484 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO.

1, JANUARY 2006

A Platform for Validation of FACTS Models

Shan Jiang, Student Member, IEEE, U. D. Annakkage, Senior Member, IEEE, and A. M. Gole, Senior Member, IEEE

Abstract—The paper presents a platform system for the incor-

poration of flexible ac transmission systems (FACTS) devices. The
platform permits detailed electromagnetic transients simulation as
it is of manageable size. It manifests some of the common problems
for which FACTS devices are used such as congestion management,
stability improvement, and voltage support. The platform can be
valuable for the validation of reduced order models such as small
signal or transient stability models. The paper presents details on
the development and validation of a small signal based model with
the inclusion of a Unified Power Flow Controller. The validated
model is then used successfully for the design of a feedback con-
troller for improved damping.
Index Terms—Electromagnetic transients simulation, FACTS,
Prony analysis, small signal analysis, test platform, validation,
UPFC.
Fig. 1. One line diagram of the 12-bus power system.

I. INTRODUCTION transient stability modeling, only the electrical machines, con-

trols, and prime movers are represented using time domain dif-
T HE expansion of power transfer capability of transmis-
sion systems has been a major problem over the past two
decades. This, together with the advancement of solid-state
ferential equations. A simplified fundamental frequency phasor
equivalent of the ac network is used, which permits larger time-
technology, has paved way to a series of new Power Electronic steps of up to 10 ms, thereby making possible the representation
devices which are capable of extending the power transfer ca- of larger networks. Alternatively, Small Signal Analysis takes a
pability limits of transmission systems through their flexibility markedly different approach in which the network and electro-
and response speed. Among them, Static Var Compensators mechanical device equations are converted at any given oper-
(SVCs), static synchronous compensators (STATCOMs), and ating point into a set of linear differential equations. Eigenvalue
Thyristor Controlled Series Compensators (TCSCs) have been and Spectral Analysis applied to these equations yields useful
widely accepted in the industry; whereas applications using the information on stability margins and damping and permits stan-
Unified Power Flow Controller (UPFC) are recently emerging dard controller design techniques to be effectively applied.
[1]. The most recent device in the family is the Interline Power One of the concerns that arises is whether the simplified
Flow Controller (IPFC), which has only been installed on an models such as the Small Signal or Transient Stability models
experimental basis. adequately represent the detailed system. Usually, simplified
A large amount of research effort has gone into designing representations are developed and benchmarked against a very
these devices and studying the impact of these devices on the small emtp-type model, which essentially includes the device
performance of the power system. Modeling plays an important connected to a single machine or an infinite bus [2]. Similarly
role in such design and application studies. At the most detailed other comparisons have been made that benchmark a small signal
level, electromagnetic transients simulation (emtp-type) based model with a transient stability model. This type of comparison is
models are used. In these models, the detailed three phase repre- of limited use as it compares two simplifications with each other
sentation of the system is simulated using a 10–50 s time-step. rather than with the most accurate (emtp-type) representation.
The operation of individual switching elements and control sys- This paper aims to address the above concern of validation by
tems in the FACTS device as well as all magnetic saturation in proposing a platform system which is large enough to demon-
transformers is also fully represented. This level of modeling strate electromechanical oscillation modes; and is yet small
is useful for confirming the operation of the FACTS device in a enough to be completely realizable in an emtp-type form. This
local setting but is often considered too detailed for investigating system can be represented in different levels of detail which can
the impact of the device on the wider electrical network. Like- then be compared. The proposed platform has been designed so
wise, it does not yield information about the damping and sta- that it manifests typical transmission bottlenecks and interarea
bility margins in a straightforward manner. On the other hand, in oscillations that can be alleviated by FACTS devices. The
procedure for validation is elucidated by connecting a UPFC
into the platform (see Fig. 1) and comparing its small signal
Manuscript received November 29, 2004; revised February 22, 2005. Paper representation against the detailed emtp-type representation
no. TPWRD-00561-2004. using the technique of Prony Analysis. The UPFC is installed
The authors are with the University of Manitoba, Winnipeg R3T 2N2,
Canada. in line 7–8 of the platform system; its primary purpose is to
Digital Object Identifier 10.1109/TPWRD.2005.852301 relieve congestion in line 1–6.
0885-8977/$20.00 © 2006 IEEE
JIANG et al.: PLATFORM FOR VALIDATION OF FACTS MODELS 485

II. SMALL POWER SYSTEM FOR MODEL VALIDATION

The proposed test system, shown in Fig. 1, consists of 12
busses (six 230 kV busses, two 345-kV buses and four 22-kV
buses). The test system covers three geographical areas (Area
1, 2, and 3). Area 1 is predominantly a generation area with
most of its generation coming from hydro power. Area 2, situ-
ated between the main generation area (Area 1) and the main
load center (Area 3), has some hydro generation available but is
insufficient to meet local demand. Area 3, situated about 500 km
from Area 1, is a load center with some thermal generation
available. Furthermore, as Area 2 generation has limited energy
availability, the system demand must often be satisfied through
transmission. The transmission system consists of 230 kV trans-
mission lines with the exception of one 345-kV link between
areas 1 and 3 (between busses 7 and 8). Areas 2 and 3 have Fig. 2. Emtp-type UPFC Model.
switched shunt capacitors to support the voltage.
Power flow studies reveal that in the event of a loss of genera-
tion in area 3, or a loss of the transmission line between busses 4
and 5, line 1–6 is overloaded while the transmission capacity of
the parallel path through the 345 kV transmission line 7–8 is un-
derutilized. This congestion can be relieved by various FACTS
solutions such as TCSCs or SSSCs on line 1–2 or line 7–8;
IPFCs on lines 1–6 and 7–8; or, as in the example presented
here, a UPFC on line 7–8.
Further, the load center (Area 3) suffers from under-voltage Fig. 3. Decoupled controller of shunt converter.
problems, which makes this test system suitable for studies on
application of SVC or STATCOM. lumped pi-sections. The representation for the UPFC in the line
Small signal stability studies presented in Section IV show connecting busses 7 and 8 is shown schematically in Fig. 2.
that the platform system has poorly damped inter-area oscilla- Every switching device in each of the series and shunt con-
tion modes which can be improved by FACTS devices at various verters is modeled individually. Firing pulses for the converter’s
locations. The study presented later in this paper shows one such switches are generated by the sinusoidal pulse width modulator
application. (SPWM) that eliminates low-frequency harmonics and makes
It is also possible to use the platform to investigate the use of the output waveform conform to the desired fundamental
FACTS devices to strengthen the network for the integration of frequency voltage waveform. The reference waveform for the
wind generation in Area 2. For example series FACTS devices SPWM modulator is generated from direct and quadrature
on lines 1–6 and/or line 6–4 and/or line 7–8 could make stronger components Vd, Vq. As shown in Fig. 3, in the shunt converter,
connections from Area 2 (with wind generation) to the other they are selected to provide the appropriate real and reactive
areas. powers from the shunt element using a decoupled controller
Thus, it can be seen that the proposed platform can be useful [3]. The real power order is generated from a dc bus voltage
for studying FACTS device applications for congestion relief, controller and the reactive power is directly ordered (often set
voltage support, transmission stability and integration of wind to zero for unity power factor operation). The decoupled control
generation. One such application, that of designing a UPFC- system ensures that a change in the real power order can be
based damping controller is discussed in this paper. It also be- implemented without any transient in the reactive power and
comes possible to validate reduced models such as Small Signal vice versa. The series converter’s controller can be modeled in
Stability models against detailed emtp-type simulation for such similar detail if necessary; however in this particular study, the
FACTS applications. These aspects will be covered in the re- d and q components of the series injected voltage are directly
mainder of the paper. ordered to ensure the required load flow.

III. DETAILED ELECTROMAGNETIC TRANSIENT SIMULATION IV. SMALL SIGNAL STABILITY MODEL
MODEL OF UPFC
A. Generator Model
Because the electromagnetic transient simulation model In the small signal model, the generators with their exciters
represents the system in extreme detail, it provides a means for are represented by the typical fourth-order dynamic model
validating the small signal and other reduced models for the (third-order generator plus a first-order exciter)
network. For this purpose, the system in Fig. 1 is represented
in full, with detailed models for the synchronous machines (1)
(including full representation of sub-transient effects), exciters.
The lines can be represented with distributed parameters or as (2)
486 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 1, JANUARY 2006

(3)

(4)

These equations can be linearized and expressed in the com-

pact form for number of generators, as shown in (5) and (6).
The derivation of this model is found in [4] and [5] Fig. 4. UPFC small signal model.

(5)
The voltage and current relationship for the UPFC are given by
(6) (10)
where

is the change in current injected by generator to the

network, and is the voltage of the network node to which
the generator is connected. (10)

B. UPFC Model where and are, respectively,

Charging of the capacitor on the dc link of UPFC is modeled the sending end and receiving end line currents expressed as
as two-dimensional (2-D) vectors consisting of the real and imag-
inary parts of the current phasor. Similarly, and are 2-D
vectors representing the sending and receiving end voltages, and
(7)
is a 2-D vector representing the boost voltage or the voltage
injected in series with the transmission line. The matrix in
where is the capacitance of the dc bus capacitor, is (10) is the dq to xy transformation matrix. Equations (7)–(10)
the voltage of the dc bus, is the real power drawn into the can be linearized to obtain the set of differential and algebraic
dc link from the shunt (exciter) branch of UPFC, and is the equations (11)–(13). A detailed derivation of the elements of all
real power flowing out of the dc link into the transmission line coefficient matrices is given in [5]
through the booster side of UPFC. The PI Controller that main-
tains at the reference setting of is modeled as
(11)
(8) (12)
(13)
where is the d component of the shunt current that is in
where
phase with the voltage Vs (see Fig. 4), and are PI con-
troller gains. Any time delays in PWM and associated controls
are modeled as

is a state variable associated with the PI controller.

The voltage-current relation of the network is modeled as

(9)
(14)
where we have the following.
The state-space representation of the complete power system
1) , and are the reference control settings of the can be obtained in the standard format of (15) and (16) by elim-
UPFC, namely, the in-phase injected voltage, quadrature inating and from the differential-algebraic equations
injected voltage, and shunt current, respectively. of the dynamic devices (5), (6) and (11)–(13) and the network
2) , and are the in-phase injected voltage, equations (14).
quadrature injected voltage, and quadrature shunt cur-
rent, respectively.
3) , and are the time constants used to model (15)
the PWM time delay associated with , and . (16)
JIANG et al.: PLATFORM FOR VALIDATION OF FACTS MODELS 487

TABLE I
MODES OF OSCILLATION

TABLE II
PARTICIPATION FACTORS

Fig. 5. Mode shape of 0.75-Hz mode.

V. ANALYSIS AND VALIDATION OF SMALL SIGNAL MODEL

The small signal model yields important information about
the oscillation modes and damping. This information is vali-
dated for accuracy using the full emtp-type representation.

A. Small Signal Eigenvalues

The system has three generators with each equipped with an
exciter. Each generator introduces three state variables, and each
exciter introduces one state variable. The UPFC introduces five
state variables. Therefore, the order of the system matrix is
17. The complex conjugate pairs of eigenvalues of system ma- Fig. 6. Mode shape of 0.85-Hz mode.
trix correspond to oscillation modes. Table I shows the eigen-
values corresponding to electromechanical oscillation modes
for the selected operating point of the power system. Note that
the oscillation frequencies of the three modes are 0.75, 0.85, and
1.12 Hz.
Further information about the oscillation modes can be ob-
tained from the participation factors and mode shapes. The par-
ticipation factors represent the activity of state variables in a
given mode [4]. Table II shows the participation factors of the
state variables in the three modes of oscillation. Each column
of Table II gives the participation factors for a given mode. It
can be noticed that the activity of generator G2 dominates in
0.85 Hz, as indicated by the relatively large participation factor
(0.4632) of state . Similarly, G3 dominates in the 1.12-Hz
mode, and G4 dominates in the 0.75-Hz mode.
The elements of the right eigenvector indicate the response of
state variables when the corresponding mode is excited. There- Fig. 7. Mode shape of 1.12-Hz mode.
fore, by observing the magnitude and phase of the elements of
the eigenvector corresponding to the state variables , In Section VI, a damping controller for the UPFC is designed
and , one can predict the relative magnitudes and phase an- in order to improve the damping of oscillations. Considering
gles of the rotor oscillations when the particular mode is excited. the topology of the network in Fig. 1, the UPFC is directly in
This information can be plotted on the complex plane to obtain the path between the infinite bus and generator G3, and thus a
what is known as the mode shape [4]. The modes shapes of the damping controller for the UPFC can be expected to be most ef-
three oscillatory modes are shown in Figs. 5–7. fective in improving the damping of modes related to G3. Partic-
488 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 1, JANUARY 2006

Fig. 8. Dominant mode of G2 from Prony analysis. Fig. 10. Dominant mode of G4 from Prony analysis.

Fig. 9. Dominant mode of G3 from Prony analysis. Fig. 11. Comparison of sending end real power of small signal model versus
emtp-type model.

ipation factors suggest that this mode is the 1.12-Hz mode. The
TABLE III
mode shape diagrams also suggest similar conclusions. It is evi- PRONY ANALYSIS OF PSCAD WAVEFORMS
dent from these diagrams that the rotor angle of Generator 3 os-
cillates when the 0.75-Hz mode is excited (Fig. 5) and when the
1.12-Hz mode is excited (Fig. 7). On the other hand, Fig. 6 indi-
cates that there is negligible influence on Generator 3 when the
0.85-Hz mode is excited. As the UPFC in line 7–8 primarily af-
fects the power in Generator 3, this suggests that the damping of
the 1.12-Hz mode and/or the 0.75-Hz mode could be improved
with its installation. A more quantitative evaluation of the effec-
tiveness of controllers is presented in Section VI. The frequency
information obtained from the eigen-analysys is verified against equations. The close agreement validates the Small Signal Anal-
the simulation results from a detailed electromagnetic transient ysis model.
simulation model in Section V-B. The electromagnetic simulation can also be used to verify
the modal frequencies, damping and participation factor infor-
B. Validation Using Electromagnetic Transients Simulation mation. Prony Analysis was performed in order to validate the
An electromagnetic transient simulation of the power system small signal model. Results of Prony analysis on the waveforms
with a switching level detailed model of the UPFC was per- from emtp-type simulation of the speed of three machines
formed using PSCAD/EMTDC. The disturbance applied was a , and the sending end power of the transmission
1% increase of the reference setting of the quadrature injected line on which the UPFC is located (line 7–8) are tabulated in
voltage of the UPFC for a period of 100 ms. Figs. 8–11 show Table III. The following observations can be made by com-
the simulated waveforms for the speeds of gener- paring this information with the eigenvalues given in Tables I
ators 2–4 and the sending end power, respectively. Superposed and II, which were obtained from the small signal model.
on the emtp-type simulations are the waveforms obtained from 1) Although the 1.12-Hz mode is presented in all four wave-
time-domain simulation of the linearized Small Signal Analysis forms (Prony analysis estimates in the range 1.11 Hz to
JIANG et al.: PLATFORM FOR VALIDATION OF FACTS MODELS 489

TABLE IV TABLE V
CONTROLLABILITY INDEX EIGENVALUE SENSITIVITY TO FEEDBACK GAIN

1.13 Hz), as expected from the participation factor anal-

ysis in Table II, it is most prominent in . The damping
TABLE VI
for this mode (in ) is 3.74%, which compares well MODES OF OSCILLATION: WITH FEEDBACK CONTROL
with the 3.30% damping calculated from the small signal
model in Table I.
2) As expected from the theoretical participation factors, the
0.85-Hz mode (estimated by Prony Analysis as 0.79 Hz)
is prominent in . Although both Prony analysis based
observation and models attribute the poorest damping to
this mode, the observed value of 3.56% is significantly
different from the 1.07% predicted by theory. However,
this error may be due to the small magnitude of the ob-
served signal and due to interference with the relatively
close 0.75-Hz mode.
3) Again, as predicted by Small Signal theory, Prony anal-
ysis confirms that the 0.75-Hz mode is prominent in ,
with an observed damping of 8.245%, in comparison
to the 7.17% from Small Signal Analysis. Both theory
and simulation observations show that this mode has the
highest damping.
With the confidence in accuracy gained from the above anal-
ysis, the small signal model can now be used effectively for Con-
troller Design, as described in the next section.
Fig. 12. Damping effect on the sending end real power. Small signal model.

VI. DESIGN OF DAMPING CONTROLLER (real power of UPFC sending end) are considered as feedback
The question arises as to whether the UPFC placed in line candidates. The effectiveness of each of these signals is evalu-
7–8 for transmission congestion relief can also be modulated to ated by calculating the eigenvalue sensitivity to feedback gain
damp system oscillations. Ideally one would like to increase the [6]. These sensitivities are given in Table V. As the objective
damping of the poorly damped 0.85- and 1.12-Hz modes. Con- is to move the real part of eigenvalue in order to improve the
trollability analysis performed on the small signal model can de- damping, the sensitivity index with the largest real part identi-
termine if such modes can be damped at all, after which linear fies the most effective feedback signal, which is . However,
control design theory can be used to select the optimal feedback in a practical controller, it is usually desirable to choose a
controller. There are three reference control settings available local signal, hence the second best signal is selected. The
in the UPFC for damping control. These are , and . larger sensitivity of the real part of the eigenvalue with than
The effectiveness of using these inputs can be evaluated by cal- with implies that is the better choice for the controller
culating the controllability indices of the modes [4]. These in- input. Once the feedback structure (i.e., feedback and input) is
dices are easily obtained once the eigenvalues and eigenvectors identified, the gain of the feedback controller can be selected.
of the small signal system are evaluated. The controllability in- Using a feedback gain of 0.3 produces a new set of eigenvalues
dices are given in Table IV. The relatively large magnitude of for the system, with the three poorly damped ones shown in
the indices for the 1.12-Hz mode using inputs and indi- Table VI. Note the slight change in the frequency from 1.12
cate the suitability of using these inputs to improve the damping to 1.10 Hz and the improved damping of this mode from 3.30
of that mode. Very small indices for the 0.85-Hz mode indicate (as in Table I) to 4.35. Time domain simulation of the small
that this mode is not controllable using any of the control inputs signal model with and without feedback as in Fig. 12 shows
of the UPFC. These findings are consistent with the mode shape this additional damping.
based arguments made in Section V-A. Once again, as shown in Fig. 13, detailed emtp-type simu-
Based on the above analysis, an output feedback damping lation produces nearly identical responses to those obtained in
controller is designed to improve the damping of the 1.12-Hz Fig. 12 for the small signal model. Using Prony Analysis on the
mode by modulating either the or input. The feedback detailed simulation waveforms the frequencies and damping are
signal is generated by multiplying the selected output marked determined, as in Table VII. The damping as observed from the
for feedback with a simple proportional gain. The outputs most prominent signal for this mode is improved from 3.74
(real power of generator G1, G2, G3), and to 4.92, which compares well with the theoretical value of 4.35.
490 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 1, JANUARY 2006

TABLE IX
CONFIGURATION OF TRANSMISSION LINE

Fig. 13. Damping effect on the sending end real power. emtp-type model.

TABLE VII
PRONY ANALYSIS OF PSCAD WAVEFORMS: WITH FEEDBACK CONTROL

Fig. 14. Transmission line structure.

TABLE VIII
BUS DATA
TABLE X
BRANCH DATA (SYSTEM BASE: 100 MVA)

The Platform can be extended to the study of other FACTS

devices in a similar manner.

VII. CONCLUSION APPENDIX

The paper presents a platform that can be used for FACTS The data pertaining to the power system shown in Fig. 1 are
controller studies. The platform has been selected to demon- given in the following tables.
strate a number of problems for which FACTS devices offer po- Table VIII(a) shows the loads and shunt compensation at load
tential solutions. These include congestion relief, voltage sup- buses 1–8, and Table VIII(b) shows the specified voltage and
port, and stability improvement. The system was selected so that real power generation at generator buses 9–12. In the transient
it is large enough to show interarea oscillations and yet man- model, the loads are represented as fixed impedances.
ageable so that detailed electromagnetic transients solutions are The geometrical and physical parameters for the transmission
possible. lines are as shown in Table IX and Fig. 14. All 230-kV lines are
The platform provides a valuable tool for benchmarking re- assumed to have the same geometrical and physical parameters
duced-order models. The example in the paper shows how the (except for different lengths). The 230-kV line is based on Man-
small signal model developed for a UPFC is incorporated into itoba Hydro’s Glenboro-South to Rugby line, and the 345-kV
the small signal framework. Detailed comparisons with emtp- line has a typical structure selected from Table 2.7.1 of the EPRI
type simulations show close agreements with the Small Signal “transmission line reference book [7]. Table X shows the line
Analysis in the time domain and in the observed frequencies and lengths for each of the 230- and 345-kV lines, as well as the se-
damping as determined by Prony Analysis. ries impedances and shunt reactances resulting from the above
JIANG et al.: PLATFORM FOR VALIDATION OF FACTS MODELS 491

TABLE XI [4] P. Kundur, Power System Stability and Control. New York: McGraw-
TRANSFORMER DATA (SYSTEM BASE: 100 MVA) Hill, 1994.
[5] S. Limyingcharoen, “Application of Unified Power Flow Controllers in
Power System Stability Enhancement,” Ph.D. dissertation, Dept. Elect.
Electron. Eng., Univ. Auckland, Auckland, New Zealand, Mar. 1999.
[6] F. L. Pagola, I. J. Perez-Arriaga, and G. C. Verghese, “On sensitivities,
residues, and participations: Applications to oscillatory stability analsis
and control,” IEEE Trans. Power Systems, vol. 4, pp. 278–285, 1989.
[7] Transmission Line Reference Book (345 kV and Above), Second ed.
Palo Alto, CA: Elect. Power Res. Inst., 1987, p. 39.

TABLE XII
GENERATOR AND EXCITER DATA
Shan Jiang (S’04) received the B.Sc. and M.Sc. (Eng.) degrees in electrical
engineering from Chongqing University, Chongqing, China, in 1989 and 1993,
respectively. He is currently pursuing the Ph.D. degree with the University of
Manitoba, Winnipeg, MB, Canada.
His research interests are FACTs and power system control.

TABLE XIII
UPFC DATA U. D. Annakkage (M’95–SM’04) received the B.Sc. (Eng.) degree in electrical
engineering from the University of Moratuwa, Moratuwa, Sri Lanka in 1982
and the M.Sc. and Ph.D. degrees from the University of Manchester Institute
of Science and Technology (UMIST), Manchester, U.K., in 1984 and 1987, re-
spectively.
He is presently a Professor with the University of Manitoba, Winipeg, MB,
line geometries for an equivalent- representation, corrected for Canada. His research interests include power system stability and control, secu-
rity assessment and control, operation of restructured power systems, and power
long-line effects (see also Tables XI–XIII). system simulation.

REFERENCES
[1] L. Gyugyi, “Unified power-flow control concept for flexible ac transmis-
sion systems,” in Proc. Inst. Elect. Eng., vol. 139, 1992, pp. 323–331. A. M. Gole (M’82–SM’04) received the B.Tech. degree in electrical engineering
[2] L. Y. Dong, L. Zhang, and M. L. Crow, “A new control strategy for the from the Indian Institute of Technology, Bombay, India, in 1978 and the Ph.D.
unified controller,” in Proc. IEEE PES Winter Meet., vol. 1, Jan. 2002, degree from the University of Manitoba, Winnipeg,MB, Canada, in 1982.
pp. 562–566. He currently holds the NSERC Industrial Research Chair in Power System
[3] I. Papic, P. Zunko, D. Povh, and M. Weinhold, “Basic control of uni- Simulation at the University of Manitoba. His research interests include tran-
fied power flow controller,” IEEE Trans. Power Syst., vol. 12, no. 4, pp. sients simulation and power electronics applications in transmission.
1734–1739, Nov. 1997. Prof. Gole is a Professional Engineer in the Province of Manitoba.