ARTICLE IN PRESS

Control Engineering Practice 17 (2009) 13801387

Contents lists available at ScienceDirect

Control Engineering Practice

journal homepage: www.elsevier.com/locate/conengprac

Design of a fractional order PID controller for an AVR using particle

swarm optimization
Majid Zamani a, Masoud Karimi-Ghartemani b,, Nasser Sadati b, Mostafa Parniani b
a
b

Department of Electrical Engineering, University of California at Los Angeles, USA

Department of Electrical Engineering, Sharif University of Technology, P.O. Box 11365-9363, Tehran, Iran

a r t i c l e in f o

a b s t r a c t

Article history:
Received 21 October 2007
Accepted 21 July 2009
Available online 19 August 2009

Application of fractional order PID (FOPID) controller to an automatic voltage regulator (AVR) is
presented and studied in this paper. An FOPID is a PID whose derivative and integral orders are
fractional numbers rather than integers. Design stage of such a controller consists of determining ve
parameters. This paper employs particle swarm optimization (PSO) algorithm to carry out the
aforementioned design procedure. PSO is an advanced search procedure that has proved to have very
high efciency. A novel cost function is dened to facilitate the control strategy over both the timedomain and the frequency-domain specications. Comparisons are made with a PID controller and it is
shown that the proposed FOPID controller can highly improve the system robustness with respect to
model uncertainties.
& 2009 Elsevier Ltd. All rights reserved.

Keywords:
AVR
Optimal control
Particle swarm optimization
Fractional order PID controller
Robustness improvement

1. Introduction
Fractional calculus extends the ordinary calculus by extending
the ordinary differential equations to fractional order differential
equations, i.e. those having non-integer orders of derivatives and
integrals. Such equations can provide linear models and transfer
functions for some innite-dimensional physical systems (Canat &
Faucher, 2003). On the other hand, fractional order controllers
may be employed to achieve feedback control objectives for such
systems. Modeling and control topics using the concept of
fractional order systems have been recently attracting more
attentions because the advancements in computation power
allow simulation and implementation of such systems with
adequate precision.
Fractional order PID (FOPID) controller is a convenient
fractional order structure that has been employed for control
purposes (Ferreira & Machado, 2003; Monje, Vinagre, Feliu, &
Chen, 2008; Podlubny, 1999). An FOPID is characterized by ve
parameters: the proportional gain, the integrating gain, the
derivative gain, the integrating order and the derivative order.
Different design methods have been reported including pole
distribution (Petras, 1999), frequency domain approach (Vinagre,
Podlubny, Dorcak, & Feliu, 2000), state-space design (Dorcak,
Petras, Kostial, & Terpak, 2001) and two-stage or hybrid approach

 Corresponding author. Tel.: +98 21 66165936; fax: +98 21 66023261.

E-mail addresses: zamani@ee.ucla.edu (M. Zamani), karimig@sharif.edu

(M. Karimi-Ghartemani), sadati@sharif.edu (N. Sadati), parniani@sharif.edu
(M. Parniani).
0967-0661/$ - see front matter & 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conengprac.2009.07.005

(Chengbin & Hori, 2004) which uses conventional (integer order)

controllers design method and then improves performance of the
designed control system by adding proper fractional order
controller. An alternative design method is presented in this
paper based on particle swarm optimization (PSO) algorithm and
employment of a novel cost function which offers exible control
over time domain and frequency domain specications.
PSO is an evolutionary-type global optimization algorithm
(Kennedy & Eberhart, 1995; Liang, Qin, Suganthan, & Baskar, 2006)
which is different from well-known similar algorithms in that no
operators, inspired by evolutionary procedures, are applied to the
population to generate new promising solutions. PSO has already
been used to determine optimal solution to several power
engineering problems such as reactive power and voltage control
(Yoshida, Kawata, & Fukuyama, 2000) and state estimation (Naka,
Genji, Yura, & Fukuyama, 2001). This algorithm is employed here
to design an FOPID controller for an automatic voltage regulator
(AVR) problem. Being the main controller of an excitation system,
AVR maintains the voltage of a synchronous generator at a specic
level. The proposed controller is simulated within various
scenarios and its performance is compared with those of an
optimally designed PID controller. The results conclude that the
FOPID control is able to signicantly improve robustness of the
system with respect to system uncertainties.
The paper is organized as follows. Sections 2 and 3 overview
the concepts of fractional calculus and PSO algorithm, respectively. Design of the proposed FOPID controller for an AVR using
PSO algorithm is described in Section 4. Section 5 is devoted to
computer simulation of the proposed controller and its comparison with a PID controller. Section 6 concludes the paper.

ARTICLE IN PRESS
M. Zamani et al. / Control Engineering Practice 17 (2009) 13801387

2. Fractional calculus
Fractional calculus is a generalization of the ordinary calculus.
The chief idea is to develop a functioning operator D, associated to
an order v not limited to integer numbers, that generalizes the
ordinary concepts of derivative (for a positive v) and integral (for a
negative v) (Valerio & Sa Da Costa, 2006).
There are different denitions for fractional derivatives. The
most usual denition is introduced by Riemann and Liouville
(Oldham & Spanier, 1974) that generalizes the following denitions corresponding to integer orders:
Z x
x  tn1
n
f t dt; n 2 N
1
0 Dx f x
n  1!
c
The generalized denition of D becomes c Dvx f x. The Laplace
transform of D pursues the renowned rule L0 Dvx  sv Fs for zero
initial conditions. This means that, if zero initial conditions are
assumed, the systems with dynamic behavior described by
differential equations including fractional derivatives give rise to
transfer functions with fractional orders of s. More details are
provided in Miller and Ross (1993) and Samko, Kilbas, and
Marichev (1993).
The most common way of using, in both simulations and
hardware implementations, of transfer functions including fractional orders of s is to approximate them with usual (integer
order) transfer functions. To perfectly approximate a fractional
transfer function, an integer transfer function would have to
involve an innite number of poles and zeroes. Nonetheless, it is
possible to obtain logical approximations with a nite number of
zeroes and poles.
One of the well-known approximations is caused by Oustaloup
who uses the recursive distribution of poles and zeroes. The
approaching transfer function is given by Oustaloup (1991):
sv  k

N
Y
1 s=oz;n
n1

1 s=op;n

The approximation is legitimate in the frequency range

ol ; oh . Gain k is also regulated so that both sides of (2) shall
have unit gain at 1 rad/s. The number of poles and zeros (N) is
chosen in advance, and the desired performance of the approximation strongly depends on: low values cause simpler approximations, but may cause ripples in both gain and phase behaviors.
Such ripples can be functionally removed by increasing N, but the
approximation will become computationally heavier. Frequencies
of poles and zeroes in (2) are given by
p
oz;1 ol Z
3

op;n oz;n e;
oz;n1 op;n Z;

n 1; . . . ; N
n 1; . . . ; N  1

1381

electric circuits which can serve as exact fractional integrators and

differentiators have also been reported in Oldham and Spanier
(1974) and Oldham and Zoski (1983).

3. Particle swarm optimization (PSO)

PSO is a population-based evolutionary algorithm that was
developed from research on swarm such as sh schooling and bird
ocking (Kennedy & Eberhart, 1995). It has become one of the
most powerful methods for solving optimization problems. The
method is proved to be robust in solving problems featuring
nonlinearity and non-differentiability, multiple optima, and high
dimensionality. The advantages of the PSO are its relative
simplicity and stable convergence characteristic with good
computational efciency.
The PSO consists of a swarm of particles moving in a D
dimensional search space where a certain quality measure and
tness are being optimized. Each particle has a position
represented by a position vector Xi xi1 ; xi2 ; . . . ; xiD and a
velocity represented by a velocity vector Vi vi1 ; vi2 ; . . . ; viD ,
which is clamped to a maximum velocity Vmax vmax1 ;
vmax2 ; . . . ; vmaxD . Each particle remembers its own best position
so far in a vector Pi pi1 ; pi2 ; . . . ; piD , where i is the index of that
particle. The best position vector among all the neighbors of a
particle is then stored in the particle as a vector
Pg pg1 ; pg2 ; . . . ; pgD . The modied velocity and position of each
particle can be manipulated according to the following equations:
wvt
c1 r1 pid  xt
c2 r2 pgd  xt

vt1
id
id
id
id
xt1
xt
vt1
;
id
id
id

d 1; . . . ; D

9
10

where w can be expressed by the inertia weights approach (Shi &

Eberhart, 1998) and often decreases linearly from wmax (of about
0.9) to wmin (of about 0.4) during a run. In general, the inertia
weight w is set according to the following equation:
wmax  wmin
:iter
11
w wmax 
iter max
where iter max represents the maximum number of iterations, and
iter is the number of current iteration or generation. Also c1 and c2
are the acceleration constants which inuence the convergence
speed of each particle and are often set to 2.0 according to the past
experiences (Eberhart & Shi, 2001). Moreover r1 and r2 are random
numbers in the range of 0; 1, respectively. If Vmax is too small,
then the particles may not explore sufciently beyond local
solutions. In many experiences with PSO, Vmax is often set to the
maximum dynamic range of the variables on each dimension,
vdmax xdmax .

5
4. AVR design using FOPID controller

v=N

e oh =ol

Z oh =ol 1v=N

4.1. FOPID controller

The case vo0 can be handled by inverting (2). For jvj41, the
approximation becomes dissatisfactory. So it is common to
separate the fractional orders of s as follows:
sv sn sd ;

v n d;

n 2 Z; d 2 0; 1

and only the second term, i.e. sd , needs to be approximated.

If a discrete transfer function approximation is sought, the
above approximation in (2) may be discretized (Vinagre,
Podlubny, Hernandez, & Feliu, 2000). There are also methods that
directly provide discrete approximations (Lubich, 1986). Besides,

The differential equation of a fractional order PIl Dm controller

is described by
l
d
ut kP et kI D
t et kD Dt et

12

The continuous transfer function of FOPID is obtained through

Laplace transform and is given by
Gc s kP kI sl kD sd

13

Design of an FOPID controller involves design of three parameters

kP , kI , kD , and two orders l; d which are not necessarily integer. The
fractional order controller generalizes the conventional integer

ARTICLE IN PRESS
1382

M. Zamani et al. / Control Engineering Practice 17 (2009) 13801387

order PID controller. This expansion can provide more exibility in

achieving control objectives.
4.2. Linearized model of excitation system
Automatic voltage regulator (AVR) is the central controller
within the excitation system that maintains the terminal voltage
of a synchronous generator at a specied level. Depending on the
method of supplying DC power, different types of excitation
systems exist (IEEE standard denition for excitation systems for
synchronous machines (ansi), 1986). As an example, Fig. 1 shows
the schematic diagram of an alternator supplied controlledrectier excitation system (IEEE standard denition for
excitation systems for synchronous machines (ansi), 1986). The
DC regulator holds constant generator eld voltage and is
commonly referred to as manual control. It is primarily for
testing, start-up and to cater to situations where the AC regulator
is faulty. Closed loop voltage control is carried out through the AC
regulator. In addition to the AVR, this loop comprises ve main
components, namely amplier, exciter, excitation voltage limiters,
generator and measurement and ltering (Fig. 2). To analyze
dynamic performance of AVR, transfer functions of these
components are represented as follows (Gaing, 2004; Kundur,
1994).
Amplier model: The amplier model is given by
VR s
kA

VC s 1 tA s

14

Typical values of kA are in the range of 10400. The amplier timeconstant often ranges from 0.02 to 0.1 s.
Exciter model: Exciter model and parameters greatly depend on
its type. A simplied transfer function of a modern exciter is
VF s
kE

VR s 1 tE s

15

Typical values of kE are in the range of 0.81 and the timeconstant tE for an AC exciter is in the range of 0.51.0 s.

Generator model: The generator model used in this paper is the

well-known third order simplied model shown in Fig. 3. Values
of the parameters are selected as K1 1:591, K2 1:5, K3 0:333,
K4 1:8, K5 0:12, K6 0:3, T3 1:91, H 3, KD 0, o0 377
(Kundur, 1994).
Measurement model: The voltage measurement block, including PT, rectier and lter, is often modeled with a single time
constant
VS s
kR

Vt s 1 tR s

16

tR ranges over 0.0010.06 s.

Excitation voltage limiter: AVR and exciter output voltages are
limited by windup and non-windup limiters (Kundur, 1994). Also,
dedicated overexcitation and underexcitation limiters are employed to assure safe operation of the generator. The purpose of
the overexcitation limiter (OXL) is to protect the generator from
overheating due to prolonged eld overcurrent. OXL has delayed
operation and, thus, needs not to be modeled for AVR dynamic
performance analysis. underexcitation limiter (UXL) operates
instantaneously to prevent angle instability.
Block diagram of the AVR compensated with an FOPID
controller is shown in Fig. 2. In this gure, the combined effects
of these limiters are represented by the upper and lower limits set
to three times of the nominal value of the eld voltage.
4.3. Performance criterion
There are several performance criteria for design of controllers
such as integral of absolute error (IAE), the integral of squarederror (ISE) or integral of time-weighted-squared-error (ITSE)
(Krohling & Rey, 2001). A disadvantage of the IAE and ISE criteria
is that they may result in a response with a relatively small
overshoot but a long settling time because they weigh all errors
uniformly over time. Although the ITSE performance criterion can
overcome this drawback, but it cannot ensure to have a desirable
stability margin (Krohling & Rey, 2001). The IAE, ISE, and ITSE

Fig. 1. Alternator-rectier exciter employing controlled rectiers.

ARTICLE IN PRESS
M. Zamani et al. / Control Engineering Practice 17 (2009) 13801387

1383

Vref(s)
FOPID Controller

Amplifier

Exciter

Limiter

Generator

Vt(s)

Measurement and Filtering

Fig. 2. Block diagram of an AVR employing an FOPID controller.

4.4. Design of FOPID using PSO

The PSO algorithm is utilized to design the controller
parameters kP , kI , kD , l and d such that the controlled system
exhibits desired response and robust stability as evaluated by the
proposed performance criterion. Details of the design process are
relatively similar to the one reported in Gaing (2004) for a regular
PID controller. The ve controller parameters kP , kI , kD , l and d
compose a particle k kP ; kI ; kD ; l; d. The ve members are
assigned as real values. If there are n particles in a population
(group), then the dimension of that population is n  5. To ensure
the stability of closed loop, the cost function is penalized with a
penalty function Pk given as

if
k is unstable
P1
21
Pk
0
else

Fig. 3. The generator model.

performance criteria are dened as follows:

Z 1
Z 1
IAE
jrt  ytj dt
jetj dt
0

ISE

17

e2 t dt

18

ITSE

If the particle k does not satisfy the stability of closed loop, then k
is an unstable particle and its cost function is penalized with a
very large positive constant P1 . Consequently, k does not survive in
the evolutionary process. Otherwise, the particle k is not
penalized. Therefore, the total evaluation of a particle k, can be
obtained as
Fk Jk Pk

te t dt

19

In this paper, a new performance criterion in the time domain and

frequency domain is proposed for evaluating the FOPID controller.
This performance criterion includes the overshoot MP , rising time
tr , settling time tS , steady-state error ESS, the IAE, integral of
squared-input, gain margin (GM) and phase margin (PM). Control
over PM and GM guarantees robust stability of the AVR. The IAE
and integral of squared-input must be computed numerically and
the integral is evaluated up to T which is chosen sufciently large
so that et is negligible for t4T. Therefore, the proposed
performance criterion Jk is dened as
Jy w1 MP w2 tr w3 ts w4 ESS
Z T
w7
w8

w5 jetj w6 u2 tdt
PM GM
0

22

The design steps are expressed as follows.

(1) Randomly initialize the particles of the population including
searching points and velocities.
(2) For each initial particle k of the population, calculate the
values of the performance criterion in (22).
(3) Compare each particles evaluation value with its personal best
Pi . The best evaluation value among the Pi is denoted as Pg .
(4) Modify the member velocity of each particle k according to
wvt
c1 r1 pid  kt
c2 r2 pgd  kt
;
vt1
id
id
id
id
i 1; . . . ; n; d 1; 2; . . . ; 5
t1
ovmax
vmin
d ovid
d

23
24

20

where k is kP ; kI ; kD ; l; d. The eld voltage VF is used for ut in

(20) to prevent large values of the eld current. The error et is
the difference between the reference voltage and the output
voltage.
The proposed performance criterion (20) comprises eight
terms that the signicance of each is determined by a weight
factor wi. It is up to the user to set the weight factors properly in
order to attain the desired specication. For the current study,
selections are w1 0:1, w2 1, w3 1, w4 1000, w5 1,
w6 1, w7 50 000 and w8 30 000. An increase in wi will
result in some improvement in the corresponding feature at the
expense of degrading other features.

where the value of w is set by (11).

(5) Modify the member position of each particle k according to
kt1
kt
vt1
;
id
id
id

t1
kmin
okmax
d okid
d

25

where kmin
and kmax
represent the lower and upper bounds,
d
d
respectively, of member d of the particle k. For example, when
and kmax
d is 1, the lower and upper bounds of kP are kmin
P
P ,
respectively.
(6) If the number of iterations reaches the maximum, then go to
Step 7, otherwise go to Step 2.
(7) The latest Pg is the optimal controller parameter.
The owchart of the algorithm is also shown in Fig. 4.

ARTICLE IN PRESS
1384

M. Zamani et al. / Control Engineering Practice 17 (2009) 13801387

K3 0:333, K4 1:8, K5 0:12, K6 0:3, T3 1:91, H 3,

KD 0, o0 377, kR 1 and tR 0:06.

Start

5.2. Details of FOPID design using PSO for the AVR

Initialize Ki and its

velocity in search
space, i=1,...,n

Evaluate cost of
(22) for Ki (F(ki)),
i=1,...,n

Inspired from practical requirements, the lower bounds of the

ve controller parameters are zero and their upper bounds are set
to kPmax 1:5, kImax kDmax 1 and lmax dmax 2. The following parameters are used for carrying out the FOPID design using
PSO:

 The members of each particle are kP , kI , kD , l and d.

 Population size 30.
 Inertia weight factor w is set as (11) where wmax 0:9 and
wmin 0:4.

 The limit of change in velocity is set to maximum dynamic

Pi=Ki
i=1,...,n

F(ki)<F(Pi)
i=1,...,n








range of the variables on each dimension.

Acceleration constants c1 2 and c2 2.
Maximum iteration is set to 1500.
ol and oh in (5)(9) are set to 105 and 105 rad=s, respectively.
The order of approximation in (2) is set to N 3.
T in (20) is set to 20 s.
P1 in (21) is set 1010 .

N
5.3. First test: basic performance

Pg=best of Pi
i=1,...,n

Update velocity and position of Ki

according to (23), (24)
i=1,...,n

Ten trials are performed for the proposed controller. The best
solution is summarized in the rst row of Table 1.
The designed FOPID controller is of order six as its fractional
derivative and integral have been approximated by N 3 in (2).
The Hankel minimum degree approximation (MDA) without
balancing (Safonov, Chiang, & Limebeer, 1990) can be employed
to reduce its order to four
1:111e005s4 1:268e006s3 1:405e007s2 2:281e006s 955
G~ c s
s4 4:258e005s3 8:407e007s2 4:322e004s

26

iter<iter_max

Optimal solution=Pg

End

Fig. 4. Flowchart of the proposed method.

The Bode diagrams of the original and reduced order

controllers are shown in Fig. 5, depicting very close behaviors in
the frequency range of interest. Thick lines in Figs. 6 and 7 show
the terminal voltage step response and Bode diagram of the AVR
with the FOPID controller, respectively. Note that the simulations
are performed based on the reduced-order implementation of
(26) for the FOPID controller.
In order to emphasize advantages of the proposed FOPID
controller, a PID controller is also designed using the same
method. The same evaluation function is used and the best
solution is summarized in the third row of Table 1. The terminal
voltage step response and Bode diagram of the AVR with the PID
controller are also shown in Figs. 6 and 7 with dash lines for
comparison. The FOPID does not offer improved performance as
far as transient time characteristics of the response are concerned.
However, the FOPID type AVR has higher robust stability than the
PID type AVR as is observed from Table 1 and also attested by the
subsequent simulations.

5. Numerical results

5.4. Second test: robustness

5.1. Study system parameters

It can be observed from Fig. 7 that the phase-angle of the

transfer function with FOPID controller is at around the gain
crossover frequency:

A practical high-order AVR is used to verify the efciency of the

proposed FOPID controller. The system parameters are kA 10,
tA 0:1, kE 1, tE 0:5, Vf 0 1, K1 1:591, K2 1:5,

d
argGc joGjojoocg C0
do

27

ARTICLE IN PRESS
M. Zamani et al. / Control Engineering Practice 17 (2009) 13801387

1385

Table 1
Comparison of the evaluation values between both the PID and the proposed FOPID controller.
Controller

KP

KI

KD

MP %

tr

ts

ESS

PM

GM

Jk

FOPID
Red. FOPID
PID

0.17

0.0046

0.03

0.0268

0.014

0.18

0.97

1.38

0
0
2

27
27
17.6

52
51
27

0
0
0

95
94
69

310
309
124

750
755
1042

80
Original FOPID Controller
Reduced FOPID Controller

Magnitude (dB)

60
40
20
0
20
180

Phase (deg)

135
90
45
0
45
90
104

102

100

102

104

Fig. 5. Bode diagrams of the original and reduced-order FOPID controllers.

0.6

shown in Fig. 7, it is not possible to create such a atness using the

PID type AVR.
To examine numerically the robustness of the FOPID controller
with respect to parameter uncertainties, the following simulation
is performed. Assume that instead of K1 1:59, and due to the
change in loading condition, the actual value is K1 1. The
terminal voltage step responses of the AVR with previously
designed PID and FOPID controllers are shown in Fig. 8. The PID
controller exhibits an excessive overshoot of about 62% compared
to 18% of the FOPID.
Now assume another uncertainty in the exciter model as

0.4

VF s
1

VR s 0:5 0:5

Step Response
1.4
AVR with FOPID
AVR with PID

1.2

Amplitude

1
0.8

0.2
0
0

50

100

150

28

The step responses with both generator and exciter uncertainties

are shown in Fig. 9. The PID-controlled system exhibits large
oscillations with an overshoot of about 88% as opposed to the 20%
overshoot of the FOPID-controlled system.

Time (sec)
Fig. 6. Step responses of the AVR controlled by PID and FOPID.

This means that the system with FOPID controller is more robust,
with respect to gain changes, in comparison with the one with
using PID controller (Valerio & Sa Da Costa, 2006). This occurs due
to the presence of phase margin in performance criterion (20). As

6. Conclusion
This paper presents a design method for determining the
FOPID controller parameters using the PSO algorithm. The
proposed method involves a new time-domain and frequencydomain performance criterion. Application of the method to a
practical AVR shows that the proposed algorithm can perform an

ARTICLE IN PRESS
1386

M. Zamani et al. / Control Engineering Practice 17 (2009) 13801387

Bode Diagram

Magnitude (dB)

50
100
AVR with FOPID
AVR with PID

150
360
270
Phase (deg)

180
90
0
90
180
270
102

101

100
Frequency (rad/sec)

101

102

Fig. 7. Bode diagrams of the AVR controlled by PID and FOPID.

Step Response

Step Response

1.8

2
AVR with FOPID

1.6

1.6

1.4

1.4
Amplitude

1.2
Amplitude

AVR with FOPID

AVR with PID

1.8

AVR with PID

1
0.8
0.6

1.2
1
0.8
0.6

0.4

0.4

0.2

0.2
0

0
0

50

100

150

200

250

300

350

400

Time (sec)
Fig. 8. Step responses of the AVR controlled by PID and FOPID in the presence of
uncertainties in generator model.

efcient search for the optimal FOPID controller parameters.

Furthermore, it can be concluded from the above simulations that
the proposed FOPID controller has more robust stability and
performance characteristics than the PID controller applied to the
AVR.

References
Canat, S., & Faucher, J. (2003). Fractional order: Frequential parametric identication of the skin effect in the rotor bar of squirrel cage induction machine. In
Proceedings of design engineering technical conference.

50

100

150

200
250
Time (sec)

300

350

400

Fig. 9. Terminal voltage step responses of the AVR controlled by PID and FOPID in
the presence of generator and exciter uncertainties.

Chengbin, M., & Hori, Y. (2004). The application of fractional order pid
controller for robust two-inertia speed control. In Proceedings of 4th
international power electronics and motion control conference (pp. 14771482),
August.
Dorcak, L., Petras, I., Kostial, I., & Terpak, J. (2001). State-space controller design for
the fractional-order regulated system. In Proceedings of ICCC (pp. 1520).
Eberhart, R., & Shi, Y. (2001). Particle swarm optimization: Developments,
applications and resources. In Proceedings of IEEE international conference on
evolutionary computation (pp. 8186).
Ferreira, F., & Machado, T. (2003). Fractional-order hybrid control of robotic
manipulators. In Proceedings of 11th international conference on advanced
robotics (pp. 393398), June.
Gaing, Z. (2004). A particle swarm optimization approach for optimum design
of pid controller in avr system. IEEE Transactions on Energy Conversion, 19(2),
384391.

ARTICLE IN PRESS
M. Zamani et al. / Control Engineering Practice 17 (2009) 13801387

IEEE standard denition for excitation systems for synchronous machines (ansi)
(1986). IEEE Std 421.1.
Kennedy, J., & Eberhart, R. (1995). Particle swarm optimization. In Proceedings of
IEEE international conference on neural networks (Vol. 4, pp. 19421947).
Krohling, R. A., & Rey, J. P. (2001). Design of optimal disturbance rejection pid
controllers using genetic algorithm. IEEE Transactions on Evolutionary Computation, 5(1), 7882.
Kundur, P. (1994). Power system stability and control. New York: McGraw-Hill.
Liang, J. J., Qin, A. K., Suganthan, P. N., & Baskar, S. (2006). Comprehensive learning
particle swarm optimizer for global optimization of multimodal functions. IEEE
Transactions on Evolutionary Computation, 10(3), 281295.
Lubich, C. (1986). Discretized fractional calculus. SIAM Journal of Mathematical
Analysis, 17(3), 704719.
Miller, K. S., & Ross, B. (1993). An introduction to the fractional calculus and fractional
differential equations. New York: Wiley.
Monje, C. A., Vinagre, B. M., Feliu, V., & Chen, Y. Q. (2008). Tuning and auto-tuning
of fractional order controllers for industry applications. Control Engineering
Practice Journal, 16(7), 798812.
Naka, S., Genji, T., Yura, T., & Fukuyama, Y. (2001). Practical distribution state
estimation using hybrid particle swarm optimization. In Proceedings of IEEE
power engineering society (Vol. 2, pp. 815820).
Oldham, K. B., & Spanier, J. (1974). The fractional calculus. New York: Academia.
Oldham, K. B., & Zoski, C. G. (1983). Analogue instrumentation for
processing polarographic data. Journal of Electroanalytical Chemistry, 157(1),
2751.

1387

Oustaloup, A. (1991). La commande CRONE: commande robuste dordre non entier.

Paris: Herme.
Petras, I. (1999). The fractional order controllers: Methods for their synthesis and
application. Electrical Engineering Journal, 50(910), 284288.
l
Podlubny, I. (1999). Fractional-order systems and pi dm controllers. IEEE Transactions on Automatic Control, 44(1), 208214.
Safonov, M. G., Chiang, R. Y., & Limebeer, D. J.N. (1990). Optimal Hankel model
reduction for nonminimal systems. IEEE Transactions on Automatic Control,
35(4), 496502.
Samko, S. G., Kilbas, A. A., & Marichev, O. I. (1993). Fractional integrals and
derivatives. Yverdon: Gordon and Breach.
Shi, Y., & Eberhart, R. (1998). A modied particle swarm optimizer. In Proceedings of
IEEE international conference on evolutionary computation (pp. 6973).
Valerio, D., & Sa Da Costa, J. (2006). Tuning of fractional pid controllers
with ZieglerNichols-type rules. Signal Processing Journal, 86(10),
27712784.
Vinagre, B. M., Podlubny, I., Dorcak, L., & Feliu, V. (2000). On fractional pid
controllers: A frequency domain approach. In Proceeding of IFAC workshop.
Vinagre, B. M., Podlubny, I., Hernandez, A., & Feliu, V. (2000). Some approximations
of fractional order operators used in control theory and applications. Fractional
Calculus and Applied Analysis, 3(3), 231248.
Yoshida, H., Kawata, K., & Fukuyama, Y. (2000). A particle swarm optimization for
reactive power and voltage control considering voltage security assessment.
IEEE Transactions on Power Systems, 15(4), 12321239.