222

ON VOLTAGE AND POWER

STABILITY IN AC/DC SYSTEMS

Working group
14.05

April 2003
 ON VOLTAGE AND POWER STABILITY
IN AC/DC SYSTEMS

Working Group
14.05

Members of the Working group 14.05:

G. Andersson (Switzerland) Convener

T. Adhikari (India), J.B. Davies (Canada), M.A Eitzmann (USA), A. Golé (Canada), R. Hirvonen
(Finland), J.Makino (Japan), V. Khoudiakov (Russia), D. Kumar (India), P.C.S. Krishnayya (Canada),
H.A.D. Lee (Malaysia), W. Long (USA), R.S. Moni (India), C. Parker (Australia), C.A.O. Peixoto
(Brazil), J. Reeve (Canada), C.V. Thio (Canada)
Copyright © 2002
Tout détenteur d'une publication CIGRE sur support papier ou électronique n'en possède
qu'un droit d'usage.Sont interdites,sauf accord express du CIGRE, la reproduction totale ou
partielle autre qu'à usage personnel et privé, et toute mise a disposition de tiers, dont la
diffusion sur un réseau intranet ou un réseau d'entreprise.
Copyright © 2002
Ownership of a CIGRE publication, whether in paper form or on electronic support only
infers right of use for personal purposes..Are prohibited, except if explicity agreed by CIGRE,
total or partial reproduction of the publication for use other than personal and transfer to a
third party; hence circulation on any intranet or other company network is forbidden.
To the memory of Clarence V. Thio
Contents

Preface vii

Summary ix

1 Basic Theory and Definitions 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Simplified Explanation of Power/Voltage Instability Phenomenon 1
1.3 Definitions of Voltage Stability in AC Power Systems . . . . . . . 2
1.3.1 Definitions According to CIGRE . . . . . . . . . . . . . . 2
1.3.2 Definitions According to IEEE . . . . . . . . . . . . . . . 2
1.3.3 Other Definitions . . . . . . . . . . . . . . . . . . . . . . . 3
1.3.4 Discussion of CIGRE and IEEE Definitions . . . . . . . . 3
1.4 Analysis of AC/DC Systems . . . . . . . . . . . . . . . . . . . . . 6
1.4.1 Maximum Power Curves . . . . . . . . . . . . . . . . . . . 7
1.4.2 Voltage Sensitivity Factors . . . . . . . . . . . . . . . . . 9
1.5 Definitions for AC/DC Systems . . . . . . . . . . . . . . . . . . . 10
1.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2 Analytical Tools and Methods 13

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Power Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.1 Maximum Available Power . . . . . . . . . . . . . . . . . 14
2.2.2 Maximum Power Curve . . . . . . . . . . . . . . . . . . . 15
2.3 Voltage Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.1 Voltage Sensitivity . . . . . . . . . . . . . . . . . . . . . . 16
2.4 Control Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4.1 Control Sensitivity Index . . . . . . . . . . . . . . . . . . 17
2.4.2 CSI Computation . . . . . . . . . . . . . . . . . . . . . . . 18
2.5 Extension to Higher-Order Models . . . . . . . . . . . . . . . . . 19
2.5.1 Maximum Power . . . . . . . . . . . . . . . . . . . . . . . 20
2.5.2 Voltage Sensitivity . . . . . . . . . . . . . . . . . . . . . . 24
2.6 Incorporation of Local Devices . . . . . . . . . . . . . . . . . . . 27
2.6.1 Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.6.2 Other Devices . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.7 Relationship Between Power and Voltage Stability Methods . . . 30
2.8 Appendix: Definition of ESCR . . . . . . . . . . . . . . . . . . . 35

v
vi Contents

3 Comparison of MPC with Detailed Simulations 37

3.1 Simulations Using a Simplified Model . . . . . . . . . . . . . . . 37
3.2 Inclusion of a Synchronous Compensator at the Converter Bus . 39
3.2.1 Differences Arising from Neglecting the Zero-Power Con-
straint for the Synchronous Compensator . . . . . . . . . 39
3.2.2 Comparison with a Detailed Machine Model with and
without Excitation Control . . . . . . . . . . . . . . . . . 40
3.2.3 Inclusion of Controlled Excitation . . . . . . . . . . . . . 41
3.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.4 Appendix: Equations used in MPC Calculations . . . . . . . . . 45

4 Influence of System Operating Conditions and Converter De-

sign 47
4.1 Effect of Load Modeling on the MAP and AC Voltage . . . . . . 47
4.2 Effect of a Parallel AC Line . . . . . . . . . . . . . . . . . . . . . 50
4.3 Effect of Tap Changing . . . . . . . . . . . . . . . . . . . . . . . 51
4.4 MPC for Capacitor Commutated Converters . . . . . . . . . . . . 51
4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

5 Multi-Infeed HVDC Systems 55

5.1 Development in the Study of Multi-Infeed
HVDC Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.2 System Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3 Analysis of Voltage/Power Stability of Multi-Infeed HVDC Systems 59

6 Planning and Analysis of Systems - Principles and Concepts 63

6.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.2 Simplified Explanation of Power/Voltage Instability Phenomenon1
63
6.3 “Weak-Connection” Versus “Weak-System” In-Feed of a DC Link 64
6.4 Important Parameters of the System . . . . . . . . . . . . . . . . 65
6.5 Relationship Between DC Stability and Recovery from AC Faults 67
6.6 Effect of “Quasi-Static” Controls on Stability . . . . . . . . . . . 67
6.7 Effect of Dynamic Controls on Stability . . . . . . . . . . . . . . 69
6.8 Introducing the Concept of “Dynamic MAP” . . . . . . . . . . . 70
6.9 System Studies of Stability on Actual System . . . . . . . . . . . 72
6.10 Implications for Other Systems . . . . . . . . . . . . . . . . . . . 75

References 77

1 This section is identical with section 1.2, but since it is essential for the subsequent

discussion in this chapter it is repeated in full here.

Preface

This report is the result of work done in Cigre WG14.05 over the years 1996 -
1999. By various reasons all the various contributions have not been put together
in a single document until now. When this work started, it was not clear to the
convenor, nor to the working group members, the amount of work required to
cover the topic in a systematic way. To do this substantial additional work would
be needed. Still, the report contains significant new results, e.g. concerning the
definitions of voltage and power stability an AC/DC systems, section 1.5. This
definition is adopted to comply with the definitions introduced in power systems
without any HVDC systems. But also other chapters contain new material.
As agreed at the SC14 meeting in Paris 2000, the material available has been
put together without any excessive editing. Therefore, it might occur repetitions
and the coherency might not be complete. Despite this, the report contains a
lot of material that is deemed to be useful for the power industry.
Since the work started in 1996 new types of converters have been introduced
for high power application, i.e voltage source converters, which not are discussed
or analyzed in this report. Only line commutated current source converters and
capacitor commutated converters are covered. However, the analysis methods
and tools in the report could also be applied to systems using voltage source
converters or other types of converters.
The main contributors of the chapters in this report are:
Chapter 1: G. Andersson
Chapter 2: H.A.D. Lee
Chapter 3: A. Golé
Chapter 4: J. Reeve
Chapter 5: H.A.D. Lee
Chapter 6: C.V. Thio and J.B. Davies
During the course of this work the working group suffered a significant loss
when Clarence Thio passed away. Clarence was one of the most active members
of the working group and his contributions were essential for the progress of the
work documented in this report. His insights and experience together with his
sincere interest to understand the problems dealt with inspired all members of
the working group and had a decisive influence of the work pursued here. We
will all remember Clarence as a prominent engineer and a great friend.

vii
viii Preface

The members of WG14.05 who participated in the work were:

T. Adhikari
G. Andersson (Convenor)
J.B. Davies
M.A. Eitzmann
A. Golé
R. Hirvonen
J. Makino
V. Khoudiakov
D. Kumar
P.C.S. Krishnayya
H.A.D. Lee
W. Long
R.S. Moni
C. Parker
C.A.O. Peixoto
D. Povh
J. Reeve
C.V. Thio
As convenor I want to thank all the members of the working group for their
work and contributions, which formed the basis for this report.

Göran Andersson
Convenor WG14.05
Summary

Voltage and power instability has been a limiting factor for many HVDC projects
during the last decades. Pioneering and systematic work to understand and
analyse the interactions leading to this instability started in the early eight-
ies, and joint Cigré and IEEE working group presented the first comprehensive
overview of the topic in the early nineties, ref. [4] of this report. After that
report was published substantial work has been done, both in the academia and
the industry, to further develop analytical and simulation tools in order to im-
prove the understanding and the computational tools. These later developments
are reported in this report. A brief overview of the report is given below.
In the late eighties and early nineties extensive work was done within Cigré
and IEEE to better understand and analyse voltage stability in (pure) AC sys-
tems. Among other things definitions regarding voltage stability were formu-
lated. These definitions are reviewed in chapter 1 and definitions applicable for
AC/DC systems that are consistent with those of AC systems are formulated.
The concepts of Maximum Power Curves (MPCs) and Voltage Sensitivity
Factors (VSFs) were early introduced for the analysis of simple AC/DC systems.
In chapter 2 the theoretical foundations of these tools are further developed and
extensions to more complex systems are carried through. Systems with loads,
parallel AC lines, and multi-infeed HVDC systems are included. In chapter 5
further considerations for multi-infeed systems are discussed.
The methods dealt with in chapter 2 are based on quasi steady-state consid-
erations. The virtue is that very powerful computational tools can be developed,
which enable very fast analysis of different system conditions as well as analysis
of the influence of different design and system parameters. However, the dy-
namics of the system is not explicitly modelled and the analysis based on quasi
steady-state methods must be extended with time simulations for a complete
analysis of system. In chapter 3 comparisons between the static analytical meth-
ods and full detailed time simulations are given. The conclusion is that static
methods, i.e. the ones based on MPC and VSF, can be used for determining the
limits of voltage/power stability, but for determining the dynamic properties a
full dynamic model of the system is required.
In chapter 4 the influence of loads, parallel AC lines, etc, on the stability
limits is studied. It is also demonstrated that the stability limits of the Capacitor
Commutated Converters (CCCs) are significantly higher as compared with those
of the standard line commutated converters. (VSC converters were not included
in this analysis, see below.)
In chapter 6 a discussion how the developed tools can be applied to real
planning and design problems is given. Particularly, the interplay between the
use of the static and dynamic analysis tools are exemplified. The Nelson River
HVDC scheme is the system mostly considered in this chapter.
The analysis methods described in this report can be used for all converter
types. However, in this report the methods have only been applied to the con-
ventional line commutated converters, and Capacitor Commutated Converters
(CCCs), but not to systems with Voltage Source Converters (VSCs). The ex-
tension to this latter kind of HVDC systems should be straightforward.

ix
1
Basic Theory and Definitions

1.1 Introduction
Modern power systems are becoming increasingly complex and large. To be able
to predict the behaviour of these systems under different operating conditions
engineers are more and more dependent on sophisticated software. Since the
models of the power system and the software used in analyses are getting more
and more complex the requirements on the engineers involved in these studies
get higher. Even if the analytical tools are much more powerful today, it is still
of utmost importance that the engineers have a thorough understanding of the
basic physical processes and interactions in the system. Without this knowledge
it is not possible to understand phenomena in complex systems, and to design
appropriate control and protection systems that can alleviate possible adverse
interactions.
Another important prerequisite for the establishment of a good understand-
ing of a phenomenon is that commonly accepted and appropriate definitions
exist. This will facilitate the discussion between experts and contribute to the
development of the knowledge and understanding. It is not uncommon that
disagreements between experts on an issue are due to lack of accepted defini-
tions. The importance of clear definitions, that are commonly accepted, should
be obvious.
In this chapter a basic explanation of the power and voltage instability phe-
nomena is given. This is done by introducing a simple model of an AC/DC
system, which exhibits the basic interactions that are dominating also in real
size power systems. Definitions of power and voltage stability are also given.
These definitions are based on definitions of voltage stability for AC systems.
The chapter is organised as follows. First a brief physical explanation of the
power and voltage instability in AC/DC systems is given. Then the definitions
of voltage stability in AC systems are reviewed and discussed. A simple AC/DC
system model is then introduced and definitions of voltage and power stability
in AC/DC systems are proposed.

1.2 Simplified Explanation of Power/Voltage Insta-

bility Phenomenon
A power/voltage instability in a conventional line-commutated HVDC system
can basically arise out of an inter-relationship between the AC voltage, DC
controls and the DC current. It manifests itself in “weak” AC systems because
it depends almost entirely on the “stiffness” of the AC voltage. In simple terms,
the basic mechanism of the instability occurs as follows.
Suppose the AC voltage drops just a small amount; then the DC voltage
at the inverter drops proportionately (they are directly related). But the main

1
2 1. Basic Theory and Definitions

function of the DC controls in many DC schemes is to hold constant power order

(Pd = Ud · Id ). If the DC voltage (Ud ) goes down, the only way the controls can
maintain constant power is to increase the DC current (Id ). However, as soon
as it does this, the reactive power consumption of the DC increases because this
is directly related to the DC current. Unless this additional reactive power is
immediately available, the AC voltage will dip further and start the cycle all
over again. In other words, an unstable cycle will be started and perpetuated,
which unless it is stopped in some way can lead to a runaway condition and a
total collapse of the entire AC/DC

1.3 Definitions of Voltage Stability in AC Power

Systems
During the late eighties substantial work was done within the power industry
to develop methods and tools to analyse and predict voltage stability in power
systems. A number of working groups within CIGRE and IEEE produced a large
number of documents on the topic, and some of the work done is summarised
in refs. [1] and [2]. These references contain the definitions that are generally
accepted in the industry today. It is of interest to note that the definitions in
these two reports are not identical. These definitions are reviewed below.

1.3.1 Definitions According to CIGRE

In ref. [1] the following definitions are given:

• A power system at a given operating point is small-disturbance voltage

stable if, following any small disturbance, voltages near loads are identical
or close to the pre-disturbance values.

• A power system at a given operating state and subject to a given distur-

bance is voltage stable if voltages near loads approach post-disturbance
equilibrium values. The disturbed state is within the region of attraction
of the stable post-disturbance equilibrium.

• A power system undergoes voltage collapse if the post-disturbance equi-

librium voltages are below acceptable limits.

As seen these definitions are quite general and refer to how power system volt-
ages react to disturbances in the system.

1.3.2 Definitions According to IEEE

Another set of definitions is given in ref. [2]:

• Voltage stability is the ability of a system to maintain voltage so that when

load admittance is increased, load power will increase, and so that both
power and voltage are controllable.

• Voltage collapse is the process by which voltage instability leads to a loss

of voltage in a significant part of the system.
1.3. Definitions of Voltage Stability in AC Power Systems 3

Ze U l ∠φ
Infinite bus

U N ∠0 Load
P l + jQ l

Figure 1.1. Simple AC power system for analysis of voltage stability.

• Voltage security is the ability of a system, not only to operate stably,

but also to remain stable (as far as the maintenance of system voltage
is concerned) following any reasonable credible contingency or adverse
system change.

• A system enters voltage instability when a disturbance, increase in load,

or system change causes voltage to drop quickly or drift downward, and
operators and automatic system controls fail to halt the decay. The voltage
decay may take just a few seconds or ten to twenty minutes. If the decay
continues unabated, steady state angular instability or voltage collapse
will occur.

This set of definitions are obviously more restrictive than those of the previous
subsection. This will be illustrated in subsection 1.3.4.

1.3.3 Other Definitions

In the literature there exist many more definitions, most of them just variations
of the ones given above. It will not be elaborated on those here, but it might
be worthwhile to mention another concept that has been used, i.e. the concept
of regularity. In ref. [3] regularity is defined as:

• A power system is said to be voltage regular at a given operating point if

an injection of reactive power into a bus does not result in a decrease of
voltage magnitudes in the system.

The rationale for this definition is obvious since voltage control is mostly ac-
complished by controlling reactive power from controllable sources, such as syn-
chronous machines, reactive shunt elements, SVCs, etc. The authors of ref. [3]
suggest that in addition to the requirements given in the above definition, the
system needs also to be voltage regular in order to be voltage stable.

1.3.4 Discussion of CIGRE and IEEE Definitions

The Nose Curve of a Simple Test System
A simple system often used in voltage stability studies is the one Figure 1.1. It
is assumed that the system quantities could be adequately modelled by phasors
for the analysis done here.
4 1. Basic Theory and Definitions

Nose Curve of Simple Power System

1.2

A
1

0.8

B
Load Voltage

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Active Load Power

Figure 1.2. Nose curve of simple AC power system. Power factor of load
=1. Impedance of line = 0.4 p.u., impedance angle = 90◦ . Voltage of
infinite bus = 1.0 p.u.

If the voltage of the infinite bus is kept constant, both magnitude and phase,
and the active load demand Pl is varied, the so-called nose curve is obtained.
This curve shows how the load voltage Ul varies as Pl is changed. It is assumed
that for a given active load there corresponds a unique reactive load, for instance
by assuming a constant power factor. A typical example of a nose curve is shown
in Figure 1.2.
The curve in Figure 1.2 shows that for each load power level below the power
corresponding to the point B, there exist two solutions: One with “high” voltage
and one with “low” voltage. At the point B these two solutions coalesce, and
above this power level no solutions exist with the given data of the system.
The point corresponding to B is often referred to as the Point of Maximum
Loadability (PML).
Physically the curve in Figure 1.2 can be constructed in the following way.
At the load bus a shunt admittance (impedance) is connected, and the value
of this admittance is increased from zero, corresponding to point A, to infinity,
corresponding to point C. The ratio between the active and reactive parts of the
admittance should be kept constant to a value determined by the power factor
of the load. For any given value of the load admittance, the load voltage can
be calculated and the load power determined, which means that to any given
value of the admittance there exists a unique point on the nose curve.

Definitions Applied to Simple Test System

The definitions given above will now be applied to the system in Figure 1.1.
It is clear that with the definition of IEEE, subsection 1.3.2, operation on the
nose curve between points B and C will be unstable by definition, since in this
region load power will decrease when load admittance is increased as explained
above. Thus only operation on the upper part of the nose curve can be stable
according to the definition of IEEE.
1.3. Definitions of Voltage Stability in AC Power Systems 5

kp Ustab
≥1 0
0.75 0.45
0.5 0.58
0 0.71 (“B”)

Table 1.1. Regions of voltage regularity for different active load voltage
dependencies, kp Operating points on the nose curve of Figure 1.2 with
voltages greater than Ustab are voltage regular, i.e. V SF > 0, those with
voltages less than Ustab are not voltage regular. The transition point for
constant power loads, i.e. loads with kp = 0 corresponds to the point B
in Figure 1.2.

If the definitions according to Cigré, subsection 1.3.1, are to be applied, a

few considerations have to be made. The voltage control of all power systems is
based on the assumption that injection of reactive power increases the voltage
magnitude at the injection bus and the adjacent buses. Injection of reactive
power could be done by connection of shunt capacitors, disconnection of shunt
reactors, raising the excitation of synchronous machines, etc. This principle
could be expressed as that the quantity called the Voltage Sensitivity Factor
(V SF ) is positive, or mathematically

∆U
V SF = >0 (1.1)
∆Q

As seen this means that the system is voltage regular according to the definition
of ref. [3], and a necessary condition for stability is thus voltage regularity. It is
straightforward to calculate the VSF for the simple system above. To simplify
the calculations it is assumed that the load is purely active and it has voltage
dependence according to

U kp
P = P0 (1.2)
U0
With these assumptions it can be shown that regions of the nose curve where
VSF is positive depends on the exponent kp . It turns out that the nose curve
is divided into two regions, one where VSF is positive and one where VSF is
negative. At the transition point VSF goes to infinity. If it is assumed that the
line impedance Ze is lossless and that the voltage of the infinite bus is kept at
1 p.u., the stability regions for some exponents kp are given in Table 1.1 can be
summarized as:
For constant power loads, i.e. kp = 0, only operating points on the upper
part of the nose curve are voltage regular and the transition takes place at the
point B of the nose curve. If the voltage dependence is as constant current load
or with a higher exponent, i.e. kp ≥ 1, all points on the nose curve are voltage
regular. For values of kp in between, the border between voltage regularity and
irregularity lies on the lower part of the nose curve as indicated in Table 1.1.
Negative values of kp are deemed unphysical and not considered.
It should be noted that in principle a voltage control strategy based on that
reactive power injection decreases the voltage could be used in the regions found
voltage irregular above. However, such a voltage control scheme had to detect
6 1. Basic Theory and Definitions

-Pd + jQd
Id
Us 0
Z θ Ut δ Ud
p

τ:1
Qc

Figure 1.3. Simple AC/DC system for analysis of voltage/power stabil-

ity. The quantity p in the figure includes all parameters that are needed
to specify the HVDC system, e.g. commutation reactance, commutation
margin, etc.

the sign of VSF before acting, and such a controller is deemed to be highly
hypothetical and not to be considered in a practical power system.
Furthermore, one could here imagine that the load characteristics are such
that the initial voltage dependence ensures stability, but slower control actions
are such that the operating point moves away from the initial one, leading
to either too high voltages in the system or to a voltage collapse. Therefore,
operation on the lower part of the nose curve might be tolerable for short period
of times, but to guarantee stability, normal (steady state) operation must be on
the upper part of the nose curve.

Conclusions
Two main conclusions can be drawn from the simple example analysed above:

• According to the IEEE definition, operation on the lower part of the nose
curve is unstable.

• According to the definition of Cigré the stability depends on the voltage

characteristics of the load and for certain disturbances operation might
be stable on the lower part of the nose curve. However, for secure stable
operation the system must be on the upper part of the nose curve.

These two conclusions will be very important when we discuss stability for
AC/DC systems in the next section.

1.4 Analysis of AC/DC Systems

For analysis of the voltage/power stability1 of an AC/DC system a test system
similar to the one in Figure 1.1 has been introduced, see e.g. ref. [4]. This
system is depicted in Figure 1.3.
The system in Figure 1.3 is believed to be a good model of a real AC/DC
system in the time frame where the DC current controls of the HVDC link have
had time to respond, but the controls of the AC system have not yet responded.
It is thus assumed that no “fast” voltage controls such as SVCs are present in
the AC system, but the voltage control is basically carried out by the AVRs of
1A definition will be proposed in the next section.
1.4. Analysis of AC/DC Systems 7

the synchronous machines. It is further assumed that loads could be modelled

by impedances. These assumptions are discussed more in detail in ref. [4], where
a description on how to model synchronous machines also is included. In the
subsequent chapters the influences of these assumptions are further analyzed
and discussed. From Figure 1.3 a few basic definitions are recalled:

• Short Circuit Ratio, SCR, is the per unit admittance of the impedance
Z∠θ in Figure 1.3. The per unit value is based on the nominal MW rating
of the HVDC converters and the nominal AC voltage of the converter AC
bus.

• Effective Short Circuit Ratio, ESCR, is SCR + Yc , where Yc is the per

unit value of the admittance of the shunt reactive power compensation Qc
in Figure 1.3.

From the definitions of SCR and ESCR it is clear the these quantities are com-
plex numbers. However, in most cases the angle θ is close to 90◦ . Furthermore,
as shown in ref. [4], system performance is rather insensitive to the angle θ as
long as its value is close to 90◦ . Therefore, if not explicitly stated, it is assumed
that θ = 90◦ , and only the magnitude of SCR is given. In this case ESCR can
be written as ESCR = SCR − Qc .
The discussion in the following will be based on the model system of Fig-
ure 1.3, and the assumptions briefly reviewed above are adopted.

1.4.1 Maximum Power Curves

A very central concept in analysis of the behaviour of AC/DC systems in the
time scale indicated above is the Maximum Power Curve (MPC). This curve
shows many similarities with the nose curve of the simple AC system of Fig-
ure 1.1, and in fact the MPC can be converted into a nose curve with the load
being an HVDC converter, as will be shown later. A fairly detailed description
of how an MPC for an HVDC inverter is derived will be given below.
Firstly, define a pre-disturbance steady state operation point of the system.
This could in principle be any operating point within the rating of the equipment
and operating limits of the system, but for simplicity it is assumed that the
pre-disturbance operating point corresponds to nominal conditions, i.e. Ut =
1.0 p.u., Id = 1.0 p.u., Pd = 1.0 p.u. Other conditions of the HVDC, e.g.
γ (commutation margin) and Ud , are also at nominal values, and the reactive
compensation Qc is also at a given (nominal) value. By solving the load flow and
HVDC equations, this would uniquely determine the values of the transformer
tap ration, τ , magnitude of the source voltage, Us , and the phase angle of the
converter bus voltage, δ 2 . All these values determine then the initial steady
state operating point of the system.
The MPC of an inverter in constant commutation margin, γ, operation3
is now derived as follows. The values of Us , τ , Qc , and γ are kept constant,
2 We have here set the phase angle of the source voltage to zero, which means that this

voltage phasor is the phase angle reference of the system.

3 One could in principle allow other modes of operation as well. The only requirement

being that when the DC current is varied that the controls uniquely determine the states of
the system so that the load flow equations could be solved. Other control modes that are
conceivable are constant firing angle control, constant DC voltage control, etc. see section 1.6
8 1. Basic Theory and Definitions

MAXIMUM POWER CURVE

1.4

1.2 A.C.−Voltage MAP

DC POWER & AC−VOLTAGE 1

0.8

D.C.−Power
0.6

0.4

SCR = 3.5

0.2

0
0 0.5 1 1.5 2 2.5
DIRECT CURRENT

Figure 1.4. Maximum Power Curve (MPC) of the system in Figure 1.3.
SCR = 3.5, γ = 17◦ . Initial operating point: Id = 1 p.u. and Ut = 1 p.u.

“frozen”, while the DC current, Id , is varied. The variation of Id is assumed to

be so “slow” that the new state of the system could be determined by solving
the resulting load flow equations of the system in Figure 1.3. The active and
reactive power of the DC system are determined by the steady state equations
of the converter. By varying Id the magnitude of AC voltage of the converter
bus, Ut , the phase angle of this voltage, δ, the active and reactive power of the
HVDC system, Pd and Qd , will vary. A typical example of how the DC power
and terminal voltage will vary with the variations of the DC current with the
assumptions above is given in Figure 1.4.
The solid curve in Figure 1.4, i.e. the MPC, has similarities with the nose
curve of Figure 1.2. The Maximum Available Power (M AP ) is defined as the
maximum point of MPC, and this points indicates thus the theoretically maxi-
mum DC power that can be injected into the system by keeping all “controllable”
quantities in Figure 1.3 constant, varying the DC current only. The correspond-
ing DC current is called IMAP . (The variables Ut and δ will of course vary as
a consequence of the variation of Id , but these are not controllable as e.g. Us
and τ .) M AP corresponds thus to PML of the nose curve in the AC system
case. In a real system, M AP may not be feasible since the corresponding DC
current might be too large, or the corresponding low AC voltage might cause
e.g. protections to act.
The shape of the MPC depends very much of the SCR, i.e. of Z, of Fig-
ure 1.3. If the SCR now is lowered to 1.5 an MPC according to Figure 1.5 is
obtained. An interesting finding in Figure 1.5 is that the M AP is to the left
of the initial operating point. Consequently, an increase in DC current at the
nominal operating point implies a decrease in the DC power, and vice versa.
This is contrary to the objective of DC power control, and therefore this op-
erating situation should be avoided if normal power scheduling should prevail.
Otherwise, an increase of DC current order, as an attempt to increase DC power
will be unsuccessful.
1.4. Analysis of AC/DC Systems 9

MAXIMUM POWER CURVE

2

1.8

1.6

1.4
DC POWER & AC−VOLTAGE

1.2

0.8

0.6

0.4

0.2

0
0 0.5 1 1.5
DIRECT CURRENT

Figure 1.5. Maximum Power Curve (MPC) of the system in Figure 1.3.
SCR = 1.5, γ = 17◦ . Initial operating point: Id = 1 p.u. and Ut = 1 p.u.

Nose Curve for AC/DC System

1.3

1.2

1.1

Initial operating point

1
AC−Voltage

0.9

MAP
0.8

0.7

0.6

0.5

0.4
0 0.2 0.4 0.6 0.8 1 1.2 1.4
DC Power

Figure 1.6. Nose curve for the AC/DC system in Figure 1.3, SCR = 3.5.

As indicated above the nose curves of the pure AC system and the MPCs
exhibit great similarities. To verify this, curves showing the variation of AC
voltage as function of the DC power are plotted in Figure 1.6 and Figure 1.7
for the MPCs in Figure 1.4 and Figure 1.5, respectively.

1.4.2 Voltage Sensitivity Factors

The V SF could also be calculated in the nominal operating points of Figure 1.4
and Figure 1.5. When doing this the MPC per se can only give limited infor-
mation. The MPC depicts the behaviour when only the DC current is varied,
10 1. Basic Theory and Definitions

Nose Curve for AC/DC System

2

1.8

1.6

1.4
AC−Voltage

1.2 MAP

1 Initial operating point

0.8

0.6

0.4

0.2
0 0.2 0.4 0.6 0.8 1 1.2 1.4
DC Power

Figure 1.7. Nose curve for the AC/DC system in Figure 1.3, SCR = 1.5.

SCR Control Mode V SF

3.5 C.C. +
3.5 C.P. +
1.5 C.C. +
1.5 C.P. -

Table 1.2. Sign of Voltage Sensitivity Factors of the systems in Figure 1.4
and Figure 1.5 at the nominal operating point. C.C. = Constant DC
Current Control. C.P. = Constant DC Power Control.

but when calculating the V SF other quantities than the DC current could be
varied. When calculating V SF , the control mode of the HVDC is crucial to
the result. For the systems of Figures 1.4 and 1.5 the control modes constant
power control and constant DC current control will be considered. Other con-
trol modes can also be analysed, but for the discussion here it is sufficient to
consider these two modes. A simple analysis gives the results of Table 1.2.
As seen from Table 1.2 all cases are voltage regular except the last one, which
corresponds to constant DC power control in the weak AC system case. It is
thus seen that for the weak system (SCR = 1.5), constant power control implies
a negative V SF , while constant DC current control implies positive V SF .

1.5 Definitions for AC/DC Systems

When comparing the analyses of the nose curve and the MPC one can conclude
that DC current plays the same role for the AC/DC system, Figure1.3, as load
admittance does in the pure AC system, Figure1.1, see subsection 1.3.2. The
roles of load characteristics in the AC case and control mode of the HVDC in
the DC case are also similar. Therefore it is motivated that the definitions of
voltage stability given in section 1.3 be modified accordingly to cover AC/DC
1.6. Discussion 11

systems. From the discussion above the following two definitions are proposed
for an AC/DC system:

Definition of Power Stability:

Power Stability of an AC/DC system is the ability, for a given operating condi-
tion and DC control mode, to increase DC power by an incremental increase in
DC current.

Definition of Small-Disturbance Voltage Stability:

An AC/DC system at a given operating state and DC control mode, and subject
to a small disturbance, is said to be small-disturbance voltage stable if AC voltage
settles to equilibrium close to pre-disturbance value.

Definition of Voltage Stability:

An AC/DC system at a given operating state and DC control mode, and subject
to a given disturbance, is said to be voltage stable for this particular DC control
mode and disturbance, if AC voltage and other system states settle to values
that are viable for system operation.

1.6 Discussion
It is clear that operation to the left of MAP is stable according to the definitions
above.
From the proposed definitions it is obvious that operation to the right of the
MAP will not be power stable. When it comes to voltage stability it is clear
that in constant DC power control operation to the right of MAP is unstable4 .
A control action to increase the voltage magnitude, e.g. by connecting a shunt
capacitor, will result in a decrease in voltage, see Table 1.2. (It is here assumed
that the constant power control loop can control the DC power to the desired
value.) This will in a real system lead to an unstable situation.
In constant DC current control and to the right of MAP, the system is
initially voltage stable since the V SF is positive. However, the system will
not transmit the desired amount of power and any attempt to adjust the DC
current with a standard DC power controller will lead to unstable conditions.
Therefore, the system can operate transiently to the right of MAP, but in order
to use the HVDC link for power control it must be brought back to the left side
of the MAP.
The definition of voltage stability involves arbitrary disturbances, such as
earth faults and trippings of lines. For such cases analyses based on MPCs are
not sufficient, since the large signal dynamics of the system plays an important
role. The restart strategies of the HVDC after faults are very important when
analysing (large disturbance) stability of the interconnected AC/DC system.
In the examples above it has been assumed that the inverter is in constant
γ control. However, as indicated above MPCs can be computed for other types
of HVDC controls such as constant DC voltage control or other types of γ
4 It is here assumed that the DC power controller has a positive gain in its main loop, i.e.

a decrease in DC power will result in an increase in the DC current order.

12 1. Basic Theory and Definitions

MAXIMUM POWER CURVE

1.4

1.2

0.8
DC POWER

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
DIRECT CURRENT

Figure 1.8. MPC for an AC/DC system with constant DC voltage control.

control. The MPC for such a control mode will look different, but there will be
a M AP point for such modes also. The M AP will occur for the point where γ
has reached its minimum permissible value, γmin . The MPC to the right of this
point would coincide with an MPC with constant γ = γmin . The same reasoning
for these types of control as for the constant γ control applies, which means that
only steady state operation to the left of M AP is stable in a practical system.
An example of an MPC with constant DC voltage control is given in Figure 1.8.
To distinguish between MPCs plotted for different control modes, an index
indicating which DC controls that are used when calculating can be used. The
standard MPC with constant γ control should thus be denoted MPCγ and the
MPC in Figure 1.8 should be MPCUd .
2
Analytical Tools and Methods

2.1 Introduction

The basic concepts of power/voltage instability associated with weak AC/DC

interconnections have been discussed in the previous chapter. A number of
analytical tools based on these concepts, in the form of system indices, have
been proposed to facilitate qualitative appraisal of power/voltage stability [4],
[5],[6],[7],[8]. These are the Maximum Available Power (M AP ), Voltage Sen-
sitivity Factor (V SF ), and Control Sensitivity Index (CSI). Essentially these
indices are all sensitivities of system states to incremental changes in control-
ling system quantities. In this chapter, the basic formulation and application of
these analytical tools are treated.
These indices have invariably been derived based on a simplified AC/DC
system configuration as shown in Figure 2.1, commonly known as a single-
infeed configuration. However, recently emerging AC/DC system configurations
around the world have necessiated other appropriate simplified system models
that would adequately represent these new situations. In particular are system
configurations with multiple HVDC links terminating in close proximity and
point-to-point HVDC links having parallel AC lines, commonly referred to as
multi-infeed and single-infeed with parallel AC line configuration, respectively.
In this chapter, the analytical tools developed with respect to the classical single-
infeed configuration would be extended to these higher-order system models as
shown in Figure 2.8 and Figure 2.9. Further, the simplified system model of
Figure 2.1 would also be enhanced by incorporation of local system components
such as loads, compensators, etc., and the basic equations derived. With the
analytical methods applied to the various system models, there exists a relation-
ship among them. This chapter also establishes this overall relationship among
the power and voltage stability methods for the various system models.

E ∠ψ U ∠δ P d, – Q d

τ:1 Ud

z ∠θ
Xc γ, µ
Id
bc

Figure 2.1. Simplified single-infeed HVDC configuration.

13
14 2. Analytical Tools and Methods

2.2 Power Stability

2.2.1 Maximum Available Power
The concept of Maximum Available Power (M AP ) was first introduced in [5]
and [6] to the system model of Figure 2.1 with the converter operating in Con-
stant Extinction Angle (CEA) control mode. This maximum available power
is defined as the maximum of the Maximum Power Curve (MPC) which is the
converter DC power Pd as a function of the direct current Id . Mathematically,
this operating point corresponds to the condition that

dPd
=0 (2.1)
dId

Equation (2.1) is known as the M AP condition. When this condition occurs,

the prevailing DC power is the maximum the converter can deliver to the AC
system at the corresponding direct current IMAP . Beyond IMAP , the DC power
delivered actually decreases with further increase in direct current. This is due to
the larger percentage decrease in the converter AC bus voltage as compared with
the increase in direct current, resulting in a net decrease in the DC power. Such
a phenomenon corresponds with unstable system behaviour, thus the M AP
condition determines the power stability limit of the AC/DC interconnection.
The index dPd /dId is the gradient of the MPC and can be derived from the power
flow equations for the converter AC and DC buses in Figure 2.1. By linearizing
these equations with respect to the system states Id , U , δ, and eliminating U ,
δ, dPd /dId may be derived as
   
dPd rU cos(γ + µ) U 4 ESCR2 + b U 2 ESCR + Qd − Q2d − Pd2
= (2.2)
dId det JI

where
1
ESCR = − bc
z
2Ud  
r= b = 2 Qd − Pd tan(γ + µ)
cos γ + cos(γ + µ)
  
4 2 2 ∂Qd   2  ∂Pd 
det JI = U ESCR − Qd + U U ESCR + Qd − Pd Pd − U
∂U Id ∂U Id

and ∂Pd /∂UI |Id and ∂Qd /∂U |Id are the voltage dependence of the converter
active and reactive power, respectively, for the constant current control mode
(see eq. (2.9)). From eq. (2.2), the M AP condition thus occurs when
 
U 4 ESCR2 + b U 2 ESCR + Qd − Q2d − Pd2 = 0 (2.3)

If the system operating point corresponds with the M AP condition under nom-
inal conditions, then the prevailing ESCR is said to be critical, CESCR.
Eq. (2.3) then solves to

 
1 b b 2 2
CESCR = 2 − + − Qd + Pd (2.4)
U 2 2
2.2. Power Stability 15

ESCR=
2.1
1.8
Pd [p.u] 1.48

1.2

IMAP

Id [p.u]

Figure 2.2. Maximum Power Curves.

This reduces further to

1 π γ + µ 
CESCR = − Q d + Pd cot + (2.5)
U2 4 2

In eqs. (2.2) - (2.5) the Thevenin impedance is assumed to be purely reactive so

as to simplify the analysis, though this is not mandatory. The form of eq. (2.5)
with the Thevenin impedance having a resistive component can be found in [4].

2.2.2 Maximum Power Curve

The MPC is practically derived by computing the converter DC power Pd nu-
merically as a function of the direct current Id . This is done by first solving the
power flow equations with the system operating under nominal conditions, i.e.
U , Ud , Id as 1 p.u. etc. Subsequently the Thevenin AC voltage magnitude is
held fixed, when the power flow equations are solved to compute Pd for every
value of Id .

Example 1 Consider the single-infeed configuration of Figure 2.1 with the

following system parameters; γ = 17◦ , dx = 0.08 p.u., bc = 0.54 p.u., |z| = 0.576,
0.496, 0.428, 0.379 (ESCR = 1.2, 1.478, 1.8, 2.1, respectively). θ = 90◦ . The
corresponding MPC’s and converter AC bus voltages are shown in Figures 2.2
and 2.3, respectively.
For the given parameter values, it is seen from Figures 2.2 and 2.3 that the
M AP condition occurs under nominal conditions (IMAP , U , Pd are 1 p.u.) for
a CESCR of 1.4775. For ESCR values less than this, the system operates on
the right-side of the MPC maxima which implies unstable operation, e.g. the
MPC for an ESCR = 1.2 in Figure 2.2. For ESCR values greater than the
CESCR, the system will operate on the left-side of the MPC maxima which
implies stable operation, e.g. the MPC for an ESCR = 1.8 or 2.1 in Figure 2.2.
The relationship between the MPC and eq. (2.2) is illustrated in Figure 2.4. In
Figure 2.4, the zero of dPd /dId as given by eq. (2.2) coincides with the maxima
of the MPCs.
16 2. Analytical Tools and Methods

1.2
1.48
1.8
2.1
U [p.u] ESCR=

Id [p.u]

Figure 2.3. Converter AC bus voltage.

ESCR=1.4775
Pd

dPd

dI d

(equation 2.2)

Id [pu]

Figure 2.4. Relationship between eq. (2.2) and MPC.

2.3 Voltage Stability

2.3.1 Voltage Sensitivity
The Voltage Sensitivity Factor (V SF ) was first introduced in [7] for the single-
infeed configuration of Figure 2.11 . This is defined as the ratio of the incremental
change in the voltage ∆U and reactive power ∆Q at the converter AC bus, given
by
∆U
V SF = (2.6)
∆Q
The V SF is essentially the converter AC voltage sensitivity to the reactive
power incremental change at the converter AC bus, and is used as a stability
index. When the V SF is small and positive, a small change of reactive power
1 In [7] the quantity was called Voltage Stability Factor, but we will use the more general

term Voltage Sensitivity Factor.

2.4. Control Stability 17

induces a small and same directional change in the converter AC bus voltage.
This characteristic is consistent with stable system behaviour. An increasing
V SF corresponds to a decreasing voltage stability margin and when the V SF
becomes infinite, the transition to an unstable system occurs. When the V SF
is negative, a small change of the reactive power induces an opposite change
in the converter AC bus voltage. This characteristic corresponds with unstable
system behaviour.
The V SF can be derived mathematically from the same power flow equa-
tions used to formulate the gradient of the MPC in eq. (2.2). By the same
linearization of these equations with respect to U and δ, and eliminating δ, the
V SF is given by

∆U/U
V SF = =
∆Q
U 2 ESCR + Qd (2.7)

∂Qd
2  ∂Pd
U 4 ESCR2 − Q2d + U U ESCR + Qd + Pd Pd − U
∂U ∂U

where ∂Pd /∂U and ∂Qd /∂U are the voltage dependence of the converter active
and reactive power, respectively, and are given by the control mode as follows
CEA/Constant Power Control


∂Pd ∂Qd 2
=0; = Qd − Pd tan[γ + µ) (2.8)
∂U ∂U U

CEA/Constant Current Control

∂Pd 2 Pd cos γ  ∂Qd 2 Pd sin(γ + µ) 

= ; = Qd −
∂U U cos γ + cos(γ + µ) ∂U U cos γ + cos(γ + µ)
(2.9)

Constant DC Voltage/Constant Power Control

∂Pd ∂Qd 2
= 0; = Qd + Pd cot(2γ + µ)] (2.10)
∂U ∂U U

Example 2 Using the same parameter values as in Example 1 for the single-
infeed configuration of Figure 2.1, the V SF as a function of ESCR for different
HVDC control modes are shown in Figure 2.5. It is seen that HVDC systems
with constant DC voltage control have small V SF even for low ESCR, implying
that this control mode is inherently stable. In contrast, constant power control
is inherently unstable at low ESCR as shown by the negative V SF below the
CESCR.

2.4 Control Stability

2.4.1 Control Sensitivity Index
The concept of control sensitivity was introduced in [8], based on the single-
infeed configuration of Figure 2.1. This is defined as the rate of change of a
18 2. Analytical Tools and Methods

a
b

VSF
c

a. const. power
a b. const. current
c. const. DC voltage

1.4775

ESCR

Figure 2.5. V SF for various control modes.

γ
Kp
H
+
- +
H + α
KI
s L
γref
L

Figure 2.6. Inverter Constant Extinction Angle controller.

controlled parameter with respect to the controlling parameter. To illustrate,

consider an example of the inverter CEA controller shown in Figure 2.6. The
extinction angle γ (controlled parameter) is required to be controlled to specified
value by directly varying the inverter firing angle α (controlling parameter). A
suitable ratio dγ/dα may then be defined as a stability index for the controller.
Stable controller action requires that an increase in α causes the system to
respond with a decrease in γ, i.e. dγ/dα < 0. Thus a negative dγ/dα indicates
stable system characteristics with respect to the controller action, and vice
versa. For other control strategies, similar ratios may be defined. In general, a
Control Sensitivity Index (CSI) [8] commensurating with the control principle
may be defined as the stability index, those for typical controllers are as given
in Table 2.1.

2.4.2 CSI Computation

The CSI may be computed from the power flow Jacobian at the steady-state
operating point under investigation. To this end, the system steady-state op-
erating point is first obtained by a Newton-Raphson power flow solution. The
Jacobian which is derived from linearization of the power flow equations with
respect to the operating state vector, is then computed at this operating point.
2.5. Extension to Higher-Order Models 19

Rectifier Inverter CSI

const. current CEA dγ/dα
const. power CEA dPd /dId
const. current cons. voltage dU/dα

Table 2.1. Common CSI definitions.

The power flow equations comprise AC power balance equations at the converter
AC bus, and converter steady-state DC power, current, and voltage equations.
This set of six equations can be found in [8].
The operating state vector comprises system states which will vary dur-
ing dynamic operation of the system. Its composition depends on the control
mode under consideration but includes the controlled and controlling system
states. For the system model of Figure 2.1, the operating state vector vxy is
(α, φ, δ, Ut , Ud , Id , Pd ) for constant power (constant γ), (α, γ, φ, δ, Ut , Ud , Pd ) for
CEA (constant current) and constant DC voltage (constant current), control.
α and φ are the inverter firing angle and power factor at inverter AC bus,
respectively, and other symbols are self-explanatory from Figure 2.1.
Thus the Jacobian, say Jxy , computed at the operating point (power flow
solution) is given by
Jxy · ∆vxy = 0 (2.11)
Rewriting eq. (2.11) gives

Jx ∆vx + Jy ∆vy = 0 (2.12)

where vxy = (vx , vy ), Jx is a 6x6 square matrix comprising columns of Jxy

related to the vx system states and Jy is a single column of Jxy corresponding
to the single controlling system state vy . Thus from eq. (2.12)
∆vx
= −Jx−1 Jy (2.13)
∆vy
One of the elements of the vector in eq. (2.13) then defines the CSI as according
to Table 2.1, for the control mode under consideration. For example for the
CEA control mode, the first element of ∆vx /∆vy defines the CSI, i.e. dγ/dα.

Example 3 Consider the single-infeed configuration of Figure 2.1 with the

following system parameter values taken from [8]; Xc = 0.18 p.u., SCR = 1.01,
bc = 0.567 p.u., γmin = 15◦ , α = 164.7◦ (= 180◦ − γmin − µ). The computed
CSI for the constant power control modes as in Table 2.1 is shown in Figure 2.7.

2.5 Extension to Higher-Order Models

Recent works have extended the concepts of maximum power and voltage sen-
sitivity to multi-infeed and single-infeed with parallel AC line configurations as
shown in Figure 2.8 and Figure 2.9, see refs. [9], [12], [14]. In this section the
basic power and voltage stability analysis methods incorporating these concepts
are developed for these two system configurations.
20 2. Analytical Tools and Methods

dP d
d Id

ESCR

Figure 2.7. Control Sensitivity Indices.

P , –Q
E 1 ∠ψ 1 d1 d1
U 1 ∠δ 1 U d1

z 1 ∠θ 1 d x1 γ 1, µ 1
1 Id1
P12, Q12 b c1

z ∠θ
12 12 P d2, – Q d2

z ∠θ U d2
2 2
2 d x2 γ 2, µ2 Id2
E 2 ∠ψ 2 U 2 ∠δ 2 b c2

Figure 2.8. Simplified multi-infeed HVDC configuration.

U 1 ∠δ 1 1 P12, Q12 2 U 2 ∠δ 2
z 12 ∠θ 12

P d1, – Q d1 P d2, Q d2 E 2 ∠ψ 2
E 1 ∠ψ 1 Id

U U d2
z 1 ∠θ 1 γ 1, µ 1, α 1d1 µ 2, α 2 z 2 ∠θ 2

b c2 b c1

Figure 2.9. Single-infeed with parallel AC line configuration.

2.5.1 Maximum Power

Similar to the single-infeed case, the power stability methods for these higher-
order system configurations are derived from their linearized power flow equa-
tions. These equations also assume constant Thevenin voltages, thus the source
2.5. Extension to Higher-Order Models 21

buses are eliminated in the power flow solution model, given by

    
∆Pd JdI 0 JdU ∆Id
 ∆Pac  =  JP I JP δ JP U   ∆δ  (2.14)
∆Qac JQI JQδ JQU ∆U/U

where;
∆Pd is a vector of incremental DC power at the converter DC buses.

∆Pac is vector of incremental AC active power at the converter AC buses.

∆Qac is a vector of incremental AC reactive power at the converter AC buses.

∆Id , ∆δ, and ∆U/U are vectors of incremental change in the system states
comprising the DC current, converter AC bus voltage angle and relative
magnitude, respectively.

The Jacobian submatrices are the partial derivatives of the power flow equa-
tions with respect to the system states. The subscript labels d, ac denote the
quantities related to the DC power, AC active and reactive power, respectively.
Note that Id is additionally chosen as a system state so that the Jacobian is an
extended form of the well known Newton-Raphson model. The expressions for
these Jacobian submatrices are given in [17].
For no active or reactive power injections at the converter AC buses, ∆Pac
,∆Qac may be assumed to be zero. Thus eq. (2.14) may be reduced to

∆Pd = JMP C ∆Id (2.15)

where
−1
JMP C = JdI − JdU JR1 JR2 , JR1 = JQU − JQδ JP−1
δ JP U

and
JR2 = JQI − JQδ JP−1
δ JP I

Standard MPC (SMPC) approach

In eq. (2.15), the diagonal and off-diagonal elements of JMP C give the intra
and inter ∆Pd − ∆Id relationship, respectively, of the constituent converters in
Figures 2.8 and 2.9. The maximum power concept for the single-infeed case is
directly extended to a referenced constituent converter by using only its intra
∆Pd − ∆Id relationship and constraining the other non-referenced constituent
converters to be invariant. From eq. (2.15), this is equivalently given by

∆Pdi 
= JMP Cii i = 1, 2 j = i (2.16)
∆Idi ∆Id =0
j

where JMP Cii is the i-th diagonal element of JMP C . When JMP Cii vanishes,
eq. (2.16) is akin to eq. (2.1) for the single-infeed case. Thus JMP Cii is used to
determine the M AP condition for these higher- order configurations. Similar
to the single-infeed MPC, the SMPC for the higher-order configurations may
also be derived. This is done practically by numerically computing the DC
22 2. Analytical Tools and Methods

Side
P1 Front

Pd1 [pu.]
Plan
P2

IdN2 [pu.] Id1 [pu.]

Figure 2.10. Three dimensional view of the SMPC surface.

power of a referenced constituent converter as a function of its DC current while

constraining the DC current of the other non-referenced constituent converters
as constant parameters, in accordance with eq. (2.16). All voltage magnitudes of
Thevenin AC voltage sources are also assumed to be fixed at values determined
for nominal conditions, i.e. Id1 , Id2 , Ud1 , Ud2 , U1 , U2 as 1 p.u.

Example 4 To illustrate the derivation of the SMPC, consider the multi-infeed

configuration of Figure 2.8 with the following system parameters: ESCR1 = 1,
ESCR2 = 3, z12 = 0.6 p.u., Id1 is a variable and Id2 is a constant parameter
taking values from 0.1 to 1.6 p.u. Figure 2.10 shows the resulting three dimen-
sional SMPC surface which is twisted-saddle shaped. The surface is a composite
of many MPC’s, each of which is the MPC of the referenced converter with re-
spect to Id1 for a given value of Id2 . This is clearly shown in Figure 2.11 which
is the front projection of the SMPC surface of Figure 2.10. Thus, the M AP and
IMAP for every MPC may be determined as the operating point corresponding
to eq. (2.1) for each given value of Id2 . Figure 2.12 shows the plan projection
of the SMPC surface. The curve corresponds to the ridge of the SMPC surface,
i.e. the loci of the M AP of every MPC of the composite surface. This curve is
thus defined as the SMPC power stability boundary.

Modal MPC (MMPC) approach

The SMPC approach gives only a system-specific or local assessment of power
stability, by virtue of allowing only a single degree of freedom for DC current
variation. Another approach to investigate system power stability is to apply an
eigenvalue decomposition technique to JMP C . This is expected to give a more
system-wide or global assessment of system power stability since this technique
treats JMP C in its entirety and no constraints on the degrees of freedom for
DC current variation are imposed. Thus decomposing JMP C into its eigenvalue
equivalent, assuming this is always possible, gives

JMP C = ΨΛMP C η (2.17)

where
2.5. Extension to Higher-Order Models 23

P1 1.6 pu.
ESCR1=1, P2
ESCR2=3
z12=0.6 pu.
Pd1 [pu.]
IdN2=0.2 pu.

Id1 [pu.]

Figure 2.11. Front projection of the SMPC surface.

JMPCii=0 P1
Id1 [pu.]

unstable dP d1
= 0
d I d1

stable
P2

IdN2 [pu.]

Figure 2.12. Plan projection of the SMPC surface.

Ψ is the right column eigenvector matrix of JMP C .

η is the left column eigenvector matrix of JMP C .
ΛMP C is a diagonal matrix with i-th diagonal element as the i-th mode eigen-
value λi .
Applying eq. (2.17) to eq. (2.15), and noting that Ψ−1 = η gives

∆pd = ΛMP C ∆id (2.18)

where ∆pd = η∆Pd and ∆id = η∆Id are the vector of incremental modal DC
power and DC current, respectively. For the i-th eigenmode, eq. (2.18) can be
written as
∆pdi
= λi (2.19)
∆idi
where

nt
∆pdi = ηij ∆Pdj , i = 1, . . . , nt (2.20)
j=1
24 2. Analytical Tools and Methods


nt
∆idi = ηij ∆Idj , i = 1, . . . , nt (2.21)
j=1

and nt is the number of converter AC buses.

It is seen from eq. (2.19) that the sign of the eigenvalue determines the
direction of the incremental relationship. Positive eigenvalues cause positive
modal DC power changes in response to positive modal DC current variations,
and vice versa for negative eigenvalues. At the transition of signs, the minimum
eigenvalue becomes zero and eq. (2.19) becomes

∆pdi
=0 (2.22)
∆idi

This is akin to eq. (2.1) for the single-infeed case. However, this is a modal
relationship, thus corresponding to the MPC for the single-infeed configuration,
the Modal Maximum Power Curve (MMPC) may be defined using eq. (2.19).
The system operating conditions resulting in eq. (2.22) would thus represent the
Modal Maximum Available Power of the system. Alternatively, eq. (2.19) may
be expressed as
∆idi 1
= (2.23)
∆pdi λi
Again, positive eigenvalues cause positive modal DC current changes in response
to positive modal DC power order variations. This corresponds to stable DC
power control behaviour. For a zero minimum eigenvalue, an infinitesimally
small change in the modal DC power order causes an infinite change in the
modal DC current, hence instability of the DC power control. Therefore, for
stable DC power behaviour eq. (2.19) and eq. (2.23) imply

λmin > 0 ; λmin = min{λ1 , . . . , λnt } (2.24)

Eq. (2.19) may be used to derive graphical plots in parameter space to define
the MMPC stability boundary. In such case, the plots are values of parameters
for which the system minimum eigenvalue, λmin , becomes zero. Thus, the plots
are the MMPC stability boundaries that demarcate the parameter space into
stable and unstable regions.

Example 5 Consider the multi-infeed configuration of Figure 2.8 with the fol-
lowing system parameter values: ESCR1 = 1, ESCR2 = 3, z12 = 0.6 p.u. The
MMPC stability boundary in the PdN1 − PdN2 parameter space (the subscript
label N denotes rated quantities) is shown by the plot in Figure 2.13. In Fig-
ure 2.13, the region above and below the curve are thus unstable and stable
regions, respectively. The curve comprises values of the parameters for which
λmin becomes zero, thus it is the system MMPC stability boundary.

2.5.2 Voltage Sensitivity

Minimum Eigenvalue
The concept of voltage sensitivity is similarly extended to the higher-order con-
figurations, using the power flow solution model for the single-infeed case. For
2.5. Extension to Higher-Order Models 25

ESCR1=1, ESCR2=3, z12=0.6 pu.

unstable
PdN1 [pu.] λmin=0

stable

PdN2 [pu.]

Figure 2.13. MMPC stability boundary in Pd space.

the case of higher order configurations, it is given by

  
∆P JP δ JP U ∆δ
= (2.25)
∆Q JQδ JQU ∆U/U

where
∆P is a vector of incremental change in converter AC bus active powers.
∆Q is a vector of incremental change in converter AC bus reactive powers.
∆δ is a vector of incremental converter AC bus voltage angles.
∆U/U is a vector of incremental converter AC bus relative voltage magnitudes.
JP δ , JP U , JQδ , JQU are the partial derivatives of the power flow equations with
respect to the voltage angles and magnitudes.
If the active power is held constant at the converter AC buses and only
incremental reactive power is applied, then ∆P may be assumed to be zero, and
eq. (2.25) thus reduces to
∆U
∆Q = JR (2.26)
U
where JR = JQU − JQδ JP−1 δ JP U is the QU reduced Jacobian matrix. Alterna-
tively, eq. (2.26) can be expressed as
∆U −1
= JR ∆Q (2.27)
U
Though eq. (2.27) appears similar to the V SF in eq. (2.6), there is no direct
relationship for the converter AC bus in question, due to the coupling from
−1
the other non-zero off-diagonal elements of JR . To obtain a decoupled form,
an eigenvalue decomposition technique similar to the MMPC approach in sub-
−1
section 2.5.1 may be used. Thus applying the same decomposition to JR in
eq. (2.27) gives
∆qi
∆ui = (2.28)
λi
26 2. Analytical Tools and Methods

where

nt
∆Uj 
nt
∆ui = ηij ∆qi = ηij ∆Qj i = 1, . . . , nt
j=1
Uj j=1

∆ui is i-th mode of incremental modal voltage.

∆qi is i-th mode of incremental modal reactive power.
∆λi is i-th mode eigenvalue of the QU reduced Jacobian matrix, JR .
∆ηi is i-th row left eigenvector of the QU reduced Jacobian matrix, JR .
nt is the number of converter AC buses.
It is seen that there is now a direct ∆U/∆Q relationship in eq. (2.28), similar to
the V SF in eq. (2.6). However, the relationship is in modal form, thus eq. (2.28)
may be considered as the modal V SF (M V SF ), i.e. the multi-dimensional
equivalent of the single-infeed V SF .
The M V SF can be used as stability indicator, similar to the VSF. From
eq. (2.28), for positive λi the incremental modal voltage is “in phase” or changes
in the same direction with the incremental modal reactive power. This corre-
sponds with voltage stable situations. However, when λi is negative, the in-
cremental modal voltage is in “anti-phase” or changes in the opposite direction
with the incremental modal reactive power. This corresponds to voltage un-
stable situations. At the transition between the stable and unstable region of
operation, λi becomes zero meaning that any infinitesimal change in modal re-
active power causes infinite modal voltage change. Thus, the bus voltage would
also be correspondingly unstable since the incremental modal voltage is a linear
combination of the incremental bus voltages. Thus for a system to be voltage
stable, the minimum eigenvalue λmin must be positive at every operating point:

λmin > 0 λmin = min{λ1 , . . . , λnt } (2.29)

Example 6 Consider the multi-infeed configuration of Figure 2.8 with the fol-
lowing system parameters; PdN 1 = 1 p.u., PdN 2 = 0.5 p.u., z12 = 0.6 p.u.
Figure 2.14 shows the M V SF stability boundary in the ESCR1 − ESCR2
parameter space. For all ESCR values on the boundary, λmin is zero.

Participation Factors
For the modal approach, not only is the minimum eigenvalue useful in estimating
the proximity to voltage instability but the eigenvectors also contain information
concerning how critical system locations are. The eigenvectors may be used to
define a participation factor [11] given by

pij = Ψij ηji (2.30)

where
pij is the participation factor for the converter AC bus i in the j-th voltage
variation mode
∆ηji is the i-th element of the j-th row left eigenvector matrix.
2.6. Incorporation of Local Devices 27

PdN1=1 pu., PdN2=0.5pu., z12=0.6 pu.

stable
ESCR1

λmin=0

unstable

ESCR2

Figure 2.14. M V SF stablity boundary in ESCR space.

∆Ψij is the i-th element of the j-th column right eigenvector matrix.
nt is the number of converter AC buses.
Physically, Ψij is a measure of the activity of converter AC bus i in the j-th
voltage variation mode, ηji is the weighting of the contribution of this activity,
their product pij is a measure of the net participation of bus i in the j-th voltage
variation mode.
In the context of multi-infeed configurations, bus participation factors com-
puted from the eigenvectors associated with the minimum eigenvalue provide
information on the critical converter AC bus. The bus with the largest par-
ticipation factor is the critical bus. Consequently it is also the most effective
system location for implementation of remedial measures [13].

2.6 Incorporation of Local Devices

The equations developed in the preceding sections do not consider any system
component, e.g. loads, compensators, connected locally at the converter AC
bus. However, these devices could impact significantly on the power/voltage
stability of the AC/ DC interconnection [9], [15], [16].

2.6.1 Loads
In this subsection, the basic equations associated with power stability are pre-
sented for the various system configurations with static loads connected at the
converter AC buses as shown in Figures 2.15 - 2.17. It is assumed that the load
models have static load characteristics given by
P
= KP C + KP I U + KP Z U 2 + KP U U nP U (2.31)
PL
Q
= KQC + KQI U + KQZ U 2 + KQU U nQU (2.32)
QL
where
28 2. Analytical Tools and Methods

E ∠ψ U ∠δ P , –Q
d d
τ:1 U
d

z ∠θ
d
x
γ, µ
PL I
d
QL b
c

Figure 2.15. Simplified single-infeed configuration incorporating static load models.

U 1 ∠δ 1 1 2 U 2 ∠δ 2
z ∠θ
12 12
P d1, – Q d1 P d2, Q d2 E 2 ∠ψ 2
E 1 ∠ψ 1 Id

U U d2
z 1 ∠θ 1 γ 1, µ 1, α 1d1 µ 2, α 2 z 2 ∠θ 2
b c2 b c1
PL1 , QL1 PL2 , QL2

Figure 2.16. Single-infeed with parallel AC line configuration incorpo-

rating static load models.

KP I = 1 − (KP C + KP Z + KP U )

KQI = 1 − (KQC + KQZ + KQU )

KP C , KQC is the per unit of constant MVA active and reactive load, respec-
tively.

KP I , KQI is the per unit of constant current active and reactive load, respec-
tively.

KP Z , KQZ is the per unit of constant impedance active and reactive load,
respectively.

KP U , KQU is the per unit of voltage dependent active and reactive load,
respectively.

nP U , nQU is the voltage sensitivity exponent for active and reactive load,
respectively.

PL , QL is the nominal (initial) active and reactive load, respectively.

Using the same approach as in subsections 2.2.1 and 2.5.1, the power flow
equations incorporating the load models of eqs . (2.31) and (2.32) are similarly
linearized and reduced to give the following.
For the single-infeed configuration, Figure 2.15, the following holds
   
dPd rU cos(γ + µ) U 4 ESCR2 + (b + KQ) U 2 ESCR + Q  2 − Pd P d
d − Q
d
=
dId det JI
(2.33)
2.6. Incorporation of Local Devices 29

P , –Q
E ∠ψ U ∠δ d1 d1
1 1 1 1
z ∠θ U
1 1 d1

PL1 d γ ,µ
1 x1 1 1 I
QL1 b d1
c1
z ∠θ
12 12
E ∠ψ P , –Q
2 2 d2 d2
z ∠θ U
2 2 d2
U 2 ∠δ 2 2
d γ ,µ
x2 2 2 I
d2
PL2 b c2
QL2

Figure 2.17. Simplified multi-infeed configuration incorporating static load models.

where r, b are as given in subsection 2.2.1 and

 
4 2  2 ∂Qd   
d −
det JI = U ESCR − Qd + U + KQ U 2 ESCR + Q
∂U Id
 
∂Pd  ∂PL
P P −U +U
∂U  Id ∂U

The zero of eq. (2.33) solves to give

 
1 b + KQ  b + KQ 
CESCR = 2 − +  d 2 + Pd P d
−Q (2.34)
U 2 2

where

Pd = Pd − PL , P d = Pd + KP and Q
 d = Qd + QL

KP = [−KP C + KP Z U 2 + (nP U − 1)KP U U nP U ]PL

KQ = [KQI U + 2KQZ U 2 + nQU KQU U nQU ]QL

∂PL
U = [KP I U + 2KP Z U 2 + nP U KP U U nP U ]PL
∂U
The basic equations for the multi-infeed and single-infeed with a parallel AC
line configurations are identical in form, given by

 
1 b + KQ1  b + KQ1 2
CESCR1 = 2 − +  
− Qd1 + (Pd1 P d1 + P0 + Q0 )
U 2 2
(2.35)
where the quantities involved are given in chapter 8 of ref. [17].

2.6.2 Other Devices

For a general device connected at the converter AC bus of the single-infeed
configuration, a general expression defining the MAP condition may be derived
30 2. Analytical Tools and Methods

as


dPd 4 2 ∂Qd  2 
= rU cos(γ + µ) U ESCR + b + U U ESCR + Qd
dId ∂U
  (2.36)
2 ∂Pd
− Qd − Pd Pd − U / det JI
∂U
where r, b, and det JI are as previously defined. Here, ∂Pd /∂U , ∂Qd /∂U are
the voltage dependence of the active and reactive power output, respectively, of
the locally connected device. For example, an SVC whose power out given by
PSV C = 0 QSV C = KSV C U (2.37)
within its linear control range would have voltage dependence given by
∂PSV C ∂QSV C
=0 = KSV C (2.38)
∂U ∂U
From eq. (2.36), the M AP condition for the single-infeed configuration of Fig-
ure 2.1 with a general device locally connected may be similarly specified and
the CESCR derived.

2.7 Relationship Between Power and Voltage Sta-

bility Methods
The relationship between the power and voltage stability analysis methods for
the various system configurations is shown in this section.
It was first shown in [10], for the single-infeed configuration of Figure 2.1 with
the converter operating in CEA/constant power control mode, that the M AP
condition coincides with the V SF becoming infinite at the same operating point.
This is shown for sample system parameter values in Figure 2.18 where the V SF
tends to infinity and the dPd /dId index becomes zero at the same ESCR value,
i.e. CESCR = 1.4775.
It can be also shown, [14],[17], for the multi- infeed configuration with all con-
verters operating in CEA/constant power control, that the MMPC and M V SF
stability boundaries coincide, i.e. λMP C and λV SF become zero for the same
parameter values in the parameter space. An example is shown in Figure 2.19
where the M V SF (circle) and MMPC (solid line) stability boundaries coincide
in the ESCR2 − z12 space.
It can be further shown, [14],[17], that the MMPC and M V SF methods
for the multi-infeed configuration, formulated from eqs. (2.15) and (2.26), re-
spectively, decouples into the V SF and MPC methods for the single-infeed
configuration when z12 tends to infinity.
These relationships can be illustrated by a relationship structure as shown
in Figure 2.20. From this diagram it is seen that the modal technique extends
the power/voltage stability analysis methods for the single-infeed to those for
the multi-infeed configuration, and vice versa.
In a more general context, the relationship between the various maximum
power and voltage sensitivity based methods for the various system configura-
tions, viz. Figures 2.15 - 2.17, was shown in [15], [17]. These relationships,
summarized in Table 2.2, hold with or without load models locally connected
at the converter AC buses.
2.7. Relationship Between Power and Voltage Stability Methods 31

CESCR=1.4775
bc=0.54 pu, γ=17o
dx=0.08 pu.
VSF

dP d
d Id
VSF

ESCR

Figure 2.18. Relationship between V SF and MPC for single-infeed con-

figuration in constant power control.
ESCR2

stable

λVSF=0 unstable

λMPC=0

z12 [pu.]

Figure 2.19. Relationship between MMPC and M V SF stability bound-

ary for multi-infeed configuration in constant power control.

Example 7 These relationships are also illustrated in Figures 2.21 - 2.23 for
the various system configurations of Figures 2.15 - 2.17 having the following
sample parameters: PL1 = PL2 = 0.1 p.u. (constant impedance), QL1 = QL2 =
0.075 p.u. (constant MVA), z12 = 0.6 p.u., γ1 = γ2 = 17◦ , bc1 = bc2 = 0.54 p.u.,
dx1 = dx2 = 0.08 p.u. Initial conditions are P12 = 0.2 p.u. (for multi-infeed)
and - 0.2 p.u. (for single-infeed with parallel AC line), and Ud1 , Ud2 , Id1 , Id2 ,
U1 , U2 are all 1 p.u. For the multi-infeed and single-infeed with parallel AC line
configurations, the constituent converter of interest is the referenced system and
32 2. Analytical Tools and Methods

modal method extends MPC

single-infeed to multi-dimensional case multi-infeed

dP d modal method reduces MMPC λ MPC = 0

= 0
d Id to single-dimensional case
corresponds corresponds
with with
modal method extends VSF 1
∆U ------------ → ∞
-------- → ∞ to multi-dimensional case λVSF
∆Q

modal method reduces MVSF

to single-dimensional case

Figure 2.20. Relationship structure for voltage/power stability analysis

methods with converters in constant power control.

Configuration Maximum Power Voltage Sensitivity

Single-Infeed MPC V SF

Single-Infeed
with parallel MPC M SV F
AC line (constant power)
M V SF
Multi-Infeed SMPC (constant power,
constant current)
M V SF
Multi-Infeed MMPC (constant power,
constant power)

Table 2.2. Relationship between power and voltage stability methods

with and without load models.

the neighbouring constituent converter is then the influencing system. For the
single-infeed configuration, the parameter subscript indices are omitted and only
the relevant parameters apply. Note that the relationship for the multi-infeed
configuration without incorporating load models is not shown in Figure 2.22,
but the correspondence is similar to Figure 2.19.
2.7. Relationship Between Power and Voltage Stability Methods 33

dashed: no load models

solid: with load models

dPd
1 d Id
-----------
VSF

ESCR

Figure 2.21. Coincidence of V SF and MPC for single-infeed configura-

tion with and without load models.
Referenced system CESCR1

MVSF (circle), MMPC (solid)

MVSF (circle), SMPC (solid)

λVSF (circle) and dPd1/dId1 (solid)

multi-infeed configuration in const.DC voltage mode

Influencing system ESCR2

Figure 2.22. Stability boundaries for multi-infeed configuration with load models.
34 2. Analytical Tools and Methods

Referenced system CESCR1

MVSF(circle) ,MPC(solid)

with load models

without load models

Influencing system ESCR2

Figure 2.23. Stability boundaries for single-infeed configuration with

parallel AC line with and without load models.
2.8. Appendix: Definition of ESCR 35

2.8 Appendix: Definition of ESCR

Since the single-infeed configuration of Figure 2.1 is a single converter system,
the system MVA base could be uniquely chosen as the converter rated DC
power. Consequently, the ESCR for the system could be uniquely defined. For
higher-order configurations, the situation is somewhat ambiguous since there are
multiple constituent converters, each with its individual rated DC powers. Thus,
the ESCR for a constituent converter with a DC power rating different from
the common MVA base, cannot be defined with the usual meaning, i.e. with
respect to its own DC power rating. To resolve this, one constituent converter,
say the i-th, can be chosen as the reference. Then the ESCR of this reference
constituent AC/DC system, ESCRi , is defined in the usual meaning as
1
ESCRi = − bci (2.39)
zi
where zi , bci are in p.u. with the rated DC power PdN i , rated AC voltage UN ,
as the common system MVA and voltage base, respectively. Now, a DC Power
Base Ratio (P BRj ) is defined as the rated DC power of a constituent converter,
say j-th, with rated DC power PdN j , relative to the reference system and given
by
PdN j
P BRj = (2.40)
PdN i
Then, the ESCR of the j-th constituent AC/DC-system, ESCRj , can be
defined as
1 1 
ESCRj = − bcj (2.41)
P BRj zj
where ESCRj is thus defined with respect to its own MVA base, i.e. PdN j ,
consistent with the usual definition. However, zj , bcj are still in p.u. on the
common MVA and voltage base, i.e. PdN i and UN respectively. In this chapter,
references to ESCR associated with the higher-order configurations are made
in the context of the above definitions. Also dx is defined as follows
Xc IdN
dx = √ (2.42)
2UN τN
where Xc , τN are the converter transformer reactance and turns ratio, respec-
tively. The subscript label N denotes rated quantities.
36 2. Analytical Tools and Methods
3
Comparison of MPC with Detailed
Simulations

Several indices have been developed for the stability analysis of HVDC Systems.
Ainsworth [5] introduced the concept of Maximum Available Power (M AP ).
This index defines the stability limit for an HVDC System with the rectifier
controlling the DC power and the inverter operating at constant extinction
angle. Hammad [7] has introduced the Voltage Stability Factor (V SF ) in which
the change in sign of the ratio (∆Q/∆V ); or incremental reactive power to
incremental voltage determines the onset of instability. Similar indices have
been introduced by Franken and Andersson [10], and by Nayak et al [8]. These
indices are typically based on the steady-state DC equations and thus do not
represent the full dynamic behavior of the system.
More recently, several investigators have attempted to determine the accu-
racy of these indices. Franken et al [10] have used stability programs, as Pilotto
[18], Nayak [8] have attempted to use electromagnetic transient (EMTP-type)
programs. These analyses generally confirm the usefulness of the steady state
indices. However, recent studies by Reeve et al [9] that consider AC and DC
lines in parallel and a more detailed load representation, show variations from
results obtained from the straightforward calculation of these indices.
Similar discrepancies may arise when synchronous compensators are present
at the converter bus. In this case the effective short circuit ratio (ESCR) is
calculated assuming the transient impedance (Xd ) of the synchronous compen-
sator to be lumped with the impedances of the external AC network and filter.
This ESCR value is later used in the calculation of the MPC. The details of
the excitation control system are also usually ignored. In this section the MPCs
obtained with the simplified model are compared with those obtained using a
detailed electromagnetic transient simulation model.
The study system is derived from the first CIGRE HVDC benchmark [19]
(60 Hz). The following changes are made to the benchmark: Inclusion of a local
load at the inverter bus of 300 MVA, 0.95 power factor local load at the inverter
bus. Inclusion of a synchronous compensator rated -165/+300 Mvar on the bus.
The resultant system is shown in Figure 3.1.

3.1 Simulations Using a Simplified Model

As a first step, the MPC was calculated for this system using the steady-state
DC and AC equations. The machine was represented by its transient reactance
behind a fundamental frequency AC source as is customary in most studies.
The transient simulation model was also similarly simplified resulting in the
equivalent setup shown in Figure 3.2.
The AC sources were selected to provide rated bus voltage at rated DC
current. The MPC in this case is a plot of the delivered DC power v/s the DC

37
38 3. Comparison of MPC with Detailed Simulations

Filters/Ac
Local Load
Capacitors
2.5 Ω, 0.5 H 2.5 Ω, 0.5 H (300 MVA,
0.95 pf)

21.67 µF
Ac Filters (300 Mvar)

Syn. Con.

System Equiv.
ESCR 2.5 , 75 System Equiv.
deg. ESCR 2.5 , 80
deg.

Figure 3.1. Modified CIGRE benchmark model [19].

Z2 (local load)

Zf (filter)
Vd= 500
kV
Zs

Xd’ (or Xd”)

Figure 3.2. Simple test system.

ESCR=1.554
Theoretical MAP with Xd’ Theoretical Vt
Simulation MAP with Xd’ Simulation Vt
1.25
Pdc (pu), Vt (pu)

1.125
1
0.875
0.75
0.6 0.7 0.8 0.9 1 1.1 1.2
Idc (pu)

Figure 3.3. Comparison of theoretical and simulated MPC assuming a

simple network equivalent.

current (see solid curve in Figure 3.3). A system in power control is stable when
its operating point lies to the left of the peak of the resultant inverted parabola
shaped curve. This is because the power controller effects an increase in power
by requesting an increase in DC current, consistent with the rising curve to the
left of the peak. Obviously, to the right of the peak, a current increase causes a
power decrease resulting in the instability of the power control mode.
Figure 3.3. shows the theoretically calculated curves for DC power and AC
bus voltage obtained from steady-state equations (see Appendix, section 3.4)
3.2. Inclusion of a Synchronous Compensator at the Converter Bus 39

superposed on the simulated curve, obtained from a detailed electromagnetic

transient simulation. The selected ESCR of 1.554 is the critical ESCR , i.e. the
onset of instability occurs at the rated DC current for this value. The simulated
curve was obtained by ramping the DC current over 30 s (in a quasi steady-
state manner) and recording the DC power. It should be noted that the MPC
characteristic is derived using (simplified) steady-state equations, whereas the
simulated model uses a far more detailed representation, with the individual
switching of each thyristor and other non-linearities. Nevertheless, the two
approaches yield identical results as seen in Figure 3.3; thereby indicating that
the MPC approach is useful in the case where the AC system can be represented
by a simple Thevenin equivalent.

3.2 Inclusion of a Synchronous Compensator at the

Converter Bus

3.2.1 Differences Arising from Neglecting the Zero-Power Con-

straint for the Synchronous Compensator

This investigation was carried out in order to check whether the MPC approach
can be used to determine the stability limits of inverters with synchronous com-
pensators on the AC bus bar. In this particular study, the exciter response was
not modelled and it was assumed that the field voltage of the machine remained
at a fixed value. As the flatness of the MPC makes it difficult to determine the
exact point at which the peak occurs, another equivalent approach is to plot
the index CSI = dP/dId as a function of the parameter being varied. The
point at which this quantity crosses the horizontal axis is thus the transition
point from stable to unstable operation. This index is also referred to as the
controller sensitivity index for the constant power mode of operation [8], see
also section 2.4.
It can be argued that lumping the synchronous compensator with the net-
work equivalent as in Model 1 (Figure 3.1) is an oversimplification. This is
because the actual synchronous compensator does not absorb or deliver any
real power in the steady state. No such assumption is made in the traditional
M AP which means that real power may be supplied or delivered by the syn-
chronous machine. The equations in the Appendix, section 3.4, can be modified
to include an additional constraint for the synchronous compensator so that it
produces zero power. A new set of M AP (or equivalently CSI) curves now
results as shown in Figure 3.4. The method for obtaining the simulated curves
is as described in section 3.1.
In simplifying the machine to an equivalent, two assumptions are possible:
Assume a synchronous reactance of Xd or Xd for the machine. As can be seen,
the added assumption makes a significant difference to the CSI and hence to
the instability point on the M AP characteristic. In comparison with a detailed
model the curves generated with the Psyn = 0 summation match the observed
response more closely.
40 3. Comparison of MPC with Detailed Simulations

Machine replaced by Xd’ reactance

Psyn equal to 0 Psyn not equal to 0
Zero
100
50
CSI
0
-50
-100
1.4 1.45 1.5 1.55 1.6 1.65 1.7 1.75 1.8
ESCR(Xd’)

Machine replaced by Xd" reactance

Psyn equal to 0 Psyn not equal to 0
Zero
300
200
CSI

100
0
-100
1.4 1.6 1.8 2 2.2 2.4
ESCR(Xd")

Figure 3.4. CSIP curves with and without Psyn = 0 assumption.

Z2 (local load)

Zf (filter)
Vd= 500
kV

M Deatiled Model of
Sync machine
with fixed Efield

Zs

Figure 3.5. Detailed machine model with fixed excitation.

3.2.2 Comparison with a Detailed Machine Model with and

without Excitation Control
In this comparison the curves obtained via theoretical calculations were com-
pared with a full electromagnetic transient model of the machine shown in Fig-
ure 3.5. As before, the field voltage was assumed fixed, but in the simulation
the machine was represented by a two-axis theory based model. The resultant
curves are as shown in Figure 3.6. for a typical ESCR of 1.6.
The curves in Figure 3.6 show the DC power and AC bus voltage regulation
as a function of the DC current. The analytically calculated curves for two
3.2. Inclusion of a Synchronous Compensator at the Converter Bus 41

ESCR=1.597
Theoretical with Xd’ Theoretical with Xd"
Simulation with H=10pu
1.05
Pdc (pu) 0.98
0.91
0.84
0.77
0.7
0.6 0.7 0.8 0.9 1 1.1 1.2
Idc (pu)

Theoretical with Xd’ Theoretical with Xd"

Simulation with H=10pu
Bus voltage, Vt (pu)

1.25
1.14
1.03
0.92
0.81
0.7
0.6 0.7 0.8 0.9 1 1.1 1.2
Idc (pu)

Figure 3.6. Comparison of theoretical MPC and with detailed machine

model based simulation (Ef d constant).

assumed reactance values (Xd and Xd ) are superposed on the simulated char-
acteristic. The simulated characteristic (obtained as described in section 3.1)
appears to follow the Xd based analytical result more closely at least at lower
DC current levels. After the peak is reached, the simulated curve deviates from
the Xd based characteristic and appears to be closer to the Xd based character-
istic. However it can be seen that the peak of the actual curve is more closely
approximated by the Xd based curve.
The oscillations in the simulated waveform, particularly in the area where
M AP is exceeded are due to the electrical and mechanical transients excited in
the model by the slow ramp-up procedure.

3.2.3 Inclusion of Controlled Excitation

The impact of modeling excitation controls on the simulated MPC is now pre-
sented. The model used in the simulation is as shown in Figure 3.7. A sim-
ple Proportional-integral control system was assumed with an exciter ceiling of
±12 p.u., which is typical for modern fast acting excitation systems.
For such a system, the MPC (a steady state concept), can be derived by
assuming the bus voltage of the synchronous compensator to be constant. A
properly acting excitation system (droop ignored) would indeed achieve this.
With such an assumption, the short circuit capacity of the inverter’s AC bus
approaches infinity and the bus can be regarded as an infinite bus. The resulting
42 3. Comparison of MPC with Detailed Simulations

Z2 (local load)

Zf (filter)
Vd= 500
kV

Exciter
Vref
M _+
Detailed
M/c model

Zs

Figure 3.7. Detailed machine model with excitation system included.

ESCR=1.597
Pd without Vt reg. Pd with Vt reg.
Vt without reg. Ideal Vt
3
Pdc (pu), Vt (pu)

2.5
2
1.5
1
0.5
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Idc (pu)

Figure 3.8. MPC and bus voltage characteristics with and without ideal excitation.

M AP now has a very large value and occurs at a very large DC current as shown
in Figure 3.8. The MPC and voltage characteristics for the earlier case (as in
Figure 3.2) are also superposed for comparison.
However, when a detailed simulation is carried out, this expected result does
not occur. The system follows the theoretical curve as seen in Figure 3.9 (solid
and dashed traces). Note that only a portion of the x-axis range of Figure
3.8 is plotted. The system becomes unstable at Id = 1.08 p.u. This is due to
instability caused by the excitation system dynamics, which were ignored in the
theoretical calculation. The exact onset level is also seen to vary depending on
the controller gains. This shows that the results obtained from a simple MPC
analysis could be excessively optimistic when excitation control is ignored.
If instead of modeling the machine in detail, a simplified exciter-machine
model as shown in Figure 3.10 is used, the results are in agreement with the
curve derived in Figure 3.8. In this approach the exciter-machine system is
modeled as a Proportional-Integral controller whose output is the Thevenin
voltage behind the machine impedance Xd . The resulting curves are shown as
the dotted lines in Figure 3.9. In this case the instability does not occur as
in the detailed model, but the curve flattens out after a current of 1.1 p.u. is
reached due to the attainment of a limit in the DC current controller. This
3.3. Conclusions 43

Theoretical MAP Simulation MAP

Simulation MAP (mach.&exciter)
1.2
Pdc (pu) 1.1
1
0.9
0.8
0.7
0.6
0.6 0.71 0.82 0.93 1.04 1.15
Idc(pu)

Thoretical Vt Simulation Vt
Simulation Vt (mach.&exciter)
Bus voltage, Vt(pu)

1.1
1
0.9
0.8
0.7
0.6
0.5
0.6 0.71 0.82 0.93 1.04 1.15
Idc(pu)

Figure 3.9. Theoretical and simulated MPC with excitation controls.

Zs Vref
Kp+Ki/s _+

Vpu

Figure 3.10. Simplified machine-exciter model.

proves that the instability observed with the detailed machine model is indeed
due to dynamics which are ignored in the theoretical calculation of MPC.

3.3 Conclusions
The following observations can be made based on the above results:

• The MPC characteristic agrees with detailed transient simulation when

machine dynamics are not modeled.

• When the Synchronous Compensator is included but the exciter dynam-

ics are ignored, the theoretically obtained graphs (with Xd or Xd for the
44 3. Comparison of MPC with Detailed Simulations

impedance) do not follow the simulated MPC over the entire range. How-
ever, a reasonable figure for the M AP can be arrived at using Xd in the
theoretical derivation.
• When Modelling Synchronous Compensators in MPC analysis, the addi-
tional condition of P = 0 gives more accurate results.
• Inclusion of an exciter results in an actual response that tracks the M AP
result, but only up to a point. The margin of stability is not improved.
3.4. Appendix: Equations used in MPC Calculations 45

3.4 Appendix: Equations used in MPC Calculations

With Simple A.C. Thevenin Equivalent Source
√
2Xc Id
Ut = (3.1)
τ (cos α + cos γ)
cos γ − cos α
cos φ = (3.2)
2
√
3 2 3
Ud = τ Ut cos γ − Xc Id (3.3)
π π
Pd = Id Ud (3.4)
Zs jXd
Pd ejφ Ut Ut 
E sys = Ut − − − (3.5)
Zs + jXd Ut cos φ Zf Zll

With Constraint on Real Power into Synchronous Compensator Eqs. (3.1)

- (3.4) are used together with a modified equation 5 as shown below and an
additional constraint equation

P ejφ Ut Ut Qsyn 
d
E sys = Ut − Zs − − −j (3.6)
Ut cos α Zf Zll Ut
√
E cos δ − U 
f t
Qsyn = 3Ut (3.7)
Xd
The variables used in the above equations are thus defined:
Ut : Line to line voltage on commutating AC busbar (kV)
Xc : Commutating impedance (Ω)
Ud , Id , Pd : DC voltage (kV), DC current (kA) and DC power (MW)
τ : Transformer turns ratio (converter side: ac network side)
α : Delay angle of inverter (deg)
γ : Extinction angle of converter (deg)
φ : Power factor angle of converter (deg)
Zf : Complex impedance of AC filters
Zll : Complex impedance of local load
Esys , Zs : Thevenin equivalent voltage (kV) and impedance (Ω) of AC network
δ: Angle of synchronous machine equivalent, with respect to the commutating
AC busbar voltage (deg)
Qsyn : Reactive power from synchronous compensator (positive for capacitive
operation) (Mvar)
Ef : Voltage behind synchronous impedance (proportional to excitation volt-
age) (kV)
Xd , Xd : Transient and subtransient impedances of synchronous compensator
(p.u.)
46 3. Comparison of MPC with Detailed Simulations
4
Influence of System Operating Con-
ditions and Converter Design

It is common to assess the AC/DC interconnection by a Short Circuit Ratio

(SCR) or by an Effective Short Circuit Ratio (ESCR) where the AC system is
represented by a Thevenin equivalent voltage source of fundamental frequency
behind an equivalent impedance, refs. [4] and [20]. However, the term SCR
being useful in preliminary planning, does not account for location of loads.
Similarly, the concept of Maximum Available Power (M AP ), as characterized
by Maximum Power Curves, MPCs, is influenced by load characteristics as well
as operational tap-changing strategies and interaction between the rectifier and
inverter AC buses via parallel AC path [21].
The M AP concept and its quantification have been largely based on the
established technology of naturally commutated converters. The different re-
active power behavior of emerging voltage-sourced converters and capacitor-
commutated converters has a corresponding influence on the MPC and the AC
voltage profiles as DC power is varied.

4.1 Effect of Load Modeling on the MAP and AC

Voltage
The change in the DC current influences the AC system voltages and conse-
quently the load voltages. In [9], the authors have investigated the influence of
the load model on the MPC for a modified version of Test System T-2 in [22],
see Figure 4.1. The Test System is modified such that values of ESCR on the
inverter side of 2.60 and 1.26 have been achieved.
For the first case, an ESCR of 2.60 produces a nominal operation below
MAP. The load flow program at a nominal operating point determines the con-
verter transformer which are kept constant for operation different from nominal.
With constant current control at the rectifier and minimum γ at the inverter,
the DC current order is increased from zero to the point where the load flow fails
to converge. The MPC are obtained by repetitive runs of the EPRI IPFLOW
2.1 load flow using the procedure described above and for different load models:

1. Test System without the inclusion of load models and with voltage sources
generating and absorbing the transmitted power for validation with a
Thevenin source AC model used in basic SCR (ESCR) calculation;

2. Constant P Q load model, as often used in steady state analyses but as-
sumed to be overly pessimistic in voltage stability studies;

3. Impedance load model which is sensitive to changes in the AC voltage

magnitude and consequently to the changes in DC current order;

47
48 4. Influence of System Operating Conditions and Converter Design

4. Voltage dependent load as found appropriate in voltage stability studies;

a polynomial load model has been used:
PL = Pn (0.83U 0 − 0.30U 1 + 0.47U 2 ) (4.1)
QL = Qn (6.70U 0 − 15.3U 1 + 9.60U 2) (4.2)

The approach has been to normalize the loads to provide the same P and Q at
the nominal operating point of the DC system while permitting the interaction
away from nominal to respond to above mentioned load models. The curves in
Figure 4.2 indicate that M AP and IMAP are sensitive to the assumption of load
model and the operation above rated DC current shows a spread in the MPCs.
For the Test System of ESCR of 1.26, the ESCR is less than the critical
value for which nominal operation coincides with maximum power. Figure 4.3
shows the MPCs for different load models, provided that the nominal operating
point for all load models used is the same.

1 2

3
8 11 12

5 6

10 9 13 7

Figure 4.1. Test System T-2 [22].

1.2

1
Pdi & Vac (pu)

0.8

0.6

0.4
−. voltage dep. loads
* impedance loads
− no loads
0.2 − − PQ loads

0
0 0.2 0.4 0.6 0.8 1 1.2
Idi (pu)

Figure 4.2. MPC sensitivity to load model (ESCR = 2.60) [9].

4.1. Effect of Load Modeling on the MAP and AC Voltage 49

1.2

0.8
Pdi (pu)

0.6

− no load
0.4 −. voltage dep. loads
− − impedance loads
* PQ loads
0.2

0
0 0.2 0.4 0.6 0.8 1 1.2
Idi (pu)

Figure 4.3. MPC sensitivity to load model (ESCR = 1.26) [9].

1.6

1.5

1.4

1.3

1.2
Vac (pu)

1.1
− no loads
1
−. voltage dep. loads
0.9 * impedance loads
− − PQ loads
0.8

0.7

0.6
0 0.2 0.4 0.6 0.8 1
Pdi (pu)

Figure 4.4. Voltage variations for different load models (ESCR=1.26) [9].

The discrepancies between the curves become more marked as the ESCR of
the inverter is reduced, as shown in Figure 4.3. A comparison of Figure 4.2 and
Figure 4.3 suggests that load representation is not a critical need in high SCR
situations but is progressively significant as the SCR reduces.
With constant current control rather than constant power control at the
rectifier, the load flow was able to converge in all four cases for a current larger
than and a common nominal operating point above was achievable. It is notable
from Figure 4.3 that the maximum available powers for the P Q and voltage-
dependent loads of 1.086 and 1.037, respectively, are higher than for the case
with no load included when the M AP was 1.01.
The four curves in Figure 4.4, so-called nose curves as commonly referred to
in voltage stability studies, show AC voltage variations at the inverter AC bus for
50 4. Influence of System Operating Conditions and Converter Design

1.6

1.4 AC voltage

1.2
MAP curve
1
Pdi & Vac (pu)

0.8

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Idi (pu)

Figure 4.5. MPC and AC voltage for test system with no parallel AC
line (ESCR=1.69) [9].

different DC powers. The turning point between stable and unstable operation
(between dUac /dPd ≤ 0 on the upper branch and dUac /dPd ≥ 0 on the lower
branch) corresponds to the different M AP values in Figure 4.3. The spread
between AC voltages is particularly noticeable below the nominal operating
point of 1.0 pu for DC current and power. This suggests that provisions for
voltage control during start up, and scheduled changes in DC power, need to
take into consideration different load models in assessing stability issues.

4.2 Effect of a Parallel AC Line

When the parallel AC line is removed from the test system the calculated ESCR
at the inverter decreases from 2.60 to 1.69, due to the reduced short-circuit level.
Note that the required shunt compensation is correspondingly reduced; a new
AC/DC load flow was run to provide nominal AC and DC conditions. The MPC
and AC voltage profile at the inverter bus, with a P Q load model, are shown in
Figure 4.5. Compared with the equivalent curve in Figure 4.2, it has the same
maximum power of 1.21 pu, contrary to usual expectations, at a DC current of
1.15 pu compared to 1.11 in Figure 4.5. This also indicates that the apparent
strengthening of the inverter AC bus by the parallel AC line does not produce
a correspondingly increased M AP . This is due to the complex reactive power
interactions of the rectifier and inverter AC buses via the AC line as the DC
power is varied. The reduced need for capacitive compensation at the inverter
bus, when the AC line is disconnected, is also a contributing factor.
The SCR and ESCR values do not characterize any specific AC coupling
between the rectifier and the inverter and therefore are not sensitive to the
AC configuration nor to any changes. Consequently, other than for very pre-
liminary analysis, it would be prudent to incorporate as much actual system
representation, as is known, in planning studies relaying a M AP concept.
4.3. Effect of Tap Changing 51

1.4

1.2 o
C
o
B
1 o
A

0.8
Pdi (pu)

0.6
− Gamma=18 (TAP=1.23)
−. Gamma=18 (TAP=1.17)
... Gamma=22 (TAP=1.17)
0.4 − − Constant Vdc

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Idi (pu)

Figure 4.6. Effect of tap changing on power transfer [9].

4.3 Effect of Tap Changing

The three curves in Figure 4.6 illustrate the increase in M AP as a result of
operating at a reduced minimum extinction angle (22◦ to 18◦ ) and also as a
result of tap changing. The straight line of inverter control to maintain constant
DC voltage has its nominal operating point at A for γ = 22◦ . As DC current
increases, the margin of voltage control reaches the γ = 18◦ limit at B with a
tap of 1.17. Tap changing to 1.23 permits the control margin to be restored (by
efficiently switching the MPC) until, with a further increase in DC current, a
new limit of minimum γ is reached at C.
The transition A-B-C permits DC power to increase while the three selected
operating points are progressively below and above the respective prevailing
MPCs. Since such an operating mode is typical of low SCR situations, the
consequences of tap changing need to be included in performance evaluation.
The enhancement of M AP by tap changing, under operational conditions,
would apply to slowly changing system conditions within the time-response ca-
pabilities of the tap-changing mechanism. This would be typical of the response
to load changes, as opposed to abrupt changes such as line switching.
The corresponding voltage profiles at the inverter AC bus are shown in
Figure 4.7. Tap changing permits DC voltage control (and hence power control
in conjunction with rectifier DC current control) to be extended to increase
power by 20%. At this limit, increasing DC current reduced DC power as the
AC voltage starts to collapse.

4.4 MPC for Capacitor Commutated Converters

The MPCs presented so far have all assumed that the converters are line com-
mutated (LCC). Recently Capacitor Commutated Converters (CCC) of current
source type have been put into operation, see refs. [23] and [24], which have
52 4. Influence of System Operating Conditions and Converter Design

1.6

1.4

1.2

1
Vac (pu)

0.8

− Gamma=18 (TAP=1.23)
0.6
−. Gamma=18 (TAP=1.17)
... Gamma=22 (TAP=1.17)
0.4
* Constant Vdc (TAP=1.17)
− − Constant Vdc (TAP=1.23)
0.2

0
0 0.2 0.4 0.6 0.8 1 1.2
Pdi (pu)

Figure 4.7. AC voltage against DC power with tap changing [9].

Type SCR Qc (p.u.) UT OV (p.u.) Figure

CCC 2.0 0.13 1.07 4.8
LCC 2.0 0.50 1.27 4.9
CCC 0.93 0.13 1.71 4.10
LCC 0.93 0.50 3.18 4.11

Table 4.1. Data and for the systems which MPCs are given in Figures 4.8
to 4.11. It should be noted that the UT OV values given refer to complete
load rejection without taking any non-linear effects, e.g. transformer sat-
uration and arrester action, into account.

drastically better properties when it comes to voltage and power stability as

will be shown in this section.
MPCs for cases with data according to Tabel 4.1 are shown in Figures 4.8
to 4.11. From these it is clearly seen that CCC exhibits superior performance.

4.5 Conclusions
The study has shown the necessity of plotting maximum power curve and ob-
taining the maximum available power for the particular AC/DC system and
particular control mode at the inverter. It has also been shown that the basic
SCR calculation, while still very helpful in the first stage of design, is insuffi-
cient in itself for evaluating the power transfer capability and power stability
of the DC system. The presence and characteristics of AC loads as well as
the presence of any parallel AC lines influence the maximum available power
transfer and make the nose curves more diverse. Furthermore, the power trans-
fer capability is affected by the adopted strategy for tap changing. It was also
shown that CCC has superior performance when it comes to voltage and power
stability.
4.5. Conclusions 53

Maximum Power Curve Capacitor Commutated Converter, gamma = const.

1.8

1.6

1.4
DC Power (p.u.) & AC−voltage (p.u.)

1.2
AC Voltage

0.8

0.6

MPC
0.4

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
DC Current (p.u.)

Figure 4.8. Maximum Power Curve (MPC) of a system with CCC and SCR = 2.0.

Maximum Power Curve of Line Commutated Converter, gamma = const.

1.8

1.6

AC Voltage
1.4
DC Power (p.u.) & AC Voltage (p.u.)

1.2

0.8

0.6
MPC

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
DC Current (p.u.)

Figure 4.9. Maximum Power Curve (MPC) of a system with LCC and SCR = 2.0.
54 4. Influence of System Operating Conditions and Converter Design

Maximum Power Curve of Capacitor Commutated Converter, gamma = const.

1.8
AC voltage
1.6

1.4
DC Power (p.u.) & AC−voltage (p.u.)

1.2

0.8

0.6

0.4
MPC

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
DC Current (p.u.)

Figure 4.10. Maximum Power Curve (MPC) of a system with CCC and
SCR = 0.93.

Maximum Power Curve of Line Commutated Converter, gamma = const.

1.8

AC Voltage
1.6

1.4
DC Power (p.u.) & AC Voltage (p.u.)

1.2

0.8
MPC

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
DC Current (p.u.)

Figure 4.11. Maximum Power Curve (MPC) of a system with LCC and
SCR = 0.93.
5
Multi-Infeed HVDC Systems

5.1 Development in the Study of Multi-Infeed

HVDC Systems
Traditionally, HVDC links are built as point-to-point interconnections or single-
infeed HVDC systems as they are commonly known, although other system
configurations such as back-to-back, multi-terminal, HVDC links have also been
built. When the HVDC link is terminated at an AC system location of low short
circuit capacity relative to the power rating of the HVDC link, the AC/DC in-
terconnection is said to be weak. Concomitant problems relating to system sta-
bility and performance are typically voltage/power instability, high temporary
over-voltages, low frequency resonances and harmonic instabilities, long fault
recovery times, and susceptibility to commutation failures [25], [26]. These sit-
uations can result in equipment damage or lead to graver scenarios as system
collapse. Therefore a fundamental understanding of their nature, developing an-
alytical techniques, and finding countermeasures, are crucial to enable successful
planning, design, and operation of power systems exhibiting such problems. A
comprehensive and systematic analysis of all important AC/DC interactions for
the single-infeed HVDC system is given in [4], while a survey and review of
AC/DC interconnections exhibiting such problems are presented in [27], [28].
As the use of HVDC transmission continues to develop, situations will also
arise where two or more HVDC converters terminate at AC system locations
that are electrically in close proximity, for example a common AC area fed by
point-to-point HVDC links. In a general context, these converters in close prox-
imity can be any combination of inverters and rectifiers [29], [30]. Such system
configurations are loosely termed as multi-infeed HVDC systems. Situations
as these arise particularly in regions where the power systems of neighboring
countries are dominantly interconnected by HVDC links. Already such sys-
tem configurations are emerging or in existence such as those in Scandinavia,
Southern California, Manitoba, and Quebec. These system configurations are
anticipated to give rise to new phenomena concerning interactions between the
constituent HVAC and HVDC systems, especially when one or more of the
AC/DC interconnections are relatively weak. Hitherto, due to the recent origin
of multi-infeed HVDC systems there has only been rudimentary investigations
on the interaction phenomena and potential associated problems. However the
reality and practicality of such system configurations have spurred growing in-
terest in this area. As far we are aware, pioneering investigations only began as
recent as the works in [29], [30], though some aspects of these system configu-
rations had been investigated in an earlier work [31]. In [29] various categories
of multi-infeed HVDC systems were defined followed by a simulator study on a
practical system representing one of the categories. The study found intimate
interactions between the various constituent systems and control coordination
among the constituent HVDC links was particularly important. In [30] an in-

55
56 5. Multi-Infeed HVDC Systems

sight into various potential problems arising from multi-infeed HVDC systems
was given. These include small-signal instability due to control interactions
among the constituent HVDC links, voltage instability and collapse, increased
commutation failures in one constituent HVDC link due to AC faults occurring
in the vicinity of a neighboring one, transient AC voltage depression due to
simultaneous recovery of constituent HVDC links after an AC fault. Similar to
[29], control coordination among the constituent HVDC links or modification of
the HVDC control modes were identified as potential solutions to overcome the
associated problems. In [32] a digital simulator study was performed to investi-
gate the transient response and recovery of multi-infeed HVDC systems based
on different control strategies of the constituent HVDC links. Transient recovery
of the studied system was found to be highly system dependent although certain
control strategies could provide robust system response. In [33] synchronizing
and damping torque modulation applied to the controls of constituent HVDC
links of a multi-infeed HVDC system were investigated. Cooperative coordi-
nation of controls among the constituent HVDC links was shown to provide
effective damping and synchronizing torque in a de-coupled manner, to achieve
overall system integrity with respect to electromechanical stability.
As seen, a broad spectrum of system phenomena could potentially arise in
multi-infeed HVDC systems [29], [30], [32], [33]. Conceivably these system phe-
nomena are related to those arising from single-infeed HVDC systems, since
multi-infeed system configurations historically evolved from the latter. More-
over, voltage/power instability problems clearly hold special interest since these
had been among the most important technical concerns for single-infeed HVDC
systems. In this respect it is appropriate to know how close these system phe-
nomena correspond, under what conditions they differ, and whether the ana-
lytical techniques developed for the single-infeed situation could be similarly
applied to the multi-infeed case. A host of other fundamental questions arise
and there is clearly a need to address them. These questions essentially provide
the main motivation for the work in [12] - [17], and [35] - [37]. Specifically these
works aim to achieve the following objectives;

• deriving a fundamental understanding of the voltage/power instability and

interaction phenomena in multi-infeed HVDC systems.

• developing techniques to analyze them.

• finding viable countermeasures to overcome the associated problems.

5.2 System Models

In the literature, study of multi-infeed HVDC systems have generally used sys-
tem configurations with two [29], [12], [30], [32], [33] or three [13], [34] point-
to-point HVDC links. However, the AC network representation in these system
configurations varies considerably in complexity and topology. For example, in
[30] the HVDC links are embedded in a very large AC system while in [29] the
AC network is a mere 5-bus system interconnected by HVDC links whose re-
mote converter terminals are interconnected by islanded single-bus AC systems.
Thus the diversity in system representation offers no ground to abstract a bench-
mark multi-infeed HVDC system model. Nevertheless, in general a benchmark
5.2. System Models 57

1st
1st

2nd

N-th 2nd

N-th

3rd

Chain configuration Ring configuration

Figure 5.1. Multi-infeed HVDC system configuration constituted by

point-to-point HVDC links.

1st 2nd N-th

Figure 5.2. Multi-infeed HVDC system configuration constituted by

multi-terminal HVDC links.

system model suitable for systematic investigations should have the following
characteristics;
• A system topology that derives its roots from the classical single-infeed
HVDC system configuration. This strong analogy between the single- and
multi-infeed HVDC system configuration is essential to allow analytical
comparisons between them.
• A system model that is simple but representative of the multi-infeed
HVDC system configuration. This allows simulation studies based on
a realistic system model, at the same time theoretical analysis on it can
be done without being distracted by intractable mathematical rigour.
These desired characteristics motivated much use of the system configuration
as shown in Figure 5.1 for the work in [12] - [17], and [35] - [37]. Figure 5.1
shows two variants of this system configuration, namely the chain and ring
multi-infeed HVDC system configuration, but their essential characteristics are
58 5. Multi-Infeed HVDC Systems

Figure 5.3. Multi-infeed HVDC system configuration constituted by

multi-terminal and point-to-point HVDC links.

that they comprise point-to-point HVDC links and all the converters are in-
verters. This is perceived to be the most common and onerous situation with
respect to voltage and power instability. Other system configurations and con-
verter type combinations are of course possible, such as shown in Figure 5.2
where the multi-infeed HVDC system configuration is constituted by multi-
terminal HVDC links. Ultimately a most general multi-infeed HVDC system
configuration would conceivably be as shown in Figure 5.3, being constituted by
point-to-point as well as multi-terminal HVDC links. This system configuration
could potentially come about as a result of the Baltic Ring, East-West Europe,
and southern Scandinavia HVDC projects.
As for converter type combinations, in general there are a number of possi-
bilities as categorized in [29], described below and schematically shown in Figure
5.4;
• Type 1A: converters of the same type connected to different HVDC sys-
tems.
• Type 1B: converters of the same type connected to a common HVDC
system.
• Type 2A: converters of mixed type connected to different HVDC systems.
5.3. Analysis of Voltage/Power Stability of Multi-Infeed HVDC Systems 59

Type 1A Type 1B Type 2A

Figure 5.4. Three categories of converter type combinations for multi-

infeed HVDC system configurations.

5.3 Analysis of Voltage/Power Stability of Multi-

Infeed HVDC Systems
Voltage stability of AC power systems has been studied from static (or quasi-
static) and dynamic approaches. Recent works have also treated voltage sta-
bility from a nonlinear dynamical system theoretic perspective. Likewise for
HVDC systems, voltage and power stability investigations have followed similar
approaches but using analysis techniques and concepts that are pertinent to
HVDC systems. These approaches and analysis techniques can be organized
into a framework as shown in Figure 5.5. The following discussion on past work
concerning stability analysis of multi-infeed HVDC systems is with reference to
this framework.

Static Approach
Sensitivity Technique The underlying assumption for the static approach to
study voltage/power instability in HVDC systems is that the phenomenon oc-
curs so fast that the AC/DC system dynamics do not appreciably affect it. Thus
the DC line and converter pole control dynamics, and voltage control dynamics
in the AC system are neglected. The stability of the AC/DC system is thus es-
sentially determined by the algebraic system states that govern the power flow
in the system, and sensitivity of system states to small changes in controlling
system quantities is used as a measure of system stability.
Two main concepts based on the sensitivity technique had been proposed
in the past to analyze the voltage/power stability of weak single-infeed HVDC
systems. One approach known as the Maximum Power Curve (MPC) method is
based on the concept of Maximum Available Power (MAP ) and was introduced
in [5] and [6]. The other approach, known as the Voltage Stability Factor (V SF )
method, is based on the concept of voltage sensitivity and was first presented
in [7].
Taking cue from these concepts for the single-infeed HVDC system, they
were extended to analyze the voltage and power stability of multi-infeed HVDC
systems in [12], [17], [37]. This entails use of the system power flow model
and modal analysis technique. Briefly, this mathematical technique decomposes
the multi-dimensional AC power flow Jacobian into its modal equivalent. The
60 5. Multi-Infeed HVDC Systems

Voltage and Power

Stability analysis of
multi-infeed HVDC
Systems

STATIC DYNAMIC
APPROACH APPROACH

SENSITIVITY SMALL- LARGE-

TECHNIQUE SIGNAL SIGNAL
TECHNIQUE TECHNIQUE

Figure 5.5. Frameworks for voltage and power stability analysis of multi-
infeed HVDC systems.

resulting incremental modal voltages thus relate with the modal reactive power
injections in a de-coupled form and through an eigenvalue of the corresponding
mode. This is intuitively akin to the VSF of the single-infeed case and can be
interpreted as the modal voltage sensitivity factor MVSF for the multi-infeed
HVDC system.
Based on the above technique, a general and comprehensive method for volt-
age stability analysis of multi-infeed HVDC systems was developed in [12], [17],
[37]. The technique was further developed in [13] through use of participation
factors of the critical modes to determine the critical system location in a multi-
infeed HVDC system where the most severe system phenomena are expected to
occur, and for evaluating the effectiveness of remedial actions implemented.
Similarly the modal analysis technique incorporating the concept of max-
imum available power was proposed in [12] to investigate the power stability
of multi-infeed HVDC systems. It was shown in [12], [17] that the DC power
flow Jacobian could be decomposed into its modal equivalent and the resulting
incremental modal DC power and modal DC currents are related in a man-
ner similar to that for the single-infeed case. A limiting condition, akin to the
Maximum Available Power (MAP ) condition for the single-infeed case, is also
reached when the eigenvalue of the corresponding mode vanishes. Thus the
concept of Modal Maximum Available Power (MMAP ) for multi-infeed HVDC
systems could be similarly proposed.
Similar to the relationship between the power and voltage analysis methods
for the single-infeed case shown in [10], such relationships were also shown in
[12], [17] to exist for the multi-infeed HVDC case. It was also shown in [12],
[17] the inter-relationship between the methods for the single-infeed and multi-
5.3. Analysis of Voltage/Power Stability of Multi-Infeed HVDC Systems 61

infeed HVDC systems. From such relationship structure it was established that
the modal power/voltage analysis methods for multi-infeed HVDC systems are
the multi-dimensional equivalent of the corresponding ones for the single-infeed
case. In [16] it was further shown that these relationships are preserved when
load models are incorporated into the system models.
In [9] the influence of static load characteristics on the MPC was shown
for a HVDC link with a parallel AC line interconnection. Similar to this, the
influence of static load characteristics on the voltage/power stability of multi-
infeed HVDC systems was shown in [15], [16], [17]. In [15], [16], [17] the basic
voltage and power stability equations incorporating static load characteristics
were derived and used to show the effect of load and system parameters on
the voltage/power stability margins of multi-infeed HVDC systems. Conditions
for ascertaining the most unfavorable static load characteristics with respect to
degrading the voltage and power stability margins were derived in [16]. Indices
to characterize and estimate a bound for the effect of static load characteristics
on the system stability margins were also defined and studied in [16].

Dynamic Approach
The dynamic approach to voltage and power stability analysis of HVDC systems
considers the full or at least the most essential dynamics of the system. The
AC/DC system is thus represented by a nonlinear differential-algebraic system
model and solution techniques are applied according to the problem of interest,
that is whether the instability problem is due to a large or small-signal system
disturbance.

Small-Signal Technique For small-signal considerations, the differential-alge-

braic system model may be linearized, the algebraic states eliminated, resulting
in a small-signal differential system model. This technique has been applied to
HVDC systems to study the dynamic stability with respect to electromechani-
cal [38], [39] and DC control loop induced oscillations [40]. It was in [30] that
this technique was first applied to investigate electromechanical oscillations in
a large AC system with multiple HVDC links. In [33] synchronizing and damp-
ing torque modulation applied to the controls of constituent HVDC links of a
multi-infeed HVDC system similar to the one used in [30] were investigated.
Cooperative coordination of controls among the constituent HVDC links was
shown to provide effective damping and synchronizing torque in a de-coupled
manner, to achieve overall system integrity with respect to electromechanical
stability. It was in [30] and [33] that a controllability table, defined as the ranked
normalized frequency response of the transfer function between the modulating
and rotor speed signals, was derived. It was shown how the controllability table
could provide reliable information regarding how and where the DC link modu-
lation is most effective in a multi-infeed HVDC system to damp the troublesome
oscillatory modes.
Another aspect of small-signal technique applied to HVDC systems is the
study of non-linear phenomena associated power/voltage instability. This was
first briefly investigated in [41] for the single-infeed HVDC system. It was in
[17], [36] that a fundamental and comprehensive study of nonlinear phenomena
in HVDC systems was embarked upon. In these works, two principal volt-
age collapse mechanisms in HVDC systems arising from nonlinear phenomena,
62 5. Multi-Infeed HVDC Systems

namely the saddle-node and Hopf bifurcation, were studied for the single-infeed
and multi-infeed HVDC systems. It was also shown in [17], [36] that the quasi-
static and dynamic conditions for voltage collapse considered via saddle-node
bifurcation in the HVDC system models considered were equivalent.

Large-Signal Technique Another important aspect of system phenomena in

multi-infeed HVDC systems is the transient or large-disturbance system behav-
ior pertaining to system recovery and collapse. Past work performed in this
respect can be found in [26] for the single-infeed HVDC and [29], [30] for the
multi-infeed HVDC case. For this type of problems, the common investigation
technique is to use nonlinear time-domain simulation using an electromagnetic
transients program [29], [30].
6
Planning and Analysis of Systems -
Principles and Concepts

6.1 Background
In the actual application and installation of HVDC schemes in the last few
years, there is an impression that not so much attention has been paid to the
power/voltage stability phenomenon. This is probably because most, if not all,
recent HVDC schemes have been applied into systems where the inverter AC
system is relatively strong, that is, with a high Equivalent Short Circuit Ratio
(ESCR). In these cases, it is unlikely that power instability would be a problem
unless unusual operating conditions were imposed. However, this phenomenon
is still very real and a major issue whenever a conventional, line-commutated
thyristor-controlled HVDC inverter is applied into an AC system with a low
ESCR.
In the past, a number of DC schemes experienced the full effects of this
power instability, including in some cases the resulting consequences of system
collapse. In these systems, the problem has been mitigated by invoking special
supplementary controls on the DC or by a combination of controls and special
compensation considerations [4], [42], [43].
In recent years, new developments in HVDC technology provide the potential
to significantly reduce any manifestation of the power stability problem in most
systems. Developments such as capacitive-commutated converters (CCC) and
voltage-source inverters (VSI) now make it theoretically possible to apply DC
links into much weaker AC systems without these problems. Both the controls
and the parameters incorporated into these new technologies are such that they
present positive impacts towards the prevention or mitigation of instability.
The application development of these technologies is still progressing and a full
treatment of their theory, operation and system effects are beyond the scope
of this document. However, it is important to note that in these cases the
basic principles of the stability phenomenon do not change. But the modes of
operation and limits of DC application can change.

6.2 Simplified Explanation of Power/Voltage Insta-

bility Phenomenon1
A power/voltage instability in a conventional, line-commutated HVDC system
can basically arise out of an inter-relationship between the AC voltage, DC
controls and the DC current. It manifests itself mainly in “weak” AC systems
because it depends almost entirely on the “stiffness” of the AC voltage. In simple
1 This section is identical with section 1.2, but since it is essential for the subsequent

discussion in this chapter it is repeated in full here.

63
64 6. Planning and Analysis of Systems - Principles and Concepts

terms, the basic mechanism of the instability occurs as follows: Suppose the AC
voltage drops just a small amount; then the DC voltage at the inverter drops
proportionately (they are directly related). But the main function of the DC
controls in many DC schemes is to hold constant power order (Pd = Ud · Id ).
If the DC voltage (Ud ) goes down, the only way the controls can maintain
constant power is to increase the DC current (Id ). However, as soon as it does
this, the reactive power consumption of the DC increases because this is directly
related to the DC current. Unless the additional reactive power is immediately
(very important - this will be explained later) available, the AC voltage will dip
further and start the cycle all over again. In other words, an unstable cycle will
be started and perpetuated which, unless it is stopped in some way, can lead to
a runaway condition and a total collapse of the entire AC/DC system.

6.3 “Weak-Connection” Versus “Weak-System” In-

Feed of a DC Link

Interaction phenomena between AC and DC systems are usually discussed in the

context of weak systems in general, without qualifications about the topology
of the AC system. That is, a weak system is any system that results in a low
ESCR connection of a DC converter.
At this time, it is of great clarifying value to introduce a new definition for
the concept of a “weak-connection” of a DC converter, to differentiate from
weak systems in general for AC/DC interaction phenomena, but particularly
for the DC power stability phenomenon.
A weak system or low ESCR connection of a DC converter can result from
a number of topological circumstances. One case, which is generally considered
the classic case within this topic, is where the DC power in-feed is large com-
pared to the strength of the entire composite AC system. A sub-case of this is
where the large DC power is from multiple DC in-feeds. This classic case is the
one most likely to require the need for converters in power control (or system
frequency or damping control) and for the inverter in minimum extinction angle
control as dictated by the DC design and operating economics. Therefore, most
of the stability principles discussed herein apply to the greatest extent to this
topological case.
It is quite feasible, however, to have a low ESCR connection of a DC con-
verter to an otherwise large or ’strong’ AC system. This condition will be
defined as the weak-connection case. A number of existing HVDC schemes are
like this. The usual circumstance will be when a relatively low power DC link,
often a back-to-back link, is connected to a particularly weak part of the AC
network. This could result, for example, if the converter is connected to only
one or two long, radial AC lines, electrically remote from the strong part of the
main network. For low power weak-connection cases, it may be technically and
economically feasible to avoid power stability problems by employing controls
other than constant power, such as constant current, and by utilizing extinction
angles higher than minimum with the inverter in voltage control, for example.
6.4. Important Parameters of the System 65

6.4 Important Parameters of the System

The reason that the unstable mechanism described above manifests itself pre-
dominantly in weak AC systems can be explained by the following concept:
The amount of power that can be delivered into an AC system from a DC
link is mainly dependent on the reactive supply to the inverters and on the
AC voltage at the inverter commutating bus. When a converter is connected
to a weak system, increasing DC power will tend to lower the AC bus voltage
dependent on the ability of the system and connected elements to supply the
rising reactive power requirements of the valves. As the DC current is increased,
the DC power will also increase until a point is reached when the rising reactive
consumption begins to lower the AC voltage, and therefore the DC voltage, at a
faster rate than the increase in current. The result is that any further increase
in current will actually lead to a decrease in DC power, giving the instability.
This instability with DC is similar to the voltage instability or voltage col-
lapse mechanisms that occur in pure AC transmission networks. The instability
involves all quantities when it occurs, including power, DC current, DC voltage,
firing angles and AC voltages. The dominant feature of the instability, however,
from the system point of view, is voltage collapse at the inverter. It is therefore
often called voltage instability, similar to instability in AC systems. However,
in DC, the current can be absolutely controlled, even if the AC voltage is not
controlled or maintained. If the current ordered is too high for the system to
support, then falling AC voltage will result in reduced power, but the system
can still operate without runaway and it can be temporarily stable but unsat-
isfactory. If, on the other hand, the DC is operating in power control and with
a power order too high, then the falling DC power results in increasing current
order. This in turn then decreases the AC voltage, through larger reactive con-
sumption, to such an extent that it leads to a runaway and unstable condition.
Since the instability is thereby caused by a power closed loop control, it can be
viewed as a power instability from the DC point-of-view and a voltage instability
from the AC point-of-view. If operation in a particular system happens to be
near the critical point of instability, then small changes in the system conditions
can produce large voltage changes and possible oscillations.
It should be noted that the DC power instability, in its most simple the-
oretical form, is mathematically at zero frequency and in this form it is not
dependent on the response time of the DC line or the controls. The general
conditions of occurrence of this instability are then normally as follows:

1. The DC is in power control or, similarly, in frequency or system damping

control, where the power order could be in excess of a critical value. Even
for a previously safe power level, this same power could become critical for
a sudden reduction of the AC system strength or AC voltage, due to line
switch-out or capacitor switch-out, respectively, for example, if adequate
margin to instability is not provided to account for such perturbations.

2. The inverter is in constant (minimum) extinction angle (gamma) control,

implying increased reactive consumption with increasing current.

3. The equivalent short circuit ratio (ESCR) (per unit admittance of AC

system plus filters, capacitors and compensators on DC power base, see
section 2.8) must be less than a certain critical value, assuming that the AC
66 6. Planning and Analysis of Systems - Principles and Concepts

system stays constant; that is, before any dynamic compensation or volt-
age regulators have a chance to respond. From the AC system viewpoint,
the critical value of ESCR and hence the maximum available DC power
should then be dependent on the Thevenin voltage behind reactance and
the equivalent system reactance in the classical case of AC system steady
state stability. Real systems, however, do not act as fixed Thevenin equiv-
alents over long times beyond a few hundred milliseconds. The Thevenin
voltage in this case is not fixed. The general effect can be instability oc-
curring in the form of relaxation oscillations at a low frequency, generally
less than 5 Hz, depending on the response lags of machine voltage regu-
lators correcting AC voltages, and on communication and master control
lags in the DC The criterion for stability, however, will be practically the
same as for a fixed Thevenin equivalent. If the DC control is for constant
frequency, then the relaxation oscillation frequency will be lower because
of system inertia effects, but the stability criterion will again be similar.

The criteria for power/voltage instability addressed above does not say anything
about electro-mechanical stability because reactive power flow is the critical
quantity, while in electro-mechanical stability, phase-angle shifts and frequency
are predominately governed by active power flow. In addition, it is only valid for
fundamental frequency. Steady state stability of the AC voltage is a prerequisite
for normal system operation. However, the problem can become worse if the AC
system is weak in the inertia sense, that is, if the total system inertia constant
is so low as to cause substantial frequency drop during a fault or other event.

In other chapters, it is explained that for a given system with a finite

impedance, expressed in terms of the ESCR, there is theoretically a maximum-
available-power (MAP ) that is allowable from DC for stable operation. Assum-
ing static conditions in an AC/DC system, a general equation can be derived
that gives the MAP level. As a corollary, if the maximum DC power is given,
the equation can give the minimum strength (highest equivalent impedance or
minimum ESCR) AC system required to support such a DC power level. Such
equations are given and explained in subsection 2.2.1 and references [4], [44],
[45].

The theoretical MAP, again in the simplest form, assumes a completely

static system devoid of any dynamic voltage control. This MAP level can be
visualized as a ’snapshot’ of the system with everything temporarily fixed in
time and condition, or as the maximum steady-state operating point of that
system with no dynamics existing to improve upon it. As long as it is kept in
mind that MAP is just one point in a dynamic system, then knowing this MAP
level can provide valuable insight for planning the system and to determine what
requirements there may be for satisfactory operation.

With everything fixed in time and condition (including controls), any at-
tempt to operate that fixed system above its MAP, or, conversely, any attempt
to weaken the system when it is operating at its MAP, implicitly implies the
onset of voltage collapse in that system. The effects that the real, non-fixed,
system controls have on stability will be treated in subsequent sections below.
6.5. Relationship Between DC Stability and Recovery from AC Faults 67

6.5 Relationship Between DC Stability and Recov-

ery from AC Faults
Before discussing the effects of dynamic voltage controls, another very impor-
tant fundamental concept should be reviewed. In the planning of systems, the
frequency of faults in the interconnected AC network and their overall effects
on system design and performance usually results in these faults being one of
the central topics in system studies.
In various studies on DC links, a general conception is often conveyed that
if a DC link can recover satisfactorily from the worst-case AC faults, then it can
be assumed to be stable for all conditions. This can be very misleading.
From the equations referenced above, it can be shown that the minimum
strength system, expressed as the critical ESCR, required for power/voltage
stability will typically be in the order of about 1.5 to 2.0, depending on the
various values of the dc system parameters used in the equations. However, it is
known in the industry that DC controls and control strategies can be designed to
allow DC schemes to recover from AC faults in systems with ESCRs as low as l.0.
This is even without any dynamic voltage assistance from compensation devices.
The point here is not to discuss how this is done, but merely to point out that
this is another domain of the controls, quite separate from the power/voltage
stability and associated critical ESC R that is being addressed here.
The purpose of a fault recovery strategy is to limit the AC power being
injected into the AC system so that at each instant of time, AC voltage and
DC power are increasing toward normal steady-state values. The fault recovery
strategy and corresponding response may in some cases share some phenomena
with true power/voltage instability, but generally they cannot be considered
as the same. If a particular system is stable in the steady state, then a fault
recovery strategy can generally be devised to allow the DC link and AC system
to recover to that steady state.
It follows that when studies on a particular AC/DC system conclude that
it requires a certain ESCR to recover, all that is being done is to make a state-
ment about a specific set of circumstances in that specific system. But this
specific conclusion bears no direct relevance or does not prove anything about
the critical ESCR associated with MAP and DC stability. The later is a true
generalized theory of DC with only a few basic parameters identified. Therefore,
the recovery from AC faults is generally out of context in this topic. From the
theory and equations referenced above, there is only one steady-state critical
ESCR (or MAP ) for a given DC system. The only way it can be changed is to
change a basic parameter contained within the confines of the classical DC the-
ory equations. Note that the DC system is emphasized because the AC system
and its possible topology and components does not even enter into the equation.
The definition is from the DC side, not the AC side.

6.6 Effect of “Quasi-Static” Controls on Stability

Notwithstanding that the voltage collapse at an AC/DC interface is fundamen-
tally similar to the classic voltage collapse in an AC system, there are some
important related control aspects to consider in recognizing the system dynam-
ics when a DC link is involved.
68 6. Planning and Analysis of Systems - Principles and Concepts

In this context, it is useful to first differentiate what can be called “quasi-

static” controls from continuous, fast acting, dynamic voltage controls that may
be associated with nearby reactive power compensation devices or with the DC
itself. These will be dealt with later.
With respect to the normal DC controls, if the DC link is in constant mini-
mum extinction angle control and constant power control, the onset of voltage
collapse just beyond the MAP level will result in an unstable rundown of the
AC voltage and DC power. This is, of course, assuming that no other dynam-
ics enters into the picture. The DC constant power controller, in attempting
to compensate for the collapsing AC/DC voltage, increases the DC current,
which further aggravates the AC voltage drawdown. A total rundown can be
prevented, in theory, by reverting to constant DC current control, similar to
a control that may be employed in a weak-connection converter. This is effec-
tively equivalent to forcing the DC to operate on a single, stable operating point
on a different maximum power curve (MPC). But if the constant-current level
corresponds to a power level more than the original system MAP, a stable op-
erating point may be temporarily reached with the AC voltage unsatisfactorily
low. If the DC uses some control other than constant power (with minimum
extinction angle) in normal operation, it is sometimes argued that such a link
is more stable than one controlling power. However, unless a DC link has a
deliberate over-design that is never to be used, which would usually only be
economic for low-power weak-connection cases, the maximum power that can
be obtained from the DC occurs at the minimum extinction angle at full rated
DC current. If the link ever operates there, even temporarily and no matter
what the normal control, then the implication on voltage stability is exactly
the same at that point for all cases and all schemes, regardless of other control
modes. It should be noted that, even if the constant power control closed loop
is opened, resulting in a pure constant current control, there is still a theoreti-
cal maximum power which can be supported by the AC system before voltage
collapse takes place.
The points made above lead to the contention that the theoretical MAP
and the critical minimum strength system values obtained from the “simple”
equations are much more widely applicable and appropriate than may often be
given credit.
Also, in the context of quasi-static controls, some mitigative measures can
be enacted to either prevent or stop instability.
Based on the simple explanation given previously, the most obvious way to
arrest the instability is to stop the normal controls from increasing the DC cur-
rent whenever the AC voltage drops. But the normal controls must be allowed
some margins and freedom, otherwise, other good things like regulation and
damping during disturbances would also be stopped. Therefore, some under
voltage threshold must be applied. In some systems, a Udhold or similar control
does this function. Basically, once the AC and hence DC voltage falls to a cer-
tain value, the Udhold ’fools’ the normal controls into thinking nothing further
is happening and it stops increasing the DC current. This control can at least
stop a runaway condition. But if this is the only protection, and if it is not
functioning or if it fails, then a collapse could still result. Note also that this is
equivalent to using a door-stop. It prevents the door from slamming the wall,
but it does not close the door again. Therefore, this control ’stop’ may not in
itself prevent a bad operating situation.
6.7. Effect of Dynamic Controls on Stability 69

While a Udhold control can be a good back-up protection, there are other
quasi-static or single action reactive controls, that are more effective in arresting
a potential collapse, but still within the DC controls. This is by a SUVC (System
Undervoltage Control) or by a PRBC (Power Runback Control). In principle,
once the AC voltage falls to a certain level, these controls reduce or ramp-down
the DC power by a set amount. This not only stops the collapse but it assists
the AC voltage to recover by releasing reactive power back into the AC system,
thereby returning the system back to a satisfactory operating state (it closes
the door again).
The controls described above can only have limited capability to protect the
system, and this is by design. In SUVC or PRBC controls, if the DC power
were to be reduced by a large amount whenever an AC undervoltage occurs,
then instability or collapse could almost always be prevented. But this would
be generally unsatisfactory for system operation and for system requirements.
In any system, the critical question will revert to determining how severe a con-
tingency condition, reflected on the AC bus, that these controls must cater for
and protect against. As examples, in many systems a determining contingency
may be the loss of a large filter, capacitor bank or other large reactive power
compensation device, including possible loss of nearby generating machines.

6.7 Effect of Dynamic Controls on Stability

Following on the above, in addition to any actions taken in the DC controls,
the controls and characteristics of any nearby compensation equipment such as
synchronous compensators (SCs) or Static Var Compensators (SVCs) can have a
profound and rather complex effect on the voltage stability. Their basic function
is to supply reactive power to the DC and to control the AC commutating
voltage on a fast and continuous basis.
It was stated at the outset that unless additional reactive power is imme-
diately available when the AC voltage drops, then the AC voltage will reduce
further and the instability will be perpetuated. But if a connected compensat-
ing device had an infinitely fast response and unlimited range, then no limiting
MAP would theoretically exist and system voltage collapse and DC instability
would be impossible. On the other hand, if the compensator response were
very slow or if it reached some compensating limit, then its ability to be able
to prevent a system collapse would be correspondingly limited. The important
factors are that any voltage controller must be sufficiently fast to correct any
incremental voltage variations before ’runaway’, and that it must have sufficient
dynamic range to ’cover’ for any event considered.
It is therefore established that there are two basic elements associated with
the total analysis of the MAP and the voltage stability. One is the equivalent
system impedance (usually considered as a static element and often expressed
as ESCR) and the other is the voltage control (the dynamic element).
In the consideration of SVCs applied to DC, unlike synchronous compen-
sators or other machines, SVCs do not directly contribute to system strength.
It is sometimes stated that they can indirectly provide the equivalent to system
strength by forced control action. But this appears to be a statement based
mainly on their effective over voltage control, by activating controlled reactive
absorption in the thyristor-controlled-reactor element. Power instability in a
70 6. Planning and Analysis of Systems - Principles and Concepts

weak system, however, is an AC under-voltage problem, not an over-voltage

problem. And most SVCs by virtue of their most efficient conventional design,
will cut-off their reactive absorption capability or be non-conducting in the reac-
tive control element near full DC power, when the DC valves require maximum
reactive power. But in contradiction, this is when the control is most needed for
stability. If, on the other hand, the SVCs are over-designed to prevent cut-off in
the reactor element at full DC load, then operation at normal conditions would
be very inefficient and larger capacitor banks or filters must also be installed
to counteract this design. Moreover, the reactive power loss from any static
capacitor elements, on voltage reduction, is proportional to the square of the
voltage drop, which then compounds the problem. This can create a severe
design ’chase’, whereby additional static capacitor banks can also lead to other
problems such as greater over voltages, lower resonant frequencies with the AC
system, aggravated DC operation due to voltage distortions and a decreased
commutation failure performance [46], [47].
As an alternative to the SVC scenario discussed above, if the switching-on
of static capacitors could be done fast enough in an emergency for stability,
this could be done, in principle, without a controlled reactor element of a ’full’
SVC. This would suggest that a fast-control, thyristor-switched-capacitor might
be better for DC application. However, it should be noted that in thyristor-
switched-capacitors, small finite steps are switched with response times up to
one cycle plus control circuit decision time, resulting in dead bands of operation.
This may not be able to stabilize, by itself, a DC converter and theoretically
could make things worse if it results in any continuous jumping in and out of
stable operation.
It is especially interesting to extend the above basic considerations to the sit-
uation of synchronous compensators (or other synchronous machines) connected
to or nearby an inverter AC bus. Although expensive, synchronous compen-
sators are most effective in curing instability because they genuinely contribute
to the system strength (reduce the equivalent system impedance) through their
impedance in the classic and static sense. In reality, however, this contribution
is not static, but dynamic, and this will be discussed extensively below. Be-
sides the assistance for power/voltage stability, synchronous compensators also
have the advantages of providing inertia, raising the harmonic resonance of the
system, improving the harmonic behaviour of the AC system and improving
commutation failure performance [46], [47].

6.8 Introducing the Concept of “Dynamic MAP”

In discussing the concepts to this point, the only assumption about the AC side
is that the AC bus voltage remains constant for the steady state determination
of critical ESCR or MAP. But when a dynamic variation in the AC bus voltage
occurs in a real system, including a response to this variation by any machine or
compensator controls, then the MAP point can in principle continuously vary
with time within the same system perturbation. This produces what can be
termed a “dynamic MAP ” [45].
These system dynamics could, in theory, be accounted for in the voltage
stability phenomenon in two different ways. A more complete equation-set
could possibly be developed to include the dynamics. Alternatively, the classic
6.8. Introducing the Concept of “Dynamic MAP” 71

equation-set (assuming static conditions) could be retained and the dynamic

effects of controls or machines could then be reflected in the system reactance
representation for the MAP and ESCR calculations. An example would be
the reflected machine reactance representation for synchronous compensators.
The analysis and impacts of dynamic system modelling on the power stability
of DC systems has recently been significantly advanced in research and in the
literature [35].
In the case of retaining the existing equation-set and for a given AC system
makeup, the task becomes one of finding the representation of that AC system
that, in fact, fits the already defined critical ESCR.
For the case where synchronous machines or compensators are connected
at or near to the inverter AC bus, for this stability phenomenon, the machine
representation is coupled to its control dynamics. Depending on the design of
the machine and its field controls, the machine dynamics can be significantly
influential throughout a time frame extending from the fastest power frequency
voltage changes right through to steady state. In other words, depending on
the stability situation and event, the machine representation that is most ap-
propriate for the “dynamic MAP ” or that best fits the critical ESCR can, in
principle, range from the lowest possible value of subtransient reactance, Xd ,
to the direct axis reactance, Xd .
For example, an Xd representation could apply, even during a dynamic per-
turbation, if the machine goes, or is forced to, an overexcited field current limit,
thus relinquishing dynamic AC voltage control. In this condition, a slow sys-
tem rundown or collapse can occur unless something special is done to stop it.
Actual system experience and studies [44], [45], [49] have demonstrated that
this condition, while in the past being considered as the separate phenomenon
of “steady state var starvation”, is actually just a boundary point of exactly
the same principle as the power instability mechanism in faster time frames.
Moreover, if this condition can occur in a DC scheme with nearby machines, it
could represent the worst scenario for the instability because it would require
the highest machine reactance (Xd ) to meet the minimum critical ESCR allowed
for stability.
At the other ’end’ of the of the reactance spectrum, at Xd , a different story
must be told. If nearby machines can be kept in continuous automatic-voltage-
regulator (AVR) control during a dynamic event, that is, kept away from any
rating or excitation limits, and if the machine controls can be made fast enough
to absolutely control the AC bus voltage at impending voltage collapse, then
the appropriate machine reactance that would apply to the critical ESCR could
in theory approach Xd . The entire problem then simply reduces to making
the machine controls fast or effective enough to produce a “dynamic MAP ”
which would be equivalent to a steady state MAP corresponding to a machine
representation using Xd . Stated another way, as long as the machines are in
continuous AVR control, then the faster the machine response, the lower the
equivalent reactance in that time frame (Xd the lowest) and the more it can
contribute to the equivalent system strength for stability considerations.
There has been a long standing debate in the industry over whether Xd or

Xd is the most appropriate to use for these considerations. However, the Nelson
River DC system with its synchronous compensators [43] demonstrates that it
is realizable to design modern machines with exciters fast enough such that Xd
can be the appropriate equivalent reactance determining stability. Of course,
72 6. Planning and Analysis of Systems - Principles and Concepts

where a DC system has older machines with slow exciters in close proximity,
then this is a different matter.

6.9 System Studies of Stability on Actual System

To gain an understanding of the effects of control on the power/voltage instabil-
ity in DC links, the Nelson River HVDC system has been extensively modelled
on transient stability and other programs [45], [49] . This system, which has
been well documented [42], [43] employs 6-160 plus 3-300 Mvar synchronous
compensators, in addition to 757 Mvar of filters at the inverter for the two
bipoles. The maximum rating of the two bipoles is 3854 MW. The design ESCR
is 2.5 based on Xd of the machines, and with two machines out-of-service (the
design criteria). The ESCR is, of course, much less when based on Xd or Xd
of the machines, which shows the importance of the machine representation for
various phenomena. For normal system operating conditions with all elements
in-service and normal loading, the operating ESCR is above 3.
For purposes of modelling in these studies, both bipoles were ’locked’ in
power control with no switching to current control. Both the System Under-
voltage Controller (SUVC), which reduces DC power a fixed amount when the
AC voltage falls below a set value, and the Udhold control, which “freezes” DC
current by way of the power-divide-by-voltage, were inactivated. These con-
trols had been instituted on this system after power instability and collapse
shutdowns were experienced. Also for the studies, the gains in all compensator
exciters were proportionately reduced from their normally tuned values. Com-
pensator numbers were reduced and capacitors were added to lower the ESCR to
the point of voltage instability for a particular exciter gain. The compensators
were always kept in a position where they retained non-limited AVR control.
The instability was ’triggered’ by a small increase in DC power or by a small
filter trip. The results of this investigation are shown in Figure 6.1.
To explain Figure 6.1, at each voltage regulator gain setting indicated on
the horizontal axis, the vertical-set of three “x-points” on the three curves are
derived from the single system which results in the onset of voltage instability,
when the compensators are operating with that particular gain setting. That
is, each vertical-set of three “x-points” represents one given system with a fixed
number of compensators and capacitors connected. Then, the vertical axis
values for the three “x-points” (and hence the three curves) are determined
by calculating the ESCR of that system corresponding to the three machine
reactances of Xd , Xd and Xd . Going from one gain setting to the next on the
horizontal axis, or from one set of three “x-points” to another, means going to a
different system with a different number of machines and capacitors to produce
the onset of instability.
The horizontal dashed line (ESCR = 2) represents the critical ESCR of
the DC system based on the classic steady state stability equations referenced
previously. The relatively high value of ESCR = 2 results partly from the DC
operating condition being at 1.1 p.u. loading with the AC bus at about 0.95 p.u.
initial voltage (the worst operating condition criteria). This does not mean that
such stringent operating conditions are required to produce instability - it just
means that inducement is easier. By comparing the three “x-point” levels at
each gain setting relative to the level of the critical ESCR of the dc system, it can
6.9. System Studies of Stability on Actual System 73

Onset of Voltage Instability for Various Excitation Gains

7

ESCR(X’’d) Dashed line

6 indicates critical ESCR
based on classical steady
state stability equation
5

ESCR(X’d)
4
ESCR

ESCR(Xd)

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Relative Gain of Voltage Regulator

Figure 6.1. Curves showing the onset of voltage instability for various
excitation gains on the synchronous compensators connected to the in-
verter AC bus (Nelson River HVDC system), as a comparison against the
critical ESCR based on the classic steady state stability equation.

be determined which reactance is most appropriate to use in the steady state

equivalent for calculating MAP ; or for calculating the ESCR to test against
the critical ESCR. The “x-point” closest to the critical ESCR = 2 (in this
case), based on the given system conditions, will indicate the most appropriate
reactance. For example, at the normal voltage regulator gain (fastest response),
the curves show that the ESCR based on Xd is the most appropriate to test
for stability and to compare to the critical ESCR. As the effectiveness of the
voltage regulator is changed from the normal fast control to a slower response
(smaller values of gain), the curves show that the appropriate machine reactance
for stability changes from Xd to Xd and finally, with the exciters disabled, to
Xd .
These results show the importance of control to the voltage instability phe-
nomenon. A modern, properly tuned exciter allows the use of Xd in the equa-
tions and so maximizes the contribution of the machine.
When a machine is on a field current limit, the above results show that Xd
will govern its contribution to the ESCR. Depending on the particular machine,
Xd may be ten times or more than the Xd quantity. For HVDC systems that
depend on synchronous compensators or other machines for strength (generally,
SC’s would not be installed unless strength was required), then a field current
74 6. Planning and Analysis of Systems - Principles and Concepts

limit applied to these machines during operation leads to a dramatic reduction

in the ESCR or the “dynamic MAP ”. For any operating states that produce
this limit, there is likely to be a significant violation of the voltage stability
criterion.
These considerations are important for any synchronous compensator design.
There should be enough transient capability in the machines so that corrective
action can be initiated before a field current limit is enforced.
The Nelson River DC system inverter bus is so dependent on synchronous
compensators for reactive supply, as well as strength, that it is probable (virtu-
ally certain at full DC load) that the tripping of more than one compensator, or
a combination of one compensator plus filters, will put the remaining machines
on a limit before any Xd based instability is ’triggered’. The critical reactance
would then revert to Xd , creating the condition for a slow voltage collapse mode
characteristic of a “steady state var starvation”.
Considering actual system experience, some years ago the inverter bus of this
system suffered a number of voltage collapse events, which were only arrested
by the out-of-step tripping of AC tie lines. The situation started developing
at the time of increased power levels on the DC due to new generation and
without an increase in reactive compensators. Generally, the collapse would
be precipitated by the loss of a reactive power producing element, either a
compensator or a filter, which would then force the operating compensators to
an output limit. In fact, there has been no known situation, when the machines
became limited, that did not initiate a voltage collapse. Conversely, there has
not been a situation where the machines have been in full AVR control and a
collapse initiated. These observations support the theoretical development of
the above sections.
The initiation of the voltage collapse was at the time thought to be caused by
simple “var starvation” - nothing to do with any critical ESCR. The theory for
the power/voltage instability, which gained prominence later, was considered to
be something separate. In reality, it was subsequently established, as explained
above, that the “var starvation” was power instability, with Xd as the repre-
sentative machine reactance. This is a more complete manner to describe these
events because the lack of continuous voltage control does not necessarily mean
a voltage collapse. There are many DC links that operate successfully with-
out continuous AC voltage control as long as their operating states are above a
critical ESCR (below their MAP ). It is rather the change in machine reactance
from Xd to Xd , which leads to a large change in ESCR and a resultant violation
of the voltage stability criterion.
When the 3-300 Mvar synchronous compensators were added to the Nel-
son River system for additional strength, a formula for the critical ESCR was
developed for use in the planning studies and for specifying the new machines
[43]. The critical ESCR = 2 was derived from this formula. However, due to
popular belief at the time, the new machines were specified to meet this target
ESCR based on a machine reactance of Xd instead of Xd in order to ensure
operation below MAP. It was not until later that the implications of Xd and
Xd as discussed above became known. It should be pointed out that for nominal
conditions on this link, with DC power and AC voltage of 1.0 p.u., instead of
1.1 p.u. and 0.95 p.u., respectively, the critical ESCR is about 1.66.
Besides the additional compensators, there are now two controls installed on
the Nelson River DC, which are intended to help correct or prevent the voltage
6.10. Implications for Other Systems 75

collapse problem. The first is the System Undervoltage Control (SUVC), which
performs a relatively minor (about 300 MW) reduction when a collapsing AC
voltage is detected. The second is a Udhold control, which substitutes a fixed
DC voltage into the constant-power-division circuit upon low AC voltage. While
Udhold should prevent a total collapse initiated by compensators on a limit, the
violation of the criterion can be severe enough to cause a large change in AC
voltage.
The main benefit of the SUVC is not in the small increase in ESCR from
the reduction in DC power, but rather in the return of controllability to the
compensators. Thereafter, the contribution of the compensators to the fault
level in the ESCR calculation is based on Xd instead of Xd . This is a dramatic
increase in ESCR, which now makes the system stable. The SUVC is thus
a dominant control in preventing voltage instability on this system for many
possible conditions.

6.10 Implications for Other Systems

For a general HVDC system with no synchronous compensators, SVC’s, or other
machines nearby to the inverter AC bus, the calculation of the system state rel-
ative to the voltage stability criterion is fairly straightforward. For synchronous
machines electrically further out into the AC network, the question of correct
machine reactance and dynamics is of reduced concern since intervening line
reactance will usually make the machine contribution and influence less impor-
tant.
In addition to the discussion already made on SVC’s, a system with a SVC at
an inverter bus presents some further interesting considerations. With the effects
of control ignored, the susceptance at which the SVC is operating is factored
into the ESCR calculation for power stability. A SVC acting capacitively lowers
the ESCR, which may be critical in an already weak system or even in the
case of a weak-connection DC in-feed. It is therefore probable that, if the SVC
is to be of any benefit, the control will be critical for viable operation. If the
control were infinitely fast in correcting small-signal deviations, thereby keeping
voltage absolutely constant, then it would totally ’hide’ the weakness of the
system or the DC connection point in terms relevant to stability. Of course, the
closed loop control cannot be infinitely fast and in reality will tend to be tuned
according to the minimum strength of the AC system into which it operates -
the weaker the system, the slower the response. This would likely lead to an
upper bound on the assistance a ’conventional’ SVC can give to the prevention
of voltage instability as the system becomes weaker. For DC application, the use
of STATCOM’s, which can more closely emulate the favourable characteristics
of synchronous compensators, may present some advantages. But the economics
of these devices will be critical to their viability.
When synchronous compensators are to be applied at an inverter, there
are certain aspects of the design which are particularly important for voltage
stability. The previous section has shown that keeping the AVR in full control is
necessary in order to produce an effectively ’strong’ dynamic system - in other
words, the case using Xd in the ESCR calculation. Furthermore, to maintain
control in contingency situations, it will usually be desirable and economic to
have some overload capability in the machines, to allow time for corrective
76 6. Planning and Analysis of Systems - Principles and Concepts

action before more restrictive limits are applied. To ensure a fast-acting control
system, the effective time constant of the voltage measurement circuit of the
AVR should be low, perhaps 25 milliseconds or less. For modern AVR designs,
this should not be a problem. The field-forcing positive-limit may have some
affect on stability for more severe disturbances by quickly returning the AC
voltage to a safe level. Finally, the contribution of the machine to the fault
level should be maximized from the perspective of stability, meaning the lowest
realistic total of subtransient reactance plus transformer reactance. Of course,
the down side of this is the possible implication on the fault rating of system
circuit breakers and other equipment.
From all the foregoing discussion, the impression has no doubt been made
that the worst single problem associated with incorporating an HVDC link into
low ESCR weak-connections or weak AC systems may well involve the interre-
lated issues of DC power stability and reactive compensation at the inverter.
Also, these issues are major problems from both technical and economic per-
spectives. All this applies to conventional line-commutated converters.
For relatively large power DC links, any requirement for sophisticated com-
pensation equipment such as synchronous compensators or SVCs will represent
major capital investment - a significant proportion of the total link cost. This
increases the difficulty for conventional HVDC to compete against evolving new
technologies in AC transmission systems and other supply alternatives. In some
areas, it will also increase the difficulty to compete against alternative energy
solutions that may also be aided by the restructuring and deregulation in the
electrical industry.
If new DC technology, such as capacitive-commutated-converters or voltage-
sourced converters, can eliminate the stability/compensation problems and be-
come more economic, this will greatly enhance the possibility of new, large
power HVDC links being applied in relatively weak or moderate AC systems.
References

[1] Cigré Task Force 38-02-12, “Criteria and Countermeasures for Voltage Col-
lapse”, Paris 1995.

[2] IEEE Committee Report, “Voltage Stability of Power Systems: Con-

cepts, Analytical Tools, and Industry Experience”, IEEE/PES 90TH0358-
2-PWR, 1990.

[3] Hill, D.J., Hiskens, I.A., and Popovic, D., “Load Recovery in Voltage Sta-
bility Analysis and Control”, Bulk Power System Phenomena III, Voltage
Stability, Security and Control, ECC, Davos, 1994, pp. 579 - 595.

[4] Cigré Working Group 14.07, “Guide for Planning DC Links Terminating
at AC Systems Locations Having Low Short-Circuit Capacities, Part I:
AC/DC Interaction Phenomena”, Cigré Guide no. 95, Paris 1992.

[5] Ainsworth, J.D., Gavrilovic, A., Thanawala, H.L., ”Static and Synchronous
Compensations for HVDC Transmission Converters Connected to Weak AC
Systems”, CIGRE General Session, Paris, France, Paper 31-01, 1980.

[6] Uhlmann, E., “Converters Connected to Networks with Limited Short-

Circuit Capacity”, Archiv für Elektrotechnik, Springer-Verlag, No.0003-
9039/81/0063, pp.267- 274, 1981 (original in German translated to English
by G.Andersson).

[7] Hammad, A.E., Sadek, K., Kauferle, J., “A New Approach for the Analysis
of and Solution of AC Voltage Stability Problems at HVDC Terminals”,
Proceedings of International Conference on DC Power Transmission, Mon-
treal, Canada, pp.164-170, Jun.1984.

[8] Nayak, O.B., et.al, “Control Sensitivity Indices for Stability Analysis of
HVDC Systems”, IEEE Transactions on Power Delivery, Vol.10 No.4,
pp.2054-2060, Oct.1995.

[9] Reeve, J., and Uzunovic, E., “Study of Power Transfer Capability of DC
Systems Incorporating AC Loads and a Parallel AC Line,” IEEE Trans. on
Power Delivery, Vol. 12, No. 1, pp. 426 - 432, January 1997.

[10] Franken, B., Andersson, G., “Analysis of HVDC Converters Connected

to Weak AC Systems”, IEEE Transactions on Power Systems, vol.5 no.1,
pp.235-242, Feb.1990.

77
78 REFERENCES

[11] Perez-Arriaga, I.J., Verghese, G.C., Schweppe, F.C.,“Selective Modal Anal-

ysis with Applications to Electric Power Systems, Part 1: Heuristic In-
troduction”, IEEE Transactions on Power Systems, vol.PAS-101 no.9,
pp.3117- 3125, Sep.1982.

[12] Lee, H.A.D., Andersson, G., “Voltage Stability Analysis of Multi-Infeed

HVDC Systems”, IEEE Transactions on Power Delivery, vol.12 no..3, pp.
1309-1316, Apr. 1997.

[13] Lee, H.A.D., Andersson, G., “Use of Participation Factors in Modal Voltage
Stability Analysis of Multi-Infeed HVDC Systems”, IEEE Transactions on
Power Delivery, vol.1 no..13, pp. 203-211, Jan. 1998.

[14] Lee, H.A.D., Andersson, G., “Power Stability Analysis of Multi-Infeed

HVDC Systems”, IEEE Transactions on Power Delivery, vol.3 no..13, pp.
923-931, Jan. 1998.

[15] Lee, H.A.D., Andersson, G., “Influence of Load Characteristics on the

Power and Voltage Stability Analysis of HVDC Systems, Part 1: Basic
Equations and Relationships”, IEEE Transactions on Power Delivery, vol.4
no..13, pp. 1437 - 1444, Jan. 1998.

[16] Lee, H.A.D., Andersson, G., “Influence of Load Characteristics on the

Power and Voltage Stability Analysis of HVDC Systems, Part 2: Stability
Margin Sensitivity”, IEEE Transactions on Power Delivery, vol.4 no..13,
pp. 1445 - 1452, Jan. 1998.

[17] Lee, H.A.D., “Voltage and Power Stability of HVDC Systems”, Ph.D The-
sis, Royal Institute of Technology, Sweden, TRITA-EES-9801, ISSN 1100-
1607, March 1998.

[18] Pilotto, L. A. S., Szechtman, M. and Hammad, A. E; “Transient AC voltage

related phenomena for HVDC schemes connected to weak AC systems”,
IEEE Trans. Power Delivery Vol. 7 , No.3 Jul 1992 , pp 1396-1404.

[19] Szechtman, M., et al.,(1991): “First Benchmark Model for HVDC Control
Studies” Electra 135. 1991. 54-73

[20] Reeve, J., “Short Circuit Ratio and Related Factors Characterizing AC
Systems for DC Transmission,” International Conference on Large High
Voltage Electric Systems - CIGRE, Colloquium “Interaction Between AC
and DC Systems”, England, September 1985.

[21] Gavrilovic, A., Krishnayya, P. C. S., Peixoto, C. A. O., Ainsworth, J. D.,

and Bowles, J. P., “Aspects of AC/DC System Interaction: Peak Available
Power, second Harmonic Resonance, Low Inertia Systems, Controllabil-
ity of HVDC,” IEEE Conference on HVDC Power Transmission, Canada,
September 1986.

[22] Lefebre, S., Gole, A. M., Reeve, J., Pilotto, L., Martins, N., and Bhat-
tacharya, S., “Report on Test Systems for AC/DC Interaction Studies,”
IEEE Trans. on Power Delivery, Vol. 10, No. 4, pp. 2027 - 2034, October
1995.
REFERENCES 79

[23] Jonsson, T., and Björklund, P., “Capacitor Commutated Converters for
HVDC”, Proceedings of Stockholm Power Tech, Part: Power Electronics,
pp. 44 -51, Stockholm, 1995.

[24] Diago, J.R. et al, “Garabi the Argentina - Brazil 1000 MW Interconnec-
tion”, Proceedings of VIII ERLAC, Cidade del Peste, Paraguay, 1999.

[25] Andersson, G., “Some Considerations on the Short-Circuit Capacity Re-

quirements for an HVDC Terminal”, Report of the department of Electric
Power Engineering, Royal Institute of Technology, Sweden, TRITA-EES-
9203: ISSN 1100 1607, June 1992.

[26] Nayak, O. B., et.al, “Dynamic Performance of Static and Synchronous

Compensators at an HVDC Inverter in a very Weak AC System”, IEEE
Transactions on Power Systems, Vol.9 No.3, pp.1350-1357, August 1994.

[27] Krishnayya, P.C.S., Piwko, R.J., Weaver, T.L., Bahrman, M.P., Hammad,
A.E., “DC Transmission Terminating at Low Short-Circuit Ratio Loca-
tions”, IEEE Transactions on Power Delivery, Vol.PWRD-1 No.3, pp.308-
318, July 1986.

[28] Thallam, R. S., “Review of the Design and Performance Features of HVDC
Systems Connected to Low Short-Circuit Ration AC Systems”, IEEE
Transactions on Power Delivery, Vol.7 No.4, pp.2065-2073, October 1992.

[29] Bui, L. X., Sood, V. K., Laurin, S., “Dynamic Interactions between HVDC
Systems Connected to AC Buses in Close Proximity”, , IEEE Transactions
on Power Delivery, Vol.6 No.1, pp.223-230, January 1991.

[30] Szechtman, M., et. al, “The Behavior of Several HVDC Links Terminating
in the Same Load Area”, Cigre General Session, Paris, France, Paper 14-
201, 1992.

[31] Lips, H. P., “Aspects of Multiple Infeed of HVDC Inverter Stations into a
Common AC System”, IEEE Transactions on Power Apparatus and Sys-
tems, Vol.92 No.2, 1973.

[32] Reeve, J., Lane-Smith, S. P., “Multi-Infeed HVDC Transient Response and
Recovery Strategies”, “, IEEE Transactions on Power Delivery, Vol.8 No.4,
pp.1995-2001, October 1993.

[33] Pilotto, L. A. S., et. al, “Synchronizing and Damping Torque Modulation
Controllers for Multi-Infeed HVDC Systems”, IEEE Transactions on Power
Delivery, Vol.10 No.3, pp.1505-1513, July 1995.

[34] Ärnlöv, B., et. al, “Aspects of Cooperation of Three HVDC Links between
Southern Denmark. Southern Sweden and Northern Germany/Jutland”,
Cigre Symposium, Tokyo, Paper 220-02, 1995.

[35] Lee, H.A.D., Andersson, G., “Impact of Dynamic System Modeling on the
Power Stability of HVDC Systems”, IEEE Transactions on Power Delivery,
Vol. 14, No 4 (1999), pp. 1427 - 1437
80 REFERENCES

[36] Lee, H.A.D., Andersson, G., “Nonlinear Dynamics in HVDC Systems”,

IEEE Transactions on Power Delivery, Vol. 14, No 4 (1999), pp. 1417 -
1416
[37] Lee, H.A.D., Andersson, G., “Modal Analysis of Multi-Infeed HVDC Sys-
tems”, Proceedings of the International Power Engineering Conference, Vol.
1, pp. 310 - 315, May 1997, Singapore
[38] Smed, T., Andersson, G., “Utilizing HVDC to Damp Power Oscillations”,
IEEE Transactions on Power Delivery, Vol.8 No.2, pp.620-627, April 1993.
[39] Arabi, S., et. al, “Small Signal Stability Program Analysis of SVC and
HVDC in AC Power Systems”, IEEE Transactions on Power Systems, Vol.6
No.3, pp.1147-1153, August 1991.
[40] Jovic, D., Pahalawaththa, N., Zavahir, M., “Stability Analysis of HVDC
Control Loops”, IEE Proceedings on Generation, Transmission, and Dis-
tribution, Vol.146 No.2, pp. 143-148, March 1999.
[41] Padiyar, K. R., “Dynamic Analysis of Voltage Instability in AC/DC Sys-
tems”, International Journal of Electric Power and Energy Systems, Vol.18
No.1, pp. 11-18, 1996.
[42] Thio, C.V., “Nelson River HVDC Bipole-Two Part I - System Aspects”,
IEEE Transactions on Power Apparatus and Systems, Vol. PAS-98, No. 1,
Jan./Feb. 1979.
[43] Thio, C.V., Davies, J.B., “New Synchronous Compensators For The Nelson
River HVDC System - Planning Requirements And Specification”, IEEE
Transactions on Power Delivery, April, 1991.
[44] Thio, C.V., Davies, J.B., “Preliminary Considerations Of Controls And
Dynamics Relevant To Voltage/Power Stability Of An AC/DC system”,
Report to CIGRE WG 14-05, Paris, France, August, 1992.
[45] Thio, C.V., “Power/Voltage Stability In AC/DC Systems”, Paper pre-
sented to CIGRE WG 14-05, Paris, France, September, 1998.
[46] Thio, C.V., Davies, J.B., Kent, K.L., “Commutation Failures in HVDC
Transmission Systems”, IEEE Transactions on Power Delivery, April, 1996.
[47] “Commutation Failures in HVDC Transmission Systems Due to AC System
Faults”, Report by CIGRE WG 14-02, CIGRE Electra No. 169, December,
1996.
[48] Ainsworth, J.D., “Power Limit Instability (Voltage Instability) in an HVDC
Link Connected to a Weak AC System”, Report to CIGRE SC14, Septem-
ber, 1985, Hollingbourne, England.
[49] Yin, Y.H., (EPRI China in conjunction with Manitoba Hydro), H. M.
Turanli, “Studies on Power/Voltage Instability Considering Voltage Con-
trollability and Dorsey ESCR”, Manitoba Hydro Report, May 31, 1989.