h k j ,

(13.12)

where r ij is the residue of the transfer function from the input to the summing junction of
generator i (e.g. Vref_i) to the output used as the stabilizing signal (e.g. speed for a PSS, bus
frequency for a FDS).
h

In comparing (13.8), i.e. ij = ih M i H ij h c j * v

*h

v ih G j h k j with (13.12), it

is observed in [1], [12] that the residue is

h

r ij = ih M i H ij h c j * v h v ih .
*

(13.13)

594

Interactions, effectiveness of PSSs &FDSs

Ch. 13

The expression in (13.13) contains much more information than the form of the expression
in (13.12) which is the basis for the analysis in Chapters 6 and 11. The roles of the participation factor ih and the inertia constant Hi (= Mi/2) in (13.6), the MITCs (through (13.7)),
and the P-Vr or P-Ur characteristics in the residue of (13.13), are isolated. These components will help to explain certain characteristics, including interactions, derived in this chapter.
13.2.2 Concept of interactions
Based on (13.10) the concept and implications of interactions between stabilizers and of
interactions between generators will now be defined. Let us assume for simplicity that a
gain increment is made on the PSS fitted to generator i and then on other stabilizers,
j = 1 n n + 1 z j i . The contribution to the mode shift by generator i is
h

gen_i = ih M i H ii h G i h k i
ih M i

n+z

H ij h c j * v * h v ih G j h k j

(13.14)

j = 1i

The first term in (13.14), an alternative form of (13.10), is the contribution to the mode shift
by an increment k i in the gain of the PSS fitted to generator i. However, if both the stabilizer gains k i and k j are increased, it is apparent from the second term in (13.14) that the
gain increment k j can be considered to modify the effect of the gain increment k i on the
mode shift. The net effect on the mode shift depends on the resulting sign of the real part
of second term. If the net effect of the increment k j is to enhance the damping of the
mode h then there is a positive interaction of stabilizer j with PSS i. It is important to note
that the jth stabilizer can be either a PSS or FDS. Clearly, if the increment k j degrades the
damping of the mode h then the interaction between stabilizers is a negative. Let us call
all such interactions stabilizer interactions [3].
It is also insightful to consider a scenario in which the gain of the PSS fitted to the ith generator is unchanged (i.e. k i = 0 ). Then, the contribution to the mode shift by generator i
is due only to its interactions with the other stabilizers in the system, i.e.
h

gen_i = ih M i

n+z

H ij h c j * v * h v ih G j h k j .

j = 1i

(13.15)

In the event that gain increments are restricted only to FDSs the summation in (13.15) is
restricted to j = n+1 to n+z. This reveals that FDSs contribute indirectly to the shift in mode
h by their interactions with the generators.
It is at times more informative to assess the contributions to the mode shifts by the n generators due a gain increment on stabilizer i only. If the stabilizer is a PSS, these contributions
can be expressed by an alternative form of (13.9):

Sec. 13.2

Relationship between mode shifts and stabilizer gains

595

pss_i = ih M i H ii h G i h k i

.
jh M j H ji h c j * v * h v ih G i h k i

(13.16)

j = 1i

The above equation reveals that the contribution to damping of generator i due to its own
PSS (the first term in (13.16)) may be enhanced or degraded due to contributions from, or
interactions with, the remaining generators (the second term) through the network. Let us call
such interactions generator interactions [3].
Consider now the case in which a gain change is restricted to stabilizer FDS j only. The resulting contribution to the mode shift by an increment k j in the gain of FDS j is given by
(13.9).This result reveals that FDS j acts to shift mode h only by means of its interactions
through the network with each of the n generators.
In addition to the cases associated with the three equations, (13.9) to (13.11), a case of special
interest is the contribution by generator i to the mode shift due to an increment in the damping gain ki of its PSS (assumed ideal). By substitution of (12.64) in (13.6), this self-contribution is found to be:
h

ii ih M i k i .

(13.17)

This result 1 provides a type of benchmark for the contribution of an ideal PSS to damping.
Typically if a machine participates significantly in a mode (usually a local-area mode), the
speed-state participation is about 0.5 or less. The mode shift is then directly to the left in the
s-plane and is equal to k i 2M i or k i 4H i . (The latter result is consistent with that
which was derived based on an analysis of the block diagram for a SMIB system in
Section 5.4.) Clearly from (13.17), with low participation in the speed state, the extent of the
mode shift is reduced. However, a reduced contribution to the mode shift may also be attributed in part to the effect of interactions as explained above. This will be illustrated in
Section 13.3.
Note from (13.17) that the extent of the mode shift for low-inertia generating units is greater
than that of high inertia units of the same rating, all else being equal. Fitting PSSs to the former units are likely to more effective than to the latter.
13.2.3 Relationships between mode shifts, MITCs, participation factors and stabilizer gains
h

Eventually, our aims are: (i) to determine the mode shift ij in a selected mode h due to
gain increments on a single or on a number of PSSs and/or FDSs, and (ii) to assess the rel-

1. This result is also employed in Section 5.9.3.

596

Interactions, effectiveness of PSSs &FDSs

Ch. 13

ative effectiveness and contribution of individual stabilizers to the enhancement or degradation of modal damping. However, let us firstly review some of the background equations.
Consider generator i and its PSS. From (13.6) it has been shown that incremental mode shift
h

ii is related to the incremental MITC by:

hii = ih M i T hii ,
(13.18)
that is, through a complex factor which is the inertia weighted, speed participation factor,
ih 2H i . For mode h , it has been shown using (13.7) that the incremental MITC for

generator i is in turn related to the incremental gain on PSS i, i.e.

T hii = H PVr_ii h G i h k i or,

(13.19)

for a ideally-tuned PSS, T hii = k i .

(13.20)

The three equations, (13.6), (13.18) and (13.20) are of particular interest in the following discussions.
A case study is now used to illustrate some of the physical insights provided by the theoretical analysis. In particular, based on (13.8) to (13.17), the concept and effects of interactions
between stabilizers will be discussed. Furthermore, it will be demonstrated how the method
developed above for estimating eigenvalue shifts can be used to assess the relative effectiveness of stabilizers and thereby gain some important insights which form a basis for the coordination of PSSs and FDSs.

13.3 Case Study: Contributions to MITCs/Mode Shifts by PSSs and

generators
The purpose of this study is to demonstrate the insights that the incremental MITCs and the
associated mode shifts provide in the dynamic performance of a multi-machine power system. This study illustrates a basis for the tuning and coordination of PSSs, and of PSSs and
FDSs. Recall that the basic approaches to methods for the tuning of PSSs has been discussed in the earlier Chapters 5 and 9, and for FDSs in Chapter 11.
The single-line diagram of the fourteen generator power system employed in the case study
is shown in Figure 10.1. Only Case 1, which is a heavy-load operating condition - and which
has been the subject of studies for both PSSs and FDSs in earlier chapters - is now examined. Two modes, an inter-area mode M and the local-area mode B will be the initial focus
of the studies.
For illustrative purposes all PSS damping gains are set to 5 pu on machine base; the PSSrelated parameters are listed in Tables 10.5 to 10.9. The FACTS device is the SVC, BSVC_4,
located at bus 412 in Figure 10.1; the parameters for its bus-frequency FDS are provided in

Sec. 13.3

Contributions to MITCs/Mode Shifts by PSSs and generators

597

(11.12) of Section 11.4.1.1. Based on Figure 11.7 the gain of the FDS is set to 30 pu on device base, a value at the lower end of its potential gain range. The characteristics of the associated rotor modes are given in Table 13.1.
Table 13.1

Characteristics of the electromechanical modes, Case 1;

PSSs and FDS of BSVC_4, in service.

Mode
No.

Real

Imag

A
B
C
D
E
F
G
H
I
J
K
L
M

-0.68
-0.40
-0.42
-1.00
-0.65
-0.88
-0.84
-1.24
-0.64
-0.92
-0.18
-0.14
-0.42

10.47
9.66
9.06
8.73
8.32
8.27
7.80
8.09
7.83
7.48
3.93
2.56
2.04

0.065
0.041
0.047
0.114
0.078
0.105
0.107
0.151
0.081
0.123
0.046
0.056
0.201

Mode Behaviour

Mode Type

VPS_2<-->EPS_2, BPS_2
SPS_4<-->CPS_4, GPS_4, TPS_4
BPS_2<-->EPS_2, VPS_2, TPS_4
NPS_5<-->TPS_5
CPS_4, SPS_4<-->GPS_4, TPS_4,
MPS_2, LPS_3<-->HPS_1, EPS_2, VPS_2
HPS_1,MPS_2,<-->YPS_3, EPS_2, VPS_2
TPS_4<-->GPS_4, SPS_4, EPS_2
YPS_3, MPS_2<-->LPS_3, EPS_2,
PPS_5<-->TPS_5, NPS_5
Area 3 <--> Area 5, Area 2
Area 5, Area 4 <--> Area 2
Area 5, Area 3 <--> Area 4

Local-area

Local-area
Inter-area

Inter-area

Nominal gain settings. All PSSs: 5 pu damping gain on machine MVA rating
FDS of BSVC_4: 30 pu on the device base.
<--> means ... swings against ....
- damping ratio.
In Mode Behaviour, generators or areas are listed in descending order of their participation
factors.

In order to assess the effects of the incremental changes in the MITCs and associated mode
shifts due to changes in stabilizer gains, let us increase the damping gain of all PSSs by 1 pu
(20% of 5 pu) and the gain of the FDS by 0.9 pu (3% of 30 pu), such gains being in per unit
on the device bases. Although the increases in PSS gains are 1 pu, an examination of
Figure 10.26 reveals that the left shifts in the associated modes are close to being linearly
related to the PSS gain increments. Similarly, for the FDS the left shift in the mode is likewise related to FDS gain increments in the vicinity of the nominal gain setting of 30 pu
(Figure 11.7). As foreshadowed, with the PSSs and the FDS in operation, it is of interest to
analyse for Case 1 the MITCs and mode shifts for the inter-area mode M,
M = 0.42 j2.04 , and for the local mode B, B = 0.40 j9.66 ; the generators in Area
4 participate fairly significantly in both modes. (It is shown in Table 10.14 that without a
FDS in operation on BSVC_4 these modes are respectively M = 0.14 j1.98 and
B = 0.39 j9.65 .)

598

Interactions, effectiveness of PSSs &FDSs

Ch. 13

Let us examine the state participation factors, shown in Figure 13.2, for modes B and M.
Mode B

(a) Local-area mode B

Mode M

(b) Inter-area mode M

Figure 13.2 Case 1. Magnitudes of the state participation factors for local-area mode B
and inter-area mode M. The PSSs and the FDS on BSVC_4 are in service.
For the local mode B the speed and rotor angle states of generators in Area 4 dominate the
participation of the state in the mode. In the case of mode M, the inter-area mode, a controller state in the FDS of the SVC is dominant, followed by speed and rotor angle states of
generators in Areas 3, 4 and 5. The mode shape for mode M is similar to that shown in
Figure 10.29(a) when the FDS is out of service. It reveals that machines in Areas 5 and 3
swing against those in Area 4. However, it is the complex inertia-weighted, speed participation factors that are of interest in (13.6); these participation factors are shown in
Figure 13.3(a) and (b). In this figure it is noteworthy that:

because fewer machines participate in the local mode, the magnitudes of the participation factors in Figure 13.3(a) are much greater than those for the inter-area mode in
Figure 13.3(b). (Recall that the sum of the complex participation factors for mode h
is unity.) Larger values of the speed participation factors of local-area modes are typically a characteristic which differentiates them from inter-area modes.

in Figure 13.3(b) for the inter-area mode M, the dominant speed participation factors
of generators in Areas 3, 4 and 5 are - for most purposes - real (or nearly real), and
thus so is the factor ih 2H i in (13.6).

Sec. 13.3

Contributions to MITCs/Mode Shifts by PSSs and generators

599

Mode M

Mode B

5.6 deg

10.8 deg
10.6 deg
-9.5 deg
-1.4 deg
15.9 deg

-1.7 deg

-1.6 deg
18.6 deg

28.4 deg
26.0 deg
32.7 deg
-29.5 deg
-68.9 deg

(a) Local-area mode B

(b) Inter-area mode M

Figure 13.3 Case 1. Magnitude and angle of inertia-weighted, speed participation factors,
2H , , of generators for local mode B and inter-area mode M.
13.3.1 Contributions to the MITC of each generator, local mode B
Based on (13.7) the full n n + z array of incremental MITCs for the local mode B
( B = 0.40 j9.66 ) is shown in Table 13.2 on page 601. The full array contains a number
of features. (i) For the increments in gains of all PSSs (1 pu) and of the single FDS (0.9 pu)
the array of incremental torque coefficients have a direct bearing on the associated mode shifts
in Table 13.3 on page 603. (ii) Interesting aspects of the MITCs for the local mode B will be
compared with similar aspects for the MITCs of the inter-area mode M,
M = 0.42 j2.04 . Figures 13.2 and 13.3 reveal that the generators in Area 4 participate
fairly significantly in both modes B and M.
The rows of Table 13.2, which are on the MVA ratings of the generator nominated in the
left-hand column, show the contributions to the total incremental MITC of each generator
resulting from gain increments: (i) on individual stabilizers (a PSS or the FDS) or (ii) on all
PSSs and the FDS (that is, the row sum,
h

i =

n+z

h
T ij ).

j=1

(13.21)

Alternatively, a column in the table reveals the components of the complex incremental
MITCs induced on each generator due to a gain increment on a selected stabilizer. Note: if
the MITCs in the array were expressed on system base MVA then the sum of MITCs in each
column would yield the total MITC induced by an individual stabilizer.

600

Interactions, effectiveness of PSSs &FDSs

Ch. 13

A diagonal element, which is the incremental torque coefficient induced on generator i by

the PSS on generator i, is essentially real 1, i.e. it is a pure damping coefficient; the diagonal
element is referred to as the self contribution. It is also observed that for such elements the
incremental damping torque coefficients are - in most cases - close to unity; ideally, this result is predicted by the incremental form of (13.20),
T hii k i = 1 pu.

(13.22)

Strictly speaking, the above result applies to the operating condition for which the PSS of
the particular generator is tuned, usually at or close to rated power output (typically 0.7 to
0.9 pu on generator MVA rating, see Chapters 5 or 10). Assume for the encompassing range
of operating conditions the P-Vr Design Case may lie at 0.7 pu on generator MVA rating.
Typically, over the range of power outputs the band of P-Vr characteristics lie within 3 dB
and 10 of the desired P-Vr design characteristic. At lower levels of power output, less
than 0.7 pu, the low frequency gain of the associated P-Vr characteristic is less than that of
the P-Vr Design Case. Consequently, in Case 1 for generating stations on part load, such as
HPS_1, MPS_2 and PPS_5, the value of T hii in the table is less than 1 0 pu 2. Conversely, for generators operating at rated power output such as LPS_3 and NPS_5, the gain of the
associated P-Vr characteristic typically exceeds that of the P-Vr Design Case. As a result, the
value of T hii in the table for such machines exceeds 1 0 pu.
An examination of the Table 13.2 on page 601 provides an insight, for example, why a PSS
is not contributing to damping of mode(s) to the extent expected, or by what mechanism
are FDSs contributing to damping.
Consider generator SPS_4.
1.

From the row we note that if all stabilizers are incremented by 1 pu the incremental
damping component of the total incremental MITC for SPS_4 is enhanced from
that induced by its own PSS, 0.93, by 0.21 to 1.14 pu. According to (13.6) the damping of mode B would also be enhanced, however, the extent of improvement is
determined by the inertia-weighted, speed participation factor of SPS_4.

2.

A gain increment only on the PSS of CPS_4 (col. 9) increases the damping component of the MITC of SPS_4 by 0.13 pu on the latters MVA rating - for which there
will be an associated improvement in the damping of mode B as implied by (13.6).

1. This is the objective of the PSS design procedures discussed in Chapters 5 or 10.
2. See Section 9.4.1. It is shown that with decreasing real power output (P) at constant reactive power (Q) the scalar voltage v d0 decreases. Correspondingly the low frequency gain
of the P-Vr characteristic reduces. However, v d0 also decreases at constant P as Q becomes more lagging (see Table 9.7). Both these effects influence the P-Vr characteristics.

Sec. 13.3

Contributions to MITCs/Mode Shifts by PSSs and generators

Table 13.2

601

Case 1, Mode B. MITCs on each unit due to gain increments

on PSSs/FDS

B of each generator i listed in the

Contributions to incremental modal induced torque coefficients T ij
left-hand column due to an increment in the gain of the stabilizers j listed in the column headings (pu
on generator MVA rating)

Local-area mode B, B = 0.40 j9.66

Gener HPS_ BPS_2 EPS_2 VPS_ MPS_ LPS_3 YPS_3 TPS_4 CPS_4 SPS_4 GPS_ NPS_ TPS_5 PPS_5 BSVC
ator 1 PSS PSS
PSS 2 PSS 2 PSS PSS
PSS
PSS
PSS
PSS 4 PSS 5 PSS PSS
PSS 4 FDS
Col.

HPS_
1

0.72

0.27

0.01

0.23

0.04

0.01

-j0.07

j0.82

j0.31

j0.06

-j0.18

j0.01

j0

BPS_
2

11

12

13

14

15

Row
Sum

10

16

-0.44

0.01

-0.04

0.04

0.03

0.87

j0.07

-j0.42

j0.46

-j0.18

j0

j0

j0

-j0.05

j0.20

-0.01

0.99

0.04

0.07

0.02

0.35

0.18

-0.17

0.07

-0.01

1.52

j0.01

-j0.06

-j0.07

-j0.03

-j0.02

j0

j0

j0.39

j0.36

j0.43

j0.19

j0

j0

j0

j0

-j0.43

EPS_
2

0.02

0.23

1.03

0.04

0.01

-0.13

-0.18

0.19

-0.07

1.14

-j0.04

-j0.05

-j0.13

j0.07

j0

j0

j0

j0.25

-j0.19

j0.21

-j0.09

j0

j0

j0

-j0.02

j0.02

VPS_
2

-0.02

0.02

-0.01

0.99

-0.02

0.19

-0.26

0.29

-0.13

-0.02

1.03

-j0.02

-j0.71

j0.33

-j0.03

j0.02

j0

j0

j0.29

j0.10

-j0.08

j0.02

j0

j0

j0

j0.01

-j0.08

MPS_
2

0.14

1.36

-0.17

-0.17

0.87

-0.44

-0.86

0.91

-0.38

-0.03

1.22

-j0.18

-j1.75

j0.35

j0.11

-j0.03

j0

j0

j1.27

-j0.62

j0.78

-j0.36

j0

j0

j0

-j0.03

-j0.46

LPS_
3

-1.28

-1.90

0.58

-0.05

0.35

1.04

-0.04

0.40

0.81

-0.85

0.32

0.06

-0.56

j0.45

j0.29

j0.02

j0.24

j0.09

-j0.06

j0.03

-j1.03

j0.53

-j0.65

j0.31

j0

j0

j0

j0.08

j0.30

YPS_
3

0.56

-2.41

1.21

0.30

0.74

-0.50

1.03

-0.93

1.19

-1.34

0.56

0.21

0.61

-j0.16

j2.08

j0.45

j1.05

-j0.14

j0.90

-j0.26

-j1.07

-j0.72

j0.71

-j0.21

j0

j0

j0

-j0.01

j1.74

TPS_
4

0.01

1.03

-0.06

0.08

-0.03

0.17

1.18

j0

j0

j0

j0

j0

j0

j0

j0.05

-j0.59

j0.65

-j0.21

j0

j0

j0

-j0.06

-j0.16

CPS_
4

0.02

0.99

0.15

1.16

j0

j0

j0

j0

j0

j0

j0

-j0.03

-j0.04

-j0.08

j0.01

j0

j0

j0

-j0.01

-j0.15

SPS_
4

-0.01

0.13

0.93

0.08

1.14

j0

j0

j0

j0

j0

j0

j0

j0

-j0.04

-j0.03

-j0.04

j0

j0

j0

-j0.10

GPS_
4

0.09

-0.31

0.86

0.89

1.54

j0

j0

j0

j0

j0

j0

j0

-j0.01

j0.20

-j0.71

j0.05

j0

j0

j0

-j0.02

-j0.49

NPS_
5

0.92

-3.20

1.64

0.33

0.92

-1.10

-0.16

-0.47

0.75

-0.82

0.33

1.08

-0.18

-0.02

0.21

0.21

-j2.33

j1.03

-j0.11

j1.51

j0.10

j1.02

-j0.03

-j0.59

-j0.49

j0.50

-j0.12

-j0.19

j0.15

j0.04

j0

j0.50

TPS_
5

1.18

-4.38

2.37

0.74

1.35

-1.41

-0.23

-0.95

1.03

-1.14

0.48

-0.19

1.06

-0.04

0.31

0.18

-j3.46

j2.21

-j0.48

j2.18

-j0.04

j1.66

-j0.01

-j0.77

-j0.99

j1.02

-j0.30

j0.14

-j0.15

j0.04

-j0.04

j1.02

PPS_
5

10.22 -18.00

9.02

-0.38

4.76

-7.66

-0.81

-2.29

4.24

-4.63

1.76

-1.16

-1.81

0.87

1.12

-4.73

j12.35

j1.81

j8.97

j1.84

j3.80

-j0.42

-j3.09

-j2.30

j2.36

-j0.56

j0.51

j0.52

j0.05

j0.24

j2.27

j0.88

Col. 16, the Row Sum, is given by (13.21). All PSS damping gains=5 pu, increment 1 pu on machine ratings.
Gain of FDS=30 pu, increment 0.9 pu on SVC base.
The incremental MITC induced by a generators PSS is shaded in yellow.

Consider now generators PPS_5 and LPS_3.

602
3.

Interactions, effectiveness of PSSs &FDSs

Ch. 13

An increment in the PSS gain on BPS_2 (col. 2) or LPS_3 (col. 6) significantly

degrades the incremental MITCs on PPS_5, while increments in the gain of other
PSSs enhance or degrade the incremental MITCs relatively less. The net effect is a
significant degradation in the damping component of MITC of PPS_5 (in col. 16).
However, a reference to Figure 13.3 on page 599 reveals that the inertia-weighted,
speed participation factor of PPS_5 for mode B is exceedingly small so that the
effects of any stabilizer, including its own, on the associated mode shifts is negligible.
A similar comment applies to an increment in PSS gain on generator LPS_3.

Consider SVC BSVC_4.

4.

For an increment in the gain of the FDS on BSVC_4 the induced damping coefficients are small or negligible on those generators in Area 4 which have significant
inertia-weighted, speed participation factors. Therefore their contribution to damping of mode B by the FDS will be small; for other generators - for which the latter
participation factor is negligible - the contribution will also be negligible.

13.3.2 Contributions of the mode shifts of each generator to mode B damping

Let us now consider the full array of mode shift contributions for the local mode B,
B = 0.40 j9.66 , and compare its features with those highlighted for its associated array
of MITCs. The mode shift array is shown in Table 13.3 on page 603.
Further to the observations made for the MITCs of mode B, the following comments are
offered on the associated mode shifts.
5.

The contribution to the incremental mode shift of generator i is related to the MITC
generated by its own PSS by (13.6), hii = ih M i T hii . Bearing that in mind,
the earlier observations on the MITCs (numbered 1 to 4 concerning the implications
for the associated mode shifts) are confirmed by an examination of Table 13.3.

6.

Ideally, the diagonal elements of the MITC array are T ii = 1 pu. According to
(13.6) above, the incremental mode shift for the mode is ideally the value for generator i given by the inertia-weighted, speed participation factor of Figure 13.3. Accordingly, this mode shift for unity MITC for generator SPS_4 is ideally
2H = 59.57 1.69 units ( 1 unit = 1 10

). As shown in Table 13.2 on

h

page 601 the value of the incremental MITC is actually T ii = 0.9328 j0.0293 thus
the mode shift according to (13.6) is hii = 55.49 + j3.38 or 55.59 3.49 units;
the value in Table 13.3 is hii = 55.5 + j3.40 .

Sec. 13.3

Contributions to MITCs/Mode Shifts by PSSs and generators

Table 13.3

603

Case 1. Mode B: Mode Shifts on each unit due to gain increments

on PSSs & FDS.
B

Contributions to Mode Shift ij by each generator (row) and by each PSS or FDS (col.) (x 10

Local-area mode B = 0.40 j9.66

Gener HPS_ BPS_ EPS_ VPS_ MPS2 LPS_ YPS_ TPS_ CPS_ SPS_4 GPS_ NPS_ TPS_ PPS_
ator 1 PSS 2 PSS 2 PSS 2 PSS PSS 3 PSS 3 PSS 4 PSS 4 PSS PSS 4 PSS 5 PSS 5 PSS 5 PSS
1

10

11

12

13

14

Sum SVC4
PSSs FDS
15

16

Row
Sum
17

HPS_
1

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

BPS_
2

0.01

0.01

-0.01

j0 -j0.02

j0

j0

j0

j0

j0 -j0.01

j0

j0

j0

j0

j0

j0 -j0.03

EPS_
2

j0

j0 -j0.01

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0 -j0.01

VPS_
2

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

MPS
2

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

LPS_
3

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

YPS_
3

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

TPS_
4

-0.01

-1.04

-0.14

0.15

-0.04

-1.09

-0.18

-1.27

j0

j0

j0

j0

j0

j0

j0 -j0.41

j0.62 -j0.70

j0.23

j0

j0

j0 -j0.26

CPS_
4

0.01

-0.43

-20.6

0.03

-24.3

0.08

j0

j0

j0

j0

j0

j0

j0

j0.53 -j1.10

j1.37 -j0.29

j0

j0

j0

j0.52

j0.21

j0.73

SPS_

0.44

-7.61

-55.5

-4.57

-67.2

-0.19

-67.4

j0

j0

j0

j0

j0

j0

j0 -j0.01

j2.36

j3.40

j2.38

j0

j0

j0

j8.13 -j0.25

j7.88

-0.57

1.85

-5.21

-5.50

-9.43

0.03

-9.39

j0

j0.06 -j1.31

j4.54 -j0.14

j3.30

GPS_
4

-3.33

j0 -j0.03
0

j0 -j0.01

j0.01 -j0.25
-24.3

j0

j0

j0

j0

j0

j0

j0

j0

j0

j3.15

j0.14

NPS_
5

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

TPS_
5

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

PPS_
5

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0

0.01

-1.61

-26.5

-63.9

-10.1

-102.

-0.26

-102.

0 -j0.02 -j0.01

j0

j0

j0

j0

j0.16

j0.58

j8.62

j2.18

j0

j0

j0

j11.5

j0.11

j11.6

Col.
Sum

All PSS damping gains=5 pu, increment 1 pu on machine MVA rating.

Gain of FDS=30 pu, increment 0.9 pu on device base.
The incremental mode shift induced by the generators own PSS is shaded in yellow.
The box highlights the PSSs and generators which are the main contributors to the damping of mode B.

604

Interactions, effectiveness of PSSs &FDSs

Ch. 13

7.

From the columns in Table 13.3 we can assess the contribution to the shift of mode
B due to a gain increment on any PSS or the FDS, or a selected group of stabilizers.
For instance, a 1 pu increment in the gain of the SPS_4 PSS only (col. 10) causes a
real mode shift of -55.5 units to be induced on SPS_4. However, due to an interaction between the PSS of SPS_4 and the other PSSs in Area 4, a real mode shift contribution is also induced on each of the other generators, i.e. -5.2, -3.3 and 0.15 units
on GPS_4, CPS_4 and TPS_4, respectively. Therefore, due to the increment in the
SPS_4 PSS gain, the units in Area 4 contribute a real mode shift of -64 units.

8.

The total mode shift comprising the sum of the real components of all the diagonal
(self) terms in the table is -82.6 units. Including the mode shifts induced by all PSS
interactions, the real part of total mode shift due to PSSs is -102 units (col. 15).
Interactions have thus enhanced the damping of the local mode B.

9.

From the inertia-weighted, speed participation factors of Figure 13.3 it is observed

that the generators SPS_4, CPS_4 and GPS_4, in that order, are the dominant participants in mode B; all other generators participate in small (e.g. TPS_4) or negligible amounts. It is therefore not surprising that, for a 1 pu increment in all PSS gains,
the three dominant participants contribute a real component of -101 units out of a
total real contribution of -102 units (col. 15) to the enhancement of the damping of
mode B.

10.

In assessing the effectiveness of PSSs on the damping of mode B note that, according to the column sum, a 1 pu increment in the gain of the PSS on SPS_4 is about
2.5 times and 6 times more effective than a similar change on the PSSs of CPS_4 and
GPS_4, respectively.

13.3.3 Contributions to the MITCs of each generator, inter-area mode M

Let us now assess the full array of the contributions to the MITC of each generator for the
inter-area mode M, M = 0.42 j2.04 . The array for 1 pu and 0.9 pu gain increments on
the PSSs and the FDS, respectively, is shown in Table 13.4 on page 605.
From Table 13.3 on page 599 for the inertia-weighted, speed participation factors it is noted
that generators in Areas 4, 5 and 3 are the dominant participants in the inter-area mode M.
Let us examine the MITCs for units within the latter Areas, initially ignoring the contributions from the FDS, and remembering that for each row the MVA rating is the base quantity
of the associated generator.

Sec. 13.3

Contributions to MITCs/Mode Shifts by PSSs and generators

Table 13.4

605

Case 1, Mode M. MITCs on each unit due to gain increments

on PSSs/FDS
M

Contributions to incremental modal induced torque coefficients T ij of each generator i due to an

increment in the gain of the stabilizers listed in the column headings listed in the left-hand column (pu
on generator MVA rating)
Inter-area mode M, M = 0.42 j2.04
Gener HPS_ BPS_2 EPS_2 VPS_ MPS_ LPS_3 YPS_3 TPS_4 CPS_4 SPS_4 GPS_ NPS_ TPS_5 PPS_5 BSVC
ator 1 PSS PSS
PSS 2 PSS 2 PSS PSS
PSS
PSS
PSS
PSS 4 PSS 5 PSS PSS
PSS 4 FDS
cols

HPS_
1

10

11

12

13

14

15

Row
Sum
16

0.74

-0.04

-0.02

-0.08

-0.33

-0.02

0.02

-0.01

0.26

j0

j0.01

j0.01

-j0.01

j0.01

-j0.03

j0

j0.00

j0

j0

j0

j0

j0

j0

j0.01

-j0.01

BPS_
2

0.21

0.90

-0.20

-0.11

-0.30

0.01

0.07

0.01

0.01

0.01

-0.04

0.56

j0.07

j0.10

j0.03

-j0.03

-j0.08

j0.04

j0.01

j0.10

j0.01

j0.01

j0.02

j0

j0

j0

j0.04

j0.32

EPS_
2

0.13

-0.20

1.00

-0.10

-0.08

-0.08

-0.01

0.03

0.01

-0.05

0.65

j0.21

j0.07

j0.07

j0.19

j0.12

j0.09

j0.01

j0.04

j0.01

j0.01

j0.01

j0

j0

j0

j0.01

j0.84

VPS_
2

0.31

-0.31

-0.31

1.00

-0.02

0.01

0.06

0.01

0.01

0.01

-0.07

0.68

j0.17

-j0.06

j0.01

j0.08

-j0.06

j0.09

j0.01

j0.05

j0.01

j0.01

j0.01

j0

j0

j0

j0.04

j0.37

MPS_
2

0.05

-0.18

-0.06

-0.01

0.77

-0.08

0.03

-0.02

0.50

j0.07

j0.02

j0.02

-j0.01

j0.09

-j0.02

j0

j0.02

j0

j0

j0

j0

j0

j0

j0.02

j0.22

LPS_
3
YPS_
3
TPS_
4

-0.01

-0.01

1.04

-0.12

-0.01

-0.02

-0.01

0.85

j0.02

j0

j0

j0

j0

j0.04

j0.06

j0

j0

j0

j0

j0

j0

j0

j0

j0.12

0.03

-0.01

-0.59

1.04

-0.01

-0.01

-0.01

0.46

j0.03

j0

j0

j0

j0

j0.21

-j0.07

j0

j0

j0

j0

j0

j0

j0

j0

j0.18

-0.01

0.01

-0.06

0.71

-0.14

-0.11

-0.13

0.61

0.93

j0.01

-j0.01

j0

j0

j0

j0

j0

j0.14

j0.02

-j0.04

j0.04

j0

j0

j0

j0.07

j0.11

CPS_
4

-0.11

0.92

-0.30

-0.19

0.10

0.41

j0

j0

j0

j0

j0

j0

j0

j0.16

j0.08

j0.12

j0.16

j0

j0

j0

-j0.08

j0.45

SPS_
4

0.01

-0.14

0.90

-0.24

0.14

0.67

j0

j0

j0

j0

j0

j0

j0

-j0.01

-j0.01

j0.10

-j0.01

j0

j0

j0

-j0.09

-j0.02

GPS_
4

-0.05

-0.12

-0.33

0.77

0.11

0.37

j0

j0

j0

j0

j0

j0

j0

j0

-j0.01

-j0.03

j0.07

j0

j0

j0

-j0.09

-j0.07

NPS_
5

-0.05

-0.01

0.76

-0.61

-0.32

-0.24

j0

j0

j0

j0

j0

j0.03

j0

j0

j0

j0

j0

j0.06

j0.09

-j0.03

j0

j0.15

TPS_
5

-0.07

-0.01

-0.15

1.06

-0.17

0.65

j0

j0

j0

j0

j0

j0.03

j0

j0

j0

j0

j0

j0.02

j0.06

j0.02

j0

j0.12

PPS_
5

-0.07

-0.01

-0.20

-0.47

0.74

0.00

j0

j0

j0

j0

j0

j0.01

j0

j0

j0

j0

j0

-j0.05

-j0.05

j0.12

j0

j0.02

All PSS damping gains=5 pu, increment 1 pu on machine MVA rating..

Gain of FDS=30 pu, increment 0.9 pu on device base.
The incremental MITC induced by the generators PSS is shaded in yellow.

11.

In the case of mode M, let us assume that only the PSSs gains on each machine in
Area 4 are raised 1 pu on generator MVA rating; the remaining PSS gains remain
unchanged. The off-diagonal terms for Area 4 generation in the box in Table 13.4

606

Interactions, effectiveness of PSSs &FDSs

Ch. 13

are negligible or negative - and therefore the associated interactions between the
Area 4 units are destabilizing. As is evident in col. 16, the net effect for each Area 4
generator is that the resulting real contribution to the MITC of each is less than that
contributed by its own PSS. In the case of local mode B on the other hand, and with
the exception for units in Area 5, it was found that the off-diagonal terms tend to
enhance the contribution to the damping (see Table 13.2 on page 601.)
12.

The same comments apply to generators in Areas 3 and 5 when only the PSS gains
in the same Area are incremented.

13.

An increment in gain on the FDS mainly increases the MITCs on Area 4 generators
and consequently improves the damping of mode M.

14.

The relative MITCs of both the local-area mode B and the inter-area mode M are
demonstrated in the bar chart shown in Figure 13.4 on page 607. Note (i) the charts
n+z

h
reflects the Row Sum, i = T ij , in column 16 for each generator in Tables
j=1

13.2 and 13.4; (ii) the Row Sum includes the FDS contribution; (iii) due to interach

tions the value of i hovers about 1 pu for mode B, but is significantly less for
than 1 pu for mode M.
13.3.4 Contributions of the mode shifts of each generator to the Mode M damping
As was examined for mode B in Section 13.3.2, let us now consider the full array of mode
shift contributions for the inter-area mode M bearing in mind that mode shifts are directly
related to the MITCs through (13.6). The components of contributions to the mode shifts
are listed in Table 13.5 on page 608. From the latter table the following are noted.
15.

If, for exploratory purposes, it is desirable to increase the gains on all PSSs by 1 pu
and the FDS gain by 0.9 pu, the total shift in the real part of mode M is found from
3

the table to be -40.8 units (col. 17), where 1 unit = 10 . Under the same conditions,
the total shift in the real part of mode B is -102 units, a factor of 2.5 times that of
mode M. This result emphasizes the relatively poorer damping characteristics of the
inter-area mode compared to the local mode for the same increments in stabilizer
gains.
16.

Gain increments of 1 pu on the PSSs of generators in Area 2 provide a relatively

small net improvement in damping of mode M (see cols. 2 to 5, 15).

17.

If damping of Mode M is to be improved by increasing the gains on PSSs by varying

amounts, the PSSs which ought to be selected are revealed by an examination of the
columns of Table 13.5. It is evident that 1 pu increment in the gain of the LPS_3
PSS followed by GPS_4, SPS_4 and TPS_4 PSSs provide a net greater boost to the
damping of mode M (-8.4, -6.5, -4.1 and -4.0 units, respectively) (see the Col. Sum).
In comparison, for the local mode B, the same increment in gain only on the SPS_4,
CPS_4, and GPS_4 PSSs boosts the damping of mode B (-63.9, -26.5 and -10.1

Sec. 13.3

Contributions to MITCs/Mode Shifts by PSSs and generators

607

Generator

units, respectively). To boost the damping of mode M, it may be desirable to coordinate the increase the gain, say, of LPS_3 by 3 pu, GPS_4 by 3pu and SPS_4 by 2 pu.

HPS_1

HPS_1

BPS_2

BPS_2

EPS_2

EPS_2

VPS_2

VPS_2

MPS_2

MPS_2

LPS_3

LPS_3

YPS_3

YPS_3

TPS_4

TPS_4

CPS_4

CPS_4

SPS_4

SPS_4

GPS_4

GPS_4

NPS_5

NPS_5

TPS_5

TPS_5

PPS_5
1

See
PPS_5
Note

See
Note

0
1
Real part of MITC

0
1
Imag. part of MITC

Mode M

Mode B
h

Figure 13.4 Total incremental MITCs, i , on generators produced by increments of 1

pu on all PSSs and 0.9 pu on the FDS from gain settings of 5 pu on PSSs and 30 pu on the
FDS, inter-area & local modes M & B. Note: 20% of MITC for PPS_5 for mode B is shown
(actual is -4.73+j2.27 pu on machine MVA rating).
18.

The FDS provides a relatively significant boost to the total incremental damping of
mode M. In Table 13.5 the PSSs gain increments contribute shifts of -30.7 units and
the single FDS -10.1 units, i.e about 25% of the total mode shift of -40.8 units.

19.

In Section 13.3.1 it was noted that with the FDS on BSVC_4 out of and in operation
(gain: 0 and 30 pu) the real part of mode M was enhanced from -0.14 to -0.42 Np/s,
a change of -274 units. In Table 13.5, for gain increment of 1 pu on the FDS, the
total boost of the damping of mode M is -10.1 units. Decreasing its gain by 30 pu to
zero yields an estimate of the change in the real part of mode M of
30 10.1 = 303 units. For exploratory purposes, the magnitude of these
changes are close enough to suggest that the relationship between gain increments
in the FDS and shifts in mode M is reasonable linear in the vicinity of 30 pu. Moreover, if the effect of an increase in the FDS gain on the damping of mode M is being
investigated, with some confidence it can be assumed that the boost in the real part
of mode M, say a 10 pu gain change, is likely to be from -0.42 to about -0.52 Np/s.

608

Interactions, effectiveness of PSSs &FDSs

Table 13.5

Ch. 13

Case1, Mode M. Mode Shift contributions for gain increments on

PSSs/FDS
M

Contributions to Mode Shift ij by each generator (row) and by each PSS or FDS (col.) (x 10-3)
Inter-area mode M, M = 0.42 j2.04
Gener HPS_ BPS_ EPS_ VPS_ MPS2 LPS_ YPS_ TPS_ CPS_ SPS_4 GPS_ NPS_ TPS_ PPS_
ator 1 PSS 2 PSS 2 PSS 2 PSS PSS 3 PSS 3 PSS 4 PSS 4 PSS PSS 4 PSS 5 PSS 5 PSS 5 PSS
1

HPS
_1

-0.52

0.02

0.01

j0.46 -j0.03 -j0.02

BPS_ -0.07 -0.11 -0.03

2
j0.20 j0.90 -j0.20

0.06

0.26

10

13

14

15

16

17

-0.01

-0.17

j0.01

j0

j0

j0

j0

j0

j0

j0.19 -j0.01

j0.17

0.03

-0.10

-0.01

-0.01

-0.02

-0.28

-0.33

0.08

-0.04

-0.01

0
-0.04

-0.18

j0.11 -j0.30

j0.01

j0

j0.07

j0.01

j0.01

j0

j0

j0

j0

j0.60 -j0.04

j0.55

-0.19

-0.09

-0.01

-0.04

-0.01

-0.01

-0.01

-0.89

-0.90

j0

j0

j0

j0

j0

j0.65 -j0.05

j0.60

-0.01

-0.01

-0.26

-0.28

j0.12 -j0.21

j1.02 -j0.12 -j0.09 -j0.09 -j0.01

j0.03

j0.01

-0.13

0.01

-0.04

-0.01

j0.18 -j0.19 -j0.18

j0.62 -j0.01

MPS
_2

-0.15

0.01

0.03

j0.04 -j0.29 -j0.11

j0.02

LPS_ 0.05 0.02 0.02

3
-j0.18 -j0.03 -j0.02

0.01

-0.12

12

Row
Sum

0.01

VPS
_2

0.08

11

SVC4
FDS

j0.01 -j0.06 -j0.19 -j0.01

-0.23

0.06

-0.14

EPS
_2

YPS
_3

-0.05

Sum
PSSs

-0.13

-0.11
0.04

0.01

-0.06

-0.01

j0

j0

j0.03

j0.01

j0

j0

j0

j0

j0

j0.45 -j0.05

j0.40

0.08

-0.04

-0.01

-0.01

-0.01

-0.62

-0.64

j1.2 -j0.12 -j0.01

j0.03

j0.01

j0

j0

j0

j0

j0

0.08

1.20

-0.01

0.11

0.23

0.11

j1.37 -j0.82

j0.01

j0

j0

j0

j0.01 -j0.03

j0.01

j0.28 -j0.02

j0.26

0.01

0.01

-1.27

-1.27

-0.61

j0.01 -j0.06
0.01

-11.0

0.01

1.05

-2.27

-j0.03 -j0.01

j0

j0 -j0.01 -j1.24

TPS_ 0.09 -0.10 0.01 0.03 0.03

4
-j0.05 j0.10 -j0.02 -j0.04 -j0.04

0.02

j0 -j0.01 -j0.02 -j0.01

-0.02
-0.02

j0.70 -j0.04

j0.66

-9.12

-9.13

-0.01
0

j1.50

j0

j0

j0

0.04

-5.54

1.11

0.84

0.99

j0.18

j0

j0.18

-2.50

-4.73

-7.23

j0 -j0.23 -j0.44 -j0.68

j0.01

j0 -j0.92

j0.14

j0.28

j0.29

j0

j0

CPS
_4

0.02

-0.02

0.01

-6.68

2.50

1.74

-j0.02

j0.03

j0 -j0.01 -j0.01

j0

j0 -j0.97 -j2.48 -j0.29 -j0.80

j0

j0

j0 -j4.55

j0.37 -j4.19

SPS_
4

0.22

2.04

-12.8

3.41

-2.26

j0

j0

j0

j0

j0

j0

j0

j0.04

j0.53 -j4.04

j0.87

j0

j0

j0 -j2.61

j0.95 -j1.65

GPS
_4

0.02

-0.02

0.01

0.01

0.01

0.85

1.93

5.46

-12.5

-4.29

-2.15

-6.44

j0

j0.01

j0 -j0.01 -j0.01

j0

j0

j0.24

j0.56

j1.62 -j3.61

j0

j0

j0 -j1.19

j1.18

j0.01

NPS
_5

0.32

0.03

-3.09

2.87

1.30

1.42

1.42

j0.01

j0

j0

j0

j0

j0

j0

j0

j0

j0

j0 -j2.35

j1.31

j1.02

j0

j0

-j0

TPS_ -0.01
5
j0.02

0.58

0.05

1.09

-6.79

1.20

-3.87

-3.87

j0

j0

j0

j0

j0.06

j0.01

j0

j0

j0

j0

j0.43 -j4.17

j0.51 -j3.14

0.37

0.03

0.89

2.30

-3.54

j0.01

j0

j0

j0

j0

j0.13

j0.02

j0

j0

j0

j0

j0.78

j1.47 -j2.53 -j0.12

-1.03

-0.10

-0.10

-0.18

-0.40

-8.42

-0.95

-4.01

-1.65

-4.09

-6.46

-0.99

-1.37

j0.79

j0.27

j0.45

j0.34

j0.57 -j0.05

PPS_
5
Col.
Sum

1.15

j0.68 -j1.42

-0.91

-1.28
-7.59

0.01

-0.90

-2.18
-9.85

j0 -j3.14
0

0.06

j0 -j0.11

-30.7

-10.1

-40.8

j1.21 -j2.41 -j3.24 -j1.14 -j1.44 -j1.00 -j8.79

j1.84

j6.96

All PSS damping gains=5 pu, increment 1 pu on machine MVA rating.

Gain of FDS=30 pu, increment 0.9 pu on device base.
The incremental mode shift induced by the generators own PSS is shaded in yellow

20.

Let us compare, for a significant participant in each of the modes B and M, the real
parts of the diagonal element with that of the column sum for a selected PSS. For
example, for a 1 pu gain increment in the PSS of SPS_4, Table 13.3 on page 603

Sec. 13.4

Stabilizer damping contribution diagrams

609

(col. 10) and Table 13.5 (col. 10), reveal that interactions not only increase the leftshift of the local mode B from -55.5 to -63.9 units, but also reduce significantly the
left-shift of the inter-area mode M from -12.8 to -4.1 units.

13.4

Stabilizer damping contribution diagrams

In order to assess the relative effectiveness of the stabilizers concerned, the contributions to
the mode shifts of many PSSs and FDSs may need to be examined jointly and compared.
Much of the information inherent in the arrays of the contributions to mode shifts for the
local- and inter-area modes can be meaningfully displayed graphically in stabilizer damping
contribution diagrams (SDCDs). From a SDCD it is possible to assess the joint effects of
positive, zero or negative gain increments of differing magnitudes on individual stabilizers.
Such diagrams relate either to the column or row sums of mode shift contributions in the
arrays of Tables 13.3 or 13.5 and therefore take the effect of interactions into account. Earlier, in Section 13.2.2, the concepts of stabilizer interactions and generator interactions are introduced based on the associated forms of the equations (13.14) and (13.16) which relate the
complex mode shifts to incremental stabilizer gains.
The SDCDs are now examined for the two types of interactions for three inter-area modes
and one local mode; the variety of information provided by each SDCD is illuminating.
Consider firstly the mode shift on generator i resulting from an increment in its PSS gain;
the shift is derived from (13.16):
h

pss_i = ih M i H ii h G i h k i

jh M j H ji h c j * v * h v ih G i h k i

j = 1i

As mentioned in Section 13.2.2 the first term in the above equation shows the mode shift
on generator i resulting from a gain increment k i on its own PSS. The second term reveals
that, for the same PSS gain increment k i , there are also contributions to the mode shift of
generator i by the other n 1 generators. The relevant information is provided in the columns for each PSS in Tables 13.3 and 13.5. For example, consider col. 6 in Table 13.5 for
the mode M, M = 0.42 j2.04 . Listed in col. 6 are the component mode shifts resulting
from a gain increment of 1 pu on the PSS of generator LPS_3. The mode shift hii which
3

is displayed in row 6 of col. 6 is 11.0 j1.37 units (1 unit = 10 ).The mode shifts ij
which are associated with interactions with the other generators are listed in the remaining
h

n 1 elements in col. 6. The net mode shift, pss_i , is the sum of all the contributions in
col. 6, i.e. 8.42 j 0.05 units; in this case the effect of interactions is to degrade the damping

contribution of the PSS on generator LPS_3 by 2.6 units.

610

Interactions, effectiveness of PSSs &FDSs

Ch. 13

Secondly, consider the SDCD associated with a gain increment on each stabilizer as displayed in Figure 13.5 for the inter-area modes M ( M = 0.42 j2.04 ) and L
( L = 0.14 j2.56 ). As noted above for the inter-area mode M the net mode shift,
h

pss_i , for the increment in the damping gain of the LPS_3 PSS is 8.42 j 0.05 units; this

result is also shown in the figure. The main contributors (found from the column sums in
the last row of Table 13.5) to the damping of the inter-area mode M are not only the PSSs
in Areas 3 & 4 but also the FDS on BSVC_4. Note that in col. 16 of Table 13.5 for mode
M the contribution from the FDS to the mode shift is 10.1 + j1.84 units. This is the dominant contribution which results from the increment in the FDS gain and is clearly observed
in Figure 13.5.
Mode L

Mode M

Figure 13.5 Case 1. Stabilizer damping contribution diagram for inter-area modes L & M
for a damping gain increment of 1 pu on each PSS and 0.9 pu on the FDS.
Nominal gain settings: All PSSs 5pu, FDS 30 pu. Note scale.
As an aside, consider the contribution of the FDS derived from (13.8) in which the gain of
FDS j is incremented by k j . The resulting contribution of FDS j to the mode shift is:
n

h fds_j = ih M i H ij h c j * v h v ih G j h k j ,
*
i = 1

and demonstrates that the FDS interacts with all of n generators to produce a shift in the
mode.

Sec. 13.4

Stabilizer damping contribution diagrams

611

The SDCD in Figure 13.5 further reveals that the damping of mode L is effectively improved by increasing the gains of PSSs in Areas 2 and 5 as well as the gain of the FDS.
Mode B

Mode K

Figure 13.6 Case 1. Stabilizer damping contribution diagram for local- and inter-area
modes B & K for damping a gain increment of 1 pu on each PSS and of 0.9 pu on the FDS.
Nominal gain settings: All PSSs 5pu, FDS 30 pu. Note scale.
In Figure 13.6 is shown the SDCD associated with a gain increment on each stabilizer for
the local-area mode B, B = 0.40 j9.66 , and the inter-area mode K,
K = 0.18 j3.93 . For mode B the main contributors to the enhanced damping of this

local-area mode (found from the column sums in the final row of Table 13.3 on page 603)
are the PSSs in Area 4. Useful information revealed by the SDCD for the inter-area mode
K is that increasing the PSS gain on unit LPS_3 is an effective way of boosting the damping
of that mode. Incrementing the gain of the FDS, however, is ineffective in improving the
damping of both modes. Note that the SDCDs do not show the extent of the interactions
between PSSs which may enhance or degrade the damping of the mode; such effects are revealed in the off-diagonal elements in the columns of Tables 13.3 and 13.5.
Note that the relative amounts of the increments in individual stabilizer gains, whether they
be positive, zero or negative, depend on a number of factors. In particular, before gain increments are decided upon, the effects of gain changes on the pertinent local- and inter-area
modes need to be examined for an encompassing range of operating conditions, normal and
outage.

612

Interactions, effectiveness of PSSs &FDSs

Ch. 13

An alternative set of SDCDs, shown in Figures 13.7 and 13.8, is based on (13.14), namely:
h

gen_i = ih M i H ii h G i h k i
ih M i

n+z

H ij h c j * u * h u ih G j h k j .

j = 1i

The contributions to the mode shift by generator i resulting from a gain increment on each
of the n + z stabilizers are shown in the rows for that generator in Tables 13.3 and 13.5. For
each generator in these tables the net mode shift, h

gen_i

(the Row Sum in col. 17), is

displayed in the SDCD.

Figure 13.7 shows the contributions to damping of modes L and M by each generator when
all stabilizer gains are increased.
In Figure 13.7 it is of interest to note that generator PPS_5 makes a negligible contribution
to mode M although it is a dominant participant in the state participation factors in
Figure 13.2. This is seen to apply also to mode L in Figure 10.33 when all PSS gains are 5 pu
on machine base but the FDS is out of operation. From these observations it is concluded
that strong participation in a mode does not necessarily mean that the unit contributes to
the damping of that mode - as is evident in Figure 10.32. Examination of the row for the
generator PPS_5 in Table 13.4 on page 605 for mode M reveals that the real component of
the incremental MITC of 0.74 pu due to the PSS installed on that generator is cancelled by
negative interactions with the other stabilizers, notably TPS_5, NPS_5 and LPS_3. Consequently, even though the inertia-weighted participation factor of about 0.06 pu for PSS_5 in
Figure 13.3 is not insignificant the mode shift due to PPS_5 is negligible because the net incremental MITCs for the generator are negligible.

Sec. 13.4

Stabilizer damping contribution diagrams

Mode L

613

Mode M

Figure 13.7 Case 1. Contributions to mode shifts of modes L & M by each generator for a
damping gain increment of 1 pu on all PSSs and 0.9 pu on the FDS.
Nominal gains: All PSSs 5pu, FDS 30 pu. Note scale.
The contributions to damping of modes B and K by each generator in Figure 13.8 on
page 614 appear to be similar to those in Figure 13.6 for the contributions to damping of
modes B and K by each stabilizer. This not the case in Figures 13.5 and 13.7 for modes M
and L. The observation on modes B and K is valid for those units for which the interactions
with other stabilizers are small compared to the mode shifts induced on generators by their
own PSSs. Furthermore, note that the contribution by the FDS to the damping of modes B
and K is negligible; it is significant for modes M and L.
While Tables 13.3 and 13.5 and the SDCDs provide - for tuning or exploratory purposes similar useful information on the effectiveness of stabilizers for small increments in gain, the
question arises for what size gain increments is that information valid?. The following illustrates an approach for answering the question.

614

Interactions, effectiveness of PSSs &FDSs

Mode B

Ch. 13

Mode K

Figure 13.8 Case 1. Contributions to mode shifts of modes B and K by each generator for
a damping gain increment of 1 pu on all PSSs and a gain of 0.9 pu on the FDS.
Gains: All PSSs 5pu, FDS 30 pu. Note scale.

13.5 Comparison of the estimated and actual mode shifts for increments in stabilizer gain settings
In the studies of Section 13.3 the increments in PSS and FDS gains are set to 1 pu and 0.9
pu, respectively. A comparison of the estimated modal trajectories from the SDCDs and the
actual trajectories is therefore made (i) for the local-area mode B, and (ii) for the inter-area
mode M, for gain increments k of 1, 5 and 10 pu for all PSSs, and 0.9, 4.5 and 9.0 pu for
the FDS based on the nominal gain settings of 5 and 30 pu for the PSSs and the FDS, respectively. The estimated mode shifts, est , the associated estimated mode value, est ,
and the actual mode values, act , are shown in Table 13.6. The actual values are determined
by recalculating the eigenvalues with the nominal plus the incremented values of the stabilizer gain settings.

Sec. 13.5
Table 13.6

Comparison of estimated and actual mode shifts

615

Estimated changes from the SDCDs and actual changes in modes B & M
for increasing increments in PSS and FDS gains
Local-area mode B, B = 0.40 + j9.66

Gain increments

est

est

act

act est

PSSs

col. 1

col. 2

col. 3

col. 4

0.102 + j0.012

0.501 + j9.670

0.502 + j9.669

0.001 j 0.001

100%, 5pu 15%, 4.5 pu 0.512 + j0.058

0.911 + j9.716

0.918 + j9.705

0.007 j0.011

200%, 10pu 30%, 9.0 pu 1.023 + j0.116

1.422 + j9.774

1.451 + j9.731

FDS

20%, 1pu * 3%, 0.9 pu

0.029 j0.403

* Increments in stabilizer gain settings used for Table 13.3 and for the SDCDs
Nominal gain settings for PSSs and the FDS are 5 and 30 pu, respectively.
est = est + 0.40 + j9.66

Inter-area mode M, M = 0.42 + j2.04

Gain increments

est

est

act

act est
col. 4

PSSs

FDS

col. 1

col. 2

col. 3

20%, 1pu *

3%, 0.9pu

0.041 j 0.007

0.459 + j2.032

0.459 + j2.031

0 j 0.001

100%, 5pu 15%, 4.5pu 0.204 j 0.035

0.622 + j2.004

0.635 + j1.997

0.013 j 0.007

200%, 10pu 30%, 9.0pu

0.826 + j1.969

0.887 + j1.929

0.061 j0.040

0.408 j 0.070

* Increments in stabilizer gain settings used for Table 13.5 and for the SDCDs
Nominal gain settings for PSSs and the FDS are 5 and 30 pu, respectively.
est = est + 0.42 + j2.04

From Table 13.6 it is observed that both the real and imaginary parts of the actual and estimated mode values agree within 5% for gain increments up to five times the increments of
1 and 0.9 pu for the PSS and FDS gains. As the gain increments increase towards ten times
the latter gain increments, the actual and estimated mode values agree within 7.5% for mode
M; the agreement for mode B remains within 5%. The nature of the actual trajectory for
mode M observed in Table 13.6 is consistent with that observed in Figure 11.7, i.e. from values of the FDS gain above 35 pu the change in the mode value does not increase linearly
with change in gain.
As seen in the above tables and in figures such as Figure 10.26, the predominately left shift
of the modes with increasing stabilizer gain is the objective of the tuning procedures outlined in Chapters 5, 6 and 11. Given that the stabilizer transfer functions are of the form
k i G i s , where k i is a real gain (in pu on device base) and G i s is the compensation transfer

616

Interactions, effectiveness of PSSs &FDSs

Ch. 13

function, then ideally, (i) G i s ensures the left shift of all modes over a range of modal frequencies, and (ii) the gain k i determines the extent of the mode shift.
The basis for the tuning of PSS transfer functions k i G i s using the P-Vr method is explained and applied in Chapters 5, 9 and 10. Adopting this approach, and employing the appropriate SDCDs, suggest that the coordination of stabilizers can be achieved through the
coordination of their gains, k i . This is the basis of the analysis in Chapter 14.
The acceptable extent to which the actual and estimated modal trajectory diverge as the stabilizer gains are incremented depends on the users application and objectives. For example,
in a formal procedure for the coordination of gain settings for PSSs and FDSs it is necessary
to confirm that the gain increments selected do not result in the difference between the estimated and actual mode shifts exceeding acceptable limits over a range of operating conditions [4].

13.6 Summary
13.6.1 Interactions [3] 1
The analysis of interactions is based on PSS and FDS transfer functions being of the form
k i G i s . As stated, the transfer function G i s is tuned to effect a left-shift of the rotor
modes and the gain k i determines the extent of the shift. It is shown that, for an increment
in PSS damping gain k i the self-induced modal torque coefficient on generator i and the
associated
hii

self-contribution

to

the

shift

in

mode

are

T ii k i

and

= ih M i T hii , respectively. This is consistent with tuning techniques based on the

h
method, however, due to interactions from other generators, T ii and hii may be

P-Vr
enhanced or degraded. Hence:

The machines with higher inertia-weighted participation factors, M , are the more
effective contributors to damping.

For local modes, which typically have only a few machines participating, the magnitudes of the factors M for the dominant machines are significantly larger than
those for the inter-area modes, which may have numerous machines participating.
Thus the self-contributions to damping hii by dominant machines are likely to be
less for inter-area modes than for local modes.

1.

2000 IEEE. Reprinted with permission [3].

Sec. 13.6

Summary

617

For inter-area modes, the effect of interactions is to degrade further the already lower
self-damping contribution hii of generators. As is shown in the study, and observed
in practice, the damping of inter-area modes is generally poorer than local modes and
is more difficult to improve using PSSs. The damping of local modes maybe enhanced
by interactions (as observed in [13]).

The studies show that the damping of several inter-area modes can be enhanced significantly by fitting a FACTS device placed at a suitable location with a tuned FDS
[14]. The FDSs may have little effect on the damping of local-area modes, however,
this is location, system and operating condition dependent.

For the inter-area mode M the FDS in the study induces positive damping torques on
the generators and thus contribute to damping by each generator. In this case, the
interactions between the FDS and the PSSs are positive, i.e. an increment in FDS gain
enhances the self-damping resulting from an increment in PSS gain on generator i.

It can be shown that the term r ij = ih M i H ij h c j * v

*h

v ih in (13.13) is the

residue from the voltage reference to the speed output on machine j [12]. However,
associated information on interactions provided by (13.8), which incorporates the resh

idue r ij , is not available through the Method of Residues in Chapters 6 and 11.
13.6.2

Relative Effectiveness of Stabilizers [3]

The stabilizer damping contribution diagram is a simple, productive tool for displaying simultaneously the contributions to damping by some or all of the PSSs and FDSs.
Hence those stabilizers which make the most significant contributions to the damping
of rotor modes can be identified rapidly.

Such diagrams provide the engineering insight and basis for the simultaneous coordination of PSSs with PSSs, and PSSs with FDSs [4]. These aspects have been found to
be particularly valuable in practical applications and will be employed in Chapter 14.

For practical applications the increments in stabilizer gains can be applied to one,
some or all stabilizers simultaneously. This facility permits the coordination of all stabilizers in Area 2 only, say, to investigate the improvement in the damping of the relevant local- and inter-area modes.

The disadvantage of the diagram is that it applies to small increments in stabilizer

gains. However it has been found, for example, that mode shifts due to PSS gain
increments of 5 pu on machine base are accurate typically within 5%; such information needs to be confirmed for the power system under study.

618

Interactions, effectiveness of PSSs &FDSs

Ch. 13

13.7 References
[1]

P. Pourbeik, Design and Coordination of Stabilisers for Generators and FACTS devices
in Multimachine Power Systems, PhD Thesis, The University of Adelaide, Australia,
1997.

[2]

P. Pourbeik and M. J. Gibbard, Damping and synchronizing torques induced on

generators by FACTS stabilizers in multimachine power systems, Power Systems,
IEEE Transactions on, vol. 11, pp. 1920-1925, 1996.

[3]

M. J. Gibbard, D. J. Vowles, and P. Pourbeik, Interactions between, and effectiveness of, power system stabilizers and FACTS device stabilizers in multimachine systems, Power Systems, IEEE Transactions on, vol. 15, pp. 748-755, 2000.

[4]

P. Pourbeik and M. J. Gibbard, Simultaneous coordination of power system stabilizers and FACTS device stabilizers in a multimachine power system for enhancing
dynamic performance, Power Systems, IEEE Transactions on, vol. 13, pp. 473-479,
1998.

[5]

E. Z. Zhou, Functional sensitivity concept and its application to power system

damping analysis, Power Systems, IEEE Transactions on, vol. 9, pp. 518-524, 1994.

[6]

N. Martins and L. T. G. Lima, Determination of suitable locations for power system

stabilizers and static var compensators for damping electromechanical oscillations in
large scale power systems, Power Systems, IEEE Transactions on, vol. 5, pp. 14551469, 1990.

[7]

F. L. Pagola, I. J. Perez-Arriaga, and G. C. Verghese, On sensitivities, residues and

participations: applications to oscillatory stability analysis and control, Power Systems, IEEE Transactions on, vol. 4, pp. 278-285, 1989.

[8]

E. Z. Zhou, O. P. Malik, and G. S. Hope, Theory and Method for Selection of Power System Stabilizer Location, Power Engineering Review, IEEE, vol. 11, p. 45, 1991.

[9]

E. Zhou, O. P. Malik, and G. S. Hope, Design of stabilizer for a multimachine power

system based on the sensitivity of PSS effect, Energy Conversion, IEEE Transactions
on, vol. 7, pp. 606-613, 1992.

[10]

D. R. Ostojic, Stabilization of multimodal electromechanical oscillations by coordinated application of power system stabilizers, Power Systems, IEEE Transactions on,
vol. 6, pp. 1439-1445, 1991.

[11]

H. B. Gooi, E. F. Hill, M. A. Mobarak, D. H. Thorne, and T. H. Lee, Coordinated

Multi-Machine Stabilizer Settings Without Eigenvalue Drift, Power Apparatus and
Systems, IEEE Transactions on, vol. PAS-100, pp. 3879-3887, 1981.

[12]

P. Pourbeik, M. J. Gibbard, and D. J. Vowles, Proof of the Equivalence of Residues

and Induced Torque Coefficients for Use in the Calculation of Eigenvalue Shifts,
Power Engineering Review, IEEE, vol. 22, pp. 58-60, 2002.

Sec. 13.7

References

619

[13]

M. J. Gibbard, Coordinated design of multimachine power system stabilisers based

on damping torque concepts, IEE Proceedings C Generation, Transmission and Distribution, vol. 135, pp. 276-284, 1988.

[14]

S. Hiley, Redesign of the Blackwell Static Var Compensator Power Oscillation Damper
Controller. Undergraduate Thesis Report, School of Information Technology and
Electrical Engineering, The University of Queensland, Australia, 2007.

Chapter 14
Coordination of PSSs and FDSs using
Heuristic and Linear Programming Approaches

14.1

Introduction

Various techniques have been reported in the literature for the coordination of PSSs in multi-machine power systems [1], [2], [3], [4]. Some of these techniques have used linear programming solutions for coordinating PSS gains [5], [6]. However, little attention has been
given to the simultaneous coordination of PSSs and FDSs [7], [8], [9] [10]; this aspect is the
subject of this chapter. It must be emphasized that in the current context the term coordination is used to mean coordinating the gains of stabilizers installed on generators and
FACTS devices, say, in an area of interest for the purpose of improving the damping of rotor
modes. This is as opposed to coordination in the context of coordinating controllers, e.g.
AVR-PSS coordination, within a single generating unit [11]. In the following text the damping gains of PSSs and the gains of FDSs are collectively referred to as stabilizer gains.
It has been emphasized that the predominately left shift of the modes with increasing stabilizer gain is the objective of the design procedures outlined in Chapters 5 and 10 for PSSs
and Chapter 11 for FDSs. In essence, because the stabilizer transfer functions are of the
form kG s , where k is a real gain and the transfer function G s provides the phase compensation, then ideally, (i) G s ensures the left shift of all modes over the selected range of
modal frequencies, and (ii) the gain k determines the extent of the left-shift of the mode.
This basic approach to the tuning of stabilizers provides the following rationale for the
methods of heuristic and automated coordination.

621

622

Coordination of PSSs and FDSs

Ch. 14

In both the heuristic and automated coordination procedures the stabilizer gain and
the phase compensation are the two important components which are essentially
decoupled for practical purposes. Therefore, in the coordination procedures that follow, the stabilizer gains are the adjustable quantities and the parameters of the compensation transfer functions G s remain unchanged.

For the process of stabilizers coordination the PSSs and FDSs should be robust over
an encompassing range of operating conditions, normal and outage (see Section 1.2
item 3 and Section 11.8.2, respectively).

Ideally, the incremental left-shifts of the rotor modes should be more-or-less linearly
related to increments in stabilizer gain for small changes about the nominal values.
(See 10.26 for PSSs.) For FDSs certain rotor modes may be insensitive to changes in
stabilizer gain (see Figure 11.7).

The nature of the trajectories for the inter-area modes K, L and M are shown in
Figure 11.7 for BSVC_4 (for mode M), Figure 11.12 for PSVC_5 (modes L, M), and
in Figure 14.2 for SVC2 (modes K, L). At the lower values of gain the incremental
left-shift of the nominated modes increases linearly with the increments in gain.

In Chapter 13 a basis is developed for the heuristic coordination of PSSs and FDSs employing stabilizer damping contribution diagrams (SDCDs). Nevertheless, it is apparent from the
example in Section 13.4 that the approach presented in that chapter would require a series
of calculations to determine the appropriate stabilizer gain settings to enhance the damping
of each rotor mode; the approach can therefore be tedious for larger systems where certain
damping criteria are to be met for a large number of rotor modes. Moreover, in the approach
presented in Chapter 13 it is not clear how the following constraints can be satisfied: (i) limiting any right-shift in exciter/controller modes that may occur as a result of increasing stabilizer gains; (ii) constraining the shift in the frequency of rotor modes (i.e. the imaginary
part of their eigenvalues) which may result from increases in stabilizer gains. Since it has
been established that the coordination of the stabilizers can be achieved by coordinating
their gains, then an appropriate method for automating the gain selection procedure subject
to a series of constraints is to use linear programming [7], [8], or genetic algorithms [4], for
example. Nevertheless, it is informative to reveal the insights provided through heuristic coordination.
Based on the SDCDs an example of the analysis and process of heuristic coordination of
PSSs and FDSs over six operating conditions is presented in Section 14.3.
The method for heuristic coordination is extended in Section 14.4 to the application of linear programming (LP) for stabilizer coordination. Employing this approach the calculation
of the gain settings becomes automated, and thus less tedious. Furthermore, the gain settings given by the solution of the LP problem is an optimal set; however, it is not guaranteed
that the solution is unique [12].

Sec. 14.2

The 14-generator power system

623

For both the heuristic and the automated approaches the two-stage coordination procedure
is adopted to determine the parameters of the PSS and FDS transfer functions k j G j s . In
stage one, the transfer functions G j s are designed to left-shift the relevant modes by providing the appropriate phase compensation (as explained in Chapters 5, 10 and 11). In the
heuristic approach in stage two, information on the sensitivity of the real part of a selected
mode to an increment in any stabilizer gain is derived from the SDCDs. The necessary increments in stabilizer gains can then be calculated to achieve a desired left-shift in the mode.
For the automated analysis in stage two, the stabilizer gains k j are determined by solving a
LP problem. The objective function of the LP problem is selected such that the weighted
sum of the stabilizer gains is minimised, subject to (i) satisfying a desired level of damping
for selected modes of rotor oscillation, (ii) constraining the right shift of the exciter or other
controller modes, and (iii) limiting the allowable change in the frequency of oscillation of the
rotor modes. The objective function is chosen because, for small system disturbances, low
stabilizer gains reduce not only the effect of limiting action on the output of the stabilizer,
the AVR and excitation systems, but also the reactive power swings on generators 1.
To illustrate the two methods of stabilizer coordination the studies are based on the multimachine power system used in the studies in Chapters 10, 11, and 13.

14.2

The 14-generator power system

The 14-generation system described in Section 10.2 again serves as an example to illustrate
the procedures for heuristic and automated coordination. The parameters of the PSSs are
provided in Tables 10.5 to 10.9; the transfer functions for the bus-frequency FDSs on
BSVC_4 and PSVC_5 on buses 412 and 507, respectively, are given in equations (11.12) and
(11.13).
Because separate studies have shown the need for a SVC and stabilizer to provide voltage
control and to assist in providing damping for mode L (frequency ~2.6 rad/s), a SVC (base
MVA is 200 Mvar) is installed at bus 212, a major load bus, located in the vicinity of other
load buses in Area 2. It is therefore necessary (i) to install in the power flow analysis a SVC
on the selected bus, (ii) to provide voltage regulation at bus 212, and (iii) to include voltage
droop. The procedures developed in Chapter 11 will then be employed to the evaluate the
parameters of its bus-frequency stabilizer. In the studies in Chapter 10 the SVCs, BSVC_4
and PSVC_5, are on-line under closed-loop voltage control but with their FDSs out of service. For Cases 1 to 6 the associated voltages, real and reactive power flows, and other variables in the steady-state power flows are provided in Table 10.2, together with the rotor
modes in Tables 10.11, 10.15 and 10.16. Note that these quantities will be slightly modified
with the addition of the SVC, called SVC_2. For Case 1 with SVC_2 in service under closed1. Reactive power swings on generators can occur with certain types of PSS (see
Section 8.3.1)

624

Coordination of PSSs and FDSs

Ch. 14

loop voltage control the inter-area modes are K 1.05 j3.68 , L 0.37 j2.45 , and
M 0.51 j1.80 (all FDSs out of service).
The small-signal model of the controller for SVC_2 is that shown in Figure 10.38. Its parameters are: 200 MVAr (Mbase), KS = 2.0, KA = 1000, K d = 0.005 pu on Sbase
(100 MVA) and T d = 0.005 s The parameters of its bus-frequency FDS for SVC_2 are to
be determined in the following section.
14.2.1 Evaluation of the transfer function for the SVC at bus 212.
The polar plot of the residues for modes K, L and M is shown in Figure 14.1 for the SVC
transfer function Frq h V ref h for the range of operating conditions, Cases 1 to 6.
The damping gains of PSSs are all set to 20 pu or all to 10 pu; no other FDSs are in operation. The selection of representative tuning parameters is weighted towards mode L; the residues for mode M are negligible.

.
Figure 14.1 Polar plot of the residues for the transfer function Frq V ref for SVC_2,
modes K, L and M and six operating conditions. All PSS damping gains set to 20 pu on machine base. For L the values of the modes are in the vicinity of 0.45 j2.6 .
Note: the magnitude scale is to be multiplied by 0.1.
The calculation of the FDS transfer function is based on the representative values, shown
in Table 14.1, with the PSS gains of 20 pu rather than 10 pu, because (i) after coordination
of PSSs and FDSs the PSS damping gains are expected to be greater than 10 pu and in the
vicinity of 20 pu; (ii) based on previous studies, the damping contributions of the FDS at the
lower PSS gains are unlikely to be reduced markedly.

Sec. 14.2

The 14-generator power system

Table 14.1

PSS
Gains
(pu)
20
10

625

Representative values for evaluation of compensation transfer function,

mode L
Mode &
Phase spread
(deg)

Representative phase
angle (deg)

Compensation angle
(deg)

Maximum
residue

L: 20 *
-136
-44
0.008
L: 17 *
-152
-28
0.008
* Note: The residues for mode M are small.

Representative mode
0.45 j2.6

Based on the algorithm in Appendix 6I.1, the FDS transfer function for SVC_2 when all
PSS damping gains are set to 20 pu is:
6s
F s = k fds --------------1 + 6s

1 + 0.157s
------------------------1 + 0.628s

1
------------------------- , k Rm = 781 pu;
1 + 0.033s

(14.1)

the same washout and low-pass filter time constants are employed as for the FDSs of
BSVC_4 and PSVC_5 to cover the range of the rotor modal frequencies.

Imaginary Part (rad/s)

5
4.5

Modes:

kkkkkkkkkkk

KKKKKKKKKKK

K L, Case 01,
PSS Gains 20 pu
kkkkkkkkk
kk

4
KKKKKKK
KKKK

k l, Case 01,
PSS Gains 10 pu

lllllllllll
LLLLLLLLLLL

3.5

K L, Case 04,
PSS Gains 20 pu

3
2.5

100pu

2
1.5

Figure 14.2

l l lllllllll
LLLLLLLLLLL

Kfds:

k l, Case 04,
PSS Gains 10 pu

1
0.5
Real Part (Np/s)

SVC_2. Cases 1 and 4. Trajectories of modes K and L as the stabilizer gain,

kfds, is increased from zero (shown by an arrow) to 100 pu in steps of 10 pu on the SVC base;

changes in Mode M are negligible. PSS damping gains are set to 20 and 10 pu.
To establish the effectiveness of the FDS on SVC_2 in improving the damping of the interarea modes K and L, the modes are tracked in Figure 14.2 as the FDS gain is increased from
zero to 100pu. The trajectories of the modes are illustrated for both Cases 1 and 4 (heavy
and light-load conditions), and for all PSS gains set to 20 or 10pu. Over the gain range the
modes shift more-or-less linearly with increments in FDS gain; all other modes are stable.

626

Coordination of PSSs and FDSs

Ch. 14

14.3 A Heuristic Coordination Approach

14.3.1 Coordination of stabilizers for damping the inter-area modes
The object of the coordination procedure is to satisfy the criteria for the damping and other
small-signal requirements of the system. Such criteria should include both normal and the
relevant outage conditions; for simplicity, the latter conditions are excluded from the analysis but are notionally accounted for in the formulation of performance criteria.
The dynamic performance criteria in the following procedure are (i) the damping ratio of the
inter-area modes is to be greater than 0.2 (20%), (ii) the PSS and FDS gains are be held at
low values for reasons explained earlier; e.g. initially set all PSS damping gains to 5 or 10
pu. In the following study the damping criterion has been chosen to be somewhat high so
that the system is likely to be small-signal stable for the outage of a major transmission element.
As has been emphasized, there is no unique method for the heuristic coordination of the
gain settings of those PSSs and FDSs units which are selected for the purpose. However, to
reveal the insights that can be derived, the following procedure is adopted.
For the FDSs that participate in the coordination of stabilizers the characteristics of their
eigen-trajectories should be noted, namely;

In Figure 11.7 for BSVC_4 the shift in mode M is more-or-less to the left in a gain
range of 0 to 40 pu on device base; the left-shifts in modes K & L are negligible.

For PSVC_5 it is observed in Figure 11.12 that the shift in mode L is to the left over a
gain range 0 to 100 pu on device base. There is a less extensive left shift in mode K
over a gain range 0 to 50 pu on device base.

The left shifts in modes K and L for SVC_2 are seen in Figure 14.2 to increase linearly
with gain increments from 0 to 100 pu on device base. The shifts in mode M are negligible.

Consider the operating conditions, Cases 1 to 6. Initially, using Case 1, the stabilizer gains
will be adjusted to satisfy the damping criterion for the inter-area modes. Proceeding to Case
2 the inter-area modes will be evaluated using the same gains to establish if the criterion is
infringed; if so, the gains are appropriately adjusted. This process, covered by a set of steps,
is continued on in Cases 3 to 6 until an acceptable set of gains is found that cover all cases
- possibly after several iterations. The steps, and the inputs and outcomes of each step in the
process, are listed in Tables 14.2 and 14.3.

Sec. 14.3

Heuristic Coordination

627

In Step 1 for Case 1 the effects on the inter-area modes of selecting an initial or nominal set
of stabilizer gains are examined. These studies will form the basis for the subsequent analysis. Details of the analysis in Steps 1 to 5 for Case 1 are shown in Table 14.2.
Table 14.2

Steps in the coordination procedure, Case 1

Inter-area mode
--------------------------damping ratio

Gains: Nominal, PSSs & FDSs

---------------------------------------Incremental (pu)
Step
PSSs
1a
All
1b

Gain/ BSVC PSVC SVC

###
Inc.
4
5
2
10
0

0
0

0
0

0
0

10
0

20
0

20
0

20
0

Act.

Act.

0.50 j3.91

0.18 j2.55

0.25 j1.92

0.126

0.069

0.131

0.55 j3.94

0.41 j2.59

0.55 j1.98

0.14

0.16

0.27

All

10 / 0 20 / 2 20 / 2 20 / 2 Est.

0.55 j3.94

0.44 j2.59

0.58 j1.99

All

10 / 1 0 / 0

0 / 0 Est.

0.56 j3.89

0.20 j2.54

0.28 j1.91

10 / 0 20 / 2 20/12 20/14 Est.

0.58 j3.96

0.53 j2.61

0.63 j2.00

0.58 j3.95

0.53 j2.61

0.63 j2.00

0.15

0.20

0.30

4a
4b
5a

5b

All
LPS_3,
YPS_3
Other
PSSs
Nom.
10pu

0 /0

10

22

32

34

Act.

10 / 5

22

32

34

Est

0.82 j3.89 0.53 j2.61

0.67 j1.99

15

22

32

34

Act.

0.81 j3.88
0.21

0.67 j1.99
0.32

0.53 j2.60
0.20

Notes. ## Est: Estimated values of modes from SDCD.

Act: Actual values of modes from eigen-analysis

Step 1a, Case 1. It is assumed that the required PSS damping gains will exceed 10 pu. With
the FDSs on the SVCs off, the inter-area modes and their damping ratios are calculated; the
latter are less than 0.2
Step 1b, Case 1. It is assumed that the gains of the PSSs and the FDS will exceed, respectively,
the nominal values of 10 pu and 20 pu on device bases; the gains are set initially to these
values. However, the performance criterion ( 0.2 ) is not satisfied by the nominal gains
although there is a significant left-shift in modes L and M due to the action of the FDSs.
Step 2, Case 1. In order to determine the effectiveness of the FDSs their gains are incremented by 10% (2 pu). The stabilizer damping contribution diagram (SDCD) is shown in

628

Coordination of PSSs and FDSs

Ch. 14

Figure 14.3; it is clear from the figure that the FDSs have little influence on mode K. However, BSVC_4 contributes significantly to the damping of mode M.
Mode K

Mode L

Mode M

-0.0075
-0.0205

-0.0045

-0.0100
-0.0079

Figure 14.3 Step 2, Case 1. Contributions to the mode shifts of inter-area modes K, L &
M by each SVC for a gain increment of 2 pu on the bus-frequency FDSs.
Nominal gains: All PSSs 10 pu, FDSs 20 pu on device bases. Note scale.
Step 3, Case 1. Based on Step 1b the aim is to establish, with all FDSs out of service, (i) the
relative effectiveness of the PSSs in improving damping, (ii) which PSSs influence mode K
in particular when the gains of all PSSs are incremented by 10% (1 pu). The resulting SDCD
is given in Figure 14.4.
From the figure it is observed that an increase in the gain on the PSS of LPS_3 is most effective in improving the damping of mode K. PSSs in Area 2 also contribute marginally to
the damping of modes K and L. Likewise, PSSs in Areas 3 and 4 are of some minor benefit
to mode M. In comparison, in Figure 14.3 it is demonstrated that all three SVC FDSs improve the damping of mode L.
Step 4a, Case 1. In Step 1b of Table 14.2 it is noted that the damping of mode L is improved
significantly when the FDSs are on-line with all gains set to a nominal value of 20 pu. For
the damping ratio of mode L to increase to 0.2, stabilizers must shift the real part of mode
from 0.41 j2.59 by -0.12 Np/s to 0.53 j2.6 . Based on the mode shifts for a gain increment of 2 pu produced by FDSs in Figure 14.3, gain increments in PSVC_5, SVC_2 and
BSVC_4 of 12 pu, 14 pu and 2 pu, respectively, would yield a shift in mode L of -0.12 Np/
s (i.e. 12 2 pu 0.01 + 14 2 pu 0.0075 + 2 2 pu 0.0045 ). Likewise
modes K and M should benefit by real shifts of -0.034 and -0.075 Np/s, respectively. The

Sec. 14.3

Heuristic Coordination

629

estimated real shifts are confirmed in Figure 14.5; there are small positive shifts in the imaginary components.
Mode K

-0.0417

Mode L

Mode M

-0.0143
-0.0059

Figure 14.4 Step 3, Case 1. Contributions to the mode shifts of inter-area modes
K, L & M by each PSS for a gain increment of 1 pu.
Nominal gains: All PSSs 10 pu, FDSs 0 pu. Note scale.
Step 4b, Case 1. Eigen-analysis reveals (i) that there is close agreement with the estimated
mode values from Step 4a, and (ii) that the damping ratios of modes L and M are equal to
or better than 0.2; however, the damping ratio of K is 0.15 which does not satisfy the performance criterion.
Step 5a, Case 1. The aim now is to increase the damping ratio of mode K at Step 4b from
0.15 to 0.2. In Figure 14.4 it is noted that the PSS on LPS_3 is most effective in increasing
the shift of the latter mode. An increase of 5 pu in the damping gain of the PSSs on LPS_3
and YPS_3 could shift the mode from 0.58 j3.95 by -0.24 to 0.82 j3.95 , say, to yield a
damping ratio of 0.2 (i.e 5 pu 0.0417 + 5 pu 0.0059 = 0.24 ). It is shown that
the estimated inter-area modes in Step 5a agree closely with the actual values in Step 5b in
Table 14.2 and also satisfy the criterion on the damping ratios.

630

Coordination of PSSs and FDSs

Mode K

Mode L

Ch. 14

Mode M

Figure 14.5 Step 4a, Case 1. Contributions to the mode shifts of inter-area modes K, L &
M by SVC_2, BSVC_4 and PSVC_5 for gain increments of 14 pu, 2 pu and 12 pu on the
respective FDSs. Nominal gains: All PSSs 10 pu, FDSs 20 pu on device base. Note scale.
It is now necessary to check if the stabilizer gain settings at Step 5b satisfy the performance
criterion for Case 2. The coordination of the stabilizer gains for Cases 2 to 6 continues in
Table 14.3.
Table 14.3

Steps in the coordination procedure, Cases 2 to 6

Inter-area mode
--------------------------------damping ratio

Gains: Nominal
--------------------------Incremental (pu)

Step/
Case
PSS

S6/
Gens_3**
Case 2 Other PSSs
S7/
Gens_3
Case 2 Other PSSs
Gens_3
S8/
Gens_2
Case 2
Other PSSs

Gain
15
10

BSVC PSVC SVC

##
4
5
2
22

32

34

Act.

0.58 j3.75 0.55 j2.75 0.62 j1.84

0.15

0.19

0.32

15/1.5
22/2.2 32/3.2 34/3.4 Est 0.66 j3.74 0.60 j2.73 0.70 j1.83
10/1.0
15 / 4.5
10 / 5 22 / 0 32 / 0 34 / 6 Est. 0.80 j3.71 0.60 j2.70 0.68 j1.84
10 / 0

S9/
Case 2

As above

19.5
15, 10

22

32

40

Act.

S10/
Case
3-6

As above

19.5,
15, 10

22

32

40

Act.

0.80 j3.70 0.59 j2.70 0.68 j1.85

0.21

0.21

0.34

All damping ratios exceed 0.2

Sec. 14.3

Heuristic Coordination

631

Inter-area mode
--------------------------------damping ratio

Gains: Nominal
--------------------------Incremental (pu)

Step/
Case
PSS

Gain

BSVC PSVC SVC

##
4
5
2

Check
1.06 j3.79 0.58 j2.58 0.71 j1.96
S11/
As above
As above
0.27
0.22
0.34
Case1
Notes. ## Est: Estimated values of modes from SDCD. Act: Actual values of modes from eigen-analysis.
** Gens_A implies all generators in the area number A,

Step 6, Case 2. Eigen-analysis based on the stabilizer gains in Step 5b reveals that the damping
ratios of mode K and L are less than 0.2.
Step 7, Case 2. In order to ascertain the relative effectiveness of the stabilizers in this case, all
PSS and FDS gains are increased by 10%. The resulting contributions to the mode shifts of
the inter-area modes are revealed in Figure 14.6.
Mode K
-0.0053
-0.0036
-0.0029
-0.0062
-0.0178
-0.0281
-0.0069

Mode L

Mode M

-0.0145

-0.0325

-0.0138

Figure 14.6 Step 7, Case 2. Contributions to the mode shifts of inter-area modes K, L &
M by all stabilizers for gain increments of 10% in nominal gains. Note scale.

632

Coordination of PSSs and FDSs

Ch. 14

From Step 6, Case 2 it is noted that is necessary to shift mode K from 0.58 j3.75 by -0.21
Np/s to 0.79 j3.75 to yield a damping ratio of 0.2. It is observed from Figure 14.6 for
Step 7, Case 2, that the combined real shift in mode K is -0.0350 Np/s for a 10% gain increment (1.5 pu) on the PSSs of LPS_3 and YPS_3. Likewise, for a 10% increment (1.0 pu)
on the PSSs of Gens_2 1 the total real shift is -0.0180 Np/s. Furthermore, the gain increment of 3.4 pu on SVC_2 (nominal gain 34 pu) produces a real shift in mode K of -0.0178
Np/s. To achieve the desired mode shift in mode K let us increase the gain on the stabilizers
from the nominal values as follows:
Gens_3 by 3x1.5=4.5 pu from 15 pu. Real shift = 3x(-0.0350) =
-0.1050 Np/s
Gens_2 by 5x1.0=5.0 pu from 10 pu. Real shift = 5x(-0.0180) =
-0.0900
SVC_2 by 6.0 pu from 34 pu.
Real shift = (6/3.4)x(-0.0178) = -0.0314
Total real shift = -0.2264
The new stabilizer gain settings are:
Gens_3: 15+4.5 = 19.5 pu; Gens_2: 10+5 = 15 pu; SVC_2: 34+6 = 40 pu.
Unchanged: Other PSSs: 10 pu; BSVC_4: 22 pu; PSVC_5: 32 pu.
Although the resulting FDS gain increases on SVC_2 and PSVC_5 are comparable with the
original nominal gains of 20 pu for each FDS, the left shift of the inter-area modes are closely linearly related to the increments in gain over the gain ranges (see the modal trajectories
for these modes in Figures 11.7, 11.12 and 14.2).
Step 8, Case 2. Using the SDCD shown in Figure 14.7 the mode shifts and modes are estimated with the gain increments proposed in Step 7; note that the required mode shift of 0.2264 Np/s for mode K is achieved. The damping ratios of all inter-area modes in Case 2
now satisfy the dynamic performance criterion.
Step 9, Case 2. Eigen-analysis confirms the validity of the results of Step 8.
Step 10, Cases 3 to 6. Based on the stabilizer gains confirmed in Step 9 the damping performance of the inter-area modes is validated using eigen-analysis, i.e. in these Cases the criterion ( 0.2 ) is satisfied.
Step 11, Case 1. With the same gain settings of Step 9 the values of the modes and damping
ratios for Case 1 are recalculated and compared with those in Step 5b of Table 14.2. The
damping of the inter-area modes calculated in this Step is an improvement over that in Step
5b.

1. Gens_m covers all generators in Area m.

Sec. 14.3

Heuristic Coordination

Mode K

633

Mode L

Mode M

Figure 14.7 Step 8, Case 2. Contributions to the mode shifts of inter-area modes K, L &
M for the stabilizer gains increments given in Table 14.3. Note scale.
In comparing the eigenvalues K, L, M in Step 1a, Case 1, at the start of the procedure of
coordination, with those for the final stabilizer settings in Step 11, Case 1, it is observed:
(i) the real parts of eigenvalues have left-shifted by some 100 to 200% 1;
(ii) the imaginary parts have remained within a band of about 3%.
These observations are consistent with the rationale of the P-Vr approach to the tuning
PSSs and of the Method of Residues for tuning FDSs.
14.3.2 Coordination of local-area modes.
The effect of the increases in stabilizer gains on the local-area modes, which are well
damped, has so far been ignored. However, it may be the case that certain modes have been
degraded during the above procedure designed to ensure the damping criterion is satisfied
for the inter-area modes. Also of interest is which local-area modes are affected by the FDSs,
and the extent of the resulting mode shifts.
The SCDC in Figure 14.8 for local modes I, G and H is based on the nominal gain settings
for Step 9. Increments of 1 pu on all PSSs and 4 pu on all FDSs are assumed. The SDCD
demonstrates that increases in stabilizer gains enhance the damping of the three local modes.
In particular, a 4 pu increment in the FDS of BSVC_4 causes a significant shift in mode H,
a mode in which TPS_4 (a generating unit electrically relatively close to the SVC) swings
against the other units in Area 4. The FDSs have little effect on the seven remaining local
modes apart from BSVC_4 which slightly degrades mode E (not shown - it is the second of
1. With the higher gain settings on the FDSs, a possible extension of the Mvar range and
ratings of the SVCs may be required.

634

Coordination of PSSs and FDSs

Ch. 14

the three local-area modes in Area 4). If only the gain on FDS of BSVC_4 were increased,
it may occur that the damping on mode E is more severely degraded; this should be checked
if the mode is relatively poorly damped.
Mode I

Mode H

Mode G

Figure 14.8 Case 1. Contributions to the mode shifts of local-area modes I, H & G for
increments of 1 pu on all PSSs and 4 pu on all FDSs. Nominal gains: as for Step 9.
Note scale.

14.4 Simultaneous Coordination of PSSs and FDSs using Linear Programming

14.4.1 Introduction
In Section 14.3 a heuristic approach is outlined for the coordination of PSSs and FDSs.
Nevertheless, it is apparent from the case studies in Section 14.3 that the approach requires
a series of calculations to determine the appropriate stabilizer gain settings to enhance the
damping of the selected rotor modes. This approach may be tedious for larger systems
where the criterion for damping is to be satisfied for a large number of rotor modes. Moreover, in the heuristic approach it is not clear how the following constraints can be met:
(i) limiting any right-shift in exciter/controller modes that may occur as a result of increasing stabilizer gains;
(ii) constraining the shift in the frequency of rotor modes (i.e. the imaginary part of their eigenvalues) which may result from increases in stabilizer gains.

Sec. 14.4

Automated Coordination

635

Since it has been established that the coordination of the stabilizers can be achieved by coordinating their gains, then an appropriate method for automating the gain selection procedure subject to a series of constraints is to use linear programming. Thus, the heuristic
method for coordination is extended in the following by the use of linear programming (LP).
Employing this approach the calculation of the gain settings becomes automated, and thus
less tedious. Furthermore the gain settings given by the solution of the LP problem is an
optimal set; however, it is not guaranteed that it is a unique solution [12]. Moreover, a useful
feature in the LP solution is a set of limitations or constraints that can be applied to the solution, namely:

(i) a minimal level of damping for selected rotor modes;

(ii) a right-shift in the eigenvalues associated with generator exciters or other controllers, due to increasing stabilizer gains;

(iii) changes in the left-shift of selected rotor modes;

(iv) changes in the frequency of selected rotor modes.

Through a LP solution some unexpected insights into the support provided by certain stabilizers to the damping of nominated modes may be experienced.
A two-stage coordination algorithm is developed. Stage one is the same as for heuristic coordination. That is, for the transfer functions of the form k j G j s for the PSSs and FDSs,
the transfer functions G j s are designed to provide appropriate phase compensation as explained in Chapters 5, 10 and 11. However, in stage two the stabilizer gains k j are now determined by solving a LP problem. The objective function of the LP problem is selected
such that the weighted sum of the stabilizer gains is minimised, subject to any of the above
set of constraints. For practical reasons this function is chosen because, for small system disturbances, low stabilizer gains reduce not only the effects of limiting in the stabilizer and the
AVR and excitation systems, but also reduce swings on the reactive output of generators.
14.4.2 Comment on the LP solution: optimality versus uniqueness
The algorithm used to solve the LP problem is the revised two-phase simplex algorithm [12],
[13]. The process of solving a LP problem can be summarised as follows. The search begins
at an extreme point of the solution space called a basic feasible solution. The procedure then
determines if a shift to an adjacent point in the solution space can improve the objective
function. If so, the algorithm moves the solution to the point which offers the greatest improvement. This procedure continues until an optimal solution is found or it is determined
that the problem is unbounded or infeasible. Typically it is difficult to determine a basic feasible solution by observation. Therefore to start the procedure, a set of artificial variables are
introduced into the problem. This allows us to manipulate the problem such that zero becomes a basic feasible solution. However, if another basic feasible solution were found, and
the LP algorithm started at that point, the algorithm may converge to a different optimal

636

Coordination of PSSs and FDSs

Ch. 14

solution. The practical implication of this is that other optimal solutions may exist. Moreover, it is possible to establish whether more than one optimal solution exists. Nevertheless,
having determined an optimal solution it is not generally possible to ascertain what the other
optimal solutions are.
14.4.3 Coordination of PSSs and FDSs
As stated, the aim of both the PSS and FDS design methods is to achieve a left-shift in the
modes of rotor oscillation by inducing pure damping torques on the shafts of generators.
The aim of the coordination procedure is to determine the minimum required stabilizer
gains to achieve desired damping criteria for selected rotor modes and to prevent undesirable right shifts in other modes.
Because the stabilizer transfer functions cannot provide ideal compensation, and due to the
effect of PSS interactions, the shift in rotor modes will not be a pure left-shift. Consequently
there will be changes in the frequency of oscillation of the rotor modes. Any excessive such
changes will be constrained by the LP problem in order to limit undesirable changes in synchronising torques.
Assume that the transfer functions, G j s , have been tuned for the relevant set of the n generators and z FACTS devices using the methods described in Chapters 10 and 11. Let each
stabilizer be a fixed-parameter device of the form k j G j s . For a given vector of small gain
increments, k n + z 1 , the shift in selected modes of the system can be estimated based
on (13.9), or:
h = h1 h2 h n + z k , where
n

hj =

c u

j* * h
ih
- H ij h G j h -------------- .
-----M
u
i

i = 1

ih

(14.2)
(14.3)

Based on this equation a linear programming problem is formulated for the simultaneous
coordination of the (n+z) stabilizer gains.
Let the set of m modes of interest be
= 1 2 m ,

which may be a combination of rotor modes of oscillation and exciter/controller modes.

This set is a subset of the N system eigenvalues (the system being represented by a N N
state-matrix for which there are N eigenvalues). Without loss of generality let the desired
damping criteria be achieved by left-shifting the m modes of interest by
= 1 2 m .
Also, let the changes in the modal frequencies be bounded by

Sec. 14.4

Automated Coordination

637

= 1 2 m .

Assuming each stabilizer gain is weighted by a coefficient w j the LP problem can be stated:
n+z

minimise

w j k j , subject to:

(14.4)

j=1

k Np/s,

(14.5)

k rad/s,

(14.6)

K max k K min 0 ,

(14.7)

where = hj (m n + z . This problem is solved by using the two-phase simplex algorithm. Note that the LP problem becomes infeasible if:

The set of gain limits K max are too small for the required left-shifts Np/s.

The allowable modal frequency deviations rad/s are too small. This may imply
that the tuning of some stabilizers may be poor and hence result in excessive frequency shifts in the modes of interest.

Also note that the coefficients, w j , in (14.4) can be chosen to weight all stabilizer gain increments equally (with unity values) - or they may be chosen to bias the solution in favour
of the most effective stabilizers. Furthermore, it is important to note that the accuracy of
the estimated shift in an eigenvalue, given by (14.2), diminishes as the gain k j becomes
larger. This is because (14.2) is a linear approximation to the non-linear eigenvalue trajectory.
To reduce the error due to this linear approximation, the coordination procedure may be
carried out in a number of steps.
Let the total required left shift in the modes of interest be total Np/s. Then the LP
problem can be split into n s steps, with a shift of total n s required per step. The following algorithm is proposed for the coordination of PSSs and FDSs having transfer functions of the form k j G j s .
Two-stage Coordination Algorithm Using a LP Solution

Stage 1: Determining stabiliser transfer functions, G j s

Follow the tuning procedures that have been described in Chapters 10 and 11 for PSSs and
FDSs, respectively.
Stage 2: Determining stabilizer gains, k j :

638

Coordination of PSSs and FDSs

Ch. 14

1. Set the required vectors of the total shift in the real parts and the allowable total shift in
the imaginary parts of the modes of interest to total Np/s and total rad/s, respectively, in order to achieve a desired damping ratio (say) for the modes. The vector total
also contains the constraints on right-shifts in exciter/controller modes. Set K max and
K min .
2. Set the number of steps n s to two or more; the gain weighting vector w to unity (say);
the step counter i s to zero; the initial stabilizer gains to a vector of nominal values (e.g. 5
pu) or zero.
3. Calculate based on (14.3).
4. Form the LP problem given by (14.4)-(14.7) with
= total n s , = total n s ;
K max step = K max n s and K min step = K min n s .

Initialise the vector of gains k to zero and solve the LP problem.

5. If the LP problem is infeasible then:
(a) choose a larger value of K max step , and/or
(b) allow a slightly greater value of h for the mode h which is most tightly constrained.
(If the increase in modal frequency h is too large then certain stabilizers may need retuning, i.e. return to Stage 1).
6. Set the step counter i s = i s + 1 and increment the vector of gains k = k + k .
7. If i s n s then recalculate system eigenvalues/vectors with the new gain settings and go
to 3.
There is an opportunity to view the intermediate results after line 7, following the eigen-analysis. According to (14.2) the contribution to the shift in the hth mode by the jth stabilizer (a
PSS or FDS) is
h

j = hj k j .

(14.8)

This result is based on (14.2). At the end of each of the n s steps of the coordination algorithm (i.e. following item 7 above), the user may assess the effectiveness of the contributions
of each stabilizer and each generator to the damping of a selected mode. This and other
physical insights provided by this technique will be demonstrated in the case study in the
next section.

Sec. 14.5

Case study: Automated Coodination

639

14.5 Case study: Simultaneous coordination in a multi-machine

power system of PSSs and FDSs using linear programming
The purpose of this study is to demonstrate:
1.

the method of simultaneous coordination of selected PSSs and FDSs;

2.

the insights - some unexpected - revealed by the action of the step-by-step LP procedure of determining the stabilizer gains;

3.

the contributions of selected stabilizers to the damping of the rotor modes (and the
inter-area modes, in particular);

4.

the benefits and disadvantages of the automatic process of determining stabilizer

gain settings.

The system under study in this section is the fourteen generator system employed in
Section 14.2 in which three SVCs and their FDSs are in service, namely SVC_2 at bus 212,
BSVC_4 at bus 412, and PSVC_5 at bus 507.
A study is conducted on Case 1 in which a number of scenarios are examined as the constraints on modal damping ratios, modal damping constants, and stabilizer gains are varied.
The nominal gains of the PSSs and the FDSs are all set to 5 pu on device base. Reducing the
gains from the higher values employed in Section 14.3.11 for heuristic coordination allows
for more flexibility in the optimization of the gain settings. The rotor modes are listed in
Table 14.4 for Case 1 with the stabilizers out of service and then in service with their gains
set to the nominal values.
In order to understand the action of the LP algorithm in adjusting the stabilizer gains let us
firstly consider the SDCD for Case 1 with the nominal gain settings. Because the emphasis
in the following analysis concerns the inter-area modes K, L, and M, the SDCD for these
modes is shown in Figure 14.9 to ascertain the effects on the modes of a 1 pu increment on
all stabilizers.

1. See step 1b, Table 14.2; nominal PSS gains 10 pu, FDS gains 20 pu.

640

Coordination of PSSs and FDSs

Table 14.4

Ch. 14

Rotor modes and modes shifts for heavy load condition, Case 1 1.
Nominal gains of all stabilizers 5 pu on device base.
Case 1. Heavy load

No.

A
B
C
D
E
F
G
H
I
J
K
L
M

All PSSs & FDSs in service.

Nominal stabilizer gains on
device bases: all 5.0 pu

Stabilizers off

Mode Shift

Real

Imag

Real

Imag

Real

Imag

-0.16
0.11
0.03
-0.56
-0.26
-0.68
-0.40
0.02
-0.19
-0.62
0.08
0.06
0.01

10.45
9.58
8.93
8.63
8.37
8.00
8.05
7.81
7.72
7.42
4.02
2.61
2.03

0.01
-0.01
-0.00
0.06
0.03
0.08
0.05
-0.00
0.02
0.08
-0.02
-0.02
-0.00

-0.67
-0.39
-0.43
-1.00
-0.69
-0.85
-0.85
-0.53
-0.65
-0.92
-0.22
-0.12
-0.20

10.49
9.65
9.03
8.74
8.37
8.25
7.87
7.86
7.82
7.49
3.99
2.60
1.99

0.06
0.04
0.05
0.11
0.08
0.10
0.11
0.07
0.08
0.12
0.06
0.05
0.10

-0.51
-0.50
-0.46
-0.45
-0.43
-0.17
-0.45
-0.54
-0.46
-0.31
-0.31
-0.19
-0.21

0.04
0.07
0.10
0.09
0.00
0.25
-0.18
0.05
0.10
0.07
-0.03
-0.01
-0.03

Mode Number.

is the damping ratio.

For insight, based on Figure 14.9, consider the stabilizers which dominate - in descending
order left to right - the contributions to the damping (i.e. left-shifts) for each of the interarea modes.
:
Mode
K
L
M

Stabilizer
LPS_3
PSVC_5
BSVC_4

YPS_3
SVC_2
GPS_4

EPS_2
MPS_2
PSVC_5

MPS_2
GPS_4
SPS_4

HPS_1
TPS_4

1. The values of the modes in the table differ slightly from those in Table 10.14 due to: (i)
the addition of a SVC on bus 212, and (ii) the FDSs on all three SVCs being in service at
their nominal gains of 5 pu.

Sec. 14.5

Case study: Automated Coodination

Mode L

Mode K

-0.043

641
Mode M

-0.010

Figure 14.9 Case 1. Contributions to modes shifts of inter-area modes K, L & M by each
stabilizer for a gain increment of 1 pu. Nominal gain settings for all stabilizers is 5 pu on
device base. (Note: contribution of LPS_3 is off-scale; the value is
5 8.57 j1.9 10

= 0.043 j0.01.)

Notice there are several stabilizers that can affect the contributions to the damping of more
than one mode. For example, if the gain on PSVC_5 needs to be increased by the LP algorithm to satisfy a requirement on mode L, it also produces a contribution to damping on
mode M. Similar implications apply to MPS_2 and GPS_4; such observations may help to
explain what may be unexpected results.
To examine the action and performance of the LP algorithm five scenarios are considered
in which all stabilizer weightings w j = 1 and the following constraints or limits are varied:
1.

the maximum gain on selected stabilizers,

2.

the type of mode (local- or inter-area),

3.

the real part of the mode-shift is limited by

the modal damping constant which must be less than or equal to a specified value;

642

Coordination of PSSs and FDSs

Ch. 14

the damping ratio of the rotor mode which must be greater than or equal to
a nominated limit;
the change in the damping constant which must be less than or equal to a
specified value.

In the following five scenarios shown in Table 14.5 for Case 1 the first three implement a
real left-shift on the inter-area modes K, L and M with the constraint that their damping ratios are 0.2 . The maximum stabilizer gains are reduced over Scenarios 1 to 3 to investigate the reallocation of stabilizer gains between stabilizers. In Scenario 4 the limits on modes
K, L and M require that the real parts of the modes are 0.4 with maximum stabilizer
gains of 20 pu on device base. Finally, in Scenario 5 the stabilizer gains must satisfy (i) a limit
0.4 on the real parts of modes K, L and M, (ii) the constraint 0.1 for the more
lightly damped local-area modes A, B, C, I and H.
The optimum stabilizer gains (in pu on device base) derived from the LP analysis and the
associated constraints for the five scenarios are summarized in Table 14.5
14.5.1 Scenario 1: Inter-area modes. Maximum PSS & FDS gain 40 pu.
In this scenario the damping ratios of the inter-area modes are to be equal to or greater than
0.2. From col. 2 of the table it is noted that the FDSs on two of the three SVCs are at their
limiting gains, and that the gains of only three of the fourteen PSSs are increased from their
nominal values. The condition 0.2 requires a significant left-shift on the higher frequency inter-area modes as revealed in the eigen-trajectories in Figure 14.10. Some local-area
modes are only marginally left-shifted by the increases in stabilizer gains, some significantly.
It is evident from the SDCD of Figure 14.9 that, of the three modes, for 1 pu gain increment
on any one stabilizer the contributions to the shifts in modes K and M are generally greater
than or comparable to those for mode L. Because stabilizers PSVC_5, SVC_2, MPS_2 and
GPS_4 contribute most to mode L, their gains are increased and, as a result, their contributions to modes K and M are also raised. As a consequence a lower contribution to mode K
by LPS_3 (the largest potential contributor to the mode) is required.

Sec. 14.5

Case study: Automated Coodination

Table 14.5

643

Case 1. Optimum stabilizer gains determined by LP for five scenarios.

PSS or FDS gain
Scenario 1

Scenario 2

Scenario 3

Scenario 4

Inter-area K, L, M

Modes -->

Scenario 5
K L M & local area
ABCHI

Kmax PSSs (pu) -->

40

40

20

20

20

Kmax FDSs (pu)-->

40 (All)

20 (FDS)

20 (All)

20(All)

20(All)

KLM:

KLM: 0.4

0.4

ABCIH: 0.1

Mode constraints ->

Damping ratio, KLM: 0.2

Generator / SVC
|

Gain
Gain (pu on device
(pu on device base)
base)
col. 1
col. 2
col. 3
col. 4
col. 5
col. 6
20
11.7
5
HPS_1
5
14.0
BPS_2
5
5
5
5
7.7
EPS_2
5
5
16.9
5
15.0
VPS_2
5
5
5
5
6.1
MPS_2
18.8
40
20
20
19.4
SVC_2
40
5
20
20
20
LPS_3
15.7
15.7
15.9
7.4
6.6
YPS_3
5
5
5
5
6.9
TPS_4
5
5
5
5
6.0
CPS_4
5
5
5
5
5
SPS_4
5
5
6.8
5
11.5
GPS_4
8.7
40
20
20
19.9
BSVC_4
5
5
20
5.4
6.1
NPS_5
5
18.7
17.1
5
5
TPS_5
5
34.2
20
16.6
9.4
PPS_5
5
5
12.5
5
5
PSVC_5
40
20
20
20
20
Sum of device gains
183
233
234
166
174
The yellow shading indicates that the stabilizer gain is a maximum.

The stabilizer gains are listed in col. 2. Gain limits of 40 pu occur only on the FDSs of
PSVC_5 and SVC_2 which are the main contributors to mode L.
In Figure 14.10 the estimated and actual values of the three inter-area modes at each step in
LP procedure are in close agreement. The associated initial and final values of the ten localarea modes, A to J, are also recorded.

644

Coordination of PSSs and FDSs

Ch. 14

AA

BB
CC

D
D
G

JJI

F
G

E E
I

HH

Initial and final values of local-area modes A to J

= 0.2
Inter-area modes:

= 0.1

L
M

l
Figure 14.10 Case 1, Scenario 1. In the lower box the eigen-trajectories of inter-area
modes K, L and M are plotted ( 0.2 ). The six steps of the trajectory are shown; at each
step the estimated values B to G of the mode are shown to agree closely with the actual values H to M. In the upper box the initial and final eigenvalues of the left-shifted local-area
modes A to J are also marked.
14.5.2 Scenario 2: Inter-area modes. Limits PSS gains 40 pu, FDSs 20 pu
For comparison with Scenario 1 the effect of the reallocation of gains to other stabilizers
resulting from the reduction in the maximum gain on all FDSs from 40 to 20 pu is of interest. (See col. 3 of Table 14.5.) The gain limit of 40 pu now occurs only on the PSSs of MPS_2
and GPS_4, and limit of 20 pu on the FDS of PSVC_5. It is insightful to examine the characteristics of the main contributions by the stabilizers to the real shifts in modes K, L and
M at steps one to six of the LP calculation. Such contributions are shown in Table 14.6 in
which it is noted:

(i) the relative real shifts contributed by a stabilizer to the inter-area modes correspond
closely to those shown in the SDCD in Figure 14.9 (e.g. for MPS_2 in the SDCD the
ratio of the contributions to L and K are about 2:1; this is reflected in the table);

(ii) the resulting stabilizer gain is

6

k = 1 + f j k nom ,

j=1

where f j is the incremental gain factor and k nom is the nominal gain setting of the stabilizer. The resulting gains are listed in col. 3 of Table 14.5.

Sec. 14.5

Case study: Automated Coodination

645

(iii) Also, from Table 14.6, the associated contribution to the real shift in inter-area
mode h by stabilizer j is given by
h

total =

h j .

j=1

(iv) Stabilizers MPS_2, PSVC_5, GPS_4 and TPS_5 play a major role in contributing
to the real left-shift in the mode L. In so doing, GPS_4 and PSVC_5 are also major
contributors to real shift in mode M. Similar observations also apply to LPS_3 and
MPS_2 for mode K.

14.5.3 Scenario 3: Inter-area modes. Limits: all stabilizer gains 20 pu

As to be expected in comparison with Scenario 2, additional stabilizers - namely the FDSs
on the three SVCs and PSSs on four generators - operate at the maximum gain of 20 pu.
The gain of the PSS of LPS_3 is more or less constant (15.7 - 15.9 pu) over scenarios 1 to 3
for reasons discussed in Section 14.5.1.
14.5.4 Scenario 4: Inter-area modes. Limits 0.4 ; all stabilizer gains 20 pu
The limit on the inter-area modes is much less onerous on modal damping than in Scenarios
1 to 3 for which the limit is 0.2 . Note that in latter studies the real part of the eigenvalue
for mode K is 0.8 at the limit. Consequently in this scenario the PSS gain of LPS_3 is approximately halved, but the FDS gains of SVC_2, PSVC_5 and PSSs on MPS_2 & GPS_4
are at the limit of 20 pu.
14.5.5 Scenario 5: Inter-area modes; local-area modes A, B, C, H, I
The limits for this scenario are: inter-area modes 0.4 ; the selected local-area modes
0.1 ; limit on all stabilizer gains 20 pu.
The object of this scenario is to assess, in comparison with Scenario 4, the influence of the
constraint on the selected local modes on the distribution of the gains between stabilizers.
The local-area modes are selected on the basis that they are the more lightly damped. The
eigen-trajectories for both inter- and local-area modes are plotted in Figure 14.11.
An expanded and more detailed plot of the eigen-trajectories of the local-area modes are displayed in Figure 14.12. Note that in both this and previous figure the shifts in the trajectories
are more-or-less directly to the left 1. For practical purposes in the LP procedure, the extent
of left-shift associated with increases in stabilizer gain remains decoupled from the stabilizer
phase compensation over the range of frequencies of the rotor modes.

1. By directly to the left is implied that the mode shift is j0 , 0 .

Case 1, Scenario 2. Contributions by stabilizer to the real shifts on Modes K, L and M Real left-shifts greater than 10
units (0.01 Np/s) are high-lighted.
HPS_1

Step
1
2
3
4
5
6

MPS_2

1.093
0.219
0.405
0
0
0.088

+4.0
+0.8
+1.6
0
0
+0.4

-14.0
-2.5
-4.5
0
0
-0.9

-2.1
-0.5
-2.1
0
0
-0.1

1.167
1.167
1.167
1.167
1.167
1.167

-10.6
-10.7
-10.6
-10.5
-10.3
-9.9

-19.7
-18.4
-17.9
-17.2
-16.8
-16.5

+0.6
+0.5
+0.5
+0.6
+0.8
+1.0

0.393
0.359
0.364
0.347
0.343
0.339

-84.2
-75.7
-76.7
-73.2
-72.8
-72.7

0
0
0
0
0
0

-13.1
-12.0
-12.2
-11.8
-11.8
-11.9

1.167
1.167
1.167
1.167
1.167
1.167

-1.1
-1.0
-1.1
-1.1
-1.2
-1.2

-16.7
-15.1
-14.4
-13.6
-13.2
-12.8

-32.6
-34.2
-35.1
-36.3
-37.0
-38.1

- Modal real shift (units), 1 unit = 10-3 Np/s.

TPS_5

PSVC_5

SVC_2 & BSVC_4

0
0
0
0.659
0.918
1.167

0
0
0
-2.1
-3.0
-3.9

0
0
0
-6.8
-9.3
-11.7

0
0
0
-3.4
-4.5
-5.5

0
1.167
1.167
1.167
1.167
1.167

0
-5.3
-5.2
-5.2
-5.1
-5.1

0
-14.5
-14.5
-14.5
-14.0
-13.3

0
-10.0
-9.4
-8.8
-8.5
-8.2

0.5
0.5
0.5
0.5
0.5
0.5

-3.0
-3.0
-2.9
-2.9
-2.8
-2.7

-14.9
-15.2
-14.7
-14.3
-13.5
-12.7

-13.2
-12.3
-11.7
-11.7
-10.5
-10.0

0
0
0
0
0
0

0
0
0
0
0
0

0
0
0
0
0
0

0
0
0
0
0
0

f - Stabilizer incremental gain factor.

- Modal real shift (units), 1 unit = 10-3 Np/s.

Coordination of PSSs and FDSs

NPS_5

1
2
3
4
5
6

GPS_4

f - Stabilizer incremental gain factor.

Step

LPS_3

646

Table 14.6

Ch. 14

Sec. 14.6

Concluding remarks

647

Clearly, in comparison to Scenario 4 and in satisfying the additional constraints imposed by

the local-area modes, the gains of the stabilizers change somewhat (see Table 14.5, col. 6).
The degradation in the figure of merit - i.e. the sum of the gains - increases slightly from 166
for Scenario 4 to 174.

Local-area modes (see Fig 14.12 for details)

= -0.4 Np/s
Inter-area modes: K
L
M

Figure 14.11 Case 1, Scenario 5. The eigen-trajectories are plotted of inter-area modes K,
L and M as well as local-area modes A, B, C, H and I. The six steps of the trajectories are
marked; at each step the estimated values B to G of the mode and the actual values H to M
agree closely as shown. The initial (A) and resulting final (N) eigenvalues of the remaining
local-area modes are also indicated.

14.6

Concluding remarks

The approaches for tuning PSSs in Chapters 5 and 10 and FDSs in Chapter 11 provide the
means of the coordination of stabilizers by the coordination of their gains using either the
SDCD- or LP-based procedures. The preliminary tuning of stabilizers constitutes Stage 1 in
their coordination. The purpose of Stage 2 in the either of the two procedures of coordination is to satisfy certain criteria on modal damping and on stabilizer gain values.
Heuristic coordination

In the heuristic based approach the SDCDs provide information on the extent of the leftshift available on a selected rotor mode for gain increments on a range of stabilizers, PSSs
and FDSs. Such information permits the user to estimate the gain increments required to
produce an adequate left-shift in the mode that satisfies the modal damping criteria within
a nominated range of gains.

648

Coordination of PSSs and FDSs

Ch. 14

Mode A
Mode B

Mode C

Mode D

Mode E
Mode F
Mode H
Mode G

Mode J

Mode I

Figure 14.12 Case 1, Scenario 5. Trajectories of local mode A-J. The six steps are shown
for the constrained modes A, B, C, I and H; at each step the estimated values B to G and
the actual mode values H to M are plotted. For the modes D, E, F, G and J the initial values
A and the final values N are displayed.
Having established a set of stabilizer gains required for a selected operating condition, this
set can be tested on other operating conditions to ascertain if the criteria on the modes are
satisfied or not. If not, the SDCDs for the new operating condition are invoked to establish
the changes in the stabilizer gains required to satisfy the relevant criteria. By iterating
through the range of encompassing operating conditions a set of stabilizer gains can be
found which satisfies all conditions; this is illustrated in Section 14.3. Clearly the process
may be tedious and does not lead to a unique solution as it depends on the users methodology and experience.
It is assumed that those stabilizers that are required to be in service are initially set to certain
minimal gains. In the scenarios demonstrating the LP procedure the nominal stabilizer gains
are all set to 5 pu on device base. This allows for flexibility in the procedure, e.g. it may reveal
that the gains of some stabilizers remain at the minimum value and therefore do not effectively contribute to any improvement in damping.
Automated coordination

The LP-based approach reveals aspects of the coordination process that are not obvious or
accounted for in the SDCD-based procedure. It reveals that certain stabilizers are (unexpectedly) more influential in satisfying the modal criteria than those stabilizers which may
appear to be the more obvious candidates. This is particularly the case in dealing with modes

Sec. 14.6

Concluding remarks

649

that are initially more poorly damped. Such information may fruitfully be incorporated into
the SDCD-based analysis.
The example of applying the LP-based procedure to the 14-generator system suggests alternative approaches can be adopted to the setting of stabilizer gains to satisfy the rotor modal
criteria. For example, if is intended to determine a set of gains for PSSs and FDSs for Area
2 only, the settings in other Areas remaining unchanged, a tailored approach may be required.
The LP-based procedure offers the following benefits.

(i) The criteria for damping of some rotor modes may differ from those for others;
this, together with a larger number of local- and inter-area modes, complicate the
analysis in the SDCD-based approach.

(ii) Criteria can be placed on exciter or controller modes to limit their right-shift.

(iii) The roles and merits of certain stabilizers are high-lighted, e.g. MPS_2, SVC_2;
these roles may not be readily apparent from the SDCD of Figure 14.9.

(iv) The stabilizer gains are no higher than necessary.

(v) Stabilizers that are the more critical to the support and improvement in the
damping of a poorly-damped rotor mode are revealed.

(vi) The number of trial-and-error studies required for coordination are reduced.

(vii) The comparison between scenarios of the sum of stabilizer gains (the quantity
minimized in the objective function) is an indicator of the gain loading on the
stabilizers; the larger the sum the more likely it is that further maximum stabilizers gains are imposed.

(viii) Information gleaned from LP-based procedure, such as in (iii) above, together
with the knowledge of the practical implications, can be incorporated into the
analysis based on the SDCD-based approach. The two approaches can complement each other.

The scenarios demonstrating the performance of the LP-based procedure highlight the effectiveness of the bus-frequency stabilized FDSs. In the studies their contributions to the
damping of certain inter-area modes is more extensive and effective than PSSs, and no deleterious effects are observed on the local-area modes. The potential degradation in damping
on rotor modes when SVCs or the FDSs are out of service requires investigation. Likewise,
a study would be required into the effect of the outage of generation on a difficult-to-damp
mode, e.g. the loss on MPS_2 on mode L.

650

Coordination of PSSs and FDSs

Ch. 14

The disadvantage of the LP-based procedure is that it provides a set of stabilizer gains, subject to certain constraints, that are optimum for the selected operating condition; it does not
have the facility to optimise the gains over a range of operating conditions (See [4], [6]).
However, at other encompassing operating conditions for which the stabilizers are tuned,
the rotor modes are likely to be left-shifted with increases in stabilizer gains as long as the
gains are in the acceptable range of values.
Using the automated approach, a case study such as that in Section 14.5 can used to examine
stabilizer coordination in a selected area of a larger power system, e.g. Area 2 in Figure 10.1.
This would provide some guidance on how better to improve damping on the system and
where additional PSSs and/or FDSs may be located to achieve better damping.
The SDCD- and the LP-based approaches are together a useful set of tools because they
provide information and insight into the power systems dynamic characteristics. Such information and the guidance allows the user to make judicious, practical decisions on the parameter settings of stabilizers. Of note, the automated approach to stabilizer coordination
has been extended to cover the set of encompassing operating scenarios in [6] and [16].
Studies on the coordination of the controllers for other FACTS installations such as multiple HVDC links are reported in the literature, e.g. [14], [15].

14.7 References
[1]

F. P. de Mello, P. J. Nolan, T. F. Laskowski, and J. M. Undrill, Coordinated Application of Stabilizers in Multimachine Power Systems, Power Apparatus and Systems,
IEEE Transactions on, vol. PAS-99, pp. 892-901, 1980.

[2]

M. J. Gibbard, Co-ordinated design of multimachine power system stabilizers based

on damping torque concepts, IEE Proceedings C Generation, Transmission and Distribution, vol. 135, pp. 276-284, 1988.

[3]

Z. Ao, T. S. Sidhu, and R. J. Fleming, Stability investigation of a longitudinal power

system and its stabilization by a coordinated application of power system stabilizers,
Energy Conversion, IEEE Transactions on, vol. 9, pp. 466-474, 1994.

[4]

J. N. Taranto, A. L. B. do Bomfim, D. M. Falcao, and N. Martins, Automated design

of multiple damping controllers using genetic algorithms, in Power Engineering Society 1999 Winter Meeting, IEEE, 1999, pp. 539-544 vol.1.

[5]

A. Doi and S. Abe, Coordinated Synthesis of Power System Stabilizers in Multimachine Power Systems, Power Apparatus and Systems, IEEE Transactions on, vol. PAS103, pp. 1473-1479, 1984.

[6]

R. A. Jabr, B. C. Pal, and N. Martins, A Sequential Conic Programming Approach

for the Coordinated and Robust Design of Power System Stabilizers, Power Systems,
IEEE Transactions on, vol. 25, pp. 1627-1637, 2010.

Sec. 14.7

References

651

[7]

P. Pourbeik, Design and Coordination of Stabilisers for Generators and FACTS devices
in Multimachine Power Systems, PhD Thesis, The University of Adelaide, Australia,
1997.

[8]

P. Pourbeik and M. J. Gibbard, Simultaneous coordination of power system stabilizers and FACTS device stabilizers in a multimachine power system for enhancing
dynamic performance, Power Systems, IEEE Transactions on, vol. 13, pp. 473-479,
1998.

[9]

M. J. Gibbard, N. Martins, J. J. Sanchez-Cascara, N. Uchida, V. Vittal, and W. Lei,

Recent applications of linear analysis techniques, Power Systems, IEEE Transactions
on, vol. 16, pp. 154-162, 2001.

[10]

C. Li-Jun and I. Erlich, Simultaneous coordinated tuning of PSS and FACTS damping controllers in large power systems, Power Systems, IEEE Transactions on, vol. 20,
pp. 294-300, 2005.

[11]

H. Bourles, S. Peres, T. Margotin, and M. P. Houry, Analysis and design of a robust

coordinated AVR/PSS, Power Systems, IEEE Transactions on, vol. 13, pp. 568-575,
1998.

[12]

J. P. Ignizio, Linear programming in single- and multiple-objective systems, PrenticeHall, 1982.

[13]

G. B. Dantzig and M. N. Thapa, Linear Programming, Springer-Verlag, 1997.

[14]

L. A. S. Pilotto, M. Szechtman, A. Wey, W. F. Long, and S. L. Nilsson, Synchronizing and damping torque modulation controllers for multi-infeed HVDC systems,
Power Delivery, IEEE Transactions on, v ol. 10, pp. 1505-1513, 1995.

[15]

A. M. Ersdal, L. Imsland, and K. Uhlen, Coordinated control of multiple HVDC

links using backstepping, in Control Applications (CCA), 2012 IEEE International
Conference on, 2012, pp. 1118-1123.

[16]

J. C. R. Ferraz, N. Martins, and G. N. Taranto, Coordinated Stabilizer Tuning in Large

Power Systems Considering Multiple Operating Conditions, in Power Engineering Society
General Meeting, 2007. IEEE, 2007, pp. 1-8.

Index
Stability with PSS 410

A
Automated coordination 634
Benefits 649
Constraints on solution 635
Eigen-trajectories 644, 647, 648
Optimum gains 642
PSSs and FDSs Sec. 14.4.3, Sec. 14.5
Simplex algorithm 635
Two stage algorithm 637
using Linear Programming 635
AVR - Analysis
Conversion from PI to PID Compensation 352
Phase-matching method 392
PI Compensation 351
PID compensation 327351
Rate feedback compensation 354370
see also PID compensation 329
Transient gain reduction 320327
Type 2B PID compensation 371383
AVR compensation types
PID and PI 316
Rate feedback 316
Transient gain reduction (TGR) 316

C
Characteristics of systems
Characteristic equation 41
Damping constant 41
Damping ratio 41
First- and second-order systems 39
Second-order response 42
Time constant 39
Undamped natural frequency 40
Classical Parameter Model 89
Block diagram form 146
Equation summary 147
Formulation 89, 145
Linearization em equations 155
Relation to cc parameters 141, 143
Steady-state conditions Sec. 4.2.13.3
Transient flux linkages 150
Coordination of stabilizers 621
- using Linear Programming Sec. 14.4
Heuristic approach Sec. 14.3
Coupled-Circuit Model 89
Relation to classical parameters 144

B
Benchmark for damping by ideal PSS

D
Definition
Automatic voltage regulator (AVR) 315
Excitation control system (ECS) 314
Excitation system (ES) 314
Frequency stability 10
Halving time 479
Per unit definitions for ECS 315, App. 7

595

Bode Plots 5156

Gain and frequency margins 59
Gain cross-over frequency 59
Stability Analysis 59
Brushless exciter
Frequency response 373
Models 386, App. 7I.2
Bus-frequency stabilizers 407
Degradation in mode shifts 410
PSSs 407

I.1

Power system stability 7

Rotor angle stability 8
Transient stability 8
Voltage stability 9

653

654

Index

E
Eigen-analysis
14 generator system 481
Calculation of Eigenvalues 448
Eigenvalue sensitivities 85
Eigenvalues of state matrix 74
Eigenvalues, modes and stability Sec. 3.5.2
Modified Arnoldi 448
Multiple-Shift-Point Eigenanalysis 448
QR factorisation 448
Errors in the steady-state
Alignment 46
Following error 48
F
FACTS device models 157
Base quantities at AC terminals 160
Linearized equations, AC quantities 159

163

Modelling assumptions 158

Representation of interface with AC network 158
SVC model 163165
VSC model 165169
FACTS Device Stabilizer
- called FDS 4
Configuration of controller and stabilizer

532

Estimate FDS mode shifts 542

Form of FDS transfer function 542
Improving damping Sec. 11.8.1
Method of Residues 537
Modal trajectories 544, 547, 550, 625
Mode shift in targeted mode 536
Polar plot of residues 539, 545, 549, 552,

624

Robustness of FDSs Sec. 11.8.2

See also FDS 4
Stability 543, 547, 550
SVC - tuning 538, 545
TCSC - Modal trajectories 556
TCSC - Polar plot of residues 554
TCSC - Stability 555
Theoretical basis for tuning 536
Tuning 531
Tuning of a SVC FDS 538
Upper value of gain 537
FDS
FACTS Device Stabilizer 4
Filter characteristics
Washout and low-pass filters 246
Frequency response diagram

see Bode Plot

Frequency stabilizing signal
for a PSS 407
for a SVC 539, 548
G
Generator Models
Airgap flux linkages 121
Alternative dq-axis rotor structures 111
Base values for generator quantities 98
Canay inductances 95
d-q equivalent circuits diagrams 97
d-q-axis equivalent circuits. 97
Definitions of reciprocal field quantities,
etc 115
Demagnetizing effects of saturation 105

110

Flux linkage distribution 95

Functions for saturation characteristics

120

Initial conditions 90
Interface between generator and network
frames 130
Lad-base reciprocal pu field quantities 116
Linearization of the coupled-circuit equations 107
Linearized coupled-circuit equations 133
Model code 91
Modelling assumptions 93
Modelling generator saturation 118124
Modelling magnetic saturation 122
Non-reciprocal pu field quantities 116
Open-circuit characteristic 117
Parameter and variable definitions 99
Park-Blondel reference frame 130
Park-Blondel transform 93
Per-unit linearized model 90
Per-unit model structure 92
Procedure to compute coefficients in linearized c-c equations 134
Rankin base values 97
Rrotor equations of motion in pu 114
Saturation Methods 123124
See also Classical Parameters
See also Operational Parameters
Standard parameters 93
Steady-state operating conditions 125

129

Summary of the per-unit system 96

Unequal mutual coupling, rotor windings

94

GEP method

Index

655

Algorithm - calculation of stabilizer

parameters 309, 543
Basis of tuning 302
Comparison - GEP and P-Vr characteristics 303
Conclusions 306
PSS-induced damping torque coefficient

305

H
Heuristic coordination 626
Features - Features 647
for inter-area modes Sec. 14.3.1
for local-area modes Sec. 14.3.2
Use of SDCDs 627, 634
HVDC link modelling 177180
AC current and apparent power

Sec. 4.3.8.5

Commutating reactance in TCCX 182

Commutating voltage 182
Converter model for inverter operation

Sec. 4.3.8.6

HVDC models, internal - external interface values Sec. 4.3.8.9

HVDC models, Power factor 186
HVDC models, Power factor angle 186
with VCSs 180
with Voltage commutated converters

181185

Hydro-Qubec
Multi-path, multi-band PSS 397
I
Integral-of-accelerating-power PSS 413
Action of pre-filter, no washouts 416
Analysis of tracking errors 442
Block diagram of prefilter 419
Corrupted speed signal 414
Degradation in performance of the pre-filter 429
Dynamic performance of the pre-filter

424

Effect of the washouts and integrators

419

Lack of fidelity in Prefilter 398

Ramp Tacking Filter 415
Torsional modes 414
Integrator Wind-up Limiting 391
Interactions 594
between PSSs & FDSs 589
Case Study Sec. 13.3

Concept of Sec. 13.2.2

Definition - Generator interactions 595
Definition - Stabilizer interactions 594
Summary Sec. 13.6.1
L
Laplace Transform 27
Linearized power system model
Form of linearized DAEs Sec. 4.4.1
Formof nodal current equations Sec. 4.4.2
Modular and sparse structure Sec. 4.23
Perturbation in load bus frequency 203
Structure of linearized DAEs Sec. 4.4.4
Load Models
Alternative modelling 204
DAEs 203, 204
Frequency dependence 201
Linearized equations 202
Network connection 202
Types of static loads 201
ZIP types 201
M
Method of Residues
Conclusions 300
FDS tuning 536
Form of compensation transfer function

294

Form of PSS transfer function 298

Theoretical basis 294
Tuning a speed-PSS 297
MITCs 4
Concepts 563, Sec. 12.2
Contibutions - local mode Sec. 13.3.1
Contibutions to inter area mode

Sec. 13.3.3

Definition, MITC 576

Features & significance 564
Modal torque coefficients induced by FDS

Sec. 12.2.3

Modal torque coefficients induced by PSS

Sec. 12.2.2

Reductions of system TFM model

App. 12I.2

Relationship, MITCs and stabiliser gains

Sec. 12.6.2

System response for single modal frequency App. 12I.1

Torque coefficients induced by centralized
FDS Sec. 12.5
Torque coefficients induced by centralized
PSS Sec. 12.4

656

Index

Torque coefficients induced by centralized

stabilizers Sec. 12.2.4
Torque coefficients induced by decentralized stabilizers Sec. 12.6
Mode Shape Analysis 449
Four-mass spring system 456
Normalised right speed-eigenvectors 452
Note of warning 459
Stored energy in masses and springs 454
Theoretical basis 450
Two-mass spring system 450
Mode shapes 80
14 generator system 482
Theoretical basis 80
Mode Shifts
- with stabilizer gain increments 590, 593,

P-Vr characteristics of generators 487

Parameters of PSS transfer functions 493
Parameters of synthesized P-Vr transfer
functions 493
Participation factors and mode shape ,
inter-area mode 484
Participation factors and mode shape,
local-area mode 485
Participation factors of a d-axis mode 486
PSS-induced torque coefficients 498
Synchronising and damping torques 498
Synchronous compensators/motors 478
Synthesized characteristic of P-Vrs 492
Tables of data for dynamic performance
analysis 527530
Tables of data for power-flow analysis

Analysis of variation 251

Modes
Types of modes 10
Units of Modal Frequency 18
Multi-band PSS 397
Block diagram of PSS 433
Frequency response 436
Objective of PSS structure 433
Q-filter 435
Source Hydro-Qubec 397
Transfer function analysis App. 8I.3
Multi-machine PSSs 462
Definition P-Vr transfer function 462
Derivation of PSS transfer functions 468
Inherent torques 471
P-Vr transfer function 462
Multi-machine system 475
14 generator model system 477
Behaviour and types of rotor modes 486,

Tables of PSS parameters 494496

Tracking of rotor modes. PSSs in service

Sec. 13.2.3

510

Composite generator 476

Correlation, small- & large-signal dynamic
performance 509
Dynamic performance criterion 479
Dynamic performance from eigen-analysis

500

Dynamic performance from participation,

mode-shape analysis 503
Dynamic performance from time
responses 505
Encompassing range of operating conditions 477
Inherent torque coefficients 498
Intra-station modes of oscillation 507
Line diagram 14-generator system 480

522526
502

N
Normal Form analysis 513
O
Operational Parameter Model 89
Exactly-defined parameters 89
P
Participation Factors 83
14 generator system 482
Bar charts 461
Inertia weighted 596, 599
of a four-mass spring system 460
Theoretical analysis 83
Theoretical basis 459
Performance criteria
Control systems 14
Damping performance 13
ECS 316319
Multi-machine power system 479
PI compensation 351
Conversion to PID 352
Using positive feedback 388
PID 2B compensation
Example Three-generator power system

371383

Frequency response characteristics 349

Phase-matching method 392
Purpose of example 371
Tuning of Type 2B 348
PID compensation

Index

657

Characteristics Type 1 330

Characteristics Type 2 333
Characteristics Type 3 335
Purpose 327
Theoretical background 329335
Tuning of Type 1 335
Tuning of Type 2A 337
Tuning of Type 2B 348
Types 1 to 3 329
Types 2A and 2B 334
POD 2, 4
Poincar 2, 9
Power System Stabilizer
Bus-frequency PSS 407
Comparison speed & bus frequency PSSs

409

Compensating transfer function 241, 243,

278

Concepts for speed-PSS 237

Damping gain 231, 245
Damping torque coefficient 231
Electrical power stabilizing signal 404
Ideal PSS 232, 236
Low-pass filter 246
Simple concepts 230
Synthesis of PSS compensating transfer
function 266
Transfer function of PSS 248
Washout filter 246
PSS & FDS tuning using
Method of Residues 225, 294
PSS tuning using
GEP(s) transfer function 225
P-Vr approach 225
P-Vr approach to tuning (Summary) 284
P-Vr characteristics
6th order generator model 264
Basis in multi-machine systems 463
Nature of 242
of individual generators 488491
Relative invariance of phase and magnitude 464
Selection of Design Case P-Vr characteristic 265
Synthesis of P-Vr transfer function 266,

466

Q
Q-filter 61
R
Ramp tracking filter 415

Analysis of following errors App. 8I.2

Frequency responses 415
Steady-state tracking - and errors 421
Theoretical basis App. 8I.1
Rate feedback compensation
Application of Root Locus Method 368
Applications 358368
Desired form of transfer function 356
Method of analysis 354
Tuning of Excitation System 354
Relative effectiveness of stabilizers

Sec. 13.6.2

Residues
Calculation of Residues Sec. 2.5
Method of Residues Sec. 6.2
Relation - residues and MITCs 593
Robustness
Multi-machine PSSs 516
Multi-machines robust PSSs 462
Robust Controller 3
Robust Controllers 15
Robustness of FDSs Sec. 11.8.2
S
SDCDs 609
Damping contributions by FDSs 610
Damping contributions by generators 612
Damping contributions by PSSs 609
Stabilizer damping contribution diagrams

Sec. 13.4

Shaft dynamics
Disabled 229
Disabled for GEP method (GEPSDD)

303

SMIB System
Heffron and Phillips Model 226
Inherent synchronizing and damping torques 232
K-coefficients, Heffron and Phillips
Model 288
Synchronizing and damping torques 227
Stability of linear systems 44
STATCOM model
Assumptions 172
Simplified model Sec. 4.3.4
State equations
Concept of state 67
Determination of residues 77
Forced response 74
Modes & eigenvalues Sec. 3.5.2
Natural response 71
Note on stability 76

658

Index

Solution of 71
Stability of operating point 69
SVCs
Simple model Sec. 4.3.2
Simplistic tuning 532
T
TCSC
Modulus of real power FDS 552
Tuning a FDS Sec. 11.7
Thyristor Controlled Series Capacitor
(TCSC)
Network nodal equations 193
Representation of TCSC in steady state

193

Transfer function
Poles and zeros 30
Proper /strictly proper 27
Transfer function form 93
Transfer function form
Exact definition 93
Relation to cc form of em equations 139
Transfer function matrix (TFM)

Model - centralized stabilizers 572

Transient stability studies 68, 509, 515
V
Voltage Commutated Converters
Converter control input Sec. 4.3.8.7
Extinction-angle advance 188
Extinction-delay angle 186
Firing-angle advance 188
Firing-delay angle 186
Linearized algebraic equations Sec. 4.3.8.8
Overlap angle 187
VSC model
Converter, AC and DC quantities 182
Summary linearized equations 170172
W
Washout and low-pass filters Sec. 5.8.6
Washout filters (1 or 2) 399
Characteristics of 1 & 2 filters 403
Frequency-domain responses 401
Time-domain responses 399