Sie sind auf Seite 1von 6

IEEE Transactions on Power Systems, Vol. 13, No.

1, February 1998

109

Modeling of Generators and Their Controls in Power System Simulations Using Singular Perturbations
X. Xu, stud. M., IEEE
R.M. Mathur, s.M., IEEE J. Jiang, - s.M., IEEE G.J. Rogers;,s.M.,
IEEE

P. Kundur, F., IEEE

Department of Electrical Engineering The University of Western Ontario London, Ontario, Canada N6A 5B9
Abstract-This paper presents an application of the singular perturbation method to power system simulations. The focus is on treatment of the fast states associated with generators and their controls. Models of generators, static exciters, power system stabilizers (PSS) and governors are reformulated using singular perturbation techniques. These singularly perturbed models are implemented in a two-area, four-machine power system. The results bear a close correlation to those obtained from a full model in which the fast states are fully simulated. This study contributes a new and efficient method in dealing with the fast states to speed up power system simulations.

Cherry Tree !3ci. Software Powertech Labs Inc. Colborne, Onitario Surrey, B.C. Canada KOK 1SO Canada V3W 7R7 faster, to allow system operation close to the stability limit. On the other hand, long-term stability evaluation requires very long simulations with simulation models capable of representing both transient and slower long-term phenomena. Model simplification without undue loss of accuracy is, therefore, important. The application of the singular perturbation theory to power system analysis was introduced more than a decade agoll-51. At its inception, its main use was in global coherent generator aggregation. Some model reduction was investigated[6-8], but the implementation of the reduction was not carried out to the point of application to practical power system simulations. In this paper, we extend a recent [9] formulation of singular perturbation model reduction to the problem of eliminating small time constant states in the models of generators and their controls. The model reduction is performed in such a way that the small time constant states can be eliminated or reintroduced under program control. Thus, when the rates of change are high, the dynamics of the fast states can be modeled, but when the rates of change are slow, algebraic approximations to the fast states are calculated. The contribution of this study is to provide a new and efficient technique for treatment of fast states to accelerate power system simulations. 11. SINGULAR PERTURBATIONS AND TWO-TIME SCALES A nonlinear two-time scale system can be represented by

I. INTRODUCTION
In the step-by-step simulation of power system dynamics, the time step is largely determined by fast states due to the very small time constants associated with the generators and their controls. With fixed step integration algorithms, the user tends to specify a small step size which in hisher experience gives adequate accuracy. Even when adjustable time step integration algorithms are used, small time steps are required immediately following a fault. Thus, it is tempting to ignore the small time constants. They, however, can have a considerable effect on the stability of controls and their neglect can lead to optimistic simulations. Apart from their effect on the integration time step, time constants, no matter how small, require to be represented by dynamic states. In the integration algorithms, this means the evaluation of the rate of change of these states and calculation of their values at the next time step. Thus, small time constant states increase the time required for dynamic power system simulations. The simulation speed is becoming more and more important. On one hand, economic power system operation requires contingency evaluation in close to real time if not
PE-l18-PwRS-l-O4-1997 A paper recommended and approved by the IEEE Power System Dynamic Performace Committee of the IEEE Power Engineering Society for publication in the IEEE Transactions on Power Systems. Manuscript submitted May 9, 1996; made available for printingApril 11, 1997.

-= f (X, z, E , t )
E--=

dx

dt dz

x(to) = xo

(1)
g(x,z,E,t)
Z(t0)

dt where x and z represent vectors of slow and fast states, respectively. The small parameter, E, determines the time scale separation of the states. By setting E=O in (l), the degenerate or reduced system is given by dx -= f ( x , GO, t ) .c(to> = xo dt 0 = g ( x , z,o, t > This is what is called the singularly perturbed system. In this system, the fast stateis are removed from the differential equations. These fast states follow the slow states and can be obtained by solving the algebraic equation in (2): 2 = h(x. t ) (3)

= zo

0885-8950/98/$10.00 0 1997 IEEE

110

It is a common practice in power system simulators to neglect the fast network transients, but, as shown in [9], this may lead to inaccuracies if the system frequency changes substantially. There are many other fast states in a power system dynamic model, e.g., the states associated with the generator amortisseur windings and those associated with fast automatic voltage regulators (AVR). It is not acceptable, in many cases, to neglect the dynamics of these states. For example, due to the phase shift they may introduce, ignoring the dynamics of the AVR fast states can have a considerable effect on the stability of controls; and neglecting the rates of change of the amortisseur windings states is equivalent to assuming zero amortisseur currents or no amortisseurs, an approximation which implies a reduced system damping and is clearly unsatisfactory. Therefore, a better approximation is proposed which is described in the next section. The two-time scale property due to the simultaneous presence or interaction of slow and fast dynamic phenomena results in a system said to be stiff. In general, a stiff system need not have the explicit singularly perturbed structure of equation (1). However, the power system models considered in this paper do have that form.

difference between the fast states and the values of these states calculated with E in (6) set to zero:
Z f =z-2,

z, = -AT;( A21x + B p )

(7)

where zs is called the tracking part of the slow states and zf the correction term. Taking derivative of zf =z-zs and substituting it and (7) in (6) give

X = ( A l l - Al2A;;A2,)x+ A12zf + (B1 - A12AiiB2)~


Ei, = A22zf -&is

(8)

In the long-term simulation it should not be necessary to retain any differential equations related to the fast states. However, if we completely ignore the fast states, the effect is to assume no amortisseurs. Therefore, a better approximation can be obtained by equating E if in (8) to zero. This gives (9) i.e., the correction term only depends on the rate of change of the slow states. From (7) and (9) we obtain the singular perturbation approximation to the fast states:

1 1 1 . MODELING OF GENERATOR AND ITS CONTROLS


A. Generator Rotor Circuits
Dynamic equations for the flux linkages of a generator with two rotor circuits on each axis can be written as [lo]: @=Avy+B,u (4) where y=[y&, ylq, y l d , y2,IT represents the flux linkages of those four rotor circuits, and u=[efd efq, ed, eq]T represents the components of the stator terminal voltage and field voltage. A , and B, are constant matrices if the generator saturation is neglected or functions of the states if the saturation is considered. Since the time constants associated with the d-axis amortisseur circuit and one of the q-axis amortisseur circuits are generally small, of the order of 0.03, the corresponding flux linkages can be taken as fast states and the remaining field and q-axis amortisseur flux linkages as slow states. To be more specific, let In this model, X , the rates of change of the slow states, are available from (6) in which the slow tracking part of z is used, and u , the rates of change of the generator terminal and field voltages, can be calculated numerically.
B. Generator Equation of Motion

The generators equation of motion can be written as [lo]:

6 = wOAW,
where 6 and d o , are the rotor angle and speed deviation, respectively; T,, T,, H and KO are the generator electrical and mechanical torques, inertia and damping constants, respectively; and COO is the system base frequency. These two states are treated as the slow variables and so retained in the simulation.
C. Static Exciter and Power System Stabilizer

Then, (4) can be reformulated as the two-time scale form: G = Allw-CAlzz~Blu


= A 2 1 ~ A222 B ~ u where x and z represent the slow and fast states, respectively. The small parameter E reflects the subtransient time constants TL0 and TGO that are much smaller than their transient counterparts Tho and T i o For a better approximation, we define new states as the
EZ

(6)

Fig. 1 shows the dynamic model of a static exciter with PSS [lo], corresponding to IEEE Types ST1 and PSSlA. By their time constants, the states x I , x2, x4 and x5 are fast variables and can be split into two parts in a similar manner to that used in subsection A above, Le., one being the part tracking the slow states (xj, and Am, ) and the other the correction term. The singular perturbation approximation to the fast states are expressed as:

111 models, have been implemented and tested in a two-area, four-machine power system shown in Fig. 2. This system is taken from [lo]. The system consists of two identical areas connected through a relatively weak double-circuit tie line. Each area includes two 900 MVA generators equipped with static exciters, PSS and governors. All system data are given in Appendix A. Under the steady state, 400 MW power is transferred from Area 1 tlo Area 2. For the long-term study, a severe disturbance is created as follows: at 1 second a three-phase fault is applied at node 7 for 0.15 second; at 10 seconds 40% load is rejected at node 7; at 20 seconds an additional 50 % load rejection occurs at node 7. This severe disturbance results in an over 2% steadystate frequency deviation. However, we do not account for off-nominal frequency effects in this paper. A discussion of singular perturbation analysis applied to off-nominal frequency effects is given in [9]. A 2-minute time simulation is performed, but, for clarity, the results are plotted foir a period of only 30 seconds; after this time the system is alrnost at its new steady state. A. Neglecting the Rates of Change o f Fast States

Power System Stabilizer

"""

Fig. 1 Fast Static Exciter and Power System Stabilizer Model

x 2 = xZs

+x z f

= VRef - Et - (TA

+ TR)'- dt
dE

(12)
d

+ [ ( K s A m r - x 3 ) - ( T A -T1 +T2 - T 3 + T 4 ) - ( K S A m r dt

-x

,)I
Efd

x4 =x4s +

x = ~[ ( ~K ~ -A X 3~ )-T ~ 2 - (d KsAmr dt

-x3)]

x5 = x 5 s + x s f = [ ( K S A m r - x 3 ) - ( ~ 4 -*I

+ ~ 2 ) x ( ~ s -x3)] ~ m r

and the exciter gain KA is represented in the expression KA x2.

D. Governor and Steam Turbine


Governors and turbines are more important in the longterm simulation than in the transient simulation. Their response is inherently slow. However, some of the governor time constants are relatively fast and the states associated with these fast time constants can be eliminated to produce a reduced order model for the long-term simulation. An IEEE generic model for governor and steam turbine is shown in Appendix A [lo]. The states yl, yz, y3, y5 and y6 can be treated as relatively fast variables during the long-term simulation and split into two parts. One part tracks the slow states (y4 and Amr ) and the other part is the correction term. Their combination is:

In this case, the rates of change of the fast states associated with the generators, statiic exciters, PSS, and governors, are simply set to zero. Figs. 3 and 4 show speed (frequency) and active power of the generator at node 1, with the solid line representing the full models which simulate the fast states fully and the dotted line neglecting the rates of change of the fast states. It can be seen that neglecting the rates of change of the fast states gives optimistic results which have a large error compared with those obtained from the full models and are therefore unsatisfactory.

B. Application o f Singularly Perturbed Models


Figs. 5 to 8 show the results of the singularly perturbed models, i.e., from (10) to (13), with the solid line denoting the full models and the dotted line the singularly perturbed models. It can be seen thlat these models give close results in most periods of the simulation. There are some differences around the times of the disturbances. The surges in the generator mechanical torque curve depicted in Fig.8 are due to sudden variations in the rate of change of the generator
AREA 1 400 MW

AREA 2

P=700 MW Q=185 MVAr

P=719 MW Q=176MVAr

Iv. IMPLEMENTATIONAND CASE STUDIES


Equations (IO) to (13), referred to as singularly perturbed

P=700 MW Q=235 MVAr

P=700MW Q=202 MVA,

Fig. 2 Two-Area Power iiybmm

112

59 8 1

I
5
10

450 I

I
5
10
15 TIME IN SECONDS

15
TIME IN SECONDS

20

25

30

20

25

30

solid line - full model dotted line - neglecting the rates of change of the fast states Fig. 3 Generator Speed (Frequency) at Node 1

solid line - full model dotted line - singularly perturbed model Fig.6 Generator Active Power at Node 1

E 5505
I3

500 -

O 9 t

1 I

TIME IN SECONDS

solid line - full model dotted line - neglecting the rates of change of the fast states Fig. 4 Generator Active Power at Node 1

500

I
5
10

15
TIME IN SECONDS

20

25

30

59 8

10

solid line - full model dotted line - singularly perturbed model Fig.5 Generator Speed (Frequency) at Node 1

15 TIME IN SECONDS

20

25

30

solid line - full model dotted line - singularly perturbed model Fig.8 Generator Mechanical Torque at Node 1

C. Switching Between Full and Singularly Perturbed Models


generator speed (dAcu,./dt) inoluded in (13), and oaused by the

disturbances (fault and load rejection). In order to achieve higher accuracy of the simulation around the times of the disturbances, a switching can be performed between the full model and the singularly perturbed model. This is relatively easy since the fast states are approximated but not eliminated in the singularly perturbed model.

At a switching point, the rates of change of the slow states can be checked. A sudden change in the slow states will excite fast state transients and the differential equations are solved for the fast states. As soon as changes in the slow states become smaller than a prespecified tolerance, the differential equations of the fast states are eliminated from the simulation and the algebraic tracking approximations used.

113
For the system under study, the rate of change of the generator speed (a slow state) dAco,./dt is used to determine the switching. The tolerance is chosen to be 0.001. Figs. 9 to 12 show the results with the switching. It can be observed that the accuracy of the results is improved. In practice it may be necessary to define a number of different speed indicators.
1.25,

D. Integration Method and Computation Time


For the system under study, a fourth order Runge-Kutta formulation with the Gill coefficients (RKG) is used [lo]. The maximum time step in the integration of the full model based on the smallest time constant is 0.01 second. On the other hand, the singularly perturbed model can be integrated with a maximum time step of 0.04 second, beyond which issues of accuracy and numerical instability start arising. For a 2minute simulation run on an IBMRISC6000 Workstation, the singularly perturbed model results in a substantial time saving, typically 35 %, as compared to the full model. The possibility of enhancing the time saving with alternative integration techniques is currently being investigated
6 1 . 2

solid line - full model dotted line - singularly perturbed model Fig. 11 Generator Voltage at Node 1

5 o o -

10

15
TIME IN SECONDS

20

25

solid line - full model dotted line - singularly perturbed model Fig.12 Generator Mechanical Torque at node 1

V. CONCLUSION
J
5
1 0

59.8'

15 TIME IN SECONDS

20

25

30

solid line - full model dotted line - singularly perturbed model Fig.9 Generator Speed (Frequency) at Node 1

The singularly perturbled models for generators and their controls have been devel oped in this paper. The simulation results with these models are in close agreement with those obtained with the full model. These models allow for a large integration step which saves computational time and is useful in the power system long-term simulation. The study presented in this paper provides a new and efficient technique in dealing with fast states in the power system simulation. VI. ACKNOWLEDGMENT The authors thankfully acknowledge Dr. R.K. Varma for his assistance and discussion in performing the above studies. VII. REFERENCES

4501

"

I
5
10

15

20

25

30

TIME IN SECONDS

solid line - full model dotted line - singularly perturbed model Fig.10 Generator Active Power at Node 1

[ l ] J.H. Chow, J.J. Allemong and P.V. Kokotovic, "Singular Perturbation Analysis of Systems witlh Sustained High Frequency Oscillations," Auromatica, Vol. 13, 1978, pp. 271-279. [2] P.V. Kokotovic, J.J. Allemong, J.R. Winkelman and J.H. Chow, "Singular Perturbation and Iterative Separation of Time Scales,'' Automatica, Vol. 16, 1980, pp. 23-33.

J.R. Winkelman, J.H. Chow, J.J. Allemong and P.V. Kokotovic, "Multi-Time-ScaleAnalysis of a Power System," Automatica, Vol. 16, 1980, pp. 35-43. J.H. Chow, Editor, Time Scale Modeling of Dynamic Networks with Application to Power Systems, Lecture Notes in Control and Information Sciences, Berlin: Springer-Verlag, 1982. G. Peponides, P.V. Kokotovic and J.H. Chow, "Singular Perturbations and Time Scales in Nonlinear Models of Power Systems,'' IEEE Transactions on Circuit and Systems, Vol. CAS-29, No. 11, November, 1982, pp. 758-767. S. Ahmed-Zaid, P.W. Sauer, M.A. Pai and M.K. Sarioglu, "Reduced Order Modeling of Synchronous Machines Using Singular Perturbation," IEEE Transactions on Circuit and Systems, Vol. CAS29, No. 11, November, 1982, pp. 782-786. P.W. Sauer, D.J. LaGesse, S. Ahmed-Zaid and M.A. Pai, "Reduced Order Modeling of Interconnected Multimachine Power Systems Using Time-Scale Decomposition," IEEE Transactions on Power Systems, Vol. PWRS-2, No. 2, May 1987, pp. 310-320. P.W. Sauer, S. Ahmed-Zaid and P.V. Kokotovic, "An Integral Manifold Approach to Reduced Order Dynamic Modeling of Synchronous Machines," IEEE Transactions on Power Systems, Vol. PWRS-3, No. 1, February 1988, pp. 17-23. R.K. Varma, R.M. Mathur, G.J. Rogers and P. Kundur, "Modeling Effects of System Frequency Variation in the Long-Term Stability Studies," Paper 95 SM 537-1 PWRS, Accepted for Presentation at the IEEE PES Summer Meeting, Portland, OR, USA 1995. [lo] P Kundur, Power System Stability and Control, New York: McGrawHill, 1994.

governor

steam turbine

BIOGRAPHIES
X. Xu received his BSc. and M.Sc. degrees from the Southeast University and the Electric Power Research Institute (EPRI), China, in 1982 and 1984, respectively, both in Electrical Engineering. From 1984. to 1991, he worked in EPRI (China) as a power system engineer. Since September 1991, he has been a Ph.D. candidate at the University of Westem Ontario, Canada. During 1992 - 1993, he worked with Ontario Hydro, Canada. R.M. Mathur received his BSc. degree in Science from Rajputana University, B.Tech.(Honours) degree in Electrical Engineering from Indian Institute of Technology(I1T) Kharagpur, and Ph.D. degree from the University of Leeds, UK in 1956, 1960 and 1969, respectively. From 1960 to 1966 he worked as Lecturer in India and from 1966 to 1969 he was a research student at the University of k e d s . From 1969 to 1987 he was a Professor and Head (1980-1987) of the Department of Electrical Engineering at the University of Manitoba, Canada. Since July 1987, he has been the Dean of the Faculty of Engineering Science at the University of Westem Ontario, Canada. Dr. Mathur's research interests are in Electrical Machines and Power Systems. He is recipient of the IEEE Centennial Medal. He is a member of IEE(London) and is a registered professional engineer in the province of Ontario. J. Jiang received his Ph.D. degree from The University of New Brunswick, Fredericton, New Brunswick, Canada in 1989. He joined the Department of Electrical Engineering, The University of Westem Ontario, London, Ontario, Canada in 1991 where he is currently an Associate Professor. His main research interests are control systems, power system dynamics and controls, and advanced signal processing.
6. J. Rogers graduated in Electrical Engineering with first class honours from Southampton University in 1961. From 1961 to 1964 he was employed as a consultant mathematician by AEI (Rugby) Ltd. From 1964 to 1978, he was lecturer in Electrical Engineering at Southampton University. From 1978 to 1993 he worked with Ontario Hydro from where he retired as Senior Engineer, Specialist-Controls in the Power System Planning Division. Currently, he operates Cherry Tree Scientific Software at Colbome, Canada. He also holds the positions of Associate Professor (part-time) at McMaster University and Adjunct Associate Professor at the University of Toronto. He is a Registered Professional Engineer in the Province of Ontario.

APPENDIX A

All four generators are identical and their parameters in per unit on 900 MVA and 20 kV base are as follows: xd = 1.8 X q = 1.7 X l = 0.2 X i = 0.3 X G = 055
= 0.25
X; = 0.25

Ra = 0.0025

TLo = 8.0 s

G'o = 0.4 s
Bsat = 9.6

T&b = 0.03 s
yl-1 = 0.9

q b ' = 0.05 s

Asat = 0.015

H = 6.5 KO = 0.0 where A,,,, B,,, and vTI define saturation function. The static exciters and PSS have the following data: KA = 200.0 TA = 0.0 TR = 0.01 K s = 20.0 Tw = 10.0 Ti = 0.16 T2 = 0.02 = 0.16 = 0.02 Vmax = 0.2 VmIn = -0.2 Efd = 5.0 E f d fin = -5.0
Each step-up transformer has an impedance of O+j0.15 per unit on 900 MVA and 201230kV base and an off-nominal ratio of 1.0. The transmission system nominal voltage is 230kV. The parameters of the lines in per unit on 100 MVA, 230kV base are: r=O.OOOlpu/km xL =0.001 pulkm bc=O.O0175pu/km The loads at nodes 7 and 9 are modeled as constant impedance in the dynamic system simulation.
The governors and steam turbinos have the following

dynamic model and data: K1 = 15.0 T G = ~ 0.1 T G = ~ 0.0 T G = ~0.25 T G = ~ 0.42 T G= ~ 4.25 T G = ~ 0.72 T G = ~ 0.0 K2 = 0.25 K3 = 0.25 K4 = 0.5 K5 = 0.0 K6 = 0.0 K7 = 0.0 K8 = 0.0 RMA, = 1.1 RMIN= -05 PMAX(1) = 1.0 PMIN(2) = 0.0

P. Kundur received his M.A.Sc. and Ph.D. degrees from the University of Toronto, Canada, in 1965 and 1967, respectively. He taught at Mysore and Bangalore Universities during 1967-1969. From 1969 to 1993 he worked with Ontario Hydro from where he retired as Manager of the Analytical Methods & Specialized Studies Department in the System Planning Division. He also holds the positions of Adjunct Professor at the University of Toronto, University of Westem Ontario and University of British Columbia. He is currently President and CEO of Powertech Labs Inc. at Surrey, B.C. Dr. Kundur was elected a Fellow of the IEEE in 1985 and is a member of several IEEE Working Groups and Task Forces. He is also a member of CIGRE Committee 38 and several other Task Forces.

Das könnte Ihnen auch gefallen