660

IEEE Transactions on Energy Conversion, Vol. 8, No. 4, December 1993

THE MODAL PERFORMANCE MEASURE FOR PARAMETER OPTIMIZATION OF POWER SYSTEM STABILIZERS
M.R. Khaldi, St. Member A.K. Sarkar K.Y. Lee, Sr. Member Department of Electrical and Computer Engineering The Pennsylvania State University University Park, PA 16802

Y.M. Park, Sr. Member Department of Electrical Engineering Seoul National University Seoul 151, Korea

Abstract - This paper develops an optimization approch using the modal performance measure for the selection of Power System Stabilizer parameters in multi-machine power systems. The goal of the optimization problem is to damp out the sustained low frequency oscillations in the outputs of a linearized power system. Thus. the modal performance measure optimization problem is to select a set of PSS parameters so that the area under the envelop of the oscillatory output response will be minimum. This paper also considers bounded and unbounded PSS parameters and compares the effects of bounds on the end results. Furthermore. this work also shows that the performance measure is not a convex function in the PSS parameters. That is, there exist many local minima and possiblly a global minimum. Single-machine infinite-bus and two-machine three-bus power systems are used to show the effectiveness of the proposed work.
K e v w o r d s - Parameter optimization; modal performance measure; Power System Stabilizer: local minima.
1. IIVTRODUCTION

the pole-placement technique with special emphasis on the frel quency response of the controller. However, not a l the states are available. Further analysis of the pole-placement technique for site selection and design of PSS in multimachine system was carried on by Yu and Li [6]. Even though the system damping was improved, the approach had an inherent defect in the preselection of PSS time constants. Use of eigenstructure assignment by the decentralized output feedback control scheme to improve the dynamic stability of the power system and to alter the dynamic behaviour of the system was done by Huang and Hsu [7]. Optimal setting of the PSS parameters using the sensitirity of the system eigenvalues with respect to the PSS gain, was proposed by Liu e t al. IS]. Recently the adaptive control techniques have been employed to improve the dynamic performance of the synchronous machines over a wide range of operating conditions 19-13;, This paper introduces a new measure, called the modal performance, proposed by Jung, Choo and Park [14] and Lee and Lee 1151 for the selection of PSS parameters. The problem of selecting the PSS parameters has been formulated as a param-

control. commonly referred to as Power System Stabilizer (PSS), to enhance the damping of these low frequency oscillations and thereby to improve the stability of the power system. Feasible approach to achieve this goal is to design the optimal values of the parameters of the PSS network. The present techniques of selection of the parameters of the PSS are based on the concept of synchronizing and damping torques [l-31 and the theory of phase compensation in the frequency domain to improve the damping torque in the low frequency range. However, when a higher order model has been used to design the PSS, it is difficult to apply the theory of phase compensation. Yu and Moussa [4]formulated the problem as an optimal linear regulator problem whose solution is a complete state feedback control scheme. This was further extended by Chow. and Sanchez-Gasca [5], where the PSS was designed by 93 WM 1 2 7 - 1 EC A paper recommended and approved by t h e I E E E Energy Development and Power Generation Committee of t h e I E E E Power Engineering Society f o r p r e s e n t a t i o n a t t h e IEEE/PES 1993 Winter Meeting, Columbus, OH, January 31 - February 5 , 1993. Manus c r i p t submitted January 13, 1992; made a v a i l a b l e f o r p r i n t i n g Novemember 2 3 , 1992.

2. SYSTEM MODELING Let a power system consist of q buses, m of which are generator buses. Then in its dynamic state the power system can be represented by the non-linear vector equation

where x is the state vector for m generators, and v and 9 are, respectively, the voltage magnitude and angle vectors for q buses. The power system was initially assumed to be operating at a steady state, characterised by x0. Hence, for small disturbances, the non-linear equation representing the power system can be linearized about the quiescent state x0 as

d(xo
or

dt

+ A x ) = f(xo, Bo, v) + -af x + -AB + -AV af af A ax a0 av

d x = [E]Ax+ [ B C ]

,I!:[

(2)

where

E = f ax

iB C ] = [ %

g].

From the active and reactive power balance equations,

and

0885-8969/93/$03.00 0 1993 IEEE

661
rameters of the PSS fails, since the dynamics of the PSS are combined with the system dynamics. In this paper, parameter optimization has been accomplished by minimizing the modal performance measure with respect to the PSS parameters in the system matrix A. Since outputs (or states) of a linear system are comprised of various modes, the modal performance measure has been defined as the weighted sum of squares of these modes. This performance measure gives a good control over each mode. The sensitivities of the performance with respect to the PSS parameters. have then been derived using gradient matrices.
3.1. M o d a l P e r f o r m a n c e M e a s u r e
- ~j

f o r i = 1, . . . , q . where subscripts g, 1 and n denote generator, load and network, respectively. In terms of the bus voltages v i , the bus voltage angles 8; and the network parameters, the injected active and reactive power flowing into the network can be expressed as
h,

~ ~ ; ( o=>Cv;vjp;jcos(ei ~ )
j=1 A-

- ej

eij),
i j ) ,

Q , ; ( O , ~ = Cv;vjY,jsin(Oi )
j=1

-~

Let us consider the linearized system in equation ( 7 ) , characterised by the state-space equation, A? = AAx, AX E R",

where l < j : magnitude of the admittance between buses i and j 8 i j : angle of the admittance between buses i and j. The generator active and reactive power outputs can be expressed as

-4 E R"""

( 8)

along with the output equation y = CAx, y E R",

C E Rnxn,

(9)

where q; 6;- 6';. Assuming a constant impedance load, eqns. (3) and (4) can be linearized and the equations can be combined into the the following matrix form: SAX=J
or

and the given initial condition Ax(0) = X O . The elements of the A matrix contain the PSS parameters. These parameters can be controlled in such a way that the eigenvalues of the A matrix move away from the imaginary axis and the system becomes more stable. Hence, without any loss of generality it can be assumed that the A matrix has distinct eigenvalues. The natural response of the states of the system can be expressed in terms of its modes as
n

[::I,

(5)

AX(^)

= C ( v j ~ x 0 ) uexp(Xjt) j
j=1

(10)

= J-~SA~,

where

and

where X j is the j - t h eigenvalue of the A matrix and has the form X j = uj i4j.Here, u j and vj are the corresponding n x 1 right and 1 x n left eigenvectors, respectively, and n is the dimension of Ax. Let Azj(t) be the j - t h state of the vector Ax(t); and Azj;(t) be the i-th mode of Azj(t). Let uij be the i-th entry of the j - t h right eigenvector uj. Then,

Azj(t) = ~ ( v ; A x 0 ) u exp(X;t) ji
i=l

(11)

Equation (6) gives an expression for the bus voltage magnitudes and bus voltage angles in terms of the system state variables. Substituting eqn. (6) into (2): the linearized form of the system equation, eqn. ( 7 ) , is obtained:

=
i=l

Azji(t),

where

or

A 2 = AX, where

(7)

Azj;(t) = ( v ; A x ~ ) u jexp(Xit). ; (12) Similarly, let yj(t) be the component of y(t) that depends on the j - t h state of vector A x ( t ) , and yj;(t) be the i-th mode of yj(t). Hence,
n

A = [ E + [ B CIJ-IS].
It can be seen that the system matrix A contains parameters of generators, exciters, PSS's, and the network.
3. P A R A M E T E R O P T I M I Z A T I O N

y ( t ) = CAx(t) = C C j A z j ( t ) = C y j ( t ) ,
j=1
j=1
n

(13)

where

y j ( t ) = ~ j ~ ~ = ( t y)j i ( t ) j
i= 1

Most of the earlier methods to select the parameters of the PSS are based on the theory of phase lead-lag compensation. However, when a higher order model is required to design the PSS, the phase lead-lag theory is difficult to apply 1161. In addition, the application of linear control theory to select the pa-

and yj;(t) = CjAzj;(i).

A square of the envelope of the i-th mode in the j - t h component in the output can be expressed as

662
1 r

1, if X j is real;
aj

aj Y j i Y j i >

2, if X j is complex

inCr"nts Neglecting the second and higher order terms of except those in Ax0 in eqn.(l5), AJj is obtained as

where ( )' and ( ) * denote the matrix transpose and complex conjugate, respectively. Hence, the weighted sum of the output of the squares of the i-th mode in the j - t h component in the output is given by

AJ~ =

c
i=l n

Mi $ ( u ~(viAx0) (viAxo)*uji A.,' )

{

+ $ ( ~ i ) ( v ~ A x o () v ~ A x o ) * Aujiu;; + $(c;)v ~ A x o (Avi*Axo*)uji u;i ( ) + $ ( ~ i )( A v ~ A x o (vi*Axo*)uji ) + $ ' ( n i )f l u ;

(vi*~xo*)uji
(18)

where Mj G ajCjfMiCj and Mj is the weighting matrix for the output. The performance measure Jj for the j - t h state is, therefore, defined as

where

$'(uj) =

30,

From the above equation, it is seen that the incremental performance measure is a function of the increments Auji, Avj and Auj. These increments in the eigenvalues and eigenvectors are due to the changes in the elements of the A matrix, and can be evaluated from the eigenvalues and eigenvector sensitivities [17]: AJj =

i=l

Y +

M j $(ui)viAxovi*Axo*2 Re(UjtDiVAAuiuji)
$(uj)ujiu:;2

Re (v;AAUD;VAxovi*Ailxo*)

where

+ $'(ui)Re(v;AAui)viAxovi*Axo*uj;u;i
where Re means the real part, and

This is the performance measure representing the area under the envelope of the j - t h state trajectory. Hence the performance index of the system, J, also called the modal performance measure, can be written as
n

J =
j=1

Jj.

(15)

This modal performance measure has to be minimized with respect to the parameters which are present in the system matrix A. Hence, the problem can be formulated as following: min J ( p )
P

subject to

A? = A(p)Ax,

and Uj" is the j - t h column vector of Ut matrix, i.e.: the j - t h row vector of the U matrix. In view of the commutative property of matrix traces [18], eqn.(l9) can be written as

where p represents the vector of PSS parameters. 3.2. P a r a m e t e r O p t i m i z a t i o n The gradient o the performance measure is then defined as f

Since the parameter vector p is contained in the system matrix, hence, by applying the chain rule, eqn.(l6) is obtained:

and then following the property of trace function [17], eqn.(20) is obtained.

where Akl is the element at the (k,Z) entry of matrix A, and

(g)is the (k,I ) entry of the sensitivity matrix (g).

k,,

The sensitivity of the modal performance measure Jj, with respect to the elements of the A matrix, can be derived by linearizing Jj. Let A J j , Auj, Auij and Avj be the increments of Jj, u j , uij and vj, respectively, due to the increments in the elements of matrix A. Then, the incremental performance measure is defined as A J j = J j ( A A A ) - Jj(A). (17)

Equation (20) gives the sensitivity of the performance measure with respect to the elements of the system matrix. Thus, the gradient of the performance measure with respect to the param-

663
Table 2 . Parameter Optimization for a Single-Machine System
T1
CASE # 1
I

T2 Init. 0.20
I

T
Final 0.24

Kc
Final

Init
1.50

Final 1.45
I

Init
20.00

Init
6.50

Final
I

19.29

6.62
I

3 4 5

I 1.50 1 I 1.50 I
3.00

1.34 1.38 2.97

I I

0.10 0.10 0.20

I I

0.24

I 10.00 I 14.09 I 7.00 I

I 20.00
2.90
~

7.21

0.24
0.21

122.49 3.60

I 7.00 1 7.33 I
3.35 3.51

XIO'

GENERATOR PARAMETERS

M
9.26 0.19

xd
0.973

T'do
7.76

Z ,

K,
50

0.55

0.05

LINELOAD
BUS DATA

R=-0.034 X=0.997 G=0.249 Bz0.262

P 0 = 1,

s = 0.015, o

-I&

0 = 1.05

0.5

1.5

2.5

3.5

4.5

BOUNDS

1 5 T 9 30,O 5 Kc 5 20.0.1 I T2 5 50,O.l 5 T i 5 100

TIME. SEC

Fig. 2 . Single-Machine Infinite-Bus, Case I

z -

x104

Fig. 1. Single-Machine Infinite-Bus Power System All data are in per unit value, except that M and the time constants are in seconds. The negative value of R is obtained while deriving the single-machine infinite-bus model for a multimachine system by equivalencing smaller generators by equivalent impedances with negative resistances. The frequency of the system is assumed to be 60Hz. The results of the parameter optimization are tabulated in Table 1. Cases 1-6 shown in Table 2 are the parameter optimization results for different initial values of PSS parameters. Cases 14 represent the cases when parameters are bounded by limits defined in Table 1, and cases 5 and 6 represent the cases when the bounds are ignored. It is seen that the final values of the PSS parameters for these cases are different, and each case converges to a local minimum. The system responses for diffrent initial values of the PSS parameters are shown in Fig. 2 and Fig. 3.

UNBOUNDED PSS PARAMETERS

-12
0
0.5
1

1.5

2.5

3.5

4.5

TIME, SEC

Fig. 3. Single-Machine Infinite-Bus, Case 6

The gradient values were found to be small and hence the step sizes in the steepest descent method were kept fixed and small. The optimization loop was exited whenever the gradient falls below a specified tolerance value and the corresponding parameters were said to be optimal. In Fig. 2 and Fig. 3, a comparison between the system responses obtained due to parameter optimization based on the modal performance measure

664
approach (curves ( b ) and (c)) and by the optimal linear regulator approach (curve (a)), clearly shows the improved system response that is obtained by the former approach. The comparison between curves (b) and (c) gives an interesting insight into the nature of the problem of parameter optimization for such systems. While it was reported by Jung e t al. [14] that the parameters obtained by them are optimal, responses as shown by curve (c) suggest that similar system response can be obtained for diEerent sets of final parameter values. This clearly demonstrates the existence of many local minima. This also implies that the performance measure is not a convex function in the parameter space. It also shows that the performance measure to be minimized is shallow with respect to the parameter space and hence, a global minimum is difficult to be achieved. This answers the question why the gradient is small and the movement of the parameters over the iterations is slow.
4.2. Two-Machine Power S y s t e m

In Fig. 5 , the improvement in the responses obtained can be clearly seen. The amplitude of oscillation of the shaft speed deviation of the two generators is significantly less in plot (b) than in plot (a). Moreover, the responses are stabilized much faster in plot (b) than in plot (a). These responses correspond to the parameter optimization result shown in Case 2 in Table 4. The improvement in the responses of the system, even when optimization was performed without any bounds imposed on the parameters, can be seen from Fig. 6 for Case 3. Table 4. Parameter Optimization For a Multimachine System

I
CASE #
1

GENERATOR # 1
I

TI
Init. 0.83 0.83 Final 0.83 0.83 Init. 50.00 5.80

T2 Final 20.00 5.80 Init. 0.83 0.83 0.83 0.83

T Final 1.00 1.00 Init. 1.59 1.59

K
Final 1.59 1.59 1.59

2 3

The multimachine system consists of two generators and a load bus interconnected by transmission lines. In this paper, bus 1 has been selected as the swing bus. A one-line diagram of this power system is shown in Fig. 4. Table 3 summarizes the machine data and the initial bus data for the system. Table 3. Data for a Two-Machine Three-Bus Power System
PAR.

1
I

I I

0.83 0.83

I I

0.83 0.83

I I

5.59 5.59

I 5.69 I
I
T2
5.72

I
I
T

0.83 0.79

I
I

1.59 1.59

I
K

1.59

CASE

GENERATOR # 2

Init.

I Final
186.40 86.40 86.40 86.40

Init.

I 86.10
86.10 86.10 86.10

5.57 5.57 5.57 1.59

I Final Init. I I 5.57 I 110.00 I

5.59 5.57 2.50 30.00 110.00 110.00

Final 23.90 26.80 109.96 109.78

Init.

I Final
I 4.32 I
20.00 13.00 4.46

4.32 20.00 13.00 4.32

2 BUS 1 65.45 0.0208 0.1675 0.1675 BUS 2 55.2 BUS 3 3

0.056 0.303 0.282

0.02 100 25.745 31.778 1.04

0

0.02 100

I I I

I I I
0.987 -0.046

1.03 0.002

I I

BUS I

0.02+j0.06

BUS 2
I

Fig. 4. Sample Multimachine Power System Cases 1-4 shown in Table 4 are the parameter optimization results for different initial values of PSS parameters. Cases 1 and 2 represent the cases when PSS parameters are bounded with limits as defined in the single machine data. Cases 3 and 4 represent the cases when these bounds are ignored.

5 . CONCLUSION

Optimization using the modal performance measure was used to select a set of PSS parameters. System responses were obtained to have very satisfactory low frequency oscillation for different sets of optimal PSS parameter values. This suggests the possibility of the existence of many local minima. Furthermore, bounds on the parameters play an important role in

665

*..
a .
I

*,a
.

2-

, ,

a: REFERENCE [ 141 b: OBTAINED

, .
I

,.
.a
I

21-

, ,

a: REFERENCE [141 b. OBTAINED

I .

..

..

-5

. ,

:
BOUNDED PSS PARAMETERS

0.5

15 .

25 .
TIME, SEC

3.5

4.5

6o

0.5

1.5

2.5
TIME, SEC

3.5

4.5

I 5

x10-4
8

GENERATOR # 2

x104
8

GENERATOR # 2
a: REFERENCE 1141

a: REFERENCE [ 141 b: OBTAINED

4-

... ,..
UNBOUNDED PSS PARAMETERS

BOUNDED PSS PARAMETERS

-6

0.5

1.5

25 .

3.5

4.5

5
TIME, SEC

TIME.SEC

Fig. 5. Multimachine Power Sgstem, Case 2 ascertaining the stability of the system. The system model was derived from a fixed operating condition, and based on this model an optimal set of PSS parameters was obtained. For implementation purposes, optimal set of PSS parameters can be computed, based on the proposed technique, for a number of operating regions. Should the system undergo abnormal conditions, then the selected PSS parameters would be switched to the optimal values of the new operating condition. For future study, an alternative approch would be to design
a robust controller to supplement the PSS controller designed

Fig. 6. Multimachine Power System, Case 3

under normal operating conditions [20].

ACKNOWLEDGEMENTS
This work was supported in part by the Korea Electric Power cooporation and AlleghenS Power System. Howerver, any findings, conclusions, or recommendations expressed herein are those of the authors and do not necessarily reflect the views of KEPCO or APS.

REFERENCES

[l] P. Kunder, M. Klein, G.J. Rogers, and M.S. Zywno, Application of Power System Stabilizers for Enhancement of Overall System Stability, IEEE Trans. Power Sys., Vol. 4 , May 1989.

[2] F. P. Demello and C. Concordia, Concept of synchronous machine stability as affected by excitation control , IEEE Trans. Power Appar. Sys., vol. 88, pp. 316-329, 1969. [3] A. A . Shaltout and E. A. Abu Al-Feilat, Damping and synchronizing torque computation in multimachine power systems, IEEE/PES 1991 Winter dleeting, 91 WM 218-8 PWRS. [4] 1. N. Yu and A. M. Moussa, Optimal power system sta bilization through excitation and / governor control, IEEE Trans. Power Appar. Sys., vol. PAS-91, p.1166-1174, May 1972. [5] J . H. Chow and J. J. Sanchez-Gasca, Pole-placement designs for power system stabilizers,, IEEE Trans. Power Sys., vol. 4, pp. 271-277, Feb. 1990. [SI Y. N. Yu and Q. H. Li, Pole-placement power stabilizer design of an unstable nine-machine system, IEEE Trans. Power Sys., vol. 5 , pp. 353-357, May 1990. 171 P. H. Huang and Y. Y. Hsu, Eigenstructure assignment in a longitudinal power system via excitation control, IEEE Trans. Power Sys., vo1.5, pp.96-102, Feb. 1990. [8] C. Liu, S.X. Zhou, and Z.H. Feng, Using decoupled characteristic in the synthesis of stabilizers in multi-machine systems, IEEE Trans. Power Sys., vol. PWRS-2, pp. 31-36, Feb. 1987.

666
[g] R. I. Reyes, D. R. Mudgett, and K. Y. Lee, Power System Stabilization via Adaptive Multiple ControUer,proc. the 29th Conference on Decision and Control, Honolulu, Hawaii, Dec. 1991, pp 3077-3082. [IO] Y. Y. Hsu and C. J. Wu, Design of self-tuning PID power system stabilizer for multimachine power systems, IEEE Trans. Power SYS., vol. 3,~~.1059-1064, Aug. 1988. [11] A. Chandra, O.P. Malik, and G.S. Hope, A self-tuning controller for the control of multi-machine power systems, IEEE Tmns. Power Sys., vol. 3, pp. 1065-1071, Aug. 1988. -121 A. J . Urdaneta and N. J. Bacalao, Tuning of power system stabilizers using optimization techniques, IEEE Trans. Power SYS., vol. 6, pp. 127-133, Feb. 1991. where Au, is a supplementary control signal from the PSS and the term AV is called the error term, and is defined as A V E u,,f - u t , which indicates if any corrective action is needed to maintain the desired value. The transfer function of the PSS is given by

where 1 is the number of lead stages. In this study, a single lead network has been used for simplicity. Constants K,, TI, T2 and T are the parameters to be optimized. The state equations of the PSS with a single lead network, can be written as

[13] M. J. Gibbard, Robust design of fixed parameter power system stabilisers over a wide range of operating conditions, IEEE Trans. Power Sys., vol. 6, pp. 794-800, May 1991. [14] J. W. Jung, J. B. Choo, and Y. M. Park, Optimal PSSParameter Selection Algorithm with new Performance Measure,IFAC Symposium on Power Systems and Power Plant Control,, pp. 381-386, 1989. [15] B.H. Lee and K.Y. Lee, Dynamic and Static Voltage Stability Enhancement of Power Systems, IEEE/PES 1992 Winter Meeting, 92 WM112-3 PWRS. 1161 Y. N. Yu, Electric Power System Dynamics, Academic Press, New York, 1983. [17] P. M. Anderson and A. A. Fouad, Power System Control and Stability, vol. 1, Iowa Univ. Press, Ames, 1977. 1181 T. R. Crossley and B. Porter, Eigenvalue and eigenvector sensitivities in linear systems theory, Int. J. Control, vol. 10, pp. 163-170,1969. [19] R. Bellman, Introduction to Matriz Analysis, McGraw-Hill, New York, 1960. [20] A. Ben-Abdennour, K.Y. Lee, and R.M. Edwards, Multivariable Robust Control of a Power Plant Deaerator, IEEE/PES 1992 Summer Meeting, 92 SM 514-0 EC.

where z j and Au, are internal state and output of the PSS, respectively.

BIOGRAPHIES
M.R. Khaldi received the B.S. and M.S. degrees in Electrical Engineering from the State University of New York at Binghamton, in 1987 and 1989, respectively. Since 1987, he has been appointed as Teaching Assistant and then as a Research -4ssistant. He is currently an Instructor in Electrical Engineering at the Penn state University, University Park where he is working on his Doctoral degree. His main interests include systems and control, and power systems. A.K. Sarkar received the B.S degree in Electrical Engineering from Jadavpur University, Calcutta, India, 1988, the M.S. in Electrical Engineering degree from the Penn State University, University Park, in 1991. He was with the Switch Gear Division of Grompton Greaves Co., Ltd., Bombay, India, where he was trained in marketing, design, production, engineering, and materials for machines and switch gears. His interests include power system analysis, and control.

APPENDIX
A third order synchronous generator model has been chosen for analysis purposes [16]. The model neglects the damper windings or any other inherent damping present in the machine, which would normally provide additional damping besides the damping due to PSS. Although the generator model is too simple, it is chosen in order to demonstrate the role of PSS parameters.

K.Y. Lee received the B.S. degree in Electrical Engineering from Seoul National University, Seoul, Korea, in 1964, the M.S. degree in Electrical Engineering from Korth Dakota State University, Fargo, in 1968, and the Ph.D. degree in System Science from Michigan State University, East Lansing, in 1971. He has been on the faculties of Michigan State, Oregon State, Houston, and the Penn State University, where he is now Professor The state equations of the third-order generator model can of Electrical Engineering. He is currently in charge of the Power be written as Engineering Program and the Power Systems Control Laboratory at Penn State. His interests are power systems, control, ;=-(p 1 -.:(xi - x q ) sm27) . (-41) operation and planning and intelligent power plant control. M 4 sin7 2xkxq Dr. Lee has been a senior member of IEEE control System Society, Power Engineering Society, and Systems, Man and Cybernetics Society. He also a registered Professional Engineer.

where variables are defined using standard notations. In this study, an IEEE type - 1 regulator exciter has been selected. The system can be represented by the following differential equation

Y.M. Park received the B.S., M.A. and Ph.D. degrees in Electrical Engineering from Seoul National University, Seoul, Korea, in 1956, 1959 and 1971, respectively. Since 1959, he has been with Seoul National University, where he is currently Professor of Electrical Engineering. He was a Visiting Professor at Tokyo University in 1981 and University of Houston in 1983. His area of research are power system control, operation and planning.