You are on page 1of 6

SICE Annual Conference 2007

Sept. 17-20, 2007, Kagawa University, Japan

Discrete Time Models of a Continuous Power System Stabilizer


G. Shabib1 and N. Hori2
Graduate School of Systems and Information Engineering, University of Tsukuba, Tsukuba, Japan
(1 Tel: +81-29-853-5566, Fax: +81-29-853-5207 E-mail: gabershabib@yahoo.com)
(2 Tel: +81-29-853-5139, Fax: +81-29-853-5207 E-mail: hori@kz.tsukuba.ac.jp)

Abstract: The rise of electricity demand in a power system requires the use of discrete-time devices. They are widely
spread and play an essential task in the operation and control of power systems. Several kinds of digital controlled
devices have been put into practical use in power systems for the last decade, such as power system stabilizer PSS,
automatic voltage regulator AVR, and proportional-integral plus derivative PID controllers. In this paper, an online
digital PSS is designed for single machine infinite-bus systems. In this study the power system (plant) is represented by
nonlinear model while the PSS (controller) is represented by a three-pole three-zero, lead-lag type PSS. Tustin’s
discretization method is used to discretize the transfer function of the PSS. The effect of sampling intervals on the
proposed digital PSS parameters is examined through simulations. They show that this approximation yields a good
digital PSS for sufficiently small intervals. It was also found that, to counter larger disturbances, smaller sampling
intervals are desired (about 2ms-8ms). For larger sampling intervals, fine-tuning of digital PSS parameters is required.
The digital PSS is applied to one machine infinite-bus system, tested for different load conditions and disturbances, and
found to be satisfactory.

Keywords: Discrete time, Power system stabilizer, dynamic stability, single-machine with infinite-bus

1. NOMENCLTURE availability of low cost digital computers. [4-5]. Digital


control systems form a class of control system which is
Z Shift operator probably the most popular in current practice. In a
Ke Exciter gain. digital control system, a control algorithm is
Te Exciter time constant. implemented in a software form, which requires a
Efd Exciter output voltage. discrete-time formulation. The generating unit consists
Vt , VB Terminal and infinite-bus voltage of a set of first order differential equations representing
respectively. the models of the synchronous machine windings, the
Vref Reference voltage. excitation system, the mechanical system, and another
D Damping coefficient. set of differential equations representing the PSS. This
δ Torque angle between q-axis and infinite- paper aims to replace the analog PSS by its digital
bus. version that is represented by a set of difference
id , iq Currents in direct and quadrature axis. equations and has the same performances as the analog
ikd , ikq Currents in dampers winding. type.
if Current in field circuit. As explained above, power systems involve both
Xd , Xq Direct and quadrature axis reactance. continuous and discrete variables. Their evaluation is
Xkd , Xkq Direct and quadrature axis damper given by equations that generally depend on both. Our
winding reactances. aim is to discretize the analog PSS model to obtain a set
Xmd , Xmq Mutual direct and quadrature reactance. of difference equations representing the PSS. When
Xfd Field circuit reactance. converting a model in continuous S domain to one in
re , Xe Equivalent resistance and reactance of discrete Z domain, a filter is designed in a continuous
the transmission line domain and implemented it in digital form. In digital
r , rfd , rkd , Resistances of armature, field and direct, model-based control implementation, a Z domain model
rkq and quadrature damper winding circuits is also needed. Moreover, performances of discrete-time
respectively. system depend largely on how it is discretized [6]. It is
ω , ωb Angular and based speed respectively. important, therefore, to perform discretization in a
proper manner; to replace the analog PSS by its digital
equivalent having similar performances.
2. INTRODUCTION There are a number of approaches known for
converting a model between S and Z domains [7-10].
Analog PSS have been used widely in the field of Some of the commonly used methods are Tustin’s
power systems to improve dynamic performances and approximation, zero-order hold equivalent, and
disturbance-rejection properties of a synchronous matched-pole method [11], with Tustin’s method being
generator [1-3]. However, it has been witnessed that very popular [6]. To relate S and Z domains transfer
digital control devices are replacing analog ones during function using Tustin’s method, the following
the past decade or so due to their versatility and the approximation is used

- 800 -
PR0001/07/0000-0800 ¥400 © 2007 SICE
1 2 ( z − 1) The IEEE Type ST1 excitation system is considered in
s= ln z ≈ (1) this study [5]. It can be represented as follows
T T ( z + 1)
1 Ke (3)
where T is the sampling interval of the discrete system. E fd = − E fd + ( V re f − V t )
Te Te
An important aspect in digital control systems is the
The output must be limited to prevent the PSS acting to
choice of sampling intervals. With digital controllers
counter action of AVR. The limits of field signals are
that emulate continuous time algorithms, this choice is
taken as ± 5.0 pu in this study.
simple; basically sample as fast as possible. This is
The mechanical shaft is represented by a second order
because of the approximations that are used to generate
swing equation given by
the difference equations describing the controllers. ωb
Smaller sampling intervals mean that the properties of ω = ( P m − Pe − D ω ) (4)
2H
the underlying controller design will be less distorted, δ = ω − ω b
hence more predictable and better performances. This where Pm and Pe are the accelerating power and the
paper proposes the design of a discrete PSS represented electrical power of the synchronous generator,
by linear approximation for single-machine infinite-bus respectively.
system represented by nonlinear differential equations, The steam - turbine - governor system is represented by
by simply discretizing the transfer function of the PSS fifth order [13]. The set of the differential equations
using Tustin’s discretization method. The performance describing the steam - turbine - governor system are
and the stability of the proposed PSS at different given in appendix 2.
operating points and different kinds of disturbances are Eqs. (2)-(4) and the set of the steam - turbine- governor
then verified through computer simulations. system differential equations represented by Eq. (A2.1)
This paper is organized as follows. In section 3 a can be organized in the form.
nonlinear model for the power system is considered. In
section 4 design of the discrete PSS using Tustin’s x = f( x , u ,t) (5)
− − −
mapping model is derived. In section 5 the evaluation of
the discrete PSS and the results of the computer where
simulations are considered to verify the performance x a vector of the state variables (13 × 13).
and stability of the proposed discrete PSS. The effect of u an input vector representing the output of the exciter
sampling intervals on the proposed discrete PSS Efd.
parameters is examined in section 6. The conclusions f a set of non-linear functions describing the
are given in section 7. differential equations of synchronous generator SG.

3. POWER SYSTEM MODEL VB


Vt
The power system considered in this study is a single T TL
-machine that is connected to an infinite-bus through a SG
transmission line as shown in fig. 1. A thirteen order
model including the electrical network, shaft, excitation
system and mechanical part is employed in order to
express in detail the system dynamic behavior which is
described by a set of Parks d-q differential equations
with reference frame based on the rotor [12, 13]. The Exciter&
analog PSS is represented by third order model. It AVR
consists of a washout block, and a lead-lag block.
The input signal to the analog PSS is the rotor
speed deviation ∆ω [5]. The block diagram of
the analog PSS is shown in fig. 2. Digital
PSS
The power system considered in this paper can be
represented by a set of differential equations as follows: Fig 1 One line diagram of a power system model with
Machine windings is represented by fifth order and Discrete PSS
given by:
x w = X 1− 1 [ ω b V1 − ( ω b R 1 + G 1 ) x w ] (2) 4. DESIGN OF DISCRETE POWER SYSTEM
where xw is a state vector represent the state variables of STABILIZER USING TUSTIN’S MAPPING
the machine windings which is X TW = [ i d i q i kd MODEL
i kq i f ]T, V1T = [ VB cosδ VB sinδ 0 0 Efd]T, The analog PSS type is commonly lead-lag structure
while X1 , R1 and G1 are parameter matrices given in used in this study [12]. The transfer function of an
appendix 1. analog, three-pole three-zero, lead-lag type, power
system stabilizer is given by

- 801 -
2
signals are made of increments or differences. In this
KS  1 + ST 
U PSS =  2
 y (6) study the limits UMAX and UMIN of the discrete PSS are
1 + ST1  1 + ST  taken as ± 0.1 pu respectively.
 3 
The eigenvalues of the system matrix of Eq. (11) are
where UPSS and y are respectively, the output and input − T + 2T1 −T + 2T3
signals, K a stabilizer gain, T1 a washout time constant (single) and (double), which are
T + 2T1 T + 2T3
(s) commonly should be relatively large, and T2 and T3
are lead-lag stabilizer time constants (s), of the stabilizer. always stable for any T and positive T1 and T3 , and
The washout circuit is included in cascade with the converge to unity as approaches zero.
lead-lag term to eliminate any unwanted signal in the
steady state. The stabilizer gain determines the amount
of damping produced by the stabilizer. The phase UMAX
compensation lead-lag transfer function used to 2
compensate for the phase lag between the excitation ∆ω  KS   1+ST2 
   
voltage and electrical torque of the synchronous  1+ST1   1+ST3  UPSS
machine. The PSS input signal y can either machine
speed deviation ∆ω or its accelerating power Pa=Pm-Pe UMIN
(difference between the mechanical power and the
electrical power). In this study the input signal to the
stabilizer is the speed deviation of the synchronous Fig. 2 Block diagram of analog PSS model
generator ∆ω . Eq. (6) can be arranged in continuous
form as 5. EVALUATION OF THE DISCRETE PSS
x n = A x n + b υ (7) AND SIMULATION RESULTS
where xn is state variables of the analog PSS and are
y1 , y2 and Upss , states variables y1 and y2 are internal For numerical computation, x(t) are calculated from
state of the analog PSS and υ= SΔω. the numerical integration of the set of differential
A and B are PSS parameters matrices and by matrix equations Eq.(5) by using fourth order Runge-Kutta
manipulation can be written as method, while xn are obtained by solving the set of
  difference equation Eq.(11) by simply updating it.
1 To evaluate the performance of the discrete PSS the
 − 0 0 
 T1  system response of the proposed discrete PSS is
  compared with the cases when no PSS and with an
  1 T2  1
A=  − − 0  analog PSS type in the system. The comparison is
 T TT  T  carried out under different kinds of disturbances as
  3 1 3  3

T  1  (8) follows:
2 T 2   1 T 2
 1
  −   − 2 − 
 T3  T3 T1T3   T3 T3  T3  A. Three phase short circuit at the terminal of the
synchronous generator for 100 ms duration with
K KT2 KT22  self clearing fault.
bT =  .
 T1 T1T3 T1T32  B. 15% step increase in the active power of the load.
C. 15% step increase in the active power of the load
The Laplace transform of Eq. (7) is with the power system operating at leading power
factor 0.98.
Sx n = Ax n + bS∆ω (9) The values of the parameters of the generating unit and
Substituting by Tustin rule from Eq. (1) into Eq. (9) and the connected power system are given in appendix 3.
rearrange yields the approximate difference equation
which can be written in the form A. Simulation Results[ P=0.8 , Q=0.45 , PF=0.87]:
x n (k+1) = [ I - AT ]-1 [ I + AT ] x n Fig. 3 and fig. 4 respectively, are the comparison of the
2 2 dynamic responses under 100ms three phase short
+[I- AT -1
] B [ ∆ω (k+1) + ∆ω (k)] (10) circuit fault of the studied system starting at t= 3s. It can
2 been seen that the analog PSS has better damping of the
Using the difference operator Z-1 Eq. (10) can be written rotor angle and speed deviation than when there is no
in the form as PSS in the system but the proposed discrete PSS has
 T 
−1
 T  approximately closed amount of positive damping
x n (k ) =  I − A   I + 2 A  x n (k − 1) compared with the analog type. The sampling interval in
 2    (11)
−1
the case of analog PSS is 0.001s while it modified in the
 T  case of discrete PSS to 0.008s to simulate the real
+I − A b {∆ω(k) + ∆ω(k − 1)}
 2  system. This modification in sampling intervals leads to
where xn(k-1) and ∆ω (k-1) denote the delayed output modify the parameters of the discrete PSS to K=0.08,
and input state of the discrete PSS. These delayed T1= 10s, T2=0.15s, T3=0.06s with a sampling period of

- 802 -
0.008s. It is clear that the discrete PSS has the same 75
effect as the analog PSS when comparing the output
stabilizing signals of two controllers as shown in fig. 5. 70

65

Rotor angle ( deg )


B. Simulation Results[ P=0.8 , Q=0.45 , PF=0.87]:
60
Fig. 6 and fig. 7 are the comparison of the system
responses to 15% step increase in the load power 55 Without PSS
( 0.8 pu to 0.92 pu ) at t=3s. The parameters of the With analog PSS
50
discrete PSS is modified to K=0.08, T1= 10s, T2=0.15s, With discrete PSS
T3=0.07s with a sampling period of 0.008s. The rotor 45
angle and the speed deviation settle to the new operating 40
point after the second swing which reveal that the 0 2 4 6 8 10 12
Time (sec)
amount of the positive damping of the discrete PSS is
very close to the amount obtained from the analog PSS. Fig. 3 Rotor angle responses to a three phase short
Again, fig. 8 shows that the output control signals of the circuit for 100ms.
two controllers are approximately closes in action to 2.5
each other which reveal that the digital PSS can be
2
replaced the existing analog PSS.

Rotor speed deviation ( rad/sec )


1.5

C. Simulation Results[ P=0.8 , Q=-.126 , PF=0.98 1

leading]: 0.5
Fig. 9 and fig. 10 are the comparison of the rotor angle 0
and speed deviation responses due to 15% step change -0.5 0 2 4 6 8 10 12
in active load power with leading power factor. This test -1 Without PSS
is necessary to explain the validity of the proposed -1.5 With analog PSS
discrete PSS to this type of disturbances. The power With discrete PSS
-2
system is operated near the unstable region and the
-2.5
result of the power system without controller was Time ( sec )
unbounded. The parameters of the discrete PSS are
Fig. 4 Speed deviation responses to a three phase short
modified to K=0.08, T1= 10s, T2=0.15s, T3=0.07s and a
circuit for 100ms.
sampling period of 0.004s is used. In this type of
disturbances the sampling period is reduced which 0.1
reveal that the rate of sampling must be fast. In this kind 0.08
of disturbances four samples each period is done to 0.06

make the discrete PSS match closely to the analog PSS. 0.04
control signal ( pu )

Again, fig. 11 shows the output control signals of the 0.02

two controllers in the case of the power system 0


-0.02 0 2 4 6 8 10 12
operating with leading power factor 0.98.
-0.04 With analog PSS
-0.06 With discrete PSS
6. EFFECT OF SAMPLING INTERVALS ON -0.08
DISCRETE PSS AND DYNAMIC -0.1
PERFORMANCE Time (sec )

Fig. 5 Control signals of the controllers to a three phase


For the design of a discrete PSS, the sampling short circuit for 100 ms.
interval T plays an important role. Fig. 12 shows the
72
dependence of the dynamic performances on sampling
interval. It is notice from fig. 12 that gradual 70
deterioration in performance take place as T increased.
The amount of positive damping obtained from the 68
Rotor angle ( deg )

digital PSS in the case of sampling period 0.002s is 66


better than sampling interval 0.004s or 0.008s. For this
Without PSS
type of disturbance approximately eight samples per 64
With analog PSS
period are done to obtain better results. In the case of 62 With discrete PSS
large disturbance we recommend the use of small
sampling interval, so the properties of the discrete PSS 60
will be less distorted, hence more predictable and better 58
performance obtained. In order to achieve 0 2 4 6 8 10 12
approximately high degree of accuracy with Tustin’s Time ( sec )

discretization method, a very small interval is required. Fig. 6 Rotor angle responses to a 15% Step increase in
load

- 803 -
1 0.02

0.8 0.015
Rotor speed deviation ( rad/sec )

0.6 0.01

Control signal ( pu )
0.4 0.005

0.2 0
0 2 4 6 8 10 12
0 -0.005 With analog PSS
0 2 4 6 8 10 12 With discrete PSS
-0.2 -0.01
Without PSS
-0.4
With analog PSS -0.015
With discrete PSS Time (sec)
-0.6
Fig. 11 Control signals of the controllers to a 15% Step
-0.8
Time (sec) increase in load with leading PF
Fig. 7 Rotor speed deviation responses to a 15% Step
increase in load. 105

0.025 95

Rotor angle ( deg)


0.02
85
0.015
Sampling interval 0.008s
75
Co n tro l sign al ( pu )

Sampling interval 0.004s


0.01
Sampling interval 0.002s
65
0.005
55
0
0 2 4 6 8 10 12
0 2 4 6 8 10 12 Time (sec)
-0.005 With analog PSS
With discrete PSS Fig.12 Effects of sampling intervals in the dynamics
-0.01 performances in case of three phase short circuit for
leading PF.
-0.015
Time ( sec )

Fig. 8 Control signals of the Controllers to a 15% Step 7. CONCLUSION


increase in load.
100 A digital PSS is designed by discretizing a third-order
analog lead-lag type PSS using Tustin’s mapping
95 method. The method is simple to implement. Dynamic
simulations were carried out using the single- machine
Rotor angle ( deg )

90
system connected to an infinite-bus to validate the
effective of the proposed digital PSS. The simulation
85
results show that this approximation yields a good
80
Without PSS discrete-time PSS for sufficiently small intervals. It was
With analog PSS
With discrete PSS
also found that, to counter larger disturbances, smaller
75 sampling intervals are desired (about 2ms - 8ms).
0 2 4 6
Time ( sec )
8 10 12
Smaller sampling intervals mean that the properties of
Fig. 9 Rotor angle responses to a 15% Step increase in the discrete PSS will be less distorted, hence more
load with leading PF predictable and better i n performances. For larger
sampling intervals, fine-tuning of digital PSS
2
parameters is required. The choice of the sampling rate
1.5 is very important and affects significantly the stability
of the power system. With these precautions, the digital
Rotor speed deviation ( rad/sec )

0.5
PSS designed from an analog original is simple and
very useful. It is found that the digital PSS provides as
0
0 2 4 6 8 10 12
good damping enhancement as the analog PSS type, and
-0.5 results in a better response behavior to damp out the
-1
oscillations for various operating points of a single
Without PSS
With analog PSS
machine-system connected to an infinite-bus.
-1.5
With discrete PSS

-2
Time (sec )
ACKNOWLEDGEMENT
Fig. 10 Rotor speed deviation responses to a 15% Step
increase in load with leading PF The authors wish to acknowledge the facilities provided
at the University of Tsukuba, Digital Control Laboratory,

- 804 -
1-1-1 Tennoudai, Ibaraki, 305-8573, Tsukuba, Japan.  −( X + X ) 
0 X mq 0 0
 q e

REFERENCES  0 −( X d + X e ) 0 Xmd X md 
 
Xmq 0 Xkq 0 0 
X1 = 
[1] R. Gupta, B. Bandyopadhyay, A. M. Kulkarni  0 X md 0 Xkd X md 
 
“Design of power system stabilizer for single machine  X 2
X 2md Xfd Xmd 
system using robust fast output sampling feedback  0 − md 0 
technique”, Electr. Power System Res. 65, 2003,  rfd rfd rfd 
247-257.
[2]Y. M. Park, W. Kim “Discrete-time adaptive sliding  0 −ω( Xd −Xe ) 0 −ωXmd −ωXmd 
mode power system stabilizer with only input/output  
measurements”, Electrical Power& Energy Systems, Vol. −ω( Xq −Xe ) 0 −ωXmq 0 0 
 
18, No. 8, pp. 509-517, 1996. G1 =  0 0 0 0 0 
[3] M. A. Abido, Y. L. Abdel-Magid “Analysis and 
 0 0 0 0 0 
design of power system stabilizers and FACTS based
 0 0 0 0 0 
stabilizers using genetic algorithms”, 14th power
systems computation conference PSCC-2002, session
14, paper 4, Seville, Spain, June 24-28, 2002 Appendix 2: The set of differential equations for
[4] Y. Yamamoto, “A function space approach to Turbine & Governor Representation
sampled data control systems and tracking problems”,
IEEE Trans. On AC, VOl. 39(4), pp 703-713, 1994. The differential equations of the steam - turbine -
[5] L. Chen, H. Tanaka, K. Katou, Y. Nakamura
governor system may be derived as follows:
“Stability analysis for digital controls of power systems”,
Electr. Power System Res. 55, 2000, 79-86. 1
x HP = (G VM P0 − x HP )
[6] N. Hori, T. Mori, P. N. Nikiforuk, “A new TCH
perspective for discrete-time models of a continuous 1
-time systems”, IEEE Trans. On AC, VOl. 37-7, pp x RH = (x HP − x RH )
TRH
1013-1017, 1992.
[7] Qing Wang, Qiang Bi, Xue-Ping Yang “High 1
x IP = (G VM x RH − x IP ) (A2.1)
performance conversions between continuous and TIP
discrete time systems” Signal Processing 81, 2001, 1
1865-1877 x LP = (x IP − x LP )
TCO
[8] W. T. Baumann, “Discrete-time control of
 1
continuous-time nonlinear system”, Int. J. control 53(1), G VM = (G − G VM )
1991, 113-128 TGVM
[9] N. Rafee, T. Chen, O. Malik “A technique for TM = FHP YHP + FIP YIP + FLP YLP
optimal digital redesign of analog controllers”, IEEE
Trans. Control System Technology 5(1), 1997, 89-99. HP, IP and LP stand for high, intermediate and low
[10] W. Lin, C. I. Byrnes “Design of discrete-time pressure in per unit respectively. VM is the control
nonlinear control systems via smooth feedback”, IEEE valve.
Trans. AC 39(11), 1994, 2340-2346.
[11] J. W. Grizzle, P. V. Kokotovic “Feedback Appendix 3: Parameters of the generating unit
linearization of sampled-data systems”, IEEE Trans. AC and the connected power system
33, 1988, 857-859.
[12] P. M. Anderson, and A. A. Fouad, “Power system Generator ωb =377 rad/s, Xd =2.0 pu, Xq=1.91pu,
control and stability”, IEEE press, 1993.
[13] Kunder P. “Power system stability and control”, Xfd=1.97 pu, Xkd=1.94 pu, Xkq=1.9 pu, r=0.005 pu,
McGraw Hill, New York 1994. r fd =0.0015 p u, r k d =0.0078 pu, r k q =0.0084 pu,
H =3.25, D =0.0
Appendix 1: Exciter Te=0.01s, Ke=100, −5. ≤ E FD ≤ 5. pu
Turbine and governor system
The parameter matrices R1, X1 and G1 are FHP= 0.24, FIP= 0.34, FLP= 0.42, THP= 0.3s, TRH=10s
 − ( r + re  ,TIP= 0.3s, Po =1.2, TGVM= 0.1s, TGVI=0 .1s
) 0 0 0 0
  Maximum opening and closing rates for both intercept
 0 − ( r + re ) 0 0 0  and inlet valves are restricted to =6.7 pu /s.
R1 =  0 0 rk q 0 0 
Transmission line re=0.063pu, xe =0.4 pu
 
 0 0 0 rk d 0  Operating point
  P=0.8 pu, Q=0.45 pu, VB=1.0 pu
 0 0 0 0 X m d 
Conventional power system stabilizer
K =0.08, T1 =10s, T2 =0.15s, T3 =0.05s

- 805 -

You might also like