Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Chennai and Dr.

MGR University Second International Conference on Sustainable Energy and Intelligent System (SEISCON 2011) ,
Dr. M.G.R. University, Maduravoyal, Chennai, Tamil Nadu, India. July. 20-22, 2011.

Design of Robust Power System Stabilizer using Mixed


Sensitivity based HfOutput-Feedback Control in LMI
Framework
K. D. V. Narasimha Rao1, Student Member, IEEE* and Subrata Paul
*Research Scholar, Department of Electrical Engineering, Jadavpur University,Kolkata, India.
Professor, Department of Electrical Engineering, Jadavpur University,Kolkata-700032, India.

Keywords:PSS, Hf Mixed Sensitivity, Linear Matrix Inequalities (LMIs), Pole Placement, Robustness, Stability

Abstract
This paper presents the design of robust power system stabilizer
(PSS) for damping low frequency oscillations. The proposed damping controller design is formulated as mixed sensitivity based H
output-feedback control with pole placement constraints. The synthesis problem can be formulated into a standard linear matrix inequalities (LMIs) framework which can be tractable numerically. A
single generator, 3-bus system is used to demonstrate the effectiveness and robustness of the designed controller at different system
operating conditions. The MATLAB simulation result shows good
damping and robustness.

1 Introduction
Interconnected power systems experience spontaneous system oscillations at very low frequencies (0.2 Hz to 3 Hz). These oscillations either decay gradually, or continue to grow to cause system
separation[3]. In general, generators are equipped with the high
gain automatic voltage regulators (AVR) to control their terminal
voltage and enhance the dynamic stability limits. But these AVRs
add negative damping to the system and aggravate its relative stability. Dynamic stability can be improved by providing additional
damping to such unstable oscillations of power system through
supplementary excitation control signal with the synchronous
generator excitation system[6]. The controller, which generates
this supplementary control signal and contributes signicantly in
mitigating the low frequency oscillations is called power system
stabilizer (PSS)[2,9].
Many conventional techniques have been employed to design and
improve the performance of PSS based on classical control theory
like the eigenvalue assignment, point-wise / regional pole placement, mathematical programming, gradient procedure for optimization and also the modern control theory[5]. The PSS design is
simple in these methods but the application of the PSS thus designed often faces drawbacks as the designed parameters of PSS
are to be tuned to suit the onsite system requirement, which is
very much time consuming and a complex exercise[6]. The conventional method of designing PSS has concentrated on a single
nominal operating point. However, in practical situation, varying
loading causes the synchronous generator dynamic characteristics

to vary and there is a limitation in getting required damping characteristic in the entire operating range of the generator for the PSS
designed through conventional method.
The objective of the robust PSS is to maintain dynamic stability
even if the power system congurations and / or operating conditions change.
Robust control techniques have been applied in power systems
since late 1980s[4]. Many conventional robust control tools such
as H optimization[14] and P synthesis[15] are used for designing robust PSS. But, these tools do not guarantee any damping
and transient response specication. Pure H design synthesis
deals mostly with frequency-domain aspects that only enforce
closed-loop stability and does not allow for direct placement of
the closed-loop poles in more specic regions of the left-half
plane. Since the pole location is related to the time response and
transient behaviour of the feedback system, it is often desirable
to impose additional damping and clustering constraints on the
closed-loop dynamics. This makes multi objective synthesis highly desirable in practice, and LMI theory offers powerful tools to
attack such problems. Therefore, H Controller design based on
the linear matrix inequality (LMI) formulation can produce desired result as it gives robust controller with regional pole placement[7,8,11,13,16]. This paper aims the design of a PSS using
H output feedback control with pole placement based on the
mixed sensitivity formulation. The robust synthesis is formulated
in a LMI framework[8]. An example of a single machine, 3-bus
power system model as shown in Figure 4 is used to illustrate the
method.
This paper is organized as follows: In Section 2 Mixed sensitivity
based H output feedback control with robust pole placement
design is discussed and this robust synthesis is then formulated
into LMI frame work. Section 3 presents the test power system
model and carries out model analysis; the procedure of damping
controller design for the PSS is presented in section 4. Section
5 shows the simulation results, and concluding remarks are provided in section 6.

2 LMI based Hf output feedback


controller design
Power system oscillations are due to the variation of load, action
of voltage regulator during fault, etc. These changes can be considered as disturbance d and output y. Hence, damping controllers
are developed to damp such oscillations and to minimize the impact of the disturbances on output.

403

Chennai and Dr.MGR University Second International Conference on Sustainable Energy and Intelligent System

The goal of this work is to obtain a dynamical output-feedback controller law u = K(s)y, with the following state-space description
x& k = Ak xk + Bk y
u = Ck xk + Dk y

(3)

where xk is state variable of the controller. The plant and controller K(s) are dened above. Hence, closed loop transfer function
Twz(s) from w to z is given by
Figure 1: Typical H Mixed Sensitivity Optimization Framework.

W ( s ) S ( s )

Twz ( s ) = 1

W2 ( s ) K ( s ) S ( s )
= Dcl + Ccl ( sI Acl )

Bcl

(4)

Where
A  B2 Dk C2

Acl :

Figure 2: H Mixed Sensitivity Output Feedback Control Block


Diagram.
The standard H mixed sensitivity formulation for output disturbance rejection and control effort is shown in Figure 1, and its
equivalent standard linear fractional transform is shown in Figure
2. In Figure 1, G(s) is the open loop system model; K(s) is the
controller to be designed W1(s) and W2(s) are weighting functions for shaping characteristics of the open-loop plant. Normally,
in the mixed sensitivity design, bounds (weights) are applied on
S(s)/K(s)S(s). The sensitivity function S(s) denotes the transfer
function from disturbance input d(s) to measured output y(s).
K(s)S(s) is the transfer function from disturbance input d(s) to
control input u(s). The complete details of the mixed sensitivity
control design can be found in12. The design objective of control
design is to minimize the sensitivity functions S(s) and K(s)S(s).
The state space description of the augmented plant P(s) is
x& = Ax + B1w + B2 u
z = C1 x + D11w + D12u
y = C2 x + D21w + D22 u

(1)

where x is the plant state vector, u represents the control input, w


is a state vector of exogenous input which include disturbances d,
measured noise n (not considered here), reference signals r, etc., y
is the measured output, and z is a vector of output signals.
The system (1) is assumed to be stabilizable and detectable. This
assumption ensures the existence of an internally stabilizing output feedback control law u = K(s)y. Without loss of generality, it
is assumed that D22 = 0. This assumption makes the closed-loop
state space matrices linear in the control matrices.
Let Twz(s) denotes the closed-loop transfer function from w to z
for a dynamical output-feedback law u = K(s)y.

Twz ( s ) = W1 ( s ) S ( s )

W2 ( s ) K ( s ) S ( s )

Bcl :
Ccl :

B1 B2Dk D21

Bk D21

C1  D12 Dk C2 D12Ck

Dcl : D11  D12 Dk D21

(5)

The constraint Twz ( s ) < g can be interpreted as a disturbance

rejection problem. This constraint is also useful to enforce robust


stability. In the LMI formulation8, the same objective is achieved
T > 0 such
in the suboptimal sense if there exists an X = X
that
T
Acl X + X Acl

T
Bcl

Ccl X

Bcl
g I
Dcl

T
X Ccl
T
Dcl
g I

<0

(6)

The transient response of a linear system is related to the location


of its poles. A conic sector with inner angle T = 90 and apex at the
origin is chosen as the region D for imposing the pole-placement
constraints. The transient response specications can be satised
by clustering the closed loop poles in the admissible region shown
in Figure 3. It ensures a minimum decay rate DR and a minimum
damping ratio ]min = cos(T/2). The real part of the poles should
be larger than to avoid very large controller gains. Ref.[11]
gives the details of the pole placement design. In order to achieve
good damping, it is required that the damping factor be greater
than 10% and the real part of all the poles of the system to be less
than -0.5.
The closed-loop system matrix has all its poles inside the conical
sector , if and only if there exits XD = XTD > 0, such that the following inequality is satised.

(2)

The sensitivity function S(s) = [I+G(s)K(s)]-1 is that ensures disturbance attenuation and good tracking performance. K(s)S(s) =
K(s)[I+G(s)K(s)]-1 handles the issues of robustness and constrains
the effort of the controller.

Bk C2

B2Ck

Ak

T
T
sinT ( Acl X D  X D Acl ) cosT ( Acl X D  X D Acl )
cosT ( X AT  A X ) sinT ( X AT  A X )C  0
D cl cl D
D cl
cl D

(7)

The design specications are feasible if and only if (6) and (7)
hold for some positive semi denite matrices X and XD and
some controller

404

Design of Robust Power System Stabilizer using Mixed Sensitivity based HfOutput-Feedback Control in LMI Framework

is equal to that of the generalized plant) and invertible.


Pre- and post multiplying each of the inequalities and inequality
(7) by 3T2 and 3T2 , and also pre- and post multiplying the inequality (6) by diag(3T2 ) and diag(3L, L) followed by a change of
variables according to equations (10)-(13), the following LMIs
are obtained10.
R

I
! 0
S

\11

\21

(15)

\12
< 0
\22

sinT() + ) T )

T
cosT() - ))

(16)

cosT() - ) T )
< 0
sinT() T + ))

(17)

Where
AR + B C
2

Figure 3: LMI Region

) :=

This problem is not jointly convex in Xf and XD unless it is


solved for the same matrix X. Therefore, we restrict our attention
to the sub-optimal formulation of H synthesis with pole placement which is dened as: Find X > 0 and a controller K(s) = :K
that satisfy (6) and (7) with X= Xf= XD.
Ak
:K

Ck

Bk

Dk

R
T
M

S
X 1 T
U
N

R I
I S
, 32 =
and satises the identity X 3 2
M
0
T
0 NT

31.

NAk M T  NBk C2  SB2Ck M T  S ( A  B2 Dk C2 ) R


B

NBk SB2 Dk
T

- JI

A + (A + B DC )T
2
2

\ 22 :=

SB1 + BD21

D11 + D 12 DD 21

C1R + D12 C
T

A T S + SA + B C 2 + C T2 B (C1 + D 12 DD 21 ) T
C1 + D 12 D C 2 - JI

\12 = \T21

(19)

(20)

(21)

(22)

The LMIs in (15),(16) and (17) are solved for A, B , C and as an


optimization problem. Once A, B , C and D are obtained, the actual
controller variables and can be recovered from and by solving
the (9)(12).

3 Power System Model

The new controller variables are dened as


A

(B1 + B2 DD21 )T

(9)

where R and S are the n x n symmetric matrices.


For 31 =

B1 + B2 DD21

\21 :=

The inequalities in (6) and (7) contain AclX and CclX i.e., products of X and the controller parameters. This problem is therefore
nonlinear in X. A change of controller parameters is necessary to
convert the problem into a linear one.
Let and be partitioned as:

AR + RA T + B2 C + C BT2

(8)

(18)

SA + BC2

\11 :=

A + B2 DD21

(10)
(11)

Ck M  Dk C2 R

(12)

Dk

(13)

The identity XX1 = I together with (8) gives

In this paper, a single generator, 3-bus system as shown in Figure


4 is used as study system. The generator is having static excitation system. The dynamics of the synchronous machine and excitation system can be represented by the following differential
equations.
The following third order model for the generator with static exciter system is assumed.

MNT = I RS
(14)
If M and N have full row rank, then given A, B, C , D, R, S , M and
N, we can always compute controller matrices Ak, Bk, Ck and Dk
that satisfy equations (10)-(13). Moreover Ak, Bk, Ck, and Dk are
uniquely determined if M and N are square (i.e., controller order

405

G&
&
Z
E& cq

Zs
2H

Z  Zs

[Tm  Ecq I q  X q  X cd I d I q  D ( Z  Z S )]

1
c
Tdo

[Ecq  X d  X cd I d  E fd ]

(23)
(24)
(25)

Chennai and Dr.MGR University Second International Conference on Sustainable Energy and Intelligent System

(2)

The standard practice in H mixed sensitivity design is to adopt


the weight W1(s) as a low pass lter for output disturbance rejection and W2(s) as a high pass lter to reduce the control effort
in the high frequency range. The weights adopted in the present
work are given by

Figure 4: One line diagram of a single machine, 3-bus power


system.

W1 ( s ) = 0.346

The power system thus in general can be described by a set of


differential algebraic equations of the following form:
x& = f ( x , z , u )

(26)

0 = g (x, z)

(27)

Where x is the vector of state variable, z is the vector of network (algebraic) variables and u is the vector of control inputs.
The dynamic stability is due to the small disturbances causing
small deviations from a stationary operating point. Hence, the set
of differential and algebraic equations of the system can be linearized around this operating point for the stability analysis. The
above set of equations can be written in state space form after linearization and elimination of the algebraic variables as follows1.
x& = Ax + B u
(28)
y = C x

(29)
where y is the output vector. A, B and C are the constant matrices with appropriate dimensions.
For the purpose of PSS design, the generator model considered
is of 3rd order and the static excitation system is of 1st order.
Thus the linearized state vector is given by

x = d w

E q

E fd

Formulate the generalized plant using the simplied system


model and the weighting functions.

W2 ( s ) = 0.55

(3)
(4)

(5)
(6)

(7)

(8)
(9)

( s +10 )
( s + 0.00005)

( s + 0.003)
( s +10 )

Dene the size and structure of the LMI variables


A, B , C , D, R, S , M and using lmivar function.
Set up the LMIs in (15),(16) and (17) in terms of the LMI
variables and each of the LMI terms and their respective
positions are specied using the lmiterm function.
Combine the individual LMIs into a set of LMIs using the
getlmi function.
Solve the J optimization problem with the set of LMI constraints (in the present work, mincx function of MATLAB
LMI Control Toolbox[8] has been used for the optimization).
Retrieve the optimum value of solution variables
A, B , C , D, R, S , M and from the output of the mincx function
by using the dec2mat function.
Determine the controller using the optimum value of the solution variables.
Simplify the designed controller.

(30)

Since the PSS has been designed taking speed deviation as input,
the output vector is given by
(41)
y = w

[ ]

The mechanical power is considered to be constant. A disturbance


w is added to the mathematical system model, which in the
present work has been considered as a false disturbance in the
form of a change in mechanical power input for small duration
of time. The effect of such disturbance on the power system frequency is studied.

5 Simulation Results
To investigate the effectiveness of the proposed robust PSS, simulations have been carried out with the power system shown in
Figure 4. The simulation results are shown in Figure 5 to Figure
9. The different operating conditions (nominal / off-nominal loading) and disturbances considered are mentioned with respective
gures. The details of the loading conditions and system parameters are given in Appendix.

4 Design of PSS
The system performance and robustness of controlled system is
determined by the proper selection of weighting functions W1(s)
and W2(s) in Figure 1. Since the direct objective of the robust PSS
is improving the damping of the whole system despite the variation of system operating condition, pole placement constraint can
be considered for satisfactory time response, and closed-loop
damping can often be achieved by forcing the closed-loop poles
into a suitable sub region of the left-half plane.
The procedural steps for the robust PSS design are summarized
as follows:
(1) Simplify the system model.

Figure 5: Oscillations at nominal operating condition of the system for a 10% mechanical input disturbance for 0.1sec duration

406

Design of Robust Power System Stabilizer using Mixed Sensitivity based HfOutput-Feedback Control in LMI Framework

Figure 6: Oscillations at off-nominal operating condition of the


system for a 10% mechanical input disturbance for 0.1sec duration.

Figure 9: Comparative results for disturbances of different magnitudes and durations at off-nominal operating condition

6 Conclusion
In this paper, a new robust design of PSS by applying mixed
sensitivity based H output feedback control technique in LMI
framework is proposed to mitigate low frequency oscillations.
The motivation to apply this control strategy is the simplicity and
exibility of synthesis procedure. The controller designed uses
only speed deviation of the generator as the feedback signal.
It may be concluded that the robust PSS can improve the damping
of low frequency oscillations in power system over a wide range
of operating conditions without changing the parameters of the
PSS.
Figure 7: Comparative Results for nominal and off-nominal condition for 10% mechanical input disturbance for 0.1 sec duration.

Appendix
The nominal parameters and the operating conditions of the system are presented below:
Nominal / Light Load P = 0.90 Q = 0.436
Off Nominal /Heavy Load P = 1.03 Q = 0.346
Xd = 0.300p.u Xq = 1.760p.u Xd = 1.81p.u
Xq = 1.76p.u Xl = 0.16p.u Ra = 0.003p.u
Td0 = 8.0 KD = 0 Z = 377rad/s
KA = 5 pu TA = 0.02 p.u KE = 0.3 p.u
TE = 0 H = 4.0 f = 60 Hz

References
Figure 8: Comparative results for disturbances of different magnitudes and durations at nominal operating condition.
The simulation results indicate that the low frequency oscillations
in the system are damped out quite satisfactorily with the introduction of PSS.
The robustness of the proposed PSS under varying operating conditions can also be noticed in the results. If the performance of the
system with nominal load is used as a benchmark against which
the performances of the system under off- nominal operating conditions is evaluated, then it is seen that the system response has
not been degraded too far.

[1]

[2]

[3]

[4]

407

A. Sil, T. K. Gangopadhyay, S. Paul, A. K. Maitra. Design


of robust power system stabilizer by using mixed sensitivity
technique, IEEE International Conference, 2009 pp. 1-4.
E. Larsen, D. Swan. Applying power system stabilizers,
parts I, II and III, IEEE Trans. Power Apparatus and Systems, volume. PAS-100, 1981 pp. 30173046.
F. P. Demello, C. Concordia. Concepts of synchronous
machine stability as affected by excitation control, IEEE
Trans on PAS, volume. 88, No. 5, 1969 pp 316-329.
J. F. Hauer. Robust damping controls for large power system, IEEE Cont, Sys. Magazine, volume 9, No. 1, 1989 pp.
12-18.

Chennai and Dr.MGR University Second International Conference on Sustainable Energy and Intelligent System

M. A. Abido. Pole placement technique for PSS and


TCSC-based stabilizer design using simulated annealing,
Eletrical Power and Energy Systems., volume 22, 2000 pp.
543-554.
[6] M. Klein, G.J. Rogers, P. Kundur. A fundamental study of
inter-area oscillations in power systems, IEE Trans.Power
Syst., volume 6, 1991 pp. 914-921.
[7] M. Soliman, H. Emara, A. Elshafei, A. Bahgat, O. P. Malik.
Robust output feedback power system stabilizer design:
an LMI approach, IEEE PES Meeting Conversion and
Delivery of Electrical Energy in the 21st Century., 2008
pp. 1-8.
[8] M. Chilali , P. Gahint. Desin with pole placement constraints: an LMI approach, IEEE Trans. Automat. Contr.,
volume 41, 1996 pp. 358-367.
[9] P. Kundur. Power system System Stability and Control,
McGraw-Hill, New York 1994.
[10] P. Gahint, A. Nemirovski, A. Laub, M. Chilali. The LMI
Control Toolbox for Use WITH Matlabl, The Math Works
Inc.1995.
[11] P. S. Rao, I. Sen, C. D. Author. Robust pole placement stabilizer design using linear matrix inequalities, IEEE Trans.
Power Syst., volume 15, 2000 pp. 313-319.
[12] S. Skagested, L. Postlethwite. Multivariable Feedback
control and Design, John Wiley and Sons 1996.
[13] S. Boyd, L. EI Gahoui, E. Feron, V. Balakrisnan. Linear
Matrix inequalities in Systems and Control Theory, SIAM,
Philadelphia 1994.
[5]

[14] S. Chen, O. P. Malik. Optimization-based power system


stabilizer design, IEEE Proc. Gener. Transm. Distrib., volume 142, 1995 pp. 179-184.
[15] S. Chen, O. P. Malik. Power System Stabilizer design using
Synthesis, IEEE Trans. Energy Convers., volume 10, 1995
pp. 175-181.
[16] S. Chen, P. Gahinet, M. Chilali. Multiobjective output feedback control via LMI optimization, IEEE Trans. Automat.
Contr., volume 42, 1997 pp. 896-911.

Biographies
K.D.V.Narasimha Rao has obtained his
M.Tech. Degree in Electrical and Electronics
Engineering at Jawaharlal Nehru Technological University, Kakinada, Andhra Pradesh, India in 2004. Currently, he is research scholar
at Jadavpur University, Kolkata (West Bengal), India. His current research interest include power system dynamics and control, robust controls, and
FACTs applications.

408

Subrata Paul is Professor with the Department of Electrical Engineering at Jadavpur


University, kolkata (West Bengal), India. His
work has been primarily concerned with
power system analysis, transient stability and
FACTS applications

You might also like