Automatica 46 (2010) 330–336

Contents lists available at ScienceDirect

Automatica
journal homepage: www.elsevier.com/locate/automatica

Brief paper

A power system nonlinear adaptive decentralized controller designI

Rui Yan a , ZhaoYang Dong b,∗ , T.K. Saha c , Rajat Majumder d
a
The Institute for Infocomm Research, Agency for Science, Technology and Research, Singapore 138632, Singapore
b
Department of Electrical Engineering, The Hong Kong Polytechnic University, Kowloon, Hong Kong
c
School of Information Technology and Electrical Engineering, The University of Queensland, St. Lucia, QLD, 4072, Australia
d
ABB Corporate Research, Sweden

article info abstract

Article history: In this paper, a novel excitation control is designed for improvement of transient stability of power
Received 3 April 2007 systems. The control algorithm is based on the adaptive backstepping method in a recursive way without
Received in revised form linearizing the system model. Lyapunov function method is applied in designing the controller to ensure
22 July 2009
the convergence of the power angle, relative speed of the generator and the active electrical power
Accepted 12 October 2009
Available online 30 November 2009
delivered by the generator when a large fault occurs. Compared with the existing nonlinear decentralized
control approaches, the proposed controller has no requirement for the bounds of interconnections in
Keywords:
the power system. And the new approach does not need the existence of solution of a designed algebraic
Adaptive control Riccati equation. Furthermore, the transient stability performance of power systems can also be improved
Nonlinear decentralized control by the designed control approach. The efficacy of the designed controller has been demonstrated in a
Power systems multimachine power system. Simulation results show transient stability enhancement of a power system
Transient stability in the face of a large sudden fault.
Backstepping design © 2009 Elsevier Ltd. All rights reserved.

1. Introduction The transient stability in power systems studies involve the de-
termination of whether or not synchronism is maintained after
Power systems are modeled as large nonlinear systems. Simpli- the machine has been subjected to severe disturbance. In order
fied linear models (deMello & Concordia, 1969) have been used for to improve the transient stability of a power system with large
a long time to design the excitation controller for synchronous gen-
disturbances, the crucial control objectives are to rapidly increase
erators. Linear control methodologies were well accepted in utili-
excitation and to decrease the mechanical input power at the same
ties due to their inherent simplicity in design and ease of real time
time (Wang, Hill, Middleton, & Gao, 1993). In the past few years, the
implementation. Linear power system stabilizers (PSS) (Larsen &
Direct Feedback Linearization (DFL) technique has been widely re-
Swann, 1981) are often used to provide supplementary damping
ported to design an excitation controller (Chapman & Ilic, 1993;
through excitation control to improve the dynamic stability limit.
Gao, Chen, Fan, & Ma, 1992; Guo, Wang, & Hill, 2000; Wang &
Such linear control mechanisms generally provide asymptotic sta-
Hill, 1996; Wang et al., 1993; Wang, Guo, & Hill, 1997). Further-
bility in a small region of the equilibrium and is only appropriate
more, in recent years a number of different advanced nonlinear
for the effect of small disturbances. In Fan, Ortmeyer, and Mukun-
control techniques were used to enhance the transient stability
dan (1990), an adaptive control method was designed for the im-
of power systems in Cong, Wang, and Hill (2004), Cong, Wang, and
provement of transient stability of multimachine power systems.
Hill (2005) and Wang, Zhang, and Hill (2004).
An approach was proposed in Dong, Hill, and Guo (2005) for power
Generally, decentralized controllers are used widely for mul-
system security assessment and enhancement based on the infor-
timachine systems due to physical limitations on the system
mation provided from the pre-defined system parameter space.
structure. Sometimes it might become infeasible to transfer infor-
The proposed scheme opens up an efficient way for real time secu-
mation among subsystems. So far there have been numerous con-
rity assessment and enhancement in a competitive electricity mar-
trol propositions for designing decentralized excitation control of
ket for single contingency cases.
power systems (Chapman, Ilic, King, Eng, & Kaufman, 1993; Jain,
Khorrami, & Fardanesh, 1994; King, Chapman, & Ilic, 1994; Lu, Sun,
I This paper was not presented at any IFAC meeting. This paper was
Xu, & Mochizuli, 1996; Wang et al., 1997). DFL was applied to trans-
fer a nonlinear multimachine power system model to a linear one
recommended for publication in revised form by Associate Editor Xiaohong Guan
under the direction of Editor Toshiharu Sugie.
and robust decentralized control was designed. The interconnec-
∗ Corresponding author. Tel.: +852 27666148; fax: +852 23301544. tions were considered only with linear bounds in the above works.
E-mail addresses: [email protected] (R. Yan), [email protected] This may result in more conservative controllers for global behav-
(Z. Dong). ior since the essential interconnections in the large-scale power
0005-1098/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.automatica.2009.10.020
 R. Yan et al. / Automatica 46 (2010) 330–336 331

systems are nonlinear. The nonlinear decentralized control scheme Electrical equations:
was developed to solve the problem of general nonlinear bounds
E (t ) = E 0 (t ) − (x − x0 )I (t ),
of interconnections in Guo, Hill, and Wang (2000). Both excita- qi qi di di di
Efi (t ) = kci ufi (t ),

tion control and steam valve control were designed to enhance the


 n
transient stability. Previous works are based on the known bounds
 X
( ) 0
(t )Eqj0 (t )Bij sin(δi − δj ),

P t = Eqi

ei

of parameters. However, some of them are very difficult to know 

j= 1

in practice especially when serious disturbances occurs. Recently,



 Xn
a constructive methodology utilizing the backstepping technique Qei (t ) = − 0
(t )Eqj0 (t )Bij cos(δi − δj ),

Eqi



has been developed for designing nonlinear adaptive control sys- 
 j =1
tems (Kanellakopoulos, Kokotović, & Morse, 1991; Krstić, Kanel- n
X (3)
lakopoulos, & Kokotović, 1995). This method can overcome the


 Idi (t ) = Eqj0
(t )Bij cos(δi − δj ),

relative degree one restriction and can be applied to handle power 
 j =1
n

systems with uncertainties.

 X
( ) 0
(t )Bij sin(δi − δj ),

In this paper, a novel nonlinear decentralized excitation con-


 I qi t = Eqj

 j=1
troller has been designed by applying adaptive backstepping 
Eqi (t ) q = xadi Ifi (t ),


method to improve the transient stability performance of multi-


Vti = (Eqi 0
+ x0di Idi )2 + (x0di Iqi )2 ,

machine power systems. Compared with the previous decentral-
ized excitation control in Chapman and Ilic (1993), Guo and Hill
et al. (2000), Jain et al. (1994), King et al. (1994), Lu et al. (1996) and where Qei (t ) is the reactive power, in p.u.; Ifi (t ) is the excitation
Wang et al. (1997), the requirement of the known bounds of inter- current, in p.u.; Iqi (t ) is the quadrature axis current, in p.u.; kci is
connection parameters has been relaxed in this method. Further- the gain of the excitation amplifier, in p.u.; ufi (t ) is the input of
more, the new approach does not need the existence of solution the Silicon Controlled Rectifier (SCR) amplifier of the generator, in
of a designed algebraic Riccati equation. Some comparison results p.u.; Bij is the ith row and jth column element of nodal susceptance
between the proposed approach and existing methods in Guo and matrix at the internal nodes after eliminating all physical buses, in
Hill et al. (2000) and Wang et al. (1997) will be given in Section 4. p.u.; xadi is the mutual reactance between the excitation coil and
This paper is organized as follows. In Section 2, the dynamic model the stator coil of the ith generator, in p.u.; xdi is the direct axis
of a power system is described. In Section 3, the adaptive backstep- reactance of the ith generator, in p.u.; x0di is the direct axis transient
ping design is proposed. Illustrative examples and design consider- reactance of the ith generator, in p.u.; Vti is the terminal voltage of
ations of applying the proposed controller to multimachine power the ith generator.
systems are provided in Section 4. The conclusion is drawn in It is obvious that the multimachine power system is highly
Section 5. nonlinear and interconnected by the transmission network.

2. Power system dynamical model 3. Decentralized nonlinear controller design

In this section, a power system consisting of n synchronous ma- Let δmi0 , ωmi0 and Pmi0 be the desired values for the power
chines is considered. Based on some standard assumptions, the angle δi , the relative speed ωi and the active power Pei of the ith
motion of the interconnected generators can be described by a clas- generator at the operating point. Denote δ̃i (t ) = δi (t ) − δmi0 ,
sical model with flux decay dynamics (Anderson & Fouad, 1994;
ωi (t ) = ωi (t ) − ωmi0 = ωi (t ) for ωmi0=0 and P̃ei (t ) = Pei (t ) − Pmi0 .
Bergen, 1986; Kundur, 1994; Pai, 1981). In the model, the generator
Based on the calculation in Wang et al. (1997), the differentiation
is modeled as the voltage behind direct axis transient reactance;
of Pei (t ) is
the angle of the voltage coincides with the mechanical angle rel-
ative to the synchronously rotating reference frame. The network ˙ 1
P̃ ei (t ) = [kci ufi (t ) + (xdi − x0di )Idi (t )Iqi (t )
has been reduced to internal bus representation. 0
Td0i
The dynamical model of the ith machine with excitation control
can be written as follows: −(P̃ei (t ) + Pmi0 )] − Qei (t )ωi (t )
n
Mechanical equations: X
 + 0
Eqi (t )Ėqj0 (t )Bij sin(δ̃i − δ̃j + δmi0 − δmj0 )
δ̇i (t ) = ωi (t ) j =1

Di ω0 (1) n
ω̇i (t ) = − (Pmi0 − Pei (t )),
X
2Hi
+
2Hi
− Eqi0 (t ) 0
Eqj (t )Bij cos(δ̃i − δ̃j + δmi0 − δmj0 )ωj (t ).
j =1
where δi (t ) is the power angle of the generator, in radian; ωi (t ) is
Denote
the relative speed of the generator, in rad/s; Pmi0 is the mechanical
input power, in p.u., which is a constant; Pei (t ) is the active γ1ij (t ) = Eqi0 (t )Ėqj0 (t )Bij , γ2ij (t ) = −Eqi0 (t )Eqj0 (t )Bij , δ̄ij = δ̃i − δ̃j +
electrical power delivered by the generator, in p.u.; ω0 = 2π f0 δmi0 − δmj0 . We further have
is the synchronous machine speed, in rad/s; Di is the per unit 1
˙
P̃ ei = kci Iqi ufi − Qei (t )ωi (t )
damping constant and Hi is the inertia constant in seconds. 0
Td0i
Generator electrical dynamics:
1
1 + 0 [−P̃ei − Pmi0 + (xdi − x0di )Iqi Idi ]
Ėqi (t ) =
0
[Efi (t ) − Eqi (t )] (2) Td0i
0
Td0i n
X n
X
where Eqi (t ) is the transient EMF in the quadrature axis of the ith
0 + γ1ij (t ) sin(δ̄ij ) + γ2ij (t ) cos(δ̄ij )ωj .
j=1 j =1
generator; Eqi (t ) is the EMF in the quadrature axis, in p.u.; Efi (t )
0
is the equivalent EMF in the excitation coil, in p.u. and Td0i is the Therefore, the multimachine power system dynamic model (1)–(3)
direct axis transient short circuit time constant, in seconds. can be compensated into:
332 R. Yan et al. / Automatica 46 (2010) 330–336
˙
 δ̃ i (t ) = ωi (t ), Lyapunov function method is applied to each subsystem to find a
ω0

 Di proper fictitious control as if it could be controlled independently.
ω̇i (t ) = − ωi (t ) −

P̃ei In order to deal with the time varying uncertainties γ i (t ) in the


2Hi 2Hi


third equation of (4), we denote S (x) = ρ1 arctan(ρ2 x), for any

P̃˙ ei = 1 kci Iqi ufi − Qei (t )ωi (t )


variable x, where ρ1 > 0 and ρ2 > 0 are positive constants to be

0
Td0i
1 chosen by the designer. Note that if we choose the gains ρ1 and ρ2
+ 0 [−P̃ei − Pmi0 + (xdi − x0di )Iqi Idi ] such that ρ1 tan ρ1 ≤ δ, then






 Td0i 2 1
n n
|x| ≥ δ
 
|x|
 X X
γ ( ) ( δ̄ ) γ2ij (t ) cos(δ̄ij )ωj .

+ t sin + xS (x) = xρ1 arctan(ρ2 x) ≥

1ij ij (5)
x2 /δ |x| < δ.


j =1 j =1

It is easy to verify that S (x) is continuous and differentiable. We

Remark 1. From Remark 3.4. in Wang et al. (1997), we have have the following property in Xu and Yan (2006).
|γ1ij (t )| ≤ 4|T|P0 ei ||max , |γ2ij (t )| ≤ |Pei (t )|max . It is clear that the Property 1. |x| − S (x)x ≤ δ.
d0j min
bounds of γ1ij (t ) and γ2ij (t ) depend on generator parameters Td0j 0
Now design the controller based on the adaptive backstepping
and |Pei (t )|max . A robust decentralized nonlinear controller is de- method in the following. The procedure is divided into three steps.
signed based on the known Td0j 0
and |Pei (t )|max in Guo and Hill et al.
Step 1. Design the fictitious control u1i according to the Lyapunov
(2000) and Wang et al. (1997). However, the above robust non-
function V1i .
linear controller design involves estimating nonlinearity bounds
Denote new coordinates z1i = x1i and z2i = x2i − u1i , where the
within a certain operating region. When serious disturbances oc-
fictitious control u1i = −q1i z1i with q1i > 0. Define a Lyapunov
cur which cause the system to operate in a wider range outside the
function V1i = 12 z1i 2
. The derivative of V1i is V̇1i = z1i ż1i = ż1i (z2i +
estimated one, the designed system may not perform well. There-
fore, in this work, we will relax the assumption that the bounds of u1i ) = z1i z2i − q1i z1i
2
.
γ1ij (t ) and γ2ij (t ) are known and design the controller by the adap- Step 2. Design the fictitious control u2i according to the Lyapunov
tive robust control. function V2i .
Define z3i = x3i − u2i , where the fictitious control u2i is
Remark 2. In the power systems, Pei (t ), Qei (t ) and Ifi (t ) are readily
q2i 1 a1i + q1i
measurable variables. From (3), we have Pei (t ) = Eqi 0
(t )Iqi (t ), and u2i = − z2i − z1i − x2i . (6)
Qei (t ) = −Eqi
0
(t )Idi (t ). Therefore, Idi (t ) and Iqi (t ) can be calculated a2i a2i a2i
using available variables. Furthermore, ωi (t ) can be measurable Define a Lyapunov function V2i = V1i + 12 z2i
2
. The derivative of V2i
and the power angle δi (t ) can be found in Mello (1994). should be
A practically realizable controller will be designed using only lo- 2
V̇2i = −q1i z1i − q2i z2i2 + a2i z2i z3i . (7)
cal measurements, i.e., a decentralized controller without remote
Step 3. Design the actual control ui according to the Lyapunov func-
signal transmissions. The control objective is to design a decentral-
tion Vi .
ized nonlinear feedback control law ufi , i = 1, 2, . . . , n for an ex-
Denote
citation control loop such that the systems are transiently stable
a21i + q1i a1i + 1
 
when a major fault occurs. gi = + Qi (t ) x2i + (a1i + q1i )x3i
a2i
Let xi (t ) = [x1i , x2i , x3i ]T = [δ̃i , ωi , P̃ei ]T , q2i q1i q2i
Di ω0 1 − z1i − z2i + q2i z3i
a1i = − , a2i = − , θi = 0 , a2i a2i
2Hi 2Hi Td0i θ i = [b−
i θi i ] ,
b− φ i = b−
i γ i, ξ Ti = [ξi gi ].
1 1 T 1

ξi = −x3i − Pmi0 + (xdi − xdi )Iqi Idi ,

0
Let βi = [β1i , . . . , β2ni ]T , where βji is the bound of φji (t ). Design
γ i = [γ1i1 , γ1i2 , . . . , γ1in , γ2i1 , γ2i2 , . . . , γ2in ]T , the controller ui as follows:
ηi = sin(δ̄i1 ), sin(δ̄i2 ), . . . , sin(δ̄in ),

T T T
T ui = −a2i z2i − q3i z3i − θ̂ i ξ i − S (β̂i ηi z3i )β̂i ηi , (8)
cos(δ̄i1 )x21 , cos(δ̄i2 )x22 , . . . , cos(δ̄in )x2n , where the parameter updating law is
kci
bi = 0 , ui (t ) = Iqi ufi . ˙
θ̂ i = ξ i z3i ,
˙
β̂i = |ηi z3i | − αi β̂i , (9)
Td0i
In these new defined parameters, a1i and a2i are known constants; with a damping coefficient αi > 0.
θi and bi are unknown constants; γ1ij and γ2ij , (j = 1, . . . , n) are From (4) and (6), we have
unknown time varying uncertainties and ξi , the elements of ηi are ż3i = ẋ3i − u̇2i
known nonlinear functions. = Qx2i + θi ξi + γ Ti ηi + bi ui
Based on the new denotations, (4) can be rewritten as follows:
q2i 1 a1i + q1i
 − z2i − z1i − x2i
ẋ1i = x2i a2i a2i a2i
ẋ2i = a2i x3i + a1i x2i (4)
ẋ = Q x + θ ξ + γ T η + b u . = θi ξi + γ Ti ηi + bi ui + gi
3i ei 2i i i i i i i
= bi (θ Ti ξ i + φTi ηi + ui ). (10)
The Direct Feedback Linearization (DFL) method always needs the
priori knowledge of the dynamic system. If there are some un- Substituting the controller ui in (8) into (10), we obtain
known parameters or some uncertainties in the dynamic system,
h
the backstepping method will be more suitable. In the following, ż3i = bi −a2i z2i − q3i z3i + (θ i − θ̂ i )T ξ i
an adaptive controller can be designed analytically using the Lya- T T
i
punov function method. The main feature of the design is that the + φTi ηi − S (β̂i ηi z3i )β̂i ηi . (11)
 R. Yan et al. / Automatica 46 (2010) 330–336 333

Now define the following Lyapunov function: Remark 5. From the equation in (3), Vti is a nonlinear function of
1 1 δi , Pei and the system structure. Therefore, any change in the system
Vi = V2i + 2
z3i + (θ i − θ̂ i )T (θ i − θ̂ i ) structure will cause the voltage to reach a different post-fault state
2bi 2
which is undesirable in practice although δi and Pei can return their
1 prefault steady values by using the above proposed controller.
+ (βi − β̂i )T (βi − β̂i ). (12)
2 Voltage regulation is also an important issue particularly in the
Then the derivative of Vi is post-transient period. Its basic objective is to regulate the voltage
to reach its nominal value. Differentiating equation of Vti in (3)
1 ˙ ˙ gives
V̇i = V̇2i + z3i ż3i − (θ i − θ̂ i )T θ̂ i − (βi − β̂i )T β̂i . (13)
bi ˙
Ṽ ti = θi ξi (t ) + γ Ti (t )ηi (t ) + bi fi (t )ufi
Substituting (7), (11) and the parametric updating law in (9) into
(13), we have 1 (Eqi0 + x0di Idi )Eqi
θi = 0 , ξi = q ,
˙ Td0i (Eqi0 + x0di Idi )2 + (x0di Iqi )2
2
V̇i = −q1i z1i − q2i z2i2 − q3i z3i2 − (θ i − θ̂ i )T (θ̂ i − ξ i z3i )
T T ˙ 0
Eqi + x0di Idi
+ φTi ηi z3i − S (β̂i ηi z3i )β̂i ηi z3i − (βi − β̂i )T β̂i bi =
kci
, fi (t ) = − q ,
0
T Td0i (Eqi0 + x0di Idi )2 + (x0di Iqi )2
= −q1i z1i2 − q2i z2i2 − q3i z3i2 + βTi |ηi z3i | − β̂i |ηi z3i |
T T T ˙
+ β̂i |ηi z3i | − S (β̂i ηi z3i )β̂i ηi z3i − (βi − β̂i )T β̂i γ i = [γ1i1 , . . . , γ1in , γ2i1 , . . . , γ2in ]T ,
≤ −q1i z1i2 − q2i z2i2 − q3i z3i2 + δ − αi (βi − β̂i )T β̂i ηi = [η1i1 , . . . , η1in , η2i1 , . . . , η2in ]T ,
αi γ1ij = Ėqj0 Bij , γ2ij = Eqj0 Bij ,
≤ −q1i z1i2 − q2i z2i2 − q3i z3i2 − k(βi − β̂i )k2
2 η1ij = sin δij g2 + cos δij g1 ,
αi η2ij = (cos δij g2 − sin δij g1 )(ωi − ωj ),
+δ + kβi k .
2
(14)
2 (Eqi0 + x0di Idi )x0di
V̇i is negative definite outside the compact set M = {(z1i , z2i , z3i ) : g1 = q ,
α
K (z1i , z2i , z3i ) ≤ δ + 2i kβi k2 }, K (z1i , z2i , z3i ) := q1i z1i
2
+ q2i z2i2 + (Eqi0 + x0di Idi )2 + (x0di Iqi )2
α
2
q3i z3i + 2i k(βi − β̂i )k2 . Further define an  -neighborhood of M with x0di2 Iqi
 > 0 M = {(z1i , z2i , z3i ) : K (z1i , z2i , z3i ) ≤ δ + α2i kβi k2 + }, then g2 = q . (15)
V̇i ≤ − . The state zji , j = 1, 2, 3 will enter the  -neighborhood,
(Eqi0 + x0di Idi )2 + (x0di Iqi )2
M , in a finite time,
r which implies the asymptotic convergence to In the above equation, θi and bi are unknown constants, γ i (t ) are
the region |zji | ≤
2δ+αi kβi k2
. unknown time varying vectors, ξi (t ), ηi (t ) and fi (t ) are known
nonlinear functions. Since γ i (t ) is dependent on the operating
2qji

Therefore, we have the following theorem for multimachine conditions, they are bounded. So a new robust adaptive control
systems: approach can be easily developed for the new system. The voltage
is introduced as a feedback variable in the proposed control
Theorem 1. For the multimachine systems (4), define z1i = x1i , approach. Thus the post-fault voltage is prevented from excessive
z2i = x2i − u1i and z3i = x3i − u2i , where u1i = −q1i z1i , u2i = variation. It is unnecessary to keep the power angle regulated once
− qa2i z2i − a1 z1i − a1ia+q1i x2i . The controller ui = −a2i z2i − q3i z3i −
2i 2i 2i transient stability is assured.
T T T ˙ By now it can be seen that the nonlinear controller for transient
θ̂ i ξ i − S (β̂i ηi z3i )β̂i ηi with the parameter updatingrlaw θ̂ i = ξ i z3i ,
stability and voltage controller achieve different control objective
˙ 2δ+αi kβi k2
β̂i = |ηi z3i | − αi β̂i will guarantee that |zji | ≤ 2qji
, for in different regions of the states. In Guo, Hill, and Wang (2001), a
global controller is designed to coordinate the transient stabilizer
j = 1, 2, 3. and voltage regulator. The designed controller is smooth and
robust with respect to different transient faults. The global control
Remark 3. To construct the control law ui (t ) based on the adaptive (1) (2) (1)
approach takes the form: uf = µδ uf + µr uf , where uf is the
backstepping method, the power angle δi (t ) should be available.
(2)
Since the power angle δi (t ) is difficult to measure, it is necessary nonlinear controller proposed in Theorem 1 and uf is the voltage
to show how to construct the power angle δi (t ) from the measure- controller which can be easily designed by Eq. (15). The definitions
ments available. It is a standard way to use a state observer to es- of µδ and µr are given for details in equations (25) and (26) of Guo
timate δi (t ). A simple way was proposed by constructing a device, et al. (2001).
named a δi -detector δ̂i (t ) = ωi (t ) in Wang et al. (1993).
4. Illustrative examples
Remark 4. From Theorem 1, it is clear that the bound of zji is de-
cided by the design parameters q1i , q2i , q3i , δ and αi . Therefore, the A three-generator system where generator #3 is infinite bus in
tracking error can be made sufficiently small by choosing appro- Fig. 1 is chosen to demonstrate the effectiveness of the proposed
priate values for the design parameters. adaptive nonlinear decentralized controller. The generator and the
Furthermore, from theoretical analysis, it is only required qji > transmission line parameters are listed as follows. For generator
0, j = 1, 2, 3 to get convergence performance. Similar to PID con- #1, xd = 1.863 p.u., x0d = 0.257 p.u., xT = 0.129 p.u., xad =
trol, different values of qji will affect the convergence speed. If qji 1.712 p.u., Td0
0
= 6.9 p.u., H = 4 s, D = 5 p.u. and kc = 1. And
is too small, the system needs a long time to reach the stable state. for generator #2, xd = 2.36 p.u., x0d = 0.319 p.u., xT = 0.11 p.u.,
However, if qji is too large, the control signal may be too large to xad = 1.712 p.u., Td00
= 7.96 p.u., H = 5.1 s, D = 3 p.u. and
realize in practice. The reader can find a suitable value to get a sat- kc = 1. Moreover, x12 = 0.55 p.u., x13 = 0.53 p.u., x23 = 0.6 p.u.
isfied performance by tuning values of qji . We will demonstrate the and ω0 = 314.159 rad/s. In the example, the excitation control
above discussion in Section 4. input limitations are −3 ≤ Efi (t ) ≤ 6, i = 1, 2.
334 R. Yan et al. / Automatica 46 (2010) 330–336

Fig. 1. A two-machine infinite bus power system.

The fault we consider in the simulation is a symmetrical three-

phase short circuit fault that occurs on one of the transmission lines
between generator #1 and generator #2. λ is the fraction of the
line to the left of the fault. If λ = 0, the fault is on the bus bar of
generator #1, λ = 0.5 puts the fault in the middle of the generators
#1 and #2. The fault sequence we considered is the following:
Stage 1. The system is in a prefault steady state.
Stage 2. A fault occurs at t = 0.1 s.
Stage 3. The fault is removed by opening the circuit breakers of
the faulted line at t = 0.25 s. Fig. 2. Power angle response of generator #1 and #2 with different values of qij .
Stage 4. The transmission lines are restored with the fault
cleared at t = 1.0 s.
Stage 5. The system is in a post-fault state.
For the power system in Fig. 1, we first calculate the values of
parameters for the dynamic system of generator #1 in (4). a11 =
ω0
D1
− 2H 1
= −0.250, a21 = − 2H = −39.2699 are known constants.
1
ξ1 = −P̃e1 + Pm10 + (xd1 − x0d1 )Iq1 Id1 = −P̃e1 + Pm10 + 1.6060Iq1 Id1
and η1 = [sin(δ̄11 ), sin(δ̄12 ), cos(δ̄11 )ω1 , cos(δ̄12 )ω2 ]T , δ̄ij =
δ̃i − δ̃j + δmi0 − δmj0 , are known functions. θ1 = T 01 =
d01
0.1449, b1 = = 0.1449 are unknown uncertainties. γ 1 =
kc1
0
Td01
0 0
[Eq1 Ėq1 B11 , Eq1
0 0
Ėq2 B12 , −Eq1
02
B11 , −Eq1
0 0
Eq2 B12 ]T is an unknown time
varying uncertainty.
Similarly, we can get the values of parameters for generator #2.
a12 = −0.2941, a22 = −30.7999,
ξ2 = −P̃e2 + Pm20 + 2.0410Iq2 Id2 ,
η2 = [sin(δ̄21 ), sin(δ̄22 ), cos(δ̄21 )ω1 , cos(δ̄22 )ω2 ]T ,
θ2 = 0.1256, b2 = 0.1256,
γ 2 = [Eq2
0 0
Ėq1 B21 , Eq2
0 0
Ėq2 B22 , −Eq1
0 0
Eq2 B21 , −Eq2
02
B22 ]T .
In the example system, since the Generator #3 is an infinite bus, Fig. 3. Responses of Efi (t ) for generator #1 and #2 with different values of qij .
Eq0 3 = const. = 1 < 0o . We demonstrate the performance of the
proposed excitation controller in the following operating points: control signal at the beginning stage. Due to the limit of excitation
control signal in practice, the performance based on this value is
δ10 = 30.5 , ◦
Pm10 = 0.57 p.u., Vt1 = 1.01 p.u.,
also not good. While using qij = 5 and 10, we can get the satis-
δ20 = 32.5◦ , Pm20 = 0.56 p.u., Vt2 = 1.00 p.u. fied performance for this example and the performance based on
and the fault location λ = 0.05. Furthermore, saturation of qij = 10 is better than that of qij = 5. The comparison results also
synchronous machines is also considered, so (2) becomes verify the explanation for different values qij in Remark 4.
Furthermore, the responses of generator terminal voltages for
1 generator #1 and generator #2 using qij = 10 into the proposed
0
Ėqi = 0 [Efi − Eqi − (1 − kfi )Eqi0 ],
Td0i controller are shown in Fig. 4. From the results shown above it
0 (ni −1) is obvious that the proposed controller can enhance the system
where kfi = 1 + cni (Eqi ) , with cn1 = 0.95, cd1 = 0.051, n1 =
c
di transient stability and dampen out the power angle oscillations.
8.727, cn2 = 0.935, cd2 = 0.064, n2 = 10.878.
The controller design is according to Theorem 1. Let γ = 0.001, Remark 6. The proposed control scheme is effective for changes
δ = 0.01 and ρ1 = 1, ρ2 = 156 such that ρ12 tan ρ11 ≤ δ . Now in voltage setpoint and adjustment to transformer tap changer
choose the different values of qij = 1, qij = 5, qij = 10 and qij = 15, position since they only affect the conditions of operating points.
i = 1, 2, 3 and j = 1, 2. The power angles of generator #1 and #2 The proposed control approach is effective for different operating
are given in Fig. 2. And the limited excitation control for generator points. It can be guaranteed from the theorem’s proof. Now we
#1 is also shown in Fig. 3. From Figs. 2 and 3, it can be seen that the demonstrate the performance of the proposed excitation controller
system does not reach the stable states in a short time for a small in the following operating points: δ10 = 59.9667◦ , Pm10 =
value of qij = 1. And for a large value of qij = 15, it needs a large 0.95 p.u., Vt1 = 1.05 p.u., δ20 = 62.9675◦ , Pm20 = 0.95 p.u.,
 R. Yan et al. / Automatica 46 (2010) 330–336 335

Fig. 4. Responses of Vti (t ).

Fig. 7. Power angle response of generator #1 and #2 with existing control

approaches.

5. Conclusion

In this paper, the adaptive backstepping method is used to

design a controller to enhance the transient stability for multi-
machine power systems. Nonlinear excitation controllers are de-
signed analytically using the Lyapunov function method. The main
feature of the design is that the Lyapunov function method is ap-
plied to each subsystem to find a proper fictitious control as if it
could be controlled independently. The proposed controller has
Fig. 5. Power angle response of generator #1 and #2. been applied to improve the transient stability in the face of a large
sudden fault. Furthermore, the effectiveness of controllers have
been demonstrated with a number of case studies.

References
Anderson, P. M., & Fouad, A. A. (1994). Power system control and stability. New York:
IEEE Press.
Bergen, A. R. (1986). Power systems analysis. Englewood Cliffs, New Jersey: Prentice
Hall.
Chapman, J. W., & Ilic, M. D. (1993). Stabilizing a multimachine power system
via decentralized feedback linearizing excitation control. IEEE Transactions on
Power Systems, 8, 830–839.
Chapman, J. W., Ilic, M. D., King, C. A., Eng, L., & Kaufman, H. (1993). Stabilizing a
multimachine power system via decentralized feedback linearizing excitation
control. IEEE Transactions on Power Systems, 8, 830–839.
Cong, L., Wang, Y., & Hill, D. J. (2004). Coordinated control design of generator
excitation and svc for transient stability and voltage regulation enhancement
of multi-machine power systems. International Journal on Nonlinear and Robust
Control, 14, 789–805. [special issue].
Cong, L., Wang, Y., & Hill, D. J. (2005). Transient stability and voltage regula-
tion enhancement via coordinated control of generator excitation and svc.
Fig. 6. Responses of Vti (t ). International Journal of Electric Power and Energy Systems, 27(2), 121–130.
deMello, F. P., & Concordia, C. (1969). Concepts of synchronous machine stability
as affected by excitation control. IEEE Transactions on Power Apparatus and
Systems, 88, 316–329.
Vt2 = 1.02 p.u. In this case, we consider the fault location λ = 0.5 Dong, Z. Y., Hill, D. J., & Guo, Y. (2005). A power system control scheme based
and use qij = 10 in the controller design. The performance results on security visualisation in parameter space. International Journal of Electrical
Power & Energy Systems, 27(7), 488–495.
are shown in Figs. 5 and 6. Fan, J. Y., Ortmeyer, T. H., & Mukundan, R. (1990). Power system stability im-
Finally, we will give some comparison between our proposed provement with multivariable self-tuning control. IEEE Transactions on Power
Systems, 5, 227–234.
approach and existing methods in Guo and Hill et al. (2000) Gao, L., Chen, L., Fan, Y., & Ma, H. (1992). A nonlinear control design for power
and Wang et al. (1997). As we pointed in Introduction, compar- systems. Automatica, 28(5), 975–979.
Guo, G., Wang, Y., & Hill, D. J. (2000). Nonlinear output stabilization control for
ing with the existing control approaches, the major advantages multimachine power systems. IEEE Transactions on Circuits and Systems, Part I,
of the proposed approach are: firstly, it has no requirement for 47(1), 46–53.
the bounds of interconnections by the transmission networks; sec- Guo, Y., Hill, D. J., & Wang, Y. (2001). Global transient stability and voltage regulation
for power systems. IEEE Transactions on Power Systems, 16, 678–688.
ondly, it does not need that the solution of a designed algebraic Guo, Y., Hill, D. J., & Wang, Y. (2000). Nonlinear decentralized control of large-scale
Riccati equation exists; thirdly, it can improve transient stability power systems. Automatica, 36, 1275–1289.
Jain, S., Khorrami, F., & Fardanesh, B. (1994). Adaptive nonlinear excitation control
performance. The comparison results are given in Fig. 7. It is clear of power systems with unknown interconnections. IEEE Transactions on Control
that the proposed control approach achieves the better perfor- Systems Technology, 2, 436–446.
Kanellakopoulos, I., Kokotović, P. V., & Morse, A. S. (1991). Systematic design
mance although the bounds of interconnections are unknown in of adaptive controllers feedback linearizable systems. IEEE Transactions on
the controller design. Automatic Control, 36, 1241–1253.
336 R. Yan et al. / Automatica 46 (2010) 330–336

ZhaoYang Dong obtained his Ph.D. from the University of

King, C. A., Chapman, J. W., & Ilic, M. D. (1994). Feedback linearizing excitation
control on a full-scale power system model. IEEE Transactions on Power Systems, Sydney, Australia in 1999. He is currently with Hong Kong
9, 1102–1109. Polytechnic University. He previously held academic po-
Krstić, M., Kanellakopoulos, I., & Kokotović, P. V (1995). Nonlinear and adaptive sitions with the University of Queensland, Australia and
control design. New York: Wiley. National University of Singapore. He also held industrial
Kundur, P. (1994). The EPRI power system engineering series. Power systems stability positions with Powerlink Queensland, and Transend Net-
and control. New York: McGraw-Hill. works, Tasmania, Australia. His research interests include
Larsen, E. V., & Swann, D. A. (1981). Applying power system stabilizers. IEEE power system planning, power system stability and con-
Transactions on Power Apparatus and Systems, 100, 3017–3046. trol, load modeling, electricity market, and computational
Lu, Q., Sun, Y., Xu, Z., & Mochizuli, T. (1996). Decentralized nonlinear optimal intelligence and its application in power engineering.
excitation control. IEEE Transactions on Power Systems, 11, 1957–1962.
Mello, F. P. D. (1994). Measurement of synchronous machine rotor angle from
analysis of zero sequence harmonic components of machine terminal voltage.
IEEE Transactions on Power Delivery, 9, 1770–1775.
Pai, M. A. (1981). Power systems stability (analysis by the direct method of Lyapunov).
Amsterdam: North-Holland.
T.K. Saha was born in Bangladesh and immigrated to Aus-
Wang, Y., Zhang, G. H., & Hill, D. J. (2004). Robust power system transient stability
tralia in 1989. He is a Professor in the School of Information
enhancement: A global control approach. In 6th IFAC symposium on nonlinear
control systems. (pp. 1325–1330). Technology and Electrical Engineering at the University of
Wang, Y., & Hill, D. J. (1996). Robust nonlinear coordinated control of power Queensland Australia. He is a senior member of the IEEE
systems. Automatica, 32(4), 611–618. and a Fellow of the Institution of Engineers, Australia. His
Wang, Y., Hill, D. J., Middleton, R. H., & Gao, L. (1993). Transient stability enhance- research interests include power systems, power quality,
ment and voltage regulation of power systems. IEEE Transactions on Power and equipment condition monitoring.
Systems, 8(2), 620–627.
Wang, Y., Guo, G., & Hill, D. J. (1997). Robust decentralized nonlinear controller
design for multimachine power systems. Automatica, 33(9), 1725–1733.
Xu, J.-X., & Yan, R. (2006). On repetitive learning control for periodic tracking tasks.
IEEE Transactions on Automatic Control, 51(11), 1842–1848.

Rui Yan received the B.S. degree and the M.S. degree in Rajat Majumder has received his Ph.D. in Electrical Power
the Department of Mathematics from Si Chuan University, Systems from Imperial College London, UK in January
Chengdu, China, in 1998 and 2001, and received the Ph.D. 2006. He is with ABB Corporate Research, Power Tech-
degree from the Department of Electrical and Computer nology Department, Sweden working with ABB’s future
Engineering from the National University of Singapore, in development in FACTS & HVDC Technologies. He has
2006. She is currently a Research Fellow in the Institute for also worked as a lecturer in Power & Energy Systems
Infocomm Research, Agency for Science, Technology and at University of Queensland, Brisbane, Australia between
Research, Singapore. 2006–2007. His research interest is Power Systems Dy-
Her research interests include advanced nonlinear namics & Control.
control, power system stability analysis and control, intel-
ligent control and social robotics.