350 IEEE TRANSACTIONSON CIRCUITSAND SYSTEMS, VOL. CAS-33, NO. 3, MARCH 1986 proposed extended version. In both the cases we use the ideal autocorrelation samples. We see from this figure the superiority of the extended Levinson recursion, which is able to recover exactly the poles of both the clutters. REFERENCES PI G. Martinelli and G. Orlandi, “ARMA spectrum estimation by extended lattice,” in Proc. ICASSP, Paper 13.3, San Diego, CA, Mar. 1984. VI J. A. Cadzow, “Spectral estimation: an overdetermined rational model equation approach,” Proc. IEEE, vol. 70, pp. 907-939, Sept. 1982. [31 G. Martinelli, P. Burrascano and G. Orlandi, “Simultaneous cancellation of two clutters by the extended lattice oredictor.” in Proc. Int. Svmo. Noise Clutter Rejdfion in Radars and Im&ing Sensors, Tokyo, Japan, p’p. 274-278. Oct. 1984. 141 S. Haykin, B. W. Curie and S. B. Kesler, “Maximum entropy spectral analysis of radar clutters,” Proc. IEEE, vol. 70, pp. 953-962, Sept. 1982. A Unique Method for Coordination of Excitation and Governor Controls for Stability Enhancement of a Power System CHOO-MIN LIM AND S. ELANGOVAN It should be noted that the first element of X is chosen to be the system only output or the rotor speed [4], [5], and uE and uc are the scalar excitation and governor control signals, respectively. It has been shown in [2], [3] that the feeding back of rotor acceleration, SAO, into the excitation and governor loops is suitable for improving the damping characteristics of power systems. However, the measurement of rotor acceleration is not a simple matter and alternative means have been proposed [2], [6]. Here, the power system rotor speed is chosen to be its output and the structure of the excitation and governor stabilizers are preselected to be [6] FE = K,s/( 1 + ST) (2.1) FG = K,s/( 1 + ST) (24 where T is prespecified and (K,, KG) are constant parameters which remain to be determined. It is interesting to note that in [7], it has been shown that irrespective of the values of (K,, KG) the minimizing value of, T with respect to a quadratic performance index is always zero. However, to ensure the physical implementation of each stabilizer, T should be chosen as small as possible but within physical limit [7]. Here, this approach is adopted and stabilizers thus designed will yield optimal performance. Abstract -A simple and unique method for coordinating the excitation and governor stabilizers for stability enhancement of power systems is presented. The required stabilizers are simple in structure and, most importantly, are readily realizable. I. INTRODUCTION The coordination of excitation and governor stabilizers for enhancing the stability of power systems has received much attention [l]-[3]. A primary design objective in the coordination of excitation and governor controls is to provide additional damping to a power system rotor oscillations [l]-[6], which are governed predominantly by a pair of complex-conjugate eigenvalues. Here, a simple and unique method for coordinating the excitation and governor stabilizers of a power system based on eigenvalue assignment is presented. Specifically, the parameters of both stabilizers are so chosen that the eigenvalues associated with rotor oscillations have desired values or specific damping [l], [4], [5]. As in [2], [4]-[6], the required stabilizers are simple in structure and readily realizable. II. Consider a single-machine infinite-bus power system whose dynamics may be expressed in the complex frequency domain as sX=AX(s)-tBU(s) (1.1) y=Ao (1.2) where UG]l 0 .. . b, ... 0 ... ... 0 0 bc 1 1 A 1 1 42 A= and A-:-T 211 2 (3.1) where O,,O,,,O, AA,,=[O The composite given by are null submatrices and 0 ... 0 closed-loop bEFE 0 ... 0 1 t bGFGlr. system characteristic A(s)=IsZ-(A+AA)I=O. COORDINATION OF EXCITER AND GOVERNOR STABILIZERS u=[u, The design objective may now be formulated as the determination of (K,, Z&) such that the complex-conjugate eigenvalues associated with rotor oscillations have desired values or specific damping [l], [4], [5]. To this end, the composite closed-loop system equation is obtained and expressed in the form sX= (A + AA) X where the A and AA matrices are as follows: (3.2) equation is (4) Let the desired eigenvalues associated with rotor oscillations be (sd, s:) or equivalently preselected to be (- [a,, k jw,J(lE2)). Now for (sd, s:) to be a pair of complex-conjugate roots of the closed-loop system characteristic equation, the following results hold 141,151 A(Q) = A(s$) = 0. (5) In order to reduce (5) into an equivalent and manageable form, partition A and AA matrices as shown in (3) such that the size of submatrix A, is 1 X 1 [4], [5]. Then it can be readily shown that A(s,)=~s,Z-A,-A12(~dZ-A2)-1(A21+AA21)~=0. Upon expansion and simplification Manuscript received May 6, 1985. The authors are with Ngee Ann Polytechnic, Electrical Engineering Department, Singapore 2159. IEEE Log Number 8406166. 009%4094/86/0300-0350$01.00 of (6) yields sd - cl = c,b,F,(s,) + c~b&(sct) =(c,b,K,+c,b,K,)s,/(l+s,T) 01986 IEEE (6) (7) 351 IEEE TRANSACTIONSONCIRCUITSAND SYSTEMS,VOL.CAS-33,N0. 3,MARCH 1986 +o- TIME IN SEC. Fig. 1. Rotor angle response of study system with and without coordinated stabilizers: Cl = open loop; C2 = with stabilizers, (5 = 0.5); C3 = with stabilizers, (5 = 0.3). TABLE I A SAMPLEOFRESULTSOBTAINED Stabilizer Parameters Desired Eigenvalues ( x0:1 Closed-loop ) KE = 0.0064 (-1.51 f j4.83, -0.759, -1.52 k j2.41, (t=.3,f~J,=5.06) KG = 0.0009 -5.14 t j6.08, -10.0 , -12.7, -16.0 -2.03 -1.51 t j4.83 KK = 0.0080 (-2.03 t j4.64, -0.757, -1.42 2 j2.47, lk.4,W,.5.06) KG = 0.0010 -4.44 t j6.44, -10.0 , -12.7, -16.7 -2.53 KE q 0.0095 l-2.53 ? j4.38, -0.752, -1.26 t j2.47, KG i 0.0011 -3.85 t j6.93, -10.0 , -12.8, -17.3 * j4.64 + j4.38 (k.5,Wnz5.06) where ( c1, c2, cs) are known constants [4], [5]. By equating the real and imaginary parts of both sides of (7), two equations are obtained from which (K,, K,) are readily computed for a given T, and no interative scheme is required. III. NUMERICAL EXAMPLES The effectiveness of coordinated excitation and governor stabilizers with rotor acceleration as input signal for enhancing the damping of turbogenerator has been demonstrated [2], [3]. Thus it would be interesting to examine the effectiveness of such a coordinated scheme in a hydroelectric power system. To this end, consider a hydro power system which consists of a singlemachine connected to an infinite-bus [l], [4]. For controllers design purposes, the system is represented by a fifth-order synchronous machine model equipped with a firstorder exciter [4] and a third-order governor without a dashpot [l]. The system equations are expressed in the form of (1) where x= I Aw AS A#, A$kd A#,, AErd Ah Ag A~]‘: The open-loop system eigenvalues are (0.119+ j4.64, -0.805, -8.47-t System Eigenvalues j5.17, -2.39+ j1.80, -13.1, -27.3). Clearly, the system is unstable due to the‘ first pair of complex , -26.8) , -26.7) , -26.6) eigenvalues associated with rotor oscillations [l], [2], [4]. In order to improve the system stability, excitation and governor stabilizers of the form given by (2) with T set to 0.1 are introduced and then coordinated using the above proposed method. Several sets of (sd, ~2) or (5, w,) are considered and in each case w, is kept constant [4], [5]. A sample of results obtained is summarized as shown in Table I, and in all cases considered exact eigenvalue assignment is always achieved. The rotor angle responses of the system with and without a coordinated excitation and governor stabilizers in service to small disturbance are shown in Fig. 1. Clearly, excitation and governor stabilizers coordinated using the proposed method are effective in improving the hydroelectric power system damping of its rotor oscillations. The effects of different values of T of both stabilizers are next examined using a performance index Z given by Z = jmAwZ dt. 0 A sample of results obtained is shown in Fig. 2. Clearly, the results obtained here are in good agreement with that of [7]. The robustness of stabilizers designed using the proposed method is also investigated by examining the eigenvalues of the overall power system over a wide range of operating conditions. In this study, the power system is represented by a third-order IEEE TRANSACTIONSON CIRCUITSAND SYSTEMS,VOL.CAS-33,N0. 3, MARCH 1986 352 Fig. 2. Performance index against time constant of stabilizers C2. TABLE II SYSTEMEIGENVALUESASSOCIATEDWITHROTOROSCILLATIONS (**EXACTASSIGNMENTOFTHESEEIGENVALUES) Heavy open loop Closed-loop load Nominal Light load P = 0.952 P = a i 0.45 a = 0.015 P = -0.225 Y t = 1.05 "t = 1.05 "t = 0.82 f j3.02 0.18 + j4i61 -0.14 f j4.95 -1.05 t jl.81 -0.90 f j6.25 -0.86 t j6.07 0.500 1.05 l * synchronous machine model equipped with a first-order exciter and a 3rd-order governor without a dashpot. It is observed that in order to guarantee an adequate amount of damping to the system over a wide range of operating conditions, both the stabilizer parameters should be selected using the system linearized equations corresponding to the heavy load condition. A sample of results obtained for such a case is shown in Table II which shows only the system eigenvalues associated with rotor oscillation. It should be noted that all of the remaining eigenvalues of the system not shown in Table II have negative real part. IV. load P = 1.25 CONCLUSION A unique method for coordinating the excitation and governor stabilizers for stability enhancement of power systems has been presented and demonstrated using a numerical example. The method is easy to apply since no interative scheme is required and moreover, it yields stabilizers which are simple in structure and physically realizable. REFERENCES WI H. A. M. Moussa and Y. N. Yu, “Optimal power system stabilization through excitation and/or governor control,” IEEE Trans. Power App. Sysr., vol. PAS-91, pp. 1166-1174,1972. 121 S. I. Ahson, B. W. Hogg, and R. T. Pullman, “Integrated control system for turbogenerator designed by Nyquist Array method,” IEEE Trans. Power App. Sysr., vol. PAS-98, pp. 543-553, 1979. [31 H. Lu, P. A. Hazel& and A. R. Daniels, “Co-ordinated single-variable excitation control and governing of turboalternators,” Proc. Inst. Elect. Eng., vol. 129, Pt. C, pp. 278-284, 1982. [41 C. M. Lim and S. Elangovan, “Pole assignment of SISO system using dynamic compensator with prespecificed poles,” IEEE Trans. Circuits Syst., vol. CAS-31, pp. 990-991, Nov. 1984. “Pole assignment of a class of multivariable system using localized [51 -1 dynamic output compensators with prespecified poles,” IEEE Trans. Circuits Sysr., accepted for publication. 161 Y. N. Yu and C. Siggers, “Stabilization and optimal control signals for a power system,” IEEE Trans. Power App. Syst., vol. PAS-90, pp. 1469-1481,198O. [71 V. H. Quintana, M. A. Zodhy, and J. H. Anderson, “Design of output feedback controller for power system stabilization,” IEEE Trans. Power App. Syst., vol. PAS-95, pp. 954-961, 1976. A Two-Stage Pole-Zero Predictor W. B. MIKHAEL, A. S. SPANIAS, G. KANG, ANDL.F%%NSEN Ahstruct -A two-stage pole-zero predictor is proposed which is capable of predicting minimum phase Auto Regressive Moving Average (ARMA) processes accurately with a reduced number of parameters. This is achieved by cascading the classical predictor (1st stage), and the pole-zero recursive-like structure (2nd stage). The constraint on the order of the first stage is relaxed since the unknown process is actually represented by the parameters of the second stage. Computer simulations for both low- and high-order ARMA processes are given to demonstrate the excellent performance of the proposed technique. I. INTRODUCTION The classical linear prediction problem can be stated as follows: Let y(n) be a (scalar) discrete-time stationary zero-mean process. We are interested in predicting the current value y(n), of this process from past measurements, where n is the discrete time index. A linear predictor of order N, will have the form NP F(n)= c PkYtn-k) (1) k=l where j(n) is the predicted value of y(n) based on data up to time (n - 1) and { pk, k = 1,2,. . . , NJ,} are the predictor coefficients. The difference between the actual value of the process and Manuscript received July 23, 1985. This work was supported by the Naval Research Laboratory under Grant SFRC 82 K 2047. W. B. Mikhael and A. S. Span& are with the Department of Electrical Engineering, West Virginia University, Morgantown, WV 26506.6101. G. Kang and L. Fransen are with the Naval Research Laboratory, Waahington, DC 20375. 0098-4094/86/0300-0352$01.00 01986 IEEE