This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON POWER SYSTEMS 1

Co-Ordinated PSS Tuning of Large Power Systems

by Combining Transfer Function-Eigenfunction
Analysis (TFEA), Optimization, and Eigenvalue
Sensitivity
Reza Jalayer, Member, IEEE, and Boon-Teck Ooi, Life Fellow, IEEE

Abstract—This paper shows that a combination of: 1) the com- power electronic controllers of FACTS to supplement the
putation time-saving transfer function and eigenfunction analysis capabilities of individual PSS [14]–[16]. Phase measurement
(TFEA) method, 2) eigenvalue sensitivity concept, and 3) optimiza- units (PMUs) offer wide-area measurements by which large
tion techniques makes a powerful tool in coordinated tuning of
power system stabilizers (PSS). The combination allows different power systems can be stabilized by a central controller which
PSS tuning strategies to be evaluated as objective functions of op- effectively is a “Global PSS” [16]–[18] that augments the
timization under constraints. Feasibility of the method is demon- damping of local PSS.
strated by numerical results from a 69-generator system. Two ob- For PSS tuning, eigenvalue sensitivities came to be used as
jective functions are used as illustrative examples. a result of modal analysis [19]–[21]. Linear programming [6],
Index Terms—Damping, eigenvalue, eigenvalue sensitivity, opti- [24] and methods based on optimization techniques have been
mization, parameters, power system stability, power system stabi- successfully applied to tuning [22], [32]–[37]. In regard to im-
lizers (PSS), small signal stability, tuning.
plementation, one of the important factors is computation cost.
This is because evaluating eigenfunctions by the QR method in-
I. INTRODUCTION creases with the cube of the matrix dimension.
The objective of this paper is to show that the computa-
tion time-saving transfer function and eigenfunction analysis
A S electric utilities interconnect to take advantage of
economy of scale, the size increase has resulted in
weakened dynamic systems with low frequency and lightly
(TFEA) of [23] can be combined with eigenvalue sensitivity
concepts and optimization techniques to address PSS tuning
of large power systems. In implementation and validation, the
damped oscillatory modes. Power system stabilizers (PSS)
steps are:
were developed to increase system damping. The operation
1) Apply (TFEA) method of [23] to form the reduced state
and design of a PSS for a single generator was explained
matrix of which contains all the electro-
as phase-compensation of the transfer function between the
mechanical modes only.
voltage reference input to the AVR and the electrical torque
2) Apply eigenvalue sensitivity formula to relate the damping
developed on the shaft of the generator [1]–[3]. The P-vr
coefficients of the electromechanical modes to the PSS
method, as the phase-compensation method is known, has been
gains and time constants, which form pa, the vector of tun-
extended to the robust design of PSS in multi machine systems
able system parameters.
[4]–[9]. The concept of residues of the eigenfunctions proved
3) Find pa to fulfill optimization functions under constraints.
to be useful in PSS design and tuning [4]–[12]. These advances
4) Use the pa solved in Step 3 to form matrix
have been accompanied by eigenfunction evaluations of the
and the full state matrix. Use the QR method
time invariant [A] matrix of the linearized system equations
to solve their eigenvalues. When their eigenvalues are in
[11]–[13].
close agreement, the method is validated.
With the advent of flexible AC transmission systems
5) Review if objectives and constraints are met to confirm the
(FACTS), damping improvement is sought by co-ordinating
feasibility of the combined method.
With powerful optimization algorithms available in MATLAB
Manuscript received May 14, 2013; revised September 13, 2013 and
for example, PSS tuning objectives can be cast as optimiza-
December 27, 2013; accepted March 27, 2014. This work was supported by
a strategic project grant of the Natural Sciences and Engineering Research tion objective functions under constraints. The authors have
Council (NSERC) of Canada on “Design and Operation of Trans-Canadian explored many ideas such as whether the poor damping of
Power Grid” awarded to Principal Investigator, Professor F. D. Galiana. Paper
inter-area modes can be increased at the expense of highly
no. TPWRS-00601-2013.
The authors are with the Department of Electrical Engineering, McGill Uni- damped local modes. This paper presents only two optimiza-
versity, Montreal, QC H3A 2A7, Canada (e-mail: [Link]@[Link]; tion examples: Objective Function 1 centered on minimizing
[Link]@[Link];)
the PSS gains under constraints as in [6]; Objective Function 2
Color versions of one or more of the figures in this paper are available online
at [Link] centered on maximizing the damping of the lowest frequency
Digital Object Identifier 10.1109/TPWRS.2014.2314717 modes [16].

0885-8950 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See [Link] for more information.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE TRANSACTIONS ON POWER SYSTEMS

TABLE I The perturbation parameters are

CHOOSING ACCORDING TO MODAL FREQUENCY RANGE
(6)
(7)
(8)
(9)
(10)

B. Eigenvalue Sensitivity: Change of With Respect to

The (ng-1) tuple vector of eigenvalue shift from can be
written as

(11)

Fig. 1. Exciter and PSS model. Tuning consists of adjusting the size of a perturbation pa-
rameter . From (1)–(5), the ng generators have a total of
parameters.
All approximate methods have inherent inaccuracies. In the The eigenvalues shifts due to perturbation of one parameter
TFEA method, it lies in the frequency dependency of the re- is
duced matrix representing the electromechan-
ical modes. This is because TFEA represents the exciter, the (12)
power system stabilizer (PSS) and the amortisseur windings
where the th row of
the column vector is
as transfer functions. For economic reasons, the QR method
can be applied to evaluate the eigenvalues and eigenvectors of . This expression comes
the matrix for only a few representative fre- from the eigenvalue sensitivity formula as explained in
quency . Therefore, only the modes whose frequen- Section IV-A.
cies are close to are accurate. Reference [23] has also shown For m parameters , one
that the eigenvalue sensitivity formula can be applied to im- defines a matrix whose columns are the vectors
prove predictions of eigenvalues around representative frequen- that is
cies . From the research of this paper and also from
(13)
[23], the choice of for accurate estimation of eigenvalues in
different modal frequency ranges is shown in Table I. and

(14)
II. OUTLINE OF METHOD
The PSS model used in this research is PSS2B from IEEE Std. As the interest of this paper is only in the real part of
421.5 [25], [26]. Of the 4 Settings in [25, Table I], the authors , the formula of the damping term is
have chosen Setting 1, which is shown in Fig. 1.
(15)

A. Parameter Perturbation where

The parameters for tuning are the gain and the four time
(16)
constants , and . It is assumed that all the ng gen-
erators are equipped with the same PSS model. Therefore, the Tuning is performed in two iterations. In Iteration 1, the vec-
parameters form ng-tuple vectors , and . tors of parameters in (1)–(5) are set to zero. At the end of
To allow for perturbation in the tuning study, each pa- Iteration 1, the vector of parameters is found. In Iteration
rameter is written as , where is the value 2, the values found in Iteration 1 is used as the initial pa-
of original design. Thus rameter setting as On further tuning, a new
is found which satisfy the objective function under constraints.
(1) The final answer to the tuning is where
(2) and are the solutions of Iteration 1 and 2, respectively.
(3) C. PSS Tuning by Objective Functions
(4) In this paper two objectives are considered to demonstrate the
(5) flexibility and comprehensiveness of the method:
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

JALAYER AND OOI: CO-ORDINATED PSS TUNING OF LARGE POWER SYSTEMS 3

Objective Function 1—Tuning the PSS of the 69 generators electromechanical torque is characterized by , the syn-
by minimizing the PSS gains, under the constraint that the chronizing torque coefficient with respect to rotor position ,
damping ratios for all electromechanical modes are at least a and by , the damping torque coefficient with respect to rotor
5%. The purpose of minimizing the gains is to ensure that all speed [27]:
the electrical modes are stable [2], [6]. Therefore
(23)

The dynamic properties of the damper windings, the excita-

(17)
tion system and the power system stabilizer are formulated as
transfer functions and embedded as frequency dependent coef-
% (18) ficients and [23]. The formulation of a single
generator is expanded to ng generators to form the frequency
In (18) and are the final damping and oscillation frequency dependent matrix of (24). According to (24), it
after the PSS tuning, obtained from is necessary to know the frequency dependant matrices
and to form the reduced state matrix of .
(19) This is achieved by applying small signal stability to the com-
bined generators transfer functions and power system network
where is the vector of original eigenvalues with no PSS. equation:
is the vector of eigenvalue shift obtained from eigenvalue sen-
sitivity which are due to all the optimized parameters .
Objective Function 2—PSS tuning to maximize the damping
for low frequency modes (target modes): (24)

(20)
where
where is the weighting factor of the jth mode. One can place (25)
constraints such as having at least a 5% damping ratio for other
electromechanical modes: (26)
(27)
% (21)
(28)
(29)
as in (18).
The tuning procedure is summarized as below: (30)
Iteration 1:
With initial PSS parameters of (1)–(5) set to zero , in which is the identity matrix, and is the
apply the TFEA method to form the reduced system nominal frequency (i.e., 60 Hz), and is the
and then apply the QR algorithm to obtain and the right diagonal matrix of generators inertias:
and left eigenvectors. From the sensitivity formula in (14), the
(31)
eigenvalue shift due to are found which fulfill the ob-
jective function of (17) under constraints of (18), or the objec-
According to the generator equations, their currents are
tive function of (20) under constraints such as (21).
functions of voltages, rotor speeds and angles:
Based on the solution , a new is formed
from . New eigenvalues are found by the QR (32)
algorithm. If the constraints are all satisfied, the search in pa-
rameter space is complete. Otherwise proceed to: where
Iteration 2:
Similar to Iteration 1 but the initial parameters are from (33)
Iteration 1. would be the solution of the second Iteration.
The final tuning parameters are The generator torques are functions of currents:

(22) (34)

In (32) and (34), and are the vectors of voltages,

currents and torques deviations respectively for all the ng gen-
III. REVIEW OF TRANSFER FUNCTION AND EIGENFUNCTION
erators.
ANALYSIS (TFEA) METHOD
Also, considering the network equations (i.e., bus admittance
Although the TFEA method is described fully in [23], it is matrix) [23], the generator currents and voltages are related as
necessary to summarize the method so as to make the paper
self-sufficient. For each of the ng-generators, its perturbation (35)
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON POWER SYSTEMS

Combining (32), (34) and (35), the perturbation torque can be Unlike (14) which applies to obtain eigenvalue shift
found in terms of all speed and angle deviations by inverting [C], minimization of (17) or (20) does not require
to be in explicit form. Substituting (42) in (41) is both
(36) convenient and computation time-saving.

Comparing (23) and (36), it is possible to form and V. VALIDATION TESTS

and therefore as in (24).
A. Further Validation of Accuracy of TFEA
IV. REVIEW OF EIGENVALUE SENSITIVITY In [23], the authors have used the 16-generator 68-bus system
taken from [29] to validate the accuracy of estimating eigen-
A. Eigenvalue Sensitivity to Input PSS Parameters for PSS values using the TFEA method. This paper provides further val-
Parameter Tuning idation of the TFEA method by using the much larger 69-gener-
The relationship between eigenvalue and its eigenvector ator 300-bus test system of [28]. The results show that accuracy
is does not deteriorate with increase of system order.

(37) B. Implementation of Optimization

The optimization algorithm used is from MATLAB: Ac-
Applying small perturbation to (37), one has tive-set optimization which applies a sequential quadratic
programming (SQP) method. In this method, the function
(38) solves a quadratic programming (QP) sub-problem at each iter-
ation and updates an estimate of the Hessian of the Lagrangian
Eliminating (37) and second and higher order terms, one has
at each iteration [30], [31].
(39) Optimization is based on advancing in parameter space of the
parameters of . Because the optimization
The transpose of [A] has the identical eigenvalues as [A] ex- algorithm does not converge to an optimum if parameter con-
cept that the eigenvectors are . Similar to (37), there is the re- straints are not specified, the authors have chosen the following
lationship range of PSS parameters:

(40) (44)
(45)
Pre-multiplying (39) by , on solving for , the eigenvalue
(46)
sensitivity is
(47)
(41) (48)

The limits are chosen after careful study of [25].

In applying the sensitive formula of (41), the eigenvectors of On successful convergence, the parameters obtained
are evaluated only once by the QR algorithm. from the optimization algorithm are used to construct
This feature makes for computation time saving. and . Their eigen-
values evaluated by the QR method are labeled respectively as
B. Implicit Evaluation of
“Predicted” and “Benchmark” in Tables II and III. The electro-
In order to reduce the number of computations, the formula- mechanical modes of are sorted out from
tions of (13) and (14) are not used. For the nth mode, the per- the pairs. They are in good agreement with
turbation matrix is formed from the “Predicted”. It should be noted that all other non-electro-
mechanical modes are positively damped. The paper addresses
two objectives and their results are compared.

C. Objective 1: Minimizing PSS Gains

(42) Objective Function 1: Minimize the gains as in (17) subject to
damping-to-frequency constraints of (18). This objective func-
where is the original matrix tion follows the thinking of [6] which states, “low stabilizer
before the PSS parameter perturbation and gains reduce not only the effect of limiting in the stabilizer, AVR
is the and excitation systems, but also Mvar swings on generators for
matrix after parameter perturbations. small disturbances”. The parameters are solved in 2 iterations.
The eigenvalue shift is found by forming (42) for the mode The weighting functions are all in (17).
of and putting it in (41) which results in 1) Tuning Results: Fig. 2, which plots the locations of 68
eigenvalues, summarizes the success of the method. All the
eigenvalues lie on or to the left of the red line representing the
(43) 5% constraint of (18). With regard to minimizing the gains of
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

JALAYER AND OOI: CO-ORDINATED PSS TUNING OF LARGE POWER SYSTEMS 5

TABLE II TABLE III

69-GENERATOR RESULTS FOR PSS TUNING FOR OBJECTIVE 1 69-GENERATOR RESULTS FOR PSS TUNING FOR OBJECTIVE 2

(17), the sum of the gains is 346. When averaged over 69 gen-
erators, . For comparison, in Fig. 2. 68 eigenvalues from Objective 1 tuning.
Objective Function 2. The upper limit of (44) is .
Fig. 3 shows the time domain response to a perturbation dis-
turbance. The solid line is , calculated for generator #44.
The two dashed curves are responses of the slowest and fastest
modes.
Because of space limit, Table II presents only 16 of the
68 modes of Fig. 2: with samples of low, medium and high
frequency eigenvalues. The eigenvalues of generators without
PSS are listed in Column 1 of Table II. The first six modes
are negatively damped. There are 3 columns under each it-
eration: The column under “Predicted” is obtained from the
smaller matrix and “Benchmark” from Fig. 3. Time domain damping of for Objective 1 after tuning.
the full matrix. The third column shows the
damping ratios computed from benchmark evaluations.
2) Accuracy Check on TFEA: In Fig. 2, there are many plus modes have less than 5% error, 16 modes with error between
(predicted) and circle (benchmark) symbols which overlap or 5%–10%. There are 3 modes which exceed 16% error. As to
close together, visually showing the level of accuracy. In com- frequency predictions, 39 modes have less than 5% error. The
paring damping coefficients from TFEA with benchmark, 34 maximum error is 10%.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON POWER SYSTEMS

Fig. 6. Regions connection for 69-generator system.

The sum of the gains is 1222. The average is

. The gains of 59 PSS have reached the limit of
in (44).
The two modes at the margin of the 5% line, which have
eigenvalues of and , are both in
Fig. 4. 68 eigenvalues from Objective 2 tuning. Tables II and III. This shows that Objective 1 is more effective
because it is achieved at minimum PSS gains.
The numerical results of Objective 1 and 2 are illustrations
of the proposed method for the specific system conditions and
modeling.
2) Accuracy Check on TFEA: On damping prediction by
TFEA, 29 modes have less than 5% error, 22 modes have an
error between 5%–10%. For all the 68 modes, the maximum
error does not exceed 15%. Frequency prediction is much better
since 43 modes have less than 5% error and the maximum error
is limited to 10%.

Fig. 5. Time domain damping of for Objective 2 after tuning. VI. ROBUSTNESS TEST

D. Objective Function 2: Maximizing the Damping of Low

Frequency Modes A. Robustness of 69-Generator System After PSS Tuning

Objective 2: Maximize the damping of low frequency modes When a power system encounters fault(s) and transmission
in (20) and also having at least a 5% damping ratio for other lines are disconnected, the tuned PSS should ensure that the
electromechanical modes in (21). This is a reasonable objective degraded power system continues to be stable. When a line is
because there is concern that as the power system continues to disconnected, a new load flow problem is solved to determine
grow in size, the power system weakens resulting in low fre- the new operating conditions. The TFEA method is applied to
quency modes which are usually lightly damped [25]. show that all eigenvalues are still on the left-side of the complex
1) Tuning Results: The x-axis of Fig. 4, which has to extend s-plane.
from to 0, compared to Fig. 2, which extends from to An illustrative example based on the 69-generator system
0, clearly shows that all 68 eigenvalues have shifted to the left, [28] is presented here. The system of [28] is assumed to be made
thus showing that the objective function of (20) is fulfilled. All up of 3 regions joined by transmission lines as shown in Fig. 6.
the eigenvalues lie on or to the left of the red line representing Three cases are considered: in Case 1, line 8-1 is disconnected;
the 5% constraint of (21). in Case 2, lines 8-1 and 7-2 are disconnected; in Case 3, lines
Fig. 5, which shows the response of generator #44 to pertur- 8-1, 7-1, 4-15, 5-15, and 6-15 are out of service. In the begin-
bation disturbance, confirms that damping has significantly in- ning, 40 MW is transferred from region 1 to 2 and 45 MW from
creased compared to Fig. 3. region 3 to 1. The PSS of the 69 generators have been tuned to
Table III shows the eigenvalues of the same 16 modes of satisfy Objective Function 1.
Table II in the same order. In the beginning, the modes of the As the 69-generator system is relatively small, the eigenanal-
first 8 rows are negatively or lightly damped. The target modes ysis of the full state matrix is solved. Some
in objective function (20) are the first 8 modes, . values of the low frequency electromechanical oscillations are
The weighting functions are . listed in Table IV. The results of other modes are not shown as
The interesting point is that there is no need to do a second they are almost unchanged. Because of topological change, the
Iteration because after the first Iteration all the 68 modes have a mode shapes are different from the pre-fault modes. In all the 3
minimum of 5% damping ratio. cases, all the modes are positively damped.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

JALAYER AND OOI: CO-ORDINATED PSS TUNING OF LARGE POWER SYSTEMS 7

TABLE IV TABLE V
69-GENERATOR RESULTS UNDER DIFFERENT OPERATING CONDITIONS TIME CONSTANTS FROM OPTIMIZATION RESULTS OF OBJECTIVE 2

Fig. 9. Block diagram of generator modeling dynamic operation.

Fig. 7. Time constant from the tuning results of Objective 2.

Fig. 10. of PSS, Exciter, Generator 44 of Configuration 1.

Fig. 8. Time constant from the tuning results of Objective 2.

first instance, the rest of the system is disregarded. This is by

holding the generator angle constant. The objective is to
VII. DISCUSSION study how the time constants of Table V affect the Bode dia-
grams of of the 3 generators. The transfer func-
A. Damping Design by Optimization
tions of Figs. 10–12 begin with the rotor speed deviation
In Objective Function 2, the gains of 59 of the 69 PSS reach as input—through the PSS, the exciter, the generator electrical
the limit of in (44). Figs. 7 and 8 show the dynamics blocks and ends with the electrical torque of the gen-
corresponding values of the time constants and erator as output. The “generator electrical dynamics”
. and are not shown as their block is taken from [27]. After understanding the role played
pattern is similar to those in Figs. 7 and 8. individual PSS to damping, the rest of the system is integrated
In optimization to reach objective function (20), the parame- and Figs. 14 and 15 show the differences.
ters , and seem to migrate to the upper and Although the time constants are chosen by optimization, the
lower limits set by (45)–(48). This is illustrated by Figs. 7 and 8. Bode Diagrams of the 3 samples show successful phase com-
Optimization methods tend to locate their objectives at the pa- pensation in the frequency range of interest [27], [29]. This ac-
rameter limits. But does this tendency yield meaningful tuning? cords with classical design [2], [3] that the choice of the time
To check on this, the optimized time constants , constants must provide a phase lead in the transfer function of
and of generators #1, #10, and #44 have been selected and the PSS to compensate for the phase lag in the exciter and gener-
listed in Table V. ator dynamics. With Bode diagram of Fig. 12, for a single-gen-
The time constants of Table V are used in the PSS of the erator system of Fig. 9, the damping torque-vs-frequency curve
generator shown in Fig. 9 [29]. takes the form of Fig. 13.
Fig. 9 shows a single generator connected to the power system The damping torque when the rest of the system is included
at , the generator angle and at , the stator voltage. In the is more complex. It must take into account of the interactions
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE TRANSACTIONS ON POWER SYSTEMS

Fig. 11. of PSS, Exciter, Generator 1 of Configuration 2. Fig. 15. Damping torque of generator #44 from Objective Fuction 2.

From Figs. 13–15, the general trend is that low frequency

oscillation has low damping torque. As Objective Function 2
is optimized to have increased damping in a number of target
modes at low frequencies, Fig. 15 shows marked increase in
damping torque at the low frequency region of the spectrum.

VIII. CONCLUSION
The paper has presented a computation time-saving and ac-
curate method for coordinated tuning of the PSS of large power
systems. The method consists of a combination of: 1) TFEA of
Fig. 12. of PSS, Exciter, Generator 10 of Configuration 3.
[23]; 2) eigenvalue sensitivity analysis; and 3) optimization. The
feasibility of the method has been validated by tuning a 69-gen-
erator test system. While TFEA offers computation time-saving,
the partnership of eigenvalue sensitivity and optimization offers
flexibility and convenience by which a large mix of parameters
(PSS gain and time constants) can be tuned. In the validation ex-
amples, the parameters consist of one gain and 4 time constants
for one generator, making a total of parameters
to be tuned. The prediction results after PSS tuning are in good
agreement with the benchmark results especially for the modes
close to the representative frequency.
Fig. 13. Damping torque of single generator. Design philosophies, operational and strategic planning are
evaluated by formulating Objective Functions under Con-
straints. For illustration, two objective functions have been
carried out. Objective Function 1 consists of minimization
the gains of the PSS while maintaining a damping-to-fre-
quency ratio. Objective Function 2 consists of maximizing the
damping of target modes while maintaining a damping-to-fre-
quency ratio. Results (based on eigenvalues, transient response,
damping torque spectrum) have been presented to show that
the objectives have been met under constraints.
Robustness studies show that after key transmission lines
have been cut, the eigenvalues of the degraded system are still
on the left side of the S-plane.
Fig. 14. Damping torque of generator #44 from Objective Function 1.
It can be concluded that the proposed method offers high
computational efficiency and acceptable accuracy in PSS
of the other 68 generators of the power system. It is evaluated tuning.
using the formula from (6) of [23], which is
REFERENCES
[1] F. P. de Mello and C. Concordia, “Concepts of synchronous machine
(50) stability as affected by excitation control,” IEEE Trans. Power App.
Syst., vol. PAS-88, pp. 316–329, 1969.
[2] E. Larsen and D. Swann, “Applying power system stabilizers, parts I, II
For a typical generator, such as #44, the damping torque from and III,” IEEE Trans. Power App. Syst., vol. PAS-100, pp. 3017–3046,
(50) takes the form as shown in Figs. 14 and 15. Jun. 1981.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

JALAYER AND OOI: CO-ORDINATED PSS TUNING OF LARGE POWER SYSTEMS 9

[3] F. P. de Mello, J. S. Czuba, P. A. Rushe, and J. R. Willis, “Develop- [23] R. Jalayer and B. T. Ooi, “Estimation of electromechanical modes of
ments in application of stabilizing measures through excitation con- power systems by transfer function and eigenfunction analysis,” IEEE
trol,” in Proc. CIGRE, 1986, 38–05. Trans. Power Syst., vol. 28, no. 1, pp. 181–189, Feb. 2013.
[4] M. J. Gibbard, “Co-ordinated design of multimachine power system [24] A. Doi and A. Abe, “Coordinated synthesis of power system stabilizers
stabilisers based on damping torque concepts,” Proc. Inst. Elect. Eng., in multimachine power systems,” IEEE Trans. Power Syst., vol. PAS-
vol. 136, no. 4, pt. C, pp. 276–284, Jul. 1988. 103, no. 6, pp. 1473–1479, Jun. 1984.
[5] M. J. Gibbard, “Robust design of fixed-parameter power system stabi- [25] I. Kamwa, R. Grondin, and G. Trudel, “IEEE PSS2B versus PSS4B:
lizers over a wide range of operating conditions,” IEEE Trans. Power The limits of performance of modern power system stabilizers,” IEEE
Syst., vol. 6, no. 2, pp. 794–800, May 1991. Trans. Power Syst., vol. 20, no. 2, pp. 903–915, May 2005.
[6] P. Pourbeik and M. J. Gibbard, “Simultaneous coordination of power [26] IEEE Recommended Practice for Excitation System Models for Power
system stabilizers and FACTS device stabilizers in a multimachine System Stability Studies, IEEE Std. 421.5, Aug. 1992.
power system for enhancing dynamic performance,” IEEE Trans. [27] P. Kundur, Power System Stability and Control. New York, NY,
Power Syst., vol. 13, no. 2, pp. 473–479, May 1998. USA: McGraw-Hill, 1994.
[7] P. Pourbeik, “Design and coordination of stabilizers for generators and [28] Power Systems Test Case Archive, Univ. Washington, Elect. Eng.
FACTS devices in multi-machine power systems,” Ph.D. dissertation, [29] G. Rogers, Power System Oscillations. New York, NY, USA:
Dept. Elect. Electron. Eng., Univ. Adelaide, Adelaide, Australia, 1997. Springer, Dec. 1999.
[8] M. J. Gibbard and D. J. Vowels, “Reconciliation of methods of compen- [30] R. Fletcher and M. J. D. Powell, “A rapidly convergent descent method
sation for PSSs in multimachine systems,” IEEE Trans. Power Syst., for minimization,” Comput. J., vol. 6, pp. 163–168, 1963.
vol. 19, no. 1, pp. 463–472, Feb. 2004. [31] D. Goldfarb, “A family of variable metric updates derived by varia-
[9] D. M. Lam and H. Yee, “A study of frequency responses of gener- tional means,” Math. Comput., vol. 24, pp. 23–26, 1970.
ator electrical torques for power system stabilizer design,” IEEE Trans. [32] I. Kamwa, G. Trudel, and L. Gérin-Lajoie, “Robust design and coor-
Power Syst., vol. 13, no. 3, pp. 1136–1142, Aug. 1998. dination of multiple damping controllers using nonlinear constrained
[10] J. Zhang, C. Y. Chung, and Y. Han, “A novel modal decomposition optimization,” IEEE Trans. Power Syst., vol. 15, no. 3, pp. 1084–1092,
control and its application to PSS design for damping-interarea oscil- Aug. 2000.
lations in power systems,” IEEE Trans. Power Syst., vol. 27, no. 4, pp. [33] J. C. Geromel, J. Bernussous, and P. L. Peres, “Decentralized control
326–335, Nov. 2012. through parameter space optimization,” Automatica, vol. 30, no. 10,
[11] “Impact of interactions among power system controllers,” CIGRE Pub- pp. 1565–1578, 1994.
lication, Task Force-02-16, vol. 161, 2000, Tech. Brochure. [34] K. Bhattacharya, J. Nanda, and M. L. Kothari, “Optimization and per-
[12] F. L. Pagola, I. J. Perez-Arriaga, and G. C. Verghese, “On sensitivities, formance analysis of conventional power system stabilizers,” Elect.
residues and participations: Applications to oscillatory stability anal- Power Energy Syst., vol. 19, no. 7, pp. 449–458, 1997.
ysis and control,” IEEE Trans. Power Syst., vol. 4, no. 1, pp. 278–285, [35] T. C. Yang, “Applying optimization method to power system sta-
Feb. 1989. bilizer design: Parts 1 and 2,” Elect. Power Energy Syst., vol. 19, no.
[13] P. Kundur, M. Klein, G. Rogers, and M. Zywno, “Application of power 1, pp. 29–43, 1997.
system stabilizers for enhancement of overall system stability,” IEEE [36] M. R. Khaldi, A. K. Sarkar, K. Y. Lee, and Y. M. Park, “The modal
Trans. Power Syst., vol. 4, no. 2, pp. 614–626, May 1989. performance measure for parameter optimization of power system sta-
[14] P. Pourbeik and M. J. Gibbard, “Damping and synchronizing torques bilizers,” IEEE Trans. Energy Convers., vol. 8, no. 4, pp. 660–666,
induced on generators by FACTS stabilizers in multimachine power Dec. 1993.
systems,” IEEE Trans. Power Syst., vol. 11, no. 4, pp. 1920–1925, Nov. [37] J. B. Simo, I. Kamwa, G. Trudel, and S.-A. Tahan, “Validation of a
1996. new modal performance measure for flexible controller design,” IEEE
[15] M. J. Gibbard, D. J. Vowles, and P. Pourbeik, “Interactions between, Trans. Power Syst., vol. 11, no. 2, pp. 819–828, May 1996.
and effectiveness of, power system stabilizers and FACTS device sta-
bilizers in multimachine systems,” IEEE Trans. Power Syst., vol. 15,
no. 2, pp. 748–755, May 2000.
[16] I. Kamwa, R. Grondin, D. Asber, J. P. Gingras, and G. Trudel, “Active- Reza Jalayer (M’13) received the B.S. and M.S. de-
power stabilizers for multi-machine power systems: Challenges and grees in electrical engineering from Sharif University
prospects,” IEEE Trans. Power Syst., vol. 13, no. 4, pp. 1352–1358, of Technology, Tehran, Iran, in 2006 and 2008, re-
Nov. 1998. spectively. He is currently pursuing the Ph.D. degree
[17] H. Golestani Far, H. Banakar, P. Li, C. Luo, and B. T. Ooi, “Damping at McGill University, Montreal, QC, Canada.
inter-area oscillations by multiple modal selectivity method,” in Proc. His research interests include power system sta-
IEEE PES General Meeting, 2009. bility and applications of power electronics in power
[18] I. Kamwa, R. Grondin, and Y. Hebert, “Wide-area measurement systems.
based stabilizing control of large power systems—A decentralized/hi-
erarchical approach,” IEEE Trans. Power Syst., vol. 16, no. 1, pp.
136–153, Feb. 2001.
[19] H. B. Gooi, E. F. Hill, M.A. Mobarak, D. H. Thorne, and T. H.
Lee, “Coordinated multi-machine stabilizer settings without eigen-
value drift,” IEEE Trans. Power App. Syst., vol. PAS-100, no. 8, pp.
3879–3887, 1981. Boon Teck Ooi (S’69–M’71–SM’85–F’02–LF’05)
[20] S. Lefebvre, “Tuning of stabilizers in multimachine power systems,” was born in Malaysia. He received the [Link].
IEEE Trans. Power App. Syst., vol. PAS-102, pp. 290–299, Feb. 1983. (Honors) degree from the University of Adelaide,
[21] F. L. Pagola and J. I. Perez-Arriaga, “Design of power system sta- Australia, the S.M. degree from the Massachusetts
bilizers using eigenvalue sensitivity and SMA techniques.,” in Proc. Institute of Technology, Cambridge, MA, USA, and
Power System Computation Conf., Helsinki, Finland, Aug. 1984, pp. the Ph.D. degree from McGill University, Montreal,
997–1001. QC, Canada.
[22] V. A. Maslennikov and S. M. Ustinov, “Method and software for coor- He is presently a Professor in the Department of
dinated tuning of power system regulators,” IEEE Trans. Power Syst., Electrical and Computer Engineering, McGill Uni-
vol. 12, no. 4, pp. 1419–1424, Nov. 1997. versity.