DECENTRALIZED NONLINEAR CONTROLLER DESIGN FOR

MULTIMACHINE POWER SYSTEMS VIA BACKSTEPPING

A.R. Roosta, D. Georges, N. Hadj-Said*
Laboratoire dAutomatique de Grenoble
*Laboratoire dElectrotechnique de Grenoble
ENSIEG, B.P. 46 - 38402 Saint Martin dHeres Cedex, France
e-mail:
roosta,georges
@lag.ensieg.inpg.fr
Keywords: Multimachine power systems, nonlinear control,
backstepping.

Abstract
In this paper, a backstepping-based decentralized control
scheme is proposed for transient stability enhancement of
multi-machine power systems. The design is based on two
stages: First, an equivalent single-machine infinite-bus model
(SMIB) is developed with respect to each machine. Then,
apart from each SMIB model, a decentralized nonlinear control scheme based on a backstepping technique is designed
which guarantees asymptotic stability of the overall interconnected power system. Some simulation results demonstrate the
effectiveness of this approach.

1 Introduction
Power systems are increasingly brought to operate at high
power transmission levels for some economical or environmental reasons, such as deregulation of the energy market. This
requires the control system to have the ability to compensate
potential instabilities and poorly damped power angle oscillations, as networks load is expected to increase in the future. In
lot of cases, transient stability limits are more constraining than
the steady-state limits. This requires the control system to have
the ability to regulate the system under diverse operating conditions. Unfortunately, power systems are some very nonlinear
systems; the behavior of conventional linear controllers, such
as power system stabilizers, that are designed on the basis of
some linearized power system models (small signal models)
is significantly affected by changes in operating conditions. In
this paper, we will focus our attention on the transient stability enhancement of multimachine power system by means of
a backstepping control design. This paper presents an extension to the multi-machine case of the backstepping controller
designed for SMIB systems proposed in [2].

large fault occurs, a linear controller may not preserve stability.

Recently, to overcome this problem, several authors (see for
example, [7], [8], [4]) have applied nonlinear control theory.
Most of these nonlinear controller designs for power systems
are based on differential geometry approach. The so-called Direct Feedback Linearization (DFL) approach was applied to design a voltage regulator [9], [10].
The paper is devoted to the design of a new nonlinear controller
for a multi-machine power system, by using a backstepping approach. Our goal is to improve transient stability of the overall
power system under the effect of a symmetrical three phase
short circuit fault. With the backstepping methodology, the design of both feedback control laws and associated Lyapunov
functions is systematic. Strong properties of global or regional
stability are built into the nonlinear system in a fixed number of
steps, which is never higher than the system order. While feedback linearization methods require precise models and often
cancel some useful nonlinearities, backstepping designs offer a
choice of design tools for accommodation of uncertain nonlinearities [3].
This paper is organized as follows: Section 2 describes the nonlinear dynamics of a multi-machine power systems. In section
3, we consider the design of a decentralized nonlinear control
scheme for a multi-machine power system based on both equivalent circuit (SMIB model) design and the use of backstepping.
Section 4 presents some simulation results and comparisons.
Finally section 5 sums up some conclusions.

2 Nonlinear dynamics of multi-machine power

systems
The classical model of a synchronous machine may be used
to study the dynamics of power systems when the system
dynamics largely dependent on the stored kinetic energy of the
rotating masses.
Mechanical equations of each generator 

A great deal of attention has been paid to the application of

linear control theory to power systems [11], [6], [1]. However,
power system stabilizer design based on some linearized models is not adequate in presence of large disturbances: when a

 
 



 
 


 


 

(1)
 
!"#$


(2)

h3k3uLivlnm0U

with
 

 
"#


with | ~}

power angle of machine  ;

relative speed of machine  ;
mechanical input power of machine  ;
electrical output power of machine  ;
synchronous machine speed;
per-unit damping contant of machine  ;
inertia constant(in sec) of machine  ;


 

+*

,-& 
.0/
+*

%  &  
)
.

, & 
( /

(3)

% &  A
2 742 &  6 5 
.
(
(

(

(4)

transient EMF in quadrature axis of machine  ;

transient EMF in direct quadrature axis of machine  ;
quadrature component armature current of machine  ;
direct component armature current of machine  ;
direct axis transient short-circuit time constant of
machine  ;
quadrature axis transient short-circuit time constant
of machine  ;
input of the SCR amplifier of the generator of
machine  ;
gain of excitation amplifier of machine  ;

B >@
8;:<

Additional electrical equations

% &  5 . D
.
% . &  92
% &  42
(

"#C


E . F
E C
(

ELK F M

E'N  
.

. 

(

E+N 
(

bgy#z {

m0U
b

real part of 
imaginary
part of 
3!@}

For each generator  of the network the state equations are given
by:
m U l~

V U l~
o
V
o

p'
 U

U l~

o \

7
U
 U?Dg p

W U UYT~ Uc\

Y U6
t

_ ^ h3icjckvlnm0U Y T Z
_ ^ h3k3uLivlnm0U Y Z
Y TZ v
U
bIo
b@o t be
b p
UJ`
JU `
acbed b$agf
a bed b$agf
c
p

_ ^ h3k3uLi?lnm0U Y T Z \ _ ^ h3icjck?lnm0U Y Z
Y Z Uv
b@o
b@o t be
t
b
UJ`
UJ`
acbed b$agf
acbed b$agf
U
Z U
Y[TZ U l~ \ t p t
U UY[TZ U]\ W U U
Y[Z U
V
Z TU6l~ o
o
t
p

Z U
Z U r
t
t
U
Z U
_ ^ h3k3uLivlnm0U Y T Z \ _ ^ h3icjck?lnm0U Y Z
\ t p t
bIo
b@o t be
b
Z U
J
U
`
UJ`
t
acbed b$agf
acbed b$agf
TU
Z
p TU W U UY[TZ U
Y[Z U l~ \
U UY[Z U
V
Z t U l~ o
o
t
p)
pRr
ZT U t
ZT U

% &5 
( (
& 5 7
 4GIH0J5
( (
& 5 . 3
 4GIH0J5
.

W U

b p

Dynamics of a multimachine power system

% &  1
2 . 342 &  6 5 . 798;:<<=?>@
.
(

U l~

, & 
( /

m0U

bgw$x@y

By substituting the above equations into equations (3), (4) and

(11), we obtain the state equations of a multimachine
system
  % &
% &
with respect to the following state variables: , , . and ( .

with
% &
(
% & 
.
5 
(
5 
(
, & $
.

U
r

:
:
=

~ }
~}

Electrical equations of each generator 

%' &  
)
(

b@o

(5)

(6)
(7)
p

(8)

TU

p
ZT U

ZTU

_ ^
h3icjck?lnm0U Y[T Z

b
be
UJ`
acbed b$agf

_ ^
h3k3uLivlnm0U Y Z
b@o t b o \
UJ`
acbed b$agf

ZT U

UJ7 U

(9)

with
ELK 

(
E . 

generator terminal voltage of machine  ;

K
quadrature component of O of machine  ;
K
direct component of O of machine  ;

In order to study the transient stability of a multi-machine

power system
RQ according to the Park model, it is necessary to
derive the P
components
of each generator expressed in the
4Q
coordinate frame P
. Therefore, for each direct and quadrature component of line  current, we get:
_ ^ h3icjck?lnm0U [
Y Z
U U
Y[Z U _ ^ h3k3uLivlnm0U [
Y Z
S$TUDVXW U UY[TZ U]\4
b@o T q
b@o t b
t p
b psr
UJ`
UJ`
acbed b$agf
aceb d b$agf
S t U V
r

where

_ ^ h3k3uLivlnm0U Y[T Z \
U UY[TZ U]\4
bIo
b
UJ`
acbed b$agf

h3icjck?lnm0U

b@o

VXW U

bgw$x@y

m0U

b \
r

bgy#z {

3 Decentralized nonlinear controller design

A classical study of transient stability will be presented here
on a nine-bus power system composed of three generators and
three loads. This approach can be obviously extended to a machine power system. A one-line impedance diagram for the
system is given in fig. 1. In order to apply the backstepping
method [2, 5], we compute:
/ *

_ ^ h3icjck?lnm0U [
Y Z
W U UY[Z U]\4
b@o t b
t
UJ`
acbed b$agf

m0U
b

The equivalent
circuit from point of view of bus
for
*
generator #
/ 
The equivalent circuit from point of view of bus
for

generator #
/e
The equivalent
circuit from point of view of bus
for

generator #

163 MW
6.7 Mvar

85 MW
-10.9 Mvar

Load C
0.0085+j0.072

B/2=j0.1045

9
B/2=j0.179

0.032+j0.161

1.025
9.3

3
0.039+j0.170

8
B/2=j0.153

0.017+j0.092

Admitance

'

1.2610-j0.5044
0.8777-j0.2926
0.9690-j0.3391

Load B

B/2=j0.079

B/2=j0.088

Load A

6
0.010+j0.085

1.025
4.7

Load
A
B
C

j0.0576

g
\

1 1.040
0.0

in three different situations corresponding to pre-fault, duration

fault and post-fault, as defined by fig. 2.

Rth + jX th

For this equivalent

we have used three different
K + Kcircuit,

with respect to pre-fault,

duration fault
impedances
EDK
and post-fault situations. However for
we have only considered the pre-fault value.
In order to apply the
method, we have changed
 backstepping


%
( ) of [2] into new variables: ,
,
and
the
state
variables
(

"
and . With regard to these new variables, the set of state
equations for each equivalent circuit is given by:

ml~
V l~
o
o

ql~ V
l~ \ l

l~
o
o
p #
p g
o#o
# g
q]

m
l~ V
\
Y t
6 ; f
p
ps f ] #y z {
g o


]

m
\
w$x@y \ R \ f

r f]
6 

m \
m \ Y Z
y#z {
] w$x@y
r ]
t
u
u
V LL 0

VeJ0 0

r



u
V

m
] $w x@y \
]
+

Vl

m
] w$x@y p +

Vl

f p
o #y z {

V
l

\
p
r

T \

ED

y#z {

(11)

m \

] w$x@y

o

q
t
m
] y#z {
(12)

t t ] w$x@y m \

ELK

pqrq

\9

V
f

Y
h
V
p

p

Y
V
l
\
t s
p f] #y z {
pA ZT f



l
m0 \
\
\
x y f \
o
r f] w$@

R
f

e J6


VX-ml~

m +Y
] #y z { o t

m \
m
w$x@y
'

V

r o

t lY
t f p

First, we put the state equations in a strict-feedback form:

m \
m
$w x@y
t ] y#z {

Y \

m
m \ $w x@y #y z {

t
'

'

o,

Vl~

o,

VX

l~

o,

(13)
(14)
\
f

m0

ZT

hV
# ,

(15)

# .
v"

With regard to (11) it must be noted that ( is a fonction of

and . If we define the functions D$;I$ N 6g6]@ 6 N as follows:

: transformer reactance

and
=
is the Thevenin voltage of each equivalent circuit in the pre-fault situation. We also supposed for simplicity
the

qthat

 variation of input mechanical power is zero
).
(

t V

2 K

p t

V
r p

'

V
r

v\
pRr

\
r

g

t ] y#z { m p ] w$x@y m

t f V
rq

(10)

The definition of the parameters are:

l T
T

TZ
p
t p o p t R
V!
T !
V
\
\

t
t

T \
t

where

Vth

Figure 2: Equivalent circuit (SMIB model) with respect to each gen-

m \

p t t Y
t f \

erator

f y#z {
m
t ] w$x@y \

V! t t
m \
] y#z {

in p.u. on a 100-MVA base

t

V

as follows:

\
r o

and the terminal voltage:

Figure 1: Nine-bus system impedance diagram; all impedances are

t $
v"

We can expresse

71.6 MW
27 Mvar

l
l \
r s
p
t
l \
\

p
r o

l
\

r o

l \

r o

j0.0586

0.0119+j0.1008

B/2=j0.0745

f
f l

l
f
l f

f$
f$

V
o
V
o

V
\
pA ZT f
f$ o

l
m0 \
y#z { f \
o

V
\
$f o
R
f
o

ZT

t ps

f]
l
0m
f] w$x@y f \
o

we get the following strict feedback form:

2 

2 

N
2D



2 #2
N


2
2 
 

2
2 #2D
N

]
N

 2
2
2 I
D

 2
2
2DI B >L

N
N
N
N


2

(16)

The main advantage of the backstepping method is the simultaneous derivation of both the Lyapunov function and the
control law [5]. According to [2] and the configuration of state
equations (13), (14) and (15) we can use the strict-feedback
form of backstepping method as following.
If we choose the
2 N
Lyapunov function candidate
N , we can state the virtual
[O 2
control
for (13):
. Then we seek for a control
2 law

law N 2
v which stabilizes the first two state equations,
where N is viewed as a control
 input

2
 and N where the related
N
Lyapunov function is O]
O
N

. v is such that

 N
 [2 N  
2
N
O;

.
Finally,
we
obtain
the control law

for each generator as follows:

Vf

l
p? p f o p f p

p f pR f p o
f

e ;

e;

f l
f \ p f o such that

p pR f o , and where I N
R
,
V

With
the Lyapunov
function l
V
l


p




l
o f \

\
f

o p

Under the realistic assumption that, we can determine which of

the parallel transmission lines is disconnected (by use of some
sensor pre-assigned on each of the lines), we can compute the
power angle, after disconnecting the faulted line. When we fix
equation
the terminal voltage to 1 
B , we can solve nonlinear
 ?"
(12) for each generator in order to compute a new ( remains

unchanged). We denote the post-fault power angle as / and
+ 1
! /
we define
.
m
is now introduced in the regulation error dynamics and the
m
control law instead of :

 
v
 4


 ! "

' "g 4 N  N  
 6  /  N  / % (  /

!
 eED "# $ /    / 0ED %&" /
/

'  / 
 ( :   >- >*)+  / 

 /  6 B B
Where

"

-, 

"

The three-machine system described by Fig.1 is chosen to

demonstrate the effectiveness of the proposed backstepping
controller. The system parameters used in the simulation are
the following:
T V6 87 29
87
A*B V6 87 1
C V6 1 >
ZT V6 487 : 9E9
ZT V6 87 9 ;4

87 4;:2<*4 7 = <?>
87 >

<F7 4;:>
87 :;4 87 4 =>

6 87 ;: 9 : 87 4 9 < 7 <;@*4>

6 =87 9D9 7 =87 >

6 87 87 <*=2< 87 9 >
6
< <?>
G B IH
H @ . for

V

tZ V
Z
T
t V

Z V
1
t
V
@

l~
V

o
The physical limit of the exciter is:

each generator.


f

54@
V

The case study

Fault sequence (permanent fault)

stage 1 :
stage 2 :
stage 3 :

J
J
J

stage 4 :

87 < 1?K
87 <;@ 1?K

The system is in the pre-fault steady state.

VX
on line 5-7 near the bus # 7.
A fault occurs at
The fault is removed by opening the breakers of
V9

the faulted line at

The system is in a post-fault state.

We have supposed that generator 1 is the reference generator.

The transient response obtained with the backstepping method
is improved compared to the PSS controller response (see
fig. 6,7, 8, 9). The variations of electrical power, terminal
voltage, and power angle for each machine are given by fig.
(3), ( 4) and (5) at the occurence of the short circuit when the
backstepping-based controllers are used.
2.5



(17)



Index .0/21 means that the related relations are expressed in

 u3
m
m
terms of instead of . We also compute
correm
sponding to the new equilibrium state induced by . Simulation results given hereafter, show clearly the improvement
brought by this methodology.

#2

Power electric (pu)

.
- with the here-proposed controller with f

The kinds of faults that we consider in this paper are the

symmetrical three-phase short circuit faults occurring on any
line. The fault is situated at the generator bus.

1.5

0.5

0.5

4 A three-machine example
In this section, the transient and steady state responses obtained
with the backstepping method are simulated and compared
with the responses obtained with PSS controllers. We have
simulated the closed-loop behavior of the system with two
different control schemes:
- without auxiliary controllers (only with conventional PSS
and AVR),

#3

Time (sec)

Figure 3: Electrical power output variation - short circuit between

line 5-7

If we change the position of fault on the other lines, for example line 4-6 (near bus # 4), we can see that there is always
asymptotic stability for the variables of machines. (see fig. 10
and 11).

1.2

350

1.1

#2

300

#3

PSS # 2

0.9

Power angle (deg)

Terminal voltage(pu)

0.8
0.7
0.6
0.5
0.4

250

200

150

100

Backstepping # 2

0.3
0.2

Time (sec)

50

Figure 4: Terminal voltage variations of each machine - short circuit

between line 5-7

Time (sec)

Figure 6: Power angle variations of generator # 2 - short circuit

between line 5-7

120
300
110
250

PSS # 3

Power angle (deg)

Power angle (deg)

100

90

80

#2
70

200

150

#3
100

60

Backstepping # 3
50

Time (sec)

Figure 5: Power angle variations of each machine - short circuit

between line 5-7

50

Both transient enhancement and good post-fault perfor-

Time (sec)

Figure 7: Power angle variations of generator # 3 - short circuit

between line 5-7

5 Conclusions
In this paper, the idea of transient stability enhancement via
backstepping nonlinear control of single-machine infinite-bus
power system proposed in [2] has been extended to the
multi-machine case. The backstepping technique has been
extended to the multi-machine power system via the design
of an equivalent circuit with respect to each generator. A
new power system controller has been proposed in this paper
to achieve both transient stability enhancement and good
EDK 

post-fault performance of the generator terminal voltage
.
It represents a realistic alternative to the usual AVR/PSS
scheme. A simple design procedure has been proposed. The
performance of this controller has been tested through different
simulation scenarios and in comparison with three existing
control schemes. The simulation results show that:

mance of both the generator terminal voltage power angle

and electrical power can be achieved;

The performance of this controller is independent of the

operating point;

Good transient enhancement is obtained, if the faulted line

changes.

Further researches will be devoted to the stability analysis of

this decentralized control scheme.

1.2
1.3

PSS # 2

1.1

1.2

Backstepping # 2

0.9

Terminal voltage(pu)

Terminal voltage(pu)

0.8
0.7
0.6
0.5

1.1

#3

#2
1

0.9

0.8

0.4
0.7

0.3
0.2

Time (sec)

0.6

Figure 8: Terminal voltage variations of generator # 2 - short circuit

between line 5-7

Time (sec)

Figure 10: Terminal voltage variations of each machine - short circuit between line 4-6
80

1.2

75
1.1

#2

Power angle (deg)

Terminal voltage(pu)

PSS # 3

Backstepping # 3
0.9

0.8

70

65

60

0.7

55
0.6

0.5

#3

Time (sec)

Figure 9: Terminal voltage variations of generator # 3 - short circuit

between line 5-7

References
[1] S. Ammari, Y. Besanger, N. Hadj-Said, and D. Georges.
Robust solutions for the interaction phenomena between
dynamic loads and facts controllers. IEEE Power Engineering Society Summer Meeting 2000, 1:401406, 2000.
[2] D. Georges A.R. Roosta and N. Hadj-Said. Nonlinear control for powery systems based on a backstepping
method. Proc. IEEE conf. on Decision and Control,
91:14691480, 2001.

50

Time (sec)

Figure 11: Power angle variations of each machine - short circuit

between line 4-6

[6] A.V. Machias, E.N. Dialynas, and C.D. Papageorgiou.

Application of an optimal control reduction algorithm to a
power system. Found.Control Eng, 12(4):181194, 1987.
[7] W. Mielczaraski and A. Zajaczkowski. Nonlinear controller for synchronous generator. IFAC Nonlinear Control System Design Symposium 89, 1989.
[8] W. Mielczaraski and A. Zajaczkowski. Nonlinear controller for synchronous generator. 11th IFAC World
Congress, 1990.

[3] H.K. Khalil. Nonlinear Systems. Macmillan Publishing

Company., 1992.

[9] Y. Wang and D.J. Hill. Transient stability enhancement

and voltage regulation of power system. IEEE Trans. on
Power Systems, 8(2):620626, 1993.

[4] K.L. Lo and Hamzah. Hilal. Power system transient stability analysis via lyapunov direct method including avr.
Sys.Sci, 22(1):97124, 1996.

[10] G. Guo Y. Wang and D. Hill. Robust decentralized nonlinear controller design for multimachine power system.
Automatica, 33(9):17251733, 1997.

[5] I. Kanellakopoulos M. Krstic and P. Kokotovic. Nonlinear and Adaptive Control Design. John Wiley and sons.
Inc., 1995.

[11] Y.N. Yu and G. Siggers. Stabilization and optimal control

signals for a power system. IEEE Trans PAS, 91:1469
1480, 1971.