Professional Documents
Culture Documents
1515/ijeeps-2013-0037 International Journal of Emerging Electric Power Systems 2013; 14(6): 561–570
Research Article
ability to suppress the potential instability due to poorly signals and some additional sensors and differentiators
damped power angle oscillations that can be dangerous are essential to measure these signals. Another type of
for the system stability and to compensate the voltage FBLC to damp electromechanical oscillations in power
dips and sags that can damage both utility and customer systems is proposed in Ref. [17], but this is achieved at
equipments. the expense of reducing the voltage regulation ability of
Power systems can be stabilised either by including the excitation system.
flexible alternating current transmission systems devices In Ref. [18], it is indicated that the implementation of an
which control the power flow and regulate the voltage exciter control based on the FBLC theory requires a exact
level or by controlling the excitation of synchronous parameters of the power system which is usually not well
generators [7]. As the use of excitation controller is eco- defined in practice. Furthermore, the controllers designed
nomical, this work is mainly focused on this approach. through feedback linearisation require information of the
To improve overall performances of the system, the coor- power system topology, and the states must be measurable.
dination between PSSs and excitation controllers has In practice, it is very difficult to measure all the states of a
been reported in the literature [8–11]. Some of these power system. In addition, feedback linearisation schemes
methods are based on the complex nonlinear simulation, use rotor angle as an output function to cancel the inherent
and the others are based on the linearised power system system nonlinearities. An exact FBLC is proposed in Ref. [19]
model. Excitation controllers designed based on approxi- for power systems with a dynamic loads where a dynamic
mately linearised models depend on a given set of oper- load is connected to a SMIB system. From Ref. [19], it can be
ating conditions and, therefore, may not work properly seen that the law is quite complex even for an SMIB system,
when the operating conditions change [12]. and it would be more complex for a multimachine power
As the real electric power systems have been experi- system which is difficult to implement. To overcome the
encing a dramatic change, the application of advanced difficulty associated with the state variables, an output-
control techniques in power systems has attracted a great feedback linear quadratic Gaussian (LQG) excitation control
attention in recent years to ensure secure system opera- scheme is proposed in Ref. [20]. The LQG controller is more
tion under wide operating ranges [8]. A coordinated AVR realistic, because it is designed using only the measurable
and PSS control has been presented for a single machine outputs and the state variables estimated from them. It is
infinite bus (SIMB) system which allows coordinated well known that the LQR controller provides good robust-
trade-off of voltage regulation and damping enhance- ness in terms of gain margin and phase margin [21].
ments [10]. A robust linear controller has been proposed However, LQR and LQG controllers are unable to provide
to solve the problem of control action coordination due to robustness against uncertainties with changes in operating
variations in the generator terminal voltage and electric conditions [22]. This motivates the use of robust LQG control
power [13]. A unified approach for a voltage regulator and techniques to ensure system stability for large disturbances.
PSS design based on predictive control in the s-domain is The assessment of voltage stability in power systems
presented in Ref. [11]. has recently gained increasing attention, because voltage
Since the transient stability and voltage regulation are instability has been responsible for several major network
ascribed to different causes, some recent proposed scenar- collapses [23]. In contrast to the traditional machine instabil-
ios apply a switching strategy of two different kinds of ity, which deals with the rotor dynamics, voltage instability
controller to cover different behaviours of system opera- is closely related to the load behaviour [24]. Different studies
tion during transient and post-transient periods [14, 15]. have shown the importance of load representation in vol-
The performance of these schemes essentially depends tage stability analyses [23, 25]. Currently, although static
upon the selection of switching time. Moreover, the use load models are commonly used in the power industry to
of different control surfaces through a highly nonlinear model dynamic behaviours of reactive loads, they do not
structure increases the complexity of the designed control- adequately do so [26]. For any systems to be effective practi-
lers. Feedback linearisation schemes are also widely used cally, the inclusion of dynamic load models is essential, but
in the design of robust controllers for power systems which that representation exhibits even more nonlinear beha-
linearise nonlinear power system models into a linear one. viours than the ones with constant impedance loads do.
A feedback linearising controller (FBLC) is used to design a Therefore, dynamic load models are needed in order to
controller for a synchronous generator connected to an analyse the system instability and to design controllers for
infinite bus in Ref. [16]. The direct feedback linearisation enhancing transient stability along with voltage regulation.
theory is easier to understand for power engineers, but it In this article, we consider the problem of designing a
uses rotor angle and accelerating power as feedback linear controller for a nonlinear power system model, in
Transducer
3 Test system and control task Table 1 Critical modes and participation factors
G8 IM8
37 29
IM10 G10
25 26 28
27
30 38
2 18
17 G9
1 24 IM9
3 16 IM6 G6
35
4
G1 15 21 22
5 14
39
6 12 19 23
IM1 7 31 11 20
36
8 10 34 33
G2
9 32 G7 IM7
G4 G5
IM2
IM3 G3 IM5
IM4
The reformulation proposed in this article using the exact value of ðL AÞ, but it is possible to obtain a
Cauchy remainder allows us to represent the nonlinear bound on k ðL AÞ k .
power system models as: The system (13) is of the form which allows for an
application of the minimax control design technique
_
ΔxðtÞ ¼ AΔxðtÞ þ B1 ΔuðtÞ þ B2 ðtÞ; ð9Þ as presented in Ref. [30]. To apply this technique, we
rewrite system (13) in terms of the block diagram shown
yðtÞ ¼ C2 ΔxðtÞ þ D2 ðtÞÞ; ð10Þ in Figure 3.
where
In this figure, we introduce a fictitious signal such that
2 3T
ðL AÞΔx ¼ B2 ðtÞ; ð14Þ
6@f1 @f8 7
L¼4 ;...; 5 ;
@x x¼x1 @x x¼x8 where
u ¼ u 1 u ¼ u 8
Xd Xd0 1 1 Xs Xs0 1 Xs Xs0
B2 ¼ diag 0; 0 ; ; ; 0 ; ; 0 ;
2 3T Tdo 2H Tr Tdom 2Hm Tdom
ð15Þ
6@f1 @f8 7
M ¼4
6 ;...; 7 ;
5
@u x¼x1
@u x¼x8 and
u ¼ u 1 u ¼ u 8
~ C
¼ fðtÞ ~1 Δx: ð16Þ
f ¼ ½f1 ; . . . ; f8 T ;
~1 is chosen such that
Matrix C
where ðxp ; up Þ, p ¼ 1; . . . ; 7, denote points lying in the
2 3
line segment connecting ðx; uÞ and ðx0 ; u0 Þ, and f denotes 0 0 1 0 0 0 0
6 0 0 0 0 1 0 07
the vector function on the right-hand side of the vector ~1 ¼ 6
C 7;
4 0 0 0 0 0 1 05
differential equations.
0 0 0 0 0 0 1
Letting ðx0 ; u0 Þ be the equilibrium point and defining
Δ Δ
Δx ¼ x x0 and Δu ¼ u u0 , it is possible to rewrite eq. ~ C ~1 Δx:
ðL AÞΔx ¼ B2 fðtÞ ð17Þ
(12) as follows
~
The expressions for obtaining fðtÞ can be determined in
Δx_ ¼ x_ x_ 0 ; the same way as in Ref. [28]. The system can now be
ð13Þ
¼ AΔx þ ðL AÞΔx þ ðM B1 ÞΔu þ B1 Δu; written as
@f @f Δx_ ¼ AΔx þ B1 Δu þ B2 ðtÞ: ð18Þ
where A ¼ @x , B1 ¼ @u . Eq. (13) is
x¼x0 x¼x0
u ¼ u0 u ¼ u0 In general, xp ; p ¼ 1; . . . ; 7, are not known beforehand, it
linear with respect to the control vector. Since is difficult to obtain the exact value of ðL AÞ, but it is
xp ; p ¼ 1; . . . ; 7, are not known, it is difficult to obtain possible to obtain a bound on fðtÞ shown in Figure 3.
Next, we introduce a scaling parameter α and write τ is a free parameter and the matrices X1 and Y1 are the
pffiffiffi ~ solutions to the following pair of parameter dependent
C 1 ¼ αC 1 , where α scales the magnitude of the uncertain
output algebraic Riccati equations [31]:
pffiffiffi ~
ζ ¼ α C 1 Δx : ð19Þ ðA B2 DT2 ðD2 DT2 Þ1 C2 ÞY1 þ Y1 ðA B2 DT2 ðD2 DT2 Þ1 C2 ÞT
h i 1
We write fðtÞ ¼ p1ffiffiα fðtÞ~ . The value of α is chosen such Y1 ðC2T ðD2 DT2 Þ1 C2 Rτ ÞY1 þ B2 ðI DT2 ðD2 DT2 Þ1 D2 ÞBT2 ¼ 0;
τ
ð26Þ
that the uncertainty, fðtÞ, shown in Figure 3 satisfies,
ð27Þ
k ðtÞ k2 α k C~1 Δx k2 : ð21Þ 1
þ ðRτ γτ G1 1 T
τ γτ Þ X1 ðB1 Gτ B1 B2 B2 ÞX1 ¼ 0:
T T
τ
and we recover the norm bound constraints [30],
The solutions are required to satisfy the following condi-
k ðtÞ k2 k ζ ðtÞ k2 : ð22Þ tions: Y1 > 0, X1 > 0, the spectral radius of the matrix
Condition (22) will enable us to apply the minimax LQG X1 Y1 is ρðX1 Y1 Þ < τ, Rτ γTτ G1 τ γτ 0, Rτ ¼ R þ τC1 C1 ,
T
control design methodology to obtain a controller for the Gτ ¼ G þ τDT1 D1 , and γτ ¼ τC1T D1 .
underlying nonlinear system. Robustness properties of To obtain the minimax LQG controller, the parameter
the minimax LQG controller ensure that this controller τ > 0 is chosen to minimise Vτ . A line search is carried out
stabilises the nonlinear system (9)–(11) for all instances of to find the value of τ > 0 which attains the minimum value
linearisation errors. of the cost function Vτ . This line search involves solving
Associated with the uncertain system (9)–(11), we the Riccati equations (26) and (27) for different values of τ
consider a cost functional J of the form and finding the value which gives the smallest Vτ .
ðT The minimax LQG optimal controller is given by the
1
J ¼ lim E ðΔxðtÞT RΔxðtÞ þ ΔuðtÞT GΔuðtÞÞdt; equations:
T!1 2T 0
ð23Þ
_^ 1 T 1 T 1
xc ¼ ðA B1 Gτ γτ Þ^xc ðB1 Gτ B1 B2 B2 ÞX1 ^
T
xc
where R 0 and G > 0, R 2 Rnn ; G 2 Rmm and E is the τ
1
expectation operator. 1
þ I Y1 X1
The quadratic cost (23) is particularly suited to the τ
design of an excitation controller for the power system. 1
ðY1 C2T þ B2 DT2 Þ ðD2 DT2 Þ1 y ðC2 þ D2 BT2 X1 Þ^
xc ;
Every generator has an over-excitation limiter which lim- τ
its the field voltage based on time integral of the voltage. ð28Þ
This means that one can apply large voltages as long as
they are for short duration. Thus, the quadratic cost
optimisation is much more suitable in this situation as u ¼ G1
τ ðB1 X1 þ γτ Þ^
T T
xc : ð29Þ
compared to H1 -norm based designs.
The minimax optimal control finds the controller The output matrix for the voltage controller is defined
which minimises J over all admissible uncertainties. The as C2 ¼ ½ 0; 0; 0; 1; 0; 0; 0 T . The theory requires that
cost function J satisfies the following relationship [31]: D2 D02 > 0 as discussed in Ref. [30]. This property is
required by the design procedure, but it does not repre-
sup Jðu Þ inf Vτ ; ð24Þ
kk2 kζ k2 τ sent any physical characteristics of the system, so we
choose the value of D2 as small as possible. Eqs (9)–(11)
where Vτ is given by
provide a new representation of the power system model
1 which contains the linear part and also another part with
Vτ ¼ tr Y1 Rτ þ ðY1 C2T þ B2 DT2 ÞðD2 DT2 Þ1 ðC2 Y1 þ D2 BT2 Þ
2 higher order terms. The new formulation presented in
1
1 this section is used with the minimax LQG control theory
X1 I Y1 X1 ;
τ to design the excitation controllers for the generator as in
ð25Þ Refs [4, 30].
interest. The maximum value of fðtÞ is obtained over this the designed controller is not globally stabilising, we can
region and not globally. If the maximum value of fðtÞ is be confident that it will stabilise the system for most
evaluated over the entire uncertainty region, the calcula- contingencies.
tion burden will be very high, and it will lead to a con- In this article, the PSS as shown in Figure 4 has been
servative controller. The controller is then designed as designed using the standard technique which uses the
follows: change in speed Δω as the feedback variable. PSS was
designed subsequent to the design of voltage controller.
Step 1. From the simulations of the faulted system, The PSS parameters are Tw ¼ 5, KSTAB ¼ 0:43, T1 ¼ 0:25,
obtain the range of the variation of all state variable T2 ¼ 0:02, T3 ¼ 0:0:252, and T4 ¼ 0:039.
and form a polytope Ω with corner points given by
ðx0p xfp Þ and ðxfp þ x0p Þ, p ¼ 1; . . . ; 7, where xfp is the
largest variation of the pth state variable about its equili- 5.1 Controller performance evaluation
brium value x0p . Formally, x 2 Ω if jx x0p j jxfp x0p j.
As mentioned earlier, the voltage controller and PSS have
Step 2. Obtain been designed sequentially. At first, voltage controller is
n o designed, and then the PSS is designed including voltage
α ¼ max
p
α : jjfðtÞjj2 < 1 : controller. The objective of PSS design is to damp elec-
x 2Ω
tromechanical oscillations in power systems. However,
The process to obtain α involves obtaining the max- this should not be done at the expense of reducing the
~
imum value of jjfðtÞjj over the polytope Ω. voltage regulation ability of the excitation system. Using
simulation results given below, we show that the PSS
Step 3. Check if there exists a feasible controller with does not have an adverse effect on the voltage controller.
α ¼ α . A simulation is carried out by applying a symmetrical
three phase to ground short circuit fault on the middle of
Step 4. If we obtain a feasible controller in the above
the line 16–19. The fault is cleared after 150 ms. Figures 5
step, either enlarge the polytope Ω, i.e. increase the and 6 show the terminal voltage and real power output of
operating region of the controller, or if we have arrived generator G4 with both voltage controller plus PSS and
at the largest possible polytope then perform an optimal only the voltage controller due to the three-phase fault on
search over the scalar parameter τ to get the infimum of the middle of the line 16–19. Figure 5 shows that the PSS
cost function [4]. If there is no feasible solution with the improves the voltage response. This is due to the
chosen α ¼ α , reduce the polytope, Ω and go to Step 2. improved damping of the electromechanical modes,
For the given power system model, we are able to obtain which is also visible from Figure 6.
a feasible controller with the value of α ¼ 0:925. The con- Responses of terminal voltage and reactive power
troller is stabilising for all variations of states in the polytope output of generator G2 , when a two-line to ground fault
h i
region Ω formed by corner points δ; ω 0 ; V
; E q
tr ; δm ; s; E
0
m
(2LG) occurs on the middle of the line 5–6 from phase B
h i and phase C to ground, are shown in Figures 7 and 8,
and δ; ω; E0q ; Vtr ; δm ; s; Em
0
with the values: jδ þ δ0 j ¼ 45 , respectively. From Figures 7 and 8, it can be seen that the
Phase
compensation
vs
Δωr sTW 1 + sT1 1 + sT3
KSTAB
1 + sTW 1 + sT2 1 + sT4
1.05 1.2
1 1.1
0.95
1
0.9
Voltage (pu)
Voltage (pu)
0.9
0.85
0.8
0.8
0.7
0.75
0.7 0.6
0.65 0.5
0 2 4 6 8 10 0 2 4 6 8 10
Time (s) Time (s)
Figure 5 Terminal voltage ðG4 Þ for three-phase fault in the middle Figure 7 Terminal voltage ðG2 Þ for the unsymmetrical fault in the
of line 16–19. (Solid line voltage controller þ PSS and dashed line middle of line 5–6. (Solid line voltage controller þ PSS and dashed
voltage controller only) line voltage controller only)
9.5
7
9
6
Reactive power (100 X MVAr)
8.5
5
8
Voltage (pu)
7.5
3
7
2
6.5
1
6
0
5.5
0 2 4 6 8 10 −1
0 2 4 6 8 10
Time (s)
Time (s)
Figure 6 Real power output ðG4 Þ for three-phase fault in the middle Figure 8 Reactive power output ðG2 Þ for the the unsymmetrical
of line 16–19. (Solid line voltage controller þ PSS and dashed line fault in the middle of line 5–6. (Solid line voltage controller þ
voltage controller only) PSS and dashed line voltage controller only)
performance of the voltage controller is not much are, Design A: 0:51 j0:326 and with Design B:
affected by the PSS in stabilising the voltage and 0:0091 j3:04. The closed-loop behaviour is also com-
producing reactive power output of generator. From pared for a 20% increase in load at bus 20. The terminal
these figures, it is clear that the designed controllers voltage and real power at generator G4 is shown in
also stabilise the generators under unsymmetrical faults. Figures 9 and 10. With Design B, the terminal voltage
In the preceding part of the article, the voltage con- dips to ~0.75 pu, and the real power oscillates for well
troller is designed first and then the PSS – Design A. This over 20 s with a peak of 670 MW. The response for Design
order of the controller design is compared with the design A is well damped and both terminal voltage and real
where the PSS is designed first and then the voltage power stay within a narrow and acceptable range. This
controller – Design B. The closed-loop dominant modes result can be explained by the fact that in Design B, the
1.2 1.8
1.15 1.6
1.1
1.4
1.05
1.2
Voltage (pu)
Voltage (pu)
1
0.95
0.9 0.8
0.85 0.6
0.8 0.4
0.75
0.2
0.7
0 2 4 6 8 10 0
Time (s) 0 2 4 6 8 10
Time (s)
Figure 9 Terminal voltage ðG4 Þ for 20% change in load at bus 20.
(Solid line controllers with Design A and dashed line Design B) Figure 11 Terminal voltage ðG4 Þ for the three-phase fault. (Solid
line designed controllers and dashed line AC1A exciter and IEEEST
stabiliser)
7
6.5
6 24
Real power (MW X 100)
5.5 22
5
20
4.5
18
4
Angle (º)
16
3.5
14
3
2.5 12
0 5 10 15 20
Time (s)
10
Figure 10 Real power output ðG4 Þ for 20% change in load at bus 8
20. (Solid line controllers with Design A and dashed line Design B) 0 5 10 15 20
Time (s)
Figure 12 Angle response ðG2 Þ for the three-phase fault. (Solid line
voltage controller changes the required phase lead to be designed controllers and dashed line AC1A exciter and IEEEST
provided by the PSS. stabiliser)
The performance of the designed controllers are also
compared with the performance of the conventional IEEE
AC1A exciter and IEEEST stabiliser. Here, a simulation is 6 Conclusions
performed for a severe symmetrical three-phase fault at
bus 20. The fault is cleared after 0.15 s. Figures 11 and 12 The results demonstrate that robust controller designed
show terminal voltage and angle response of the genera- in this article, with a proper bounding of the uncertainty
tors G4 and G2 , respectively. From Figures 11 and 12, we due to unmodelled nonlinear dynamics, performs well
can see that the proposed controller stabilises voltage over a large region of operation. From this work, it can
within few cycles of fault occurrence and damps out the be concluded that the conservativeness of robust control
power angle oscillations. It is clear that the proposed can be overcome with the proposed method for bounding
controller has a better performance in terms of settling the unmodelled dynamics. The bounding method relies
time, damping, overshoot, and oscillations. on the special dynamic behaviour of power systems
which favours the evolution of the dynamics about cer- feedback which is a major consideration for practical
tain trajectories. The quadratic objective function is a systems. The computational burden in the uncertainty
good choice to damp the poles of the closed-loop system. bounding and controller design is not of concern as
The chosen robust control method, minimax LQG, is able demonstrated by the design of the controller for the ten-
to design high performing controllers with output machine New England system.
References
1. Pariz N, Shanechi HM, Vaahedi E. Explaining and validating power systems via exact linearization approach. Int J Electrical
stressed power systems behavior using modal series. IEEE Power Energy Syst 2012;41:54–62.
Trans Power Syst 2003;18:778–85. 17. Ramos RA, Alberto LF, Bretas NG. Linear matrix inequality
2. Kundur P, Paserba J, Ajjarapu V, Andersson G, Bose A, based controller design with feedback linearisation: applica-
Canizares C, et al. Definition and classification of power sys- tion to power systems. IEE Proc Control Theory Appl
tem stability. IEEE Trans Power Syst 2004;19:1387–401. 2003;150:551–6.
3. Zhou X, Yi J, Song R, Yang X, Li Y, Tang H. An overview of power 18. Chapman JW, Ilic MD, King CA. Stabilizing a multi-machine
transmission systems in china. Energy 2010;35:4302–12. power system via decentralized feedback linearizing excita-
4. Hossain MJ, Pota HR, Ugrinovskii V, Ramos RA. Voltage mode tion control. IEEE Trans Power Syst 1993;8:830–8.
stabilisation in power systems with dynamic loads. Int J 19. Mahmud MA, Hossain MJ, Pota HR. Nonlinear excitation control
Electrical Power Energy Syst 2010;32:911–20. of power systems with dynamic loads via feedback lineariza-
5. Schaefer RC, Kim K. Power system stabilizer performance with tion. In 20th Australasian Universities Power Engineering
summing point type var/power factor controllers. In Conference (AUPEC), 2010:1–6.
Conference Record of the 2006 IEEE IAS Pulp and Paper 20. Seo JC, Kim TH, Park JK, Moon SI. An LQG based PSS design for
Conference, 2006:1–7. controlling the SSR in power systems with series-compensated
6. Leon JA, Taylor CW. Understanding and solving short term lines. IEEE Trans Energy Conversion 1996;11:423–8.
voltage stability problems. In IEEE Power Engineering Society 21. Anderson B, Moore JB. Optimal control: linear quadratic meth-
Summer Meeting, Chicago, IL, 2002:745–52. ods. USA: Prentice-Hall, 1990.
7. Leona AE, Solsonaa JA, Figueroaa JL, Valla MI. Optimization 22. Ray LR. Stability robustness of uncertain LQG/LTR systems.
with constraints for excitation control in synchronous genera- IEEE Trans Automatic Control 1993;38:304–8.
tors. Energy 2011;36:5366–73. 23. Taylor CW. Power system voltage stability. New York: McGraw-
8. Bevrani H, Hiyama T. Stability and voltage regulation enhance- Hill, 1994.
ment using an optimal gain vector. In IEEE Power Engineering 24. Vu KT, Liu CC, Taylor CW, Jimma KM. Voltage instability:
Society General Meeting, Canada, 2006:1–6. mechanism and control strategies [power systems]. Proc IEEE
9. Bevrani H, Hiyama T. Robust coordinated AVR-PSS design 1995;83:1442–55.
using H∞ static output feedback control. IEEJ Trans Power 25. Overbye TJ. Effects of load modelling on analysis of power
Energy 2007;127:70–6. system voltage stability. IEEE Trans Power Syst 1994;16:329–38.
10. Lawt KT, Hill DJ, Godfrey NR. Robust co-ordinated AVR-PSS 26. Li Y, Chiang H-D, Choi B-K, Chen Y-T, Lauby D-H. Load models
design. IEEE Trans Power Syst 1994;9:1218–25. for modeling dynamic behaviors of reactive loads: evaluation
11. Saidy M. A unified approach to voltage regulator and and comparison. Int J Electrical Power Energy Syst
power system stabiliser design based on predictive control in 2008;30:497–503.
analogue form. Int J Electrical Power Energy Syst 27. Hossain MJ, Pota HR, Ugrinovskii V, Ramos RA. A novel
1997;19:103–09. STATCOM control to augment LVRT capability of fixed-speed
12. Cao YJ, Jiang L, Cheng SJ, Chen DD, Malik OP, Hope GS. A induction generators. In 48th IEEE Conference on Decision and
nonlinear variable structure stabilizer for power system stabi- Control, Shanghai, China, 2009:7843–8.
lity. IEEE Trans Energy Conversion 1994;9:489–95. 28. Hossain MJ, Pota HR, Ugrinovskii V, Ramos RA. Simultaneous
13. Heniche A, Bourles H, Houry MP. A desensitized controller for STATCOM and pitch angle controls for improved LVRT capabil-
voltage regulation of power systems. IEEE Trans Power Syst ity of fixed-speed wind turbines. IEEE Trans Sustainable Energy
1995;10:1461–5. 2010;1:142–51.
14. Guo Y, Hill DJ, Wang Y. Global transient stability and voltage 29. Hossain MJ, Pota HR, Ugrinovski V. Short and long-term
regulation for power systems. IEEE Trans Power Syst dynamic voltage instability. In 17th IFAC World Congress,
2001;16:678–88. Seoul, Korea, 2008:9392–7.
15. Yadaiah N, Kumar AG, Bhattacharya JL. Fuzzy based coordi- 30. Ugrinovskii VA, Petersen IR. Minimax LQG control of stochastic
nated controller for power system stability and voltage regu- partially observed uncertain systems. SIAM J Control
lation. Electric Power Syst Res 2004;69:169–77. Optimization 2001;40:1189–226.
16. Mahmud MA, Pota HR, Hossain MJ. Full-order nonlinear obser- 31. Petersen IR, Ugrinovskii VA, Savkin AV. Robust control design
ver-based excitation controller design for interconnected using H∞ methods. London: Springer, 2000.