Froceedln@ of the 3161 Conhnnca

on k l r l o n and Control
Tucson, Arkona k e r n b a r 1982
0 WPIO 16:20 -
ROBUST NONLINEAR CONTROLLER DESIGN FOR TRANSIENT
STABILITY ENHANCEMENT OF POWER SYSTEMS
Youyi Wang, Lihua Xie David J. Hill, Richard H. Middleton
School of Electrical and Electronic Engineering Department of Electrical and Computer Engineering
Nanyang Technological University, Singapore 2263 The University of Newcastle, NSW 2308, Australia

ABSTRACT tive control techniques can be employed ([5]). When a large fault
occurs, it is very difficult for an adaptive controller to adapt
In this paper, a robust nonlinear excitation controller is proposed
to variations in reactances of transmission lines so that when a
to enhance transient stability for power systems. Robust control
fault occurs close to the generator terminal, the system using an
technique in linear systems ([l] and [2]) is extended to design adaptive controller usually can not maintain transient stability
a robust nonlinear excitation controller for power systems. The
Design of the resulting controller is indepentent of the operating (see 141).
point. Simulation results show that transient stability of a power In contrast to adaptive controllers, robust controllers have fixed
system under a large sudden fault can be improved. gains. Few papers, however, considered stability enhancement
of power systems using robust techniques ([6-81). In these design
LIST OF SYMBOLS:
approaches, an approximately linearized model is used as a design
N ( t ) = 6 ( t ) - so; model for controller design and these controllers are linear con-
6 ( t ) : the power angle of the generator; trollers which are designed for a specific operating point. In the
60 : the power angle of the generator at the operating point; case where a large fault occurs, the parameters and the operating
w ( t ) : the relative speed of the generator; point of the system change significantly, and linear controllers de-
APe(t)= P J t ) - P, signed based on an approximately linearized model generally can
P,,, : the mechanical input power; not maintain transient stability. We want to develop a new ex-
P,(t) : the active electrical power delivered by the generator; citation control law which is independent of the operating point
W O : the synchronous machine speed; W O = 2afo and can stabilize the system for all admissible parameter vari-
D : the per unit damping constant; ations. This is a difficult topic in electric power system design
H : the per unit inertia constant; due to the complexity and nonlinearity of a power system. In [3],
E:(t) : the transient EMF in the quadrature axis; we reviewed a nonlinear design approach, which is called direct
E , ( t ) : the EMF in the quadrature axis; feedback linearization (DFL) technique ( [ l l ] ) , for power system
E f ( t ) : the equivalent EMF in the excitation coil; controller design. A new DFL excitation controller has been pro-
??do : the direct a x i s transient short circuit time constant; posed which achieves both transient stability improvement and
Q e ( t ): the reactive power; voltage regulation. Using DFL technique, a power system can
I f ( t ): the excitation current; be linearized in the whole working region (0 < 6 ( t ) < 180') and
I q ( t ) : the quadrature axis current; the linearized model is independent of the operating point of the
& ( t ) : the generator terminal voltage; system. In this paper, to prevent the loss of synchronism under
k, : the gain of the excitation amplifier; a large sudden fault, robust control techniques for linear systems
u,(t) : the input of the SCR amplifier of the generator; ([1-2]),together with DFL design technique, are employed to de-
Ids = 2" + !jzL + z d ; = ZT + 2ZL t sign a robust nonlinear excitation controller for a power system.
IT : the reactance of the transformer; 2. PLANT MODEL
Z d : the direct axis reactance;
z& : the direct axis transient reactance; Electrical power systems are complex multivariable dynamic sys-
z L : the reactance of one transmission line;
tems. To simplify the discussion, as an approximation, we as-
2 8 = ZT t !jzL sume that the complex generator bus voltage is independent of
i,d the mutual reactance between the excitation coil and the
:
how our generator is adjusted, i.e. it is an infinite bus. The sim-
stator coil; plified model can be described as a single generator connected
through two parallel transmission line to a very large network
V, : the infinite bus voltage.
approximated by an infinite bus. The model is shown in Fig. 2.1.
1. INTRODUCTION
Generator Breaker x
In previous papers ([3] and [4]), we proposed a nonlinear exci-
tation controller and a nonlinear adaptive controller to enhance
transient stability for a power system, while also achieving post-
@+- Transformer
fault voltage regulation. In this paper, a robust nonlinear exci- Fault
tation controller is proposed. Robust control technique in linear
systems ([l]and [2]) is extended to design a robust nonlinear ex- Fig. 2.1 Single machine-infinite bus model
citation controller for transient stability enhancement of power
systems. The fault we will consider in this paper is a symmetrical 3-phase
Due to a variety of effects, such as lighting, severe storms, equip- short circuit fault which occurs on one of the transmission lines.
ment failures, etc., the configuration of an electric power system ZL (=0.4853) is total reactance of one transmission line. X is the

is not constant. Thus important parameters, such as effective fraction of the faulted line to the left of the fault. If A = 0, the
reactances, Z L of transmission lines, in an electric power sys- fault is at the generator bus. X = 0.5 puts the fault in the middle
tem, especially in a multi-machine network, always have some of the line, and so on.
uncertainties. To overcome the parametric uncertainties, adap-
A model we consider can be written as follows ([9-101):
CH3229-2/92/0000-1117$1 .OO 0 1992 IEEE 1117

Authorized licensed use limited to: ASU Library. Downloaded on June 22,2023 at 17:17:45 UTC from IEEE Xplore. Restrictions apply.
Mechanical Equations:

G e n e r a t o r Electrical Dynamics:

(3) where

Electrical Equations:

If we let

I.

+ Tk5Eq(t)coa6(t)w(t)
Ida - P, (12)
we can get from (11)
1
Ti0
+ Til1
A P,(t) = -- A Pe(t) - ~ t ( t ) (13)

The model (1) to (3) then have been linearized. The linearized
model is
Example system parameters used in later simulation studies are
Ab(t) = w ( t )
as follows:
&(t) = - - wD( t ) - A P.(t)
- WO
H H
WO = 314.159; D = 5.0; H = 8.0 sec; Tw = 6.9 sec; kc = 1
Zd = 1.863; Z& = 0.257; ZT = 0.127; E L = 0.4853

The physical limit of the excitation voltage is: where u f ( t )is the new input. Note that the mapping (12) from
u f ( t )to u f ( t ) is invertable, except when sinb(t) = 0 (which is
maz 1 k C q ( t )I= 1.8 p.u. not in the working region).

The operating point is R e m a r k 3.1: From the analysis above we can see that when
the compensating law, (12), is employed the linearized plant is
independent of the operating point of the system. This is of great
60 = 720; P,O = 0.9 p.u.; v,, = 1.0.'.P
importance in power system design. 0
When considering any control law, an important consideration is
In this paper we consider the following fault sequences: whether the compensating law is practically realizable. In our
Fault Sequence 1: (Temporary Fault) case, the linearizing law (12), cannot be implemented since the
Stage 1: The system is in a prefault steady-state. power angle, 6(t), cannot normally be measured. To overcome
Stage 2: A fault.occurs at t = 0.1 sec. this problem, we use equations (7) and (8) to rewrite (12) as
S t a g e 3: The fault is removed by opening the breakers of the
faulted line a t t =, 0 2 5 sec. vf(t) = Iq(t){kuf(t) + TdO(zd - z&)Iq(t)w(t))
Stage 4: T h e transmission lines are restored with the fault
cleared a t t = 1.0 e c ,
Stage I: The syrtem is in a postfault-state. (14)
From (7) we have that Zq(t) is available since P J t ) and Z f ( t )are
Fault Sequence 2: (Permanent Fault) readily measurable physical variables. Also since Qe(t) and w ( t )
Stage 1: The system is in a prefault steady-state. are available,, so the compensating law is practically realizable
Stage 2: A fault occurs at t = 0.1 sec. and from (14) we can get
Stage 3: The fault is removed by opening the breakers of the
faulted line at t = 0.25 sec.
Stage 4: The system is in a postfault-state.
3. N O N L I N E A R C O N T R O L L E R D E S I G N FOR P O W E R
SYSTEMS
From the model given in Section 2 we know that the power system
model is nonlinear. In this section we will discuss the design R e m a r k 3.2: From the analysis above we can see that the com-
principles using DFL technique to design nonlinear controllers pensatiqg law (15) is simple and practically realizable and the
for a power system. lineariration is of the whole working region. Using the nonlin-
ear compensating law (157 we can linearize the power system
Since E i ( t ) is physically unmeasurable we eliminate $ ( t ) by dif- model through the excitation loop. After linearization the plant
ferentiating (6),and using (1) to (6): becomes a linearized model. 0

1118

Authorized licensed use limited to: ASU Library. Downloaded on June 22,2023 at 17:17:45 UTC from IEEE Xplore. Restrictions apply.
When a large sudden fault occurs the reactance of the transmis- Remark 4.1: The parametric uncertainties A A and AB can be
sion line Z L changes a lot, which is one reason why the system expressed as follows:
may lose synchronism. In the next section we will introduce ro-
bust control design techniques to improve the transient stability
[ A A , A B ] = DJ'(t)[EI, E,]
of power systems. where D , E1 and E2 are constant matrices;
4. ROBUST CONTROLLER DESIGN
As we discussed above, when the parameters in the power system
are known, we can design a DFL control law to linearize the
plant. But when a large sudden fault occurs the reactance of the
transmission line Z L changes a lot. In the paper these changes
D = [ 0 , 0 , 1IT;

Ez = [O, 0, lIT;
El =
[:::]0

F ( t ) = [O, 7-r
0

P -p(t)I
are treated as parametric uncertainties. IPI
and F T ( t ) F ( t )5 aZZ( C Y > 0 ) , 7 is a constant. 0
Considering the uncertainty in z ~the
, plant model becomes
Remark 4.2: From the above analysis we see that the DFL
Ab(t) = w(t) control law (16) is well defined if and only if
D
w ( t ) = --w(t) - 5 A P,(t) U t )# 0.
H H If I,,(t) = 0, from (7), we know that the non-zero power angle
6 ( t ) = k x 180" where IC is an integer. At normal operating point
the power angle is between ' 0 and 180". Normally, if the power
angle reaches 180°, it is not possible to maintain synchronism. 0
Remark 4.3: To design a robust control law u j ( t ) to stabilize
the power system model (1)-(10) with uncertainties is equivalent
where to design a robust control law v j ( t ) to stabilize the linearized
plant (20). 0

To solve the robust control problem for the linearized plant (20)
A Z Ldenotes the uncertainty in XL, and involves solving the following algebraic Riccati equation ([l]and
(21):
( A - BR-'EFE1)'P + P ( A - BR-'EFEI)
+P ( ~ D -D B~R - * B ~ ) + E ~o
P E;(I - E ~ R - ~ E : ) = (21)
where R = ErEZ > 0 .
Our main "stability" result is as follows:
Theorem 4.1: The power system (1)-(10) under a symmetrical
3-phase short circuit fault is transiently stable via the nonlinear
we have
DFL control law

- ( z d - zL)Tmw(t)Ii(t) Pm} (22)

and
v , ( t ) = -R-'(BTP + ETE1)z(t) (23)
if and only if there exists a stabilizing solution P 2 0 for the
Riccati equation (21). The power system (1)-(10) is transiently
stable via the control law (22) and (23) means that the power
system under a symmetrical %phase short circuit fault can avoid
the loss of synchronism. Morever in the postfault period we have
Choosing the state as z T ( t ) = [A6(t), w ( t ) , APe(t)],then the that
equaions (17)-( 19) become lim 1 A6(t)I= 0
1-m

k(t)= ( A + A A ) z ( t )+ ( E + A B ) v j ( t ) (20) lim I w ( t ) I= 0

1-m

where lim I A P e ( t )I= 0

t-m

Proof: Using DFL technique, the power system (1)-( 10) can be
linearized as (20), so that to design a robust nonlieanr control law
u j ( t ) to transiently stabilize the power system (1)-(10) under a
a symmetrical 3-phase short circuit fault is equivalent to design
a robust linear control law v j ( t ) to stabilize the linearized plant
(20) with parametric uncertainties.
From the proofs in 111 and (21, we know that a linear plant in
the form (20) is quadratically stable if and only if there exists a
1119

Authorized licensed use limited to: ASU Library. Downloaded on June 22,2023 at 17:17:45 UTC from IEEE Xplore. Restrictions apply.
stabilizing solution P 2 0 for the Riccati equation (21). Morever, erator terminal voltage so that we can achieve, in addition, that
a suitable feedback law is given by
lim
1-00 I A&(t) I= 0
v f ( t )= -R-’(BTP + ErEl)Z(t)
0
0
The design procedure is as follows:
Remark 4.4: Since the physicallimit of the excitation voltageis Step 1: The fault occurs at t = t o and the robust control law is
considered, in the case where a symmetrical 3-phase short circuit
fault, as described in Section 2, occurs on one of the transmission
lines (for A not too small), the robust control system maintains
transient stability of the power system in some cases. The small-
est value of X for which transient stability is retained can be found -(zd - z&)TdOw(t)Ii(t) + pm}
by simulation. Simulation results are shown in Section 5. 0 and
Remark 4.5: In the robust control scheme proposed in this v f l ( t )= -k6 * A$(t) - ku * w ( t ) - kp * AP.(t)
section, the robust DFL control law is where [k6, kw, IEp] = R-’(BTP E r E l ) . +
Step 2: At t = tl the feedback law is switched to
1
Uf(t)= -M t ) - TAQe(t)w(t)- Z 0 4 )
kCIP(t)
-(zd - z$)TdOW(t)Ii(t) + Pm}
and
v j ( t ) = -k6 * A6(t)- ku * ~ *
( t-) kp AP,(t)
where [k6,b,&] = R-’(BTP +ErEl).
Since the power angle 6 ( t ) can not be measured, the estimate o f
* AV(t) - leu] * ~ ( t-) &I * APe(t)
vfz(t) = - k u i
6 ( t ) , k(t), can be obtained by using 6-detector ([3]) where the design procedure for h,kul and can be found in
[3] and [12]. tl is called the switch time.
%(t)= w ( t )
5. SIMULATION RESULTS
Then v f ( t ) becomes
In our example from (20) we can obtain that
* &(t) - k, * w ( t ) - kp * APe(t)

I;
Vf(t) = -k6 (24)

1
0 1 0 0 0 0
where Ak(t) = k ( t )- 60. As we show above, the control law (22)
and (24) can stabilize the following plant
A= 0 -0.625
0 0
-39.2699
-0.51635 AA=[: ;,I)]
Ak(t) = w(t) B = [0, 0, 0.51635p; AB = (0, 0, p(t)lT
D
&(t) = - # t ) - and
1
1 1 + P ( t ) l [ a O + ~TAQc(t)l
P = [jv&
Ai).(t) = -[-
TA
+At))A p=(t)+ [T
Tdo
+ ~(t)j[le,4(t)uf(t)
We have I p ( t ) 15 0.2347; I ATio 15 0.6053;
+ T ~ o (-zz ;~) I ; ( t ) ~ ( t+) (TA+ ATAO)Q.(t)w(t) 1 AT& I_<3.09050; I p 15 3.23,
+(TA + ATk)w(t)- P,} Choosing
Due to the difference of the initial values between 6 ( t ) and k(t)
it is clear that by introducing k(t) in the control law (24) we can
no longer obtain that D = [0, 0, 1IT; E1 =

lim I Aa(t) I= 0
i-m

We can only obtain that

lim I A6(t) I= conatant

7 = 0.2347 and u2 = 2 x (0.2347)’ > 0. We have
1-m
R = ErE2 = I > 0
0
and
Remark 4.6: Since we do not introduce the generator terminal F T ( t ) F ( t )5 a21
voltage, & ( t ) , in the control, we may not obtain that
Solving the Riccati equation (21) gives
lim I A&(t) I= 0
i-m
[k&:a,L,kp] = [-1.31, -15.78, 61.631
Thus in many c w s , the generator terminal voltage is not the
The robust control law is
same in the postfgult state as in the prefault state, which is un-
desirable in practice. To achieve the postfault regulation of the +
v f l ( t )= 1.31 * ( k ( t )- 60) 15.78 * w ( t ) - 61.63 * (P.(t) - Pmo)
generator terminal voltage x ( t ) ,we employ the robust controller
proposed in the section t o enhance the transient stability of the From [3) and [12],we have
power system and the DFL excitation controller design appro#& v f z ( t ) = -47.03* ( % ( t )- KO)+6.93* ~ ( t- )28.60* (P,(t) - Pmo)
proposed in [3] to improve the postfault performance of the gen-
1120

Authorized licensed use limited to: ASU Library. Downloaded on June 22,2023 at 17:17:45 UTC from IEEE Xplore. Restrictions apply.
The DFL compensating law is as in (22).
The operating point is

so = 720; '
,
P = 0.9 p.u.; KO = 1.0 p . u .

Firstly, we wiil test the influence of initial conditions of the 6-

detector. Different initial values of the &detector (60 = 58', 72'
and 86') are chosen. Fig. 5.1 shows the the results ( t l = 0.7sec).

Tune (uc)

Fig. 5.1

The simulation result shows that the results using 8 in the control
law v j l are very close to that using the real power angle 6 ( t ) in the
control law. Using the &detector, we can overcome the difficulty
of measueing the power angle.
The responses of the power angle under different fault locations
(A = 0.5, 0.1, 0.055, and 0.05) are given in Fig. 5.2 ( t l = 0.7sec).
From the results we can see that since the physical limit of the
excitation voltage is considered, when a fault occurs close to the
generator terminal ( A = 0.05) the system, using only the excita-
tion controller, can not maintain the synchronism.
Fig. 5.3b

Tim (e)

Fig. 5.2
Time (sec)

Next we will test our new robust DFL controller (A = 0.5, tl = 0.7 Fig. 5.4a
sec). Two fault sequences are considered. Fig. 5.3a and Fig. 5.3b
show the results. Fig. 5.4aand Fig. 5.4b show the results at the
defferent operating point (60= 52'):

60 = 470; P,o = 0.45 P.u.;V,, = 1.003 p . u .

1121

Authorized licensed use limited to: ASU Library. Downloaded on June 22,2023 at 17:17:45 UTC from IEEE Xplore. Restrictions apply.
 [5] Pierre, D. A., A Perspective on Adaptive Control of Power
Systems, IEEE Trans. Power Systems, PWRS-2, pp.387-
396,1988.
(61 Chow, J. H., A Pole-Placement Design Approach for Systems
with Multiple Operating Conditions, Proc. 1988 CDC,
pp.1272-1277, 1988.
[7] bthman, H., J. J. Sanchez-Gasca, M. A. Kale and J. H.
Chow, On the Design of Robust Power System Stabilizers,
Proc. of 28th CDC, pp.1853-1857, Tampa, Florida, 1989.
[8] Pai, M. A. and P. W. Sauer, A Framework for Application of
Generalized Kharitonov’s Theorem in the Robust Stability
Analysis of Power Systems, Proc. of 28th CDC, pp.1818-
1821, Tampa, Florida, 1989.
(9) Anderson, P. M. and A. A. Fouad, Power System Control and
Stability, Iowa State University Press, Ames, Iowa, 1977.

From the results, we can see that using the new robust DFL ex- (10) Bergan, A. R. Power Systems Analysis. Prentice-Hall, New
citation controller proposed in Section 4, we can achieve both Jersey, 1986.
transient stability enhancement and voltage regulation The per- (111Gao, L., L. Chen, Y. Fan and H. Ma, DFL-Nonlinear Con-
formance of the controller is independent of the operating points. trol Design with Applications in Power Systems, submitted
6. CONCLUSION for publication, 1990.

In this paper, a new robust DFL excitation controller is proposed [12]Wang, Y., D. J. Hill, L. Gao and R. H. Middleton. Transient
t o improve the transient stability and to achieve the voltage reg- Stability Enhancement and Voltage Regulation of Power
ulation of power systems. Simulation results show that the ad- Systems, IEEE Trans. Power Systems, to appear, 1992.
vantage of the controller are

0 The fault location does not need to be known and the con-
troller can overcome the variation of the reactance of the
transmission line;
Both transient stability enhancement and voltage regula-
tion can be achieve;
0 The performance of the controller is indenpendent of the
operating point;
0 The controller is simple and effective.

The robust DFL controller proposed here can easily be extended

t o the case where more parameters, such as H, D , x d , etc, are
considered unknown. Also it can easily be extended to the case
where the DFL coordinated controller ([12]) is used.
REFERENCES:

[l] Khargonekar, P. P., I. P. Petersen and K. Zhou, Robust

Stabilization of Uncertain Linear Systems: Quadratic Sta-
bilizability and H m Control Theory, IEEE Trans. Auto.
Control, AC-35, pp. 356-361, 1991.
[2] Xie, L., M. Fu and C. E. de Souza, H , Control and Quadratic
Stabilization of Systems with Parameter Uncertainty via
Output Feedback, to appear IEEE Trans. Auto. Control,
1992.
[3] W a g , Y., L. Gao and D. J. Hill. On the Design of New
Nonliear Excitation Controllers of Power Systems. Proc.
of 2nd Inter. Conf. Autom., Robotics and Compu. Vision,
Singapore, 1992.
[4] Wang, Y., D. J. Hill,R. H. Middleton and L. Gao, Tran-
sient stabilization of Power Systems with Adaptive Control
Laws, submitted for publication, 1992.

1122

Authorized licensed use limited to: ASU Library. Downloaded on June 22,2023 at 17:17:45 UTC from IEEE Xplore. Restrictions apply.