International Journal of Power and Energy Systems, Vol. 26, No.

3, 2006
A NOVEL FEEDBACK LINEARIZING
STATCOMCONTROLLER FOR
POWER SYSTEMDAMPING
N.C. Sahoo,

R. Ranjan,

P.K. Dash,

and G. Panda

Abstract
The authors present a nonlinear control approach to the multi-
input multi-output (MIMO) nonlinear model of a static synchronous
compensator (STATCOM). The proposed controller design is based
on a feedback linearization scheme. Its prime goal is the coordinated
control of AC and DC voltage for a STATCOM installed in a power
system. First, the nonlinear mathematical model of STATCOM is
derived. Then, using inputoutput feedback linearization, a state
feedback control law is derived by pole placement. The proposed
control strategy is tested on both distribution and generation system,
by digital computer simulations on the complete system for various
types of loads and/or disturbances. Comparison of these results
with those obtained in conventional cascade control architecture
establishes the elegance of this new control approach.
Key Words
STATCOM, feedback linearization, nonlinear control
1. Introduction
Due to rapid developments in power electronics technology,
static synchronous compensator (STATCOM) is acquiring
a signicant role in reactive power modulation in mod-
ern electric power networks. There are several successful
applications of STATCOM for reactive power supply [1],
load balancing [2], and rapid voltage control [3, 4]. In
1991 the rst full-scale STATCOM became operational [5].
The fundamental principle of a STATCOM installed in a
power system is the generation of a controllable AC voltage
source by a voltage source inverter (VSI) connected to a
DC capacitor. The AC voltage source, in general, appears
behind a transformer leakage reactance. The active and
reactive power transfer between the power system and the
STATCOM is caused by the voltage dierence across this

Faculty of Engineering & Technology, Multimedia University,

Bukit Beruang, 75450 Melaka, Malaysia; e-mail: nirodchan-
[email protected], [email protected]

Director, Silicon Institute of Technology, Bhubaneswar, Orissa,

India; e-mail: [email protected]

Department of Applied Electronics and Instrumentation Engi-

neering, National Institute of Technology, Rourkela 769 008,
India; e-mail: [email protected]
(paper no. 203-3566)
reactance. The AC voltage control is achieved by ring
angle control. Ideally the output voltage of the VSI is in
phase with the STATCOM-bus voltage. In steady state,
the DC side capacitance is maintained at a xed voltage
and there is no real power exchange, except for losses.
The STATCOM diers from other reactive power gen-
erating devices (such as capacitors, static var compen-
sators, etc.) in that the ability to store energy is not a
rigid necessity but is only required for system unbalance
or harmonic absorption. As a consequence, the not-so-
strict requirement for large energy storage device makes
STATCOM more robust, and it also enhances the response
speed.
Fundamentally, two control objectives are imple-
mented in the STATCOM. One is the AC voltage regula-
tion of the power system at the STATCOM bus and the
other is DC voltage control across the capacitor inside
the STATCOM. It is widely known that shunt reactive
power injection can be used to control the bus voltage. In
the conventional control scheme, two voltage regulators
are designed for these purposes: AC voltage regulator for
STATCOM bus voltage control and DC voltage regulator
for capacitor voltage control. In the simplest form, both
the voltage regulators are proportional-integral (PI) type
cascaded controllers [6, 7]. The modelling and control
design are usually carried out in the synchronous d-q
frame [8]. The shunt current is split into d-axis and
q-axis components. The reference values for these currents
are obtained by separate PI regulators from DC voltage
and AC-bus voltage errors, respectively. Subsequently,
these reference currents are regulated by another set of
PI regulators whose outputs are ring angle () and the
modulation index (m) of the PWM inverter. Although
this cascade control structure yields good performance, it
is not very eective for all operating conditions because,
in general, one chosen set of PI-gains may not be appro-
priate for all operating points. Moreover, it becomes a
very dicult task to choose the PI gains for the four PI
regulators of the cascade structure because of the inherent
coupling between the d-axis and q-axis variables. Recently,
a linear multivariable controller approach [9] was used for
the control design. However, because the complete model
is highly nonlinear, the linear approach obviously does not
lead to better dynamic decoupling.
281
This paper focuses on a nonlinear multivariable control
technique for STATCOM using feedback linearization ap-
proach [10]. The feedback linearization technique is based
on the idea of canceling the nonlinearities of the system
and imposing a desired linear dynamics to control the sys-
tem. There are reports of several successful applications of
this control design [1113]. Here, a multivariable controller
is designed for the STATCOM, which is a MIMO system,
installed in a distribution and/or generation system. The
control design is tested by computer simulations under
various types of disturbances. Comparison of these re-
sults with those of conventional cascaded control structure
reveals the supremacy of the proposed control design.
2. Modelling and Cascade Control Scheme of
STATCOM
Figure 1. Basic STATCOM connected to a load bus in
power system.
Fig. 1 shows the basic structure of a six-pulse STAT-
COM in a power system where R
p
represents the ON
state resistance of the switches including transformer leak-
age resistance, L
p
is transformer leakage inductance, and
the switching losses are taken into account by a shunt
dc-side resistance R
dc
. A VSI resides at the core of the
STATCOM. It generates a balanced and controlled three-
phase voltage V
p
. The voltage control is achieved by ring
angle control of the VSI. Under steady state, the dc-side
capacitor possesses xed voltage V
dc
and there is no real
power transfer, except for losses. Thus, the AC-bus voltage
remains in phase with the fundamental component of V
p
.
However, the reactive power supplied by STATCOM is
either inductive or capacitive depending upon the relative
magnitude of fundamental component of V
p
with respect
to V
t
. If |V
t
| >|V
p
| the VSI draws reactive power from
the AC-bus, whereas if |V
t
| <|V
p
|, it supplies reactive
power to the AC-system. The active power is transferred
from the DC capacitor to the terminal and makes the DC
link voltage drop when V
p
leads V
t
. The active power
is transferred from the terminal to the DC capacitor and
makes the DC link voltage rise when V
p
lags V
t
.
2.1 Dynamic Model of STATCOM
Fig. 2 shows the equivalent circuit model of a STATCOM
installed in a power system.
The dynamic equations governing the instantaneous
values of the three-phase voltages across the two sides of
Figure 2. Equivalent circuit of STATCOM connected in a
power system.
STATCOM and the current owing into it are given by:
_
R
p
+L
p
d
dt
_
i
p
= V
t
V
p
(1)
where:
i
p
= [i
a
i
b
i
c
]
T
,
V
t
= [V
ta
V
tb
V
tc
]
T
, (2)
V
p
= [V
pa
V
pb
V
pc
]
T
R
P
and L
P
are (3 3) diagonal matrices with diagonal
elements as R
P
and L
P
, respectively. With the assumption
of no zero sequence components, currents and voltages can
be uniquely represented by equivalent space phasors [14]
and then transformed into synchronous d-q-O frame by
applying the well known Parks Transformation. Thus, the
transformed dynamic equations are:
di
pd
dt
=
R
p
L
p
i
pd
+i
pq
+
1
L
p
(V
td
V
pd
) (3)
di
pq
dt
= i
pd

R
p
L
p
i
pq
+
1
L
p
(V
tq
V
pq
) (4)
where is the angular frequency of the source. It is to
be noted that in the d- and q-axis components of the VSI
voltage, that is, V
pd
and V
pq
, all harmonics near to/above
the VSI switching frequency are neglected. In the real-time
implementation, they should be converted into modulation
index (m) and phase angle ():
m =
_
V
2
pd
+ V
2
pq
kV
dc
and = tan
1
_
V
pq
V
pd
_
(5)
where k is a constant whose amplitude depends upon the
adopted modulation technique. For an eective dc-voltage
control, the input power should be instantaneously equal
to the sum of load power (if any) and the charging rate of
capacitor voltage. Thus, by power balance between the ac
input and the dc output, we get:
dV
dc
dt
=
3
2
V
td
i
pd
+ V
tq
i
pq
(i
2
pd
+ i
2
pq
)R
p
CV
dc

V
dc
CR
dc
(6)
282
Equation (6) models the dynamic behaviour of the dc-side
capacitor voltage. In essence, (3), (4), and (6) together de-
scribe the dynamic model of the STATCOM, summarized
below where V
pd
and V
pq
are re-expressed using (5).
d
dt
_

_
i
pd
i
pq
V
dc
_

_
=
_

R
p
L
p
i
pd
+i
pq

R
p
L
p
i
pq
i
pd
3
2CV
dc
[V
td
i
pd
+ V
tq
i
pq
(i
2
pd
+ i
2
pq
)R
p
]
V
dc
CR
dc
_

_
+
_

_
1
L
p
0
0
1
L
p
0 0
_

_
_
_
V
td
mkV
dc
cos
V
tq
mkV
dc
sin
_
_
(7)
2.2 Cascade Control of STATCOM
The STATCOM control concerns with the control of ac-
and dc-bus voltage. These objectives are met by generating
appropriate current reference (for d- and q-axis) and then
by regulating those currents. PI controllers are conven-
tionally employed for both the tasks while attempting to
decouple the d- and q-axis current regulators. In this study
the STATCOM current (i
p
) is split into real (in phase with
AC-bus voltage) and reactive components. The reference
value for the real current is decided so that the capacitor
voltage is regulated by power balance. The reference for
reactive component is determined by ac-bus voltage regu-
lator. As per the strategy, the new frame of reference (d

-q

frame) is so chosen that the terminal voltage is along the

new d

-axis. Hence V
td
=V
t
and V
tq
=0 (Fig. 3).
Figure 3. Phasor diagram showing d-q and d

-q

frame.
Thus, in d

-q

frame, the currents i

pd
and i
pq
represent
the real and reactive currents given by:
i
pd
= i
pd
cos
t
+ i
pq
sin
t
and i
pq
= i
pq
cos
t
i
pd
sin
t
(8)
Now, the procedure as outlined in [8] has been adopted by
re-expressing the dynamic equations as:
di
pd

dt
=
R
p
L
p
i
pd
+i
pq
+
1
L
p
(V
t
mkV
dc
cos ) (9)
di
pq

dt
= i
pd

R
p
L
p
i
pq
+
1
L
p
(mkV
dc
sin) (10)
where =
0
+d
t
/dt (
0
is original base frequency). The
modied dc-bus voltage dynamic equation is:
dV
dc
dt
=
3
2CV
dc
[V
t
i
pd
(i
2
pd
+ i
2
pq
)R
p
]
V
dc
CR
dc
(11)
The control architecture is shown in Fig. 4, where
K
pv
, K
iv
, K
pc
, K
ic
, K
pm
, K
im
, K
p
and K
i
are the gains of
the PI controllers.
Figure 4. Cascade control architecture for STATCOM.
This approach leads to good control. However, the
approach is not able to decouple the d-q currents com-
pletely because of the coupled equations and the frame
transformation. There are several PI controller gains to
be determined for eective control. The technique does
not take into account the coupling resulting through the
dc-capacitor. These diculties demand a better control
approach. Such an approach using feedback linearization
has been proposed in this study.
3. Feedback Linearizing Control of STATCOM
3.1 Feedback Linearizing Control
A brief review of nonlinear control using feedback lin-
earization [10] is presented. The following MIMO system
is considered:
x = f(x) + g(x)u and y = h(x) (12)
where x(
n
) is state vector, u(
m
) represents control
inputs, y(
m
) stands for outputs, f and g are smooth
vector elds, and h is a smooth scalar function. The
input-output linearization of this system is achieved by
dierentiating y until the inputs appear explicitly. Thus:
y
i
= L
f
h
i
+
m

j=1
(L
gj
h
i
)u
j
, i = 1, . . . , m (13)
where L
f
h and L
g
h represent the Lie derivatives of h(x)
with respect to f(x) and g(x), respectively. If L
gj
h
i
(x) =0
for all j, then the inputs do not appear in (13) and further
dierentiation is to be repeated as:
283
y
(ri)
i
= L
(ri)
f
h
i
+
m

j=1
(L
gj
L
(ri1)
f
h
i
)u
j
, i = 1, . . . , m (14)
such that L
gj
L
(ri1)
f
h
i
(x) =0 for at least one j. This
procedure is repeated for each output y
i
. Thus, there will
be a set of m equations given by:
_

_
y
1
(r1)
.
.
.
y
m
(r
m)
_

_
=
_

_
L
r1
f
h
1
(x)
.
.
.
L
rm
f
h
m
(x)
_

_
+E(x)
_

_
u
1
.
.
.
u
m
_

_
and E(x) =
_

_
L
g1
L
r11
f
h
1
L
gm
L
r11
f
h
1


L
g1
L
rm1
f
h
m
L
gm
L
rm1
f
h
m
_

_
(15)
If E(x) nonsingular, then the control u is given by:
u = E
1
(x)
_

_
L
r1
f
h
1
(x)


L
rm
f
h
m
(x)
_

_
+E
1
(x)
_

_
v
l
.
.
.
v
m
_

_
(16)
where [v
1
. . . v
m
]
T
are the new set of inputs dened by
the designer. The resultant dynamics of the system with
new control is easily derived as in (17) where input-output
relation is decoupled and linear:
_

_
y
(r1)
1
.
.
.
y
(rm)
m
_

_
=
_

_
v
1
.
.
.
v
m
_

_
(17)
3.2 Feedback Linearizing Control of STATCOM
Here, V
t
is taken as a state in addition to other three states
(i
pd
, i
pq
, V
dc
). The dynamic equation for V
t
is obtained
with reference to Fig. 3. V
s
is treated as a constant (strong
source).
V
t
= V
sd
R
es
(i
pd
+ i
rd
) +
0
L
es
(i
pq
+ i
rq
) (18)
dV
t
dt
=
_
R
es
R
p
L
p

2
0
L
es
_
i
pd

_

0
R
es
+

0
R
p
L
es
L
p
_
i
pq

R
es
L
p
V
t
R
es

i
rd
+
0
L
es

i
rq
+
R
es
kV
dc
mcos
L
p

0
L
es
kV
dc
msin
L
p
(19)
The complete state space mode is expressed in the form of
(12) as follows:
x =
_

_
x
1
x
2
x
3
x
4
_

_
=
_

_
i
pd

i
pq

V
dc
V
t
_

_
, u =
_
_
u
1
u
2
_
_
=
_
_
mcos
m sin
_
_
(20)
x
1
= f
1
(x) + g
1
u
1
, x
2
= f
2
(x) + g
2
u
2
, x
3
= f
3
(x),
x
4
= f
4
(x) + g
41
u
1
+ g
42
u
2
where:
f
1
(x) =
R
p
L
p
x
1
+x
2
+
1
L
p
x
4
, g
1
=
k V
dc
L
p
= g
2
,
f
2
(x) = x
1

R
p
L
p
x
2
f
3
(x) =
3
2Cx
3
_
x
1
x
4
R
p
(x
2
1
+ x
2
2
)

x
3
CR
dc
f
4
(x) =
_
R
es
R
p
L
p

2
0
L
es
_
x
1

_

0
R
es
+

0
R
p
L
es
L
p
_
x
2

R
es
L
p
x
4
R
es

i
rd
+
0
L
es

i
rq

g
41
=
R
es
k V
dc
L
p
, g
42
=

0
L
es
k V
dc
L
p
The outputs of the system are V
t
and V
dc
. Thus, y
1
=V
t
and y
2
=V
dc
. Therefore:
_
_
y
1
y
2
_
_
=
_
_
f
4
(x)
a
11
f
1
(x) + a
12
f
2
(x) + a
13
f
3
(x) + a
14
f
4
(x)
_
_
+
_
_
g
41
g
42
a
11
g
1
+ a
14
g
41
a
12
g
2
+ a
14
g
42
_
_
_
_
u
1
u
2
_
_
= A(x) +E(x)
_
_
u
1
u
2
_
_
(21)
a
11
=
3
2Cx
3
(x
4
2R
p
x
1
), a
12
=
3R
p
x
2
Cx
3
,
a
13
=
3
2x
2
3
_
R
p
(x
2
1
+ x
2
2
) x
1
x
4

1
CR
dc
, a
14
=
3x
1
2Cx
3
Thus:
_
_
u
1
u
2
_
_
= E
1
(x)
_
_
A(x) +
_
_
v
1
v
2
_
_
_
_
(22)
| E(x) | =
3
2
0
k
2
V
dc
2Cx
es
x
2
p
_
V
t
+
_
R
P
i
pd
+ 2
R
es
X
es
R
P
i
pq

__
(23)
284
The term
_
R
P
i
pd
+2
R
es
X
es
R
P
i
pq

_
is very small compared
to V
t
in the normal operating range. Thus, |E(x)| always
exists. The new control inputs v
1
and v
2
are selected as
(by PI control):
_
_
v
1
v
2
_
_
=
_
_
y
1ref
+ K
11
e
1
+ K
12
_
e
1
dt
y
2ref
+ K
21
e
2
+ K
22
e
2
+ K
23
_
e
2
dt
_
_
(24)
where e
1
and e
2
are error variables dened by:
e
1
=V
ref
t
V
t
, and e
2
= V
ref
dc
V
dc
(25)
The error dynamics are:
e
1
+ K
11
e
1
+ K
12
e
1
= 0
and e
2
+ K
21
e
2
+ K
22
e
2
+ K
23
e
2
= 0 (26)
The gain parameters K
11
, K
12
, K
21
, K
22
, and K
23
are
determined by assigning desired poles on the left-half S-
plane to achieve asymptotic tracking. From u
1
and u
2
, the
control signals in d

-q

frame are determined by:

m =
_
u
2
1
+ u
2
2
and = tan
1
u
2
u
1
(27)
Figure 5. Transient performances for 50% change in linear impedance load (G
0
=0.8, B
0
=0.4, FL(), PI( )}.
4. Simulation Results
The performance of the proposed controller is evaluated
and compared with that of cascade PI control structure by
computer simulation. Distribution and generation systems
are undertaken separately. The simulation results and
evaluations are reported below.
4.1 Distribution System
The equivalent circuit model of the system is as shown in
Fig. 2. A double-circuit transmission line with load at the
receiving end is considered. Dierent types of load models
are used: linear and nonlinear impedance load, rectier
load, and induction motor load. The load models and the
parameters of the loads and controller are given in the
Appendix. Fig. 5 shows the simulation results for linear
impedance load when G and B both change from their
nominal value after 0.1 sec. Clearly, the performance of the
feedback linearizing (FL) controller is superior to that of
the conventional PI control. The performances for ac-bus
voltage (V
t
) and DC-bus voltage (V
dc
) are compared in
Fig. 6 for nonlinear impedance load where the FL controller
outperforms PI controller both in terms of overshoot and
settling time. Figs. 7 and 8 display the simulation results
for a rectier load for change in load resistance and ring
angle, respectively. The superior performance of the FL
controller is demonstrated.
285
Figure 6. Comparison of transient responses for nonlin-
ear impedance load, FL(), PI( ); Initially, P
0
=0.8,
Q
0
=0.4, n
1
=3, and n
2
=1.5, after 0.1 sec, P=1.2P
0
and
Q=1.2Q
0
, n
1
=3; n
2
=1.5;
Figure 7. Comparison of transient performances for
50% change in linear impedance load (G
0
=0.8, B
0
=0.4,
FL(), PI( ).
Figure 8. Performance for change in ring angle (from 10

to 30

at 0.1 sec) and load resistance =7 for a rectier load,

FL(), PI( ).
Figure 9. (a) Responses for change in induction motor load
torque (increases by 50% after 0.1 sec): FL(), PI( ).
(b) Responses when induction motor load torque reduces
to zero after 0.1 sec: FL(), PI( ).
Figure 10. Transient performances for change in R
dc
(150
to 100) at 0.1 sec. FL(), PI( ).
For induction motor load, the simulations are carried
out with sudden change in load torque. The performances
are shown in Figs. 9(a) and 9(b). The eect of parameter
variation has also been evaluated. Fig. 10 displays perfor-
mances when R
dc
of the capacitor is changed, but in the
computations of the FL control law the original value of
R
dc
is taken. The performance of the FL controller with
changed parameters is better than that of PI control.
286
Figure 11. Transient performances for line switching at the
receiving end side of STATCOM bus after 0.1 sec. FL(),
PI( ).
Figure 12. Transient performances for line switching at the
sending end side of STATCOM bus after 0.1 sec. FL(),
PI( ).
The simulation study with parameter variation is fur-
ther extended to the line switching of the double circuit
transmission lines on either side of the STATCOM bus, so
that the line impedance changes. The controller uses the
original value of line impedance in its computations. The
results are shown in Figs. 11 and 12.
To test the sensitivity of the control law to parameter
changes, the performances of FL control law with original
parameters are compared with those of the changed pa-
rameters for the case of line switching at the sending-end
side of the STATCOM bus (Fig. 13). The performances
are very close to each other.
4.2 Single-Machine Innite Bus Generating Sys-
tem
Fig. 14 shows the single-machine innite-bus (SMIB)
power system considered here. The load and the STAT-
COM are connected at the load bus located between the
generator bus and the innite-bus. The mathematical
Figure 13. Comparison of transient performances for
line switching at the sending end side of STATCOM bus.
FL with changed parameter (---) and FL with original
parameter ().
Figure 14. STATCOM connected in a SMIB generating
system.
model of the generator system and the controller parame-
ters are provided in the Appendix. The same FL control
law is also used in this study. Thus, generator bus (V
s
) is
eectively treated as a constant voltage bus for the control
law calculation. In the transient simulation, V
s
varies as
per system conditions. By the assumption of constant
V
s
for the control law, any unusual rise in the control
magnitude is avoided for very severe faults in the system,
that is, three-phase fault. Figs. 15 and 16 show transient
performances for a three-phase fault at innite bus (for 0.1
sec) and line switching (between the STATCOM bus and
generator bus) for the operating point P
e
=0.8 p.u. and
Q
e
=0.4 p.u., respectively. The results clearly establish the
superiority of the FL controller.
The above simulation results demonstrate that feed-
back linearizing controller has stabilizing properties supe-
rior to those of the cascade PI controller due to its nonlin-
ear nature. It is very dicult to choose any optimal set
of PI gains for the conventional controller. But the pole
placement in FL controller is a simpler task. Moreover, the
PI controller gains for distribution system are found to be
dierent from those of generation system for the same level
of performance, whereas the same poles work well for both
generation and distribution system with the FL controller.
287
Figure 15. Transient performance of three-phase fault at the innite bus for 0.1 sec: FL(), PI( ).
Figure 16. Performance for line switching (between STATCOM and generator bus): FL(), PI( ).
5. Conclusion
A nonlinear control approach based on feedback lineariza-
tion has been proposed for high-performance control of
STATCOM in a power system. The novelty of the approach
is that the nonlinearity appearing in the system model is
eliminated by inputoutput linearization. Thus, a linear
control law is obtained via pole placement. As illustrated
by computer simulation results, the superior performance
of the proposed strategy over the conventional cascade con-
trol approach has been established for static and dynamic
loads, three-phase fault, and line switching. In the present
design, the poles are placed at xed locations throughout
the system operation. Adaptive pole placement would be
a better approach and is currently under investigation.
References
[1] L. Gyugyi, Dynamic compensation of AC transmission lines
by solid-state synchronous voltage sources, IEEE Trans. on
PWRD, 9(2), 1994, 904911.
[2] B.N. Singh, A. Chandra, & K. Al-Haddad, DSP-based indirect-
current-controlled STATCOM, Part 2: Multifunctional capa-
bilities, IEE Proc. Part B, 147(2), 2000, 113118.
[3] B.M. Han, G.G. Karady, J.K. Park, & S.I. Moon, Interaction
analysis model for transmission static compensator with EMTP,
IEEE Trans. on PWRD, 13(4), 1998, 12971302.
288
[4] H.F. Hang & F. Li, Multivariable sampled regulators for the
coordinated control of STATCOM AC and DC voltage, IEE
Proc. Part C, 14(2), 2000, 9398.
[5] S. Mori, K. Matsuno, M. Takeda, & M. Seto, Development of
a large static Var generator using self-commutated inverters
for improving power system stability, IEEE Trans. on Power
Systems, 1, 1993, 371378.
[6] M. Mohaddes, A.M. Gole, & P.G. McLaren, A neural network
controlled optimal pulse-width modulated STATCOM, IEEE
Trans. on PWRD, 14(2), 1999, 481488.
[7] K.R. Padiyar & A.M. Kulkarni, Control design and simulation
of a unied power ow controller, IEEE Trans. on PWRD,
13(4), 1998, 13481354.
[8] C. Schauder & H. Mehta, Vector analysis and control of
advanced static VAR compensators, IEE Proc. Part C, 140(4),
1993, 299306.
[9] P.W. Lehn & M.R. Iravani, Experimental evaluations of STAT-
COM closed loop dynamics, IEEE Trans. on PWRD, 13(4),
1998, 13781384.
[10] J.E. Slotine & W. Li, Applied nonlinear control (Englewood
Clis, NJ: Prentice-Hall, 1991).
[11] W. Mielczarski & A.M. Zajaczkowski, Nonlinear eld voltage
control of a synchronous generator using feedback linearization,
Automatica, 30(10), 1994, 16251630.
[12] H. Sira-Ramirez & M. Ilic-Spong, Exact linearization in
switched-mode DC-to-DC power converters, International
Journal of Control, 50(2), 1989, 511524.
[13] D.I. Kim, I.J. Ha, & M.S. Ko, Control of induction mo-
tor via feedback linearization with input-output decoupling,
International Journal of Control, 51(4), 1990, 863886.
[14] P. Vas, Vector control of AC machines (New York: Oxford
University Press, 1990).
[15] B.C. Lesieutre, P.W. Sauer, & M.A. Pai, Development and
comparative study of induction machine based dynamic P, Q
load models, IEEE Trans. on PWRS, 10(1), 1995, 182191.
[16] P.M. Anderson & A A. Fouad, Power system control and
stability (Piscataway, NJ: IEEE Press, 1994).
[17] A.R. Bergen, Power system analysis (Englewood Clis, NJ:
Prentice-Hall, 1986).
Appendix
Nomenclature:
rotor angle with respect to the innite
bus system voltage
M eective inertia constant
E

q
transient q-axis voltage
x
d
d-axis reactance
x
q
q-axis reactance
x

d
d-axis transient reactance
E
fd
direct excitation voltage
T

do
equivalent transient rotor time constant
P
m
mechanical power
P
e
, Q
e
active and reactive power
The subscript 0 indicates the initial value.
Linear Impedance Load:
Y = GjB,
Load bus voltage = V
rd
+ jV
rq
,
Load current = (GV
rd
+ BV
rq
) + j(GV
rq
BV
rd
)
Nonlinear Impedance Load:
P
L
= P
L0
_
V
r
V
r0
_
n1
,
Q
L
= Q
L0
_
V
r
V
r0
_
n2
where n
1
and n
2
are the indices (>0)
Rectier Load:
Figure A1. Structure of rectier load.
V
dc,r
=
3

V
r,LL
cos
3x
r
i
dc,r

,
i
rL
=

i
dc,r
,
and cos = cos
i
dc,r
x
r

2V
r,LL
,
=ring angle, =phase dierence between ac-side volt-
age and current. P
dc,r
=0.8, R
dc,r
=7, =10
0
, x
r
=0.1
Induction Motor Load [15]:
A third-order induction motor model is taken as given
below.
Stator V
ds
= E

d
X

I
qs
and V
qs
= E

q
X

I
ds
Rotor T

0
dE

q
dt
= E

q
+
X
2
m
X
r
I
ds
s
X
r
R
r
E

d
and T

0
dE

d
dt
= E

d
+
X
2
m
X
r
I
qs
s
X
r
R
r
E

q
Torque
Equation 2H
ds
dt
= T
L
(E

q
I
qs
+ E

d
I
ds
).
T
L
= k
0

r
= k
0

s
(1 s) = k
L
(1 s),
X

=
X
s
X
r
X
2
m
X
r
, T

0
=
X
r

s
R
r
,
s =

s

r

s
,
X
s
=1.2287, X
r
=1.2233, X
m
=1.18, R
r
=0.0053,
s
=
2f
0
, H=0.41, R
s
=0.0079, V
ds
and V
qs
correspond to
voltages of the load-bus.
Synchronous Generator Mathematical Model:
The synchronous generator is described by a third-order
nonlinear mathematical model [16, 17]:
289
d
dt
= ,
d
dt
=
1
M
[P
m
P
e
],
dE

q
dt
=
1
T

do
[E
fd
E

q
(x
d
x

d
)i
d
]
where =
0
.
AVR and PSS:
The excitation system of the generator consists of an
automatic voltage regulator (AVR) with a power system
stabilizer (PSS). The complete system is shown in Fig. A2.
Figure A2. AVR+PSS control system of the generating
system.
System Data:

0
=2f
0
, f
0
=50 Hz, R
e
=0.05 p.u., R
es
=R
e
,
0
L
e
(=X
e
) =0.15 p.u., X
es
=X
e
, R
p
=0.04 p.u.,
0
L
p
(=X
p
) =
0.1 p.u., R
dc
=150, C=5000 F, x
d
= 1.9 p.u., x
q
=1.6 p.u,
x

d
=0.17 p.u., T

do
=4.314 sec., M=0.03 p.u., K
e
=50,
T
e
= 0.1 sec, K
pw
=5, K
iw
=12, E
max
fd
=6 p.u., E
min
fd
=
6 p.u., u
max
pss
=0.01 p.u., u
min
pss
= 0.01 p.u., P
L0
=0.1 p.u.,
Q
L0
= 0.1 p.u., n
1
=n
2
=2.0
Cascade PI Controller for Distribution System:
K
pv
=0.5, K
iv
=100, K
pm
=0.5, K
im
=1,
K
pc
=1, K
ic
=10, K
p
=0.5, K
i
=0.5
Cascade PI Controller for Generation System:
K
pv
=1, K
iv
=10 , K
pm
=0.01 , K
im
=0.1,
K
pc
=1, K
ic
=2, K
p
=0.1, K
i
=0.1
Feedback Linearizing Controller (For Both Distribution
and Generation System):
Desired pole locations for computation of K
11
and K
12
are:
s
1
=100 and s
2
=10. Desired pole locations for compu-
tation of K
21
, K
22
and K
23
are: s
1
=200, s
2
=100, and
s
3
=50.
Biographies
Nirod Chandra Sahoo, formerly
employed at a Lecturer at the
National Institute of Technology,
Rourkela, India, is now with the
Faculty of Engineering and Tech-
nology, Multimedia University,
Melaka, Malaysia. His research
interests are control of electric
drives, power system optimiza-
tion and control, and application
of fuzzy logic and genetic algo-
rithm to system identication and
control.
Rakesh Ranjan has been with
the Faculty of Engineering and
Technology, Multimedia Univer-
sity, Melaka, Malaysia, since 2001.
He has published more than 50
research papers in international
journals and conferences in the
area of radial distribution system
planning, operation and control.
Pradipta Kishore Dash is the
Director of the Silicon Institute
of Technology, Bhubaneswar,
Orissa, India. Previously, he was
a Professor at the National In-
stitute of Technology, Rourkela,
India. He is a Fellow of the Na-
tional Academy of Engineering,
India. His research interests are
power system optimization and
control, power quality assessment
and control, and application of AI
tools to power system.
Ganapati Panda is a Professor at
the National Institute of Technol-
ogy, Rourkela, India. He is a Fel-
low of the National Academy of
Engineering, India. His research
interests are power system opti-
mization and control, power qual-
ity assessment and control, digital
signal processing, VLSI, telemat-
ics, and application of AI tools to
power system.
290
Reproducedwith permission of thecopyright owner. Further reproductionprohibited without permission.