Академический Документы
Профессиональный Документы
Культура Документы
(mf),
pf
(d),
pf
Z
th
R
th
jX
th
Z
1
Z
sh
Z
1
Z
sh
Here, Z
1
R
1
jX
1
and Z
sh
R
c
==jX
m
.
The input impedance of the motor is:
Z
in
Z
1
Z
sh
==Z
2
where, Z
2
is the rotor impedance and is given by:
Z
2
R
2
s
jX
2
The maximum torque, full-load torque, full-load current and power factor of the motor
are determined by the following equations:
T
max
d
3V
2
th
2v
s
R
th
R
2
th
X
th
X
2
2
_
_ _ 1
T
fl
d
3V
2
th
R
2
sv
s
R
th
R
2
=s
2
X
th
X
2
2
I
fl
d
V
ph
Z
in
3
pf
fl
d cos tan
21
X
th
X
2
R
th
R
2
=s
_ _ _ _
4
Figure 1.
Single-cage rotor model
R
1
R
2
R
th
V
th
X
th
X
2
R
2
S
S
I
1
Notes: (a) General equivalent circuit; (b) Thevenin equivalent circuit
V
ph
R
c
X
1
(a) (b)
X
m
X
2
Bio-inspired
optimization
algorithms
533
The objective of the bio-inspired optimization algorithms is to minimize the deviation
between the determined and the manufacturer data. The objective function of the
problem is formulated as follows:
Minimize F
T
max
d
T
max
mf
21
_ _
2
T
fl
d
T
fl
mf
21
_ _
2
I
fl
d
I
fl
mf
21
_ _
2
pf
fl
d
pf
fl
mf
21
_ _
2
5
The minimization of the above objective function is subjected to the following
constraints:
.
Minimum and maximum parameter limits:
X
i;min
# X
i
# X
i;max
.
Maximum torque constraint:
T
max
d 2T
max
mf
T
max
mf
# ^0:2
.
Efciency balance:
P
fl
2 I
2
1fl
R
1
I
2
2fl
R
2
P
rot
_ _
P
fl
h
fl
mf
3.2 Double-cage rotor model formulation
The problem formulation uses the starting torque, maximum torque, full-load torque,
full-load current and full-load power factor manufacturer data to estimate the
parameters of the double-cage rotor model. Figure 2 shows the equivalent circuit of a
double-cage rotor model of an induction motor. The mathematical formulation is done
as follows.
The admittance of each branch of the equivalent circuit model is calculated as
follows:
Y
1
1
R
1
jX
1
Y
m
1
jX
m
Y
21
1
R
21
=s jX
1
Y
22
1
R
22
=s jX
2
Y
tot
jY
1
kY
m
Y
21
Y
22
j
jY
1
Y
m
Y
21
Y
22
j
Figure 2.
Equivalent circuit of a
double-cage rotor model
R
1
V
ph
X
m
X
21
R
21
I
21
X
22
R
22
I
22
I
1
X
1
COMPEL
31,2
534
The starting torque, maximum torque and full-load torque of the motor are determined
as follows:
T
st
d
3V
2
th
jY
1
j
2
R
21
jY
21
j
2
R
22
jY
22
j
2
v
S
jY
1
Y
m
Y
21
Y
22
j
2
6
T
max
d
3V
2
th
Y
1
j j
2
R
21
Y
21
j j
2
R
22
Y
22
j j
2
_ _
s
max
v
s
Y
1
Y
m
Y
21
Y
22 j j
2
7
T
fl
d
3V
2
th
Y
1 j j
2
R
21
Y
21 j j
2
R
22
Y
22 j j
2
_ _
sv
s
Y
1
Y
m
Y
21
Y
22
j j
2
8
The starting current and full-load current of the motor are given by:
I
st
d V
ph
jY
tot
j
s1
9
I
fl
d V
ph
jY
tot
j 10
The power factor of the motor is given by:
pf
fl
d
RY
21
RY
22
jY
1
j R
1
jY
1
kY
m
Y
21
Y
22
j
2
jY
m
Y
21
Y
22
kY
1
Y
m
Y
21
Y
22
j
11
The following objective function should be minimized:
F
T
st
d
T
st
mf
21
_ _
2
I
st
d
I
st
mf
21
_ _
2
T
max
d
T
max
mf
21
_ _
2
T
fl
d
T
fl
mf
21
_ _
2
pf
fl
d
pf
fl
mf
21
_ _
2
I
fl
d
I
fl
mf
21
_ _
2
12
The parameter determination problem is subjected to the following constraints:
.
Minimum and maximum parameter limits:
X
i;min
# X
i
# X
i;max
.
Inner and outer cage leakage reactance constraint:
X
21
. X
22
.
Inner and outer cage rotor resistance constraint:
R
22
. R
21
.
Maximum torque constraint:
T
max
d 2T
max
mf
T
max
mf
# ^0:2
Bio-inspired
optimization
algorithms
535
.
Efciency balance:
P
fl
2 I
2
1fl
R
1
I
2
2fl
R
2
P
rot
_ _
P
fl
h
fl
mf
4. Overview of bio-inspired optimization algorithms
4.1 Genetic algorithm
GA is one of the biologically inspired algorithms developed by Holland which mimics
the principle of natural evolution (Goldberg, 1989). In GA, candidate solution for a
specic problem is called an individual or a chromosome. Each individual is made-up
from genes. GA works with a population of potential solutions. Each individual is
decided by an evaluating mechanism to obtain its tness value. Based on the tness
value and undergoing genetic operators, a new population is generated iteratively with
each successive population referred to as a generation. The GA uses three basic
operators such as reproduction, crossover, and mutation to manipulate the genetic
composition of a population (Sadeghierad et al., 2010).
Reproduction is a process in which individual strings are selected according to their
tness. The tness is determined by calculating how well each string ts an objective
function. Copy strings according to their tness value implies that strings which t the
objective function well have a higher probability of contributing one or more offspring
in the next generation.
Crossover is a two-step process that involves mating and swapping of partial
strings. Each time the crossover operator takes action; two randomly selected strings
from the mating pool are mated. Then, in the case of simple crossover, a position along
one string is selected at random, and all binary digits following the position are
swapped with the second string. The result is two entirely new strings that move on to
the next generation.
Mutation follows crossover and protects against the loss of useful genetic
information. The operator works by randomly selecting one string and one bit location,
and changing that strings bit from 1 to a 0 or vice versa. The pseudo-code for the GA
algorithm is given in Elbeltagi et al. (2005).
4.2 Particle swarm optimization
PSO is a stochastic global optimization technique which uses swarming behaviors
observed in ocks of birds, schools of sh or swarms of bees, which the intelligence
is emerged. It was developed in 1995 by James Kennedy and Russell Eberhart and
uses a number of particles that constitute a swarm moving around in an
N-dimensional search space looking for the best solution (Kennedy and Eberhart,
1995; Eberhart and Shi, 2000). Each particle keeps track of its coordinates in the
solution space which are associated with the best solution that has achieved so far
by that particle is called as personal best position (Xpbest) and the another best
value obtained so far by any particle in the neighborhood of that particle is called as
global best position (Xgbest). Each particle tries to modify its position using the
concept of velocity.
In a physical N-dimensional search space, the position and velocity of individual i are
represented as the vectors X
i
[X
i, 1
, X
i, 2
. . . X
i, N
] and V
i
[V
i, 1
, V
i, 2
. . . V
i, N
],
repectively, in the PSOalgorithm. Let Xpbest
i
(X
i, 1
Pbest
, . . . , X
i, N
Pbest
)
, and Xgbest
i
(X
i, 1
gbest
, . . . X
i, N
gbest
), respectively, be the best position of individual i and its neighbors
COMPEL
31,2
536
best position so far. Using the information, the updated velocity of individual i is
modied under the following equation in the PSO algorithm:
V
k1
i;n
W V
k
i;n
C
1
rand
1
Xpbest
k
i;n
2X
k
i;n
_ _
C
2
rand
2
Xgbest
k
n
2X
i;
k
n
_ _
i 1; 2; . . . ; S n 1; 2; . . . ; N
13
Each individual moves from the current position to the next one by the modied
velocity using the following equation:
X
k1
i;n
X
k
i;n
V
k1
i;n
X
min;i;n
# X
k1
i
# X
maxi;n
X
min;i;n
if X
k1
i;n
, X
maxi;n
X
min;i;n
if X
k1
i;n
. X
max;i;n
14
The inertia weight factor (IWA) is a linearly decreasing dynamic parameter framework
(Chaturvedi et al., 2009; Chung and Tsu, 2010) descending from W
max
to W
min
to
enhance the convergence characteristics as follows:
W W
max
2W
max
2W
min
K
K
max
15
The pseudo-code for the PSO algorithm is given in Elbeltagi et al. (2005).
4.3 BFO with adaptive chemotaxis
BFO is a newly introduced evolutionary optimization algorithm that mimics the
foraging behavior of Escherichia coli (commonly referred to as E. coli ) bacteria. BFOwas
rst introduced by Passino (2002). BFO models the movement of E. coli bacterium
moves using a pattern of two types of movements: tumbling and swimming. Tumbling
refers to a randomchange in the direction of movement, and swimming refers to moving
in a straight line in a given direction. A bacterium in a neutral medium alternates
between tumbling and swimming movements.
Suppose it is desired to search for the position Xin an N-dimensional space. Let X
i
be
the initial position of bacterium i in the search space, i 1, 2, . . . , S, where S is the
number of bacteria. In biological bacteria populations, S can be as high as 10
9
and N is
three. Let F (X
i
) represent an objective function. Let F (X
i
) , 0, F (X
i
) 0, and F (X
i
) . 0
represent the bacterium at location X
i
in nutrient rich, neutral, and noxious
environments, respectively. Chemotaxis is a foraging behavior that captures the
process of optimization, where bacteria to climb up the nutrient concentration gradient.
The chemotactic step size parameter is the main factor for controlling the search ability
of the BFO. The BFO algorithm with xed step size suffers from the following
problems:
.
If the step size is high, then the bacteria reach to optimum value quickly but
accuracy of optimum value gets low.
.
If the step size is very small, then it requires many generations to reach optimum
solution. It may not achieve global optima with less number of generations.
Bio-inspired
optimization
algorithms
537
From these perspectives, balancing the exploration and exploitation of the search could
be achieved by adjusting the step size parameter. In this paper, an adaptive step size to
perform the swim walk instead of the constant step is proposed. This function is
expressed as follows (Sakthivel et al., 2010a):
Ci
jFX
i
j
jFX
i
j a
1
1 a=jFX
i
j
16
where:
a positive constant.
FX
i
objective function of the ith bacterium.
Ci variable step size of ith bacterium.
The bacterium i at position X
i
takes a chemotactic step j with the step size C(i) and
evaluates itself for objective function F (X
i)
at each step. If at position X
i
( j 1), the
objective value F is better than at position X
i
( j), then another step of same size C(i) in the
same direction will be taken again, if that step resulted in a position with a better value
than at the previous step. This is referred to as a swimming step. Swimming is continued
until for a maximumnumber of steps N
s
. After N
c
chemotactic steps, a reproductionsteps
is taken in which the population is sorted in ascending order of the objective function
value F and least healthy bacteria are replaced by the copies of the healthier bacteria.
After N
re
reproduction steps, an elimination-dispersal step is taken. Here, a bacteriumis
eliminated and a newbacteriumis created at a randomlocation in the search space with
probability r
ed
. The optimization stops after N
ed
elimination-dispersal steps.
Bacteria create swarms by means of cell-to-cell signaling via an attractant and a
repellant. Cell-to-cell attraction for bacterium i is represented with F
cc
(X
g
, X
i
), i 1,
2, . . . , S. This is dened as follows:
F
cc
X
g
; X
i
S
i1
d
attract
exp 2v
attract
N
n1
X
g
2X
i
2
_ _ _ _
S
i1
h
repellant
exp 2v
repellant
N
n1
X
g
2X
i
2
_ _ _ _ 17
The cell-to-cell signaling F
cc
( ) helps cells to move toward other cells, but not very close
to them. In BFO algorithm, the maximum number of objective function evaluations is
given by:
N
evl
S N
c
N
s
N
re
N
ed
The detailed pseudo-code for BFO algorithm is given in the Appendix.
5. Application of bio-inspired algorithms to parameter determination of
induction motor
Bio-inspired algorithms are used to search the optimal equivalent circuit parameters of
single- and double-cage rotor models of the induction motor. The procedure of the
proposed parameter determination method is as follows:
COMPEL
31,2
538
.
Step 1. Input the manufacturer data of the motor.
.
Step 2. Initialize the parameters of a bio-inspired algorithm (GA, PSO or BFO).
.
Step 3. Specify the lower and upper boundaries of the equivalent circuit
parameters.
.
Step 4. Generate the positions of equivalent circuit parameters randomly for a
population.
.
Step 5. Evaluate the objective value using equation (5) or (12).
.
Step 6. Apply the bio-inspired algorithm to nd a set of equivalent circuit
parameters of the motor which minimize the error between the
determined and the manufacturer data.
.
Step 7. Output the optimal equivalent circuit parameters corresponding to the
overall best solution.
.
Step 8. Compute the starting, maximum and full-load torques of the motor and
compare with the corresponding values supplied by the manufacturer.
6. Simulation results and analysis
In order to compare the performances of the bio-inspired algorithms, GA, PSOand BFO
in solving the parameter determination problem, two sample motors (5 and 40 hp) are
considered. The specications of the motors are given in Table I. To validate the
parameter determination results obtained by the bio-inspired algorithms, the equivalent
circuit parameters are also determined from no-load, locked-rotor and stator resistance
tests (IEEE Std. 112-F method). Owing to the randomness of the bio-inspired
algorithms, their performances cannot be decided by the results of a single run. Many
trials should be made to reach a valid conclusion about the performance of the
algorithms. In this paper, 20 independent runs have been carried out. All the algorithms
are implemented in MATLAB 7.0 and employed on an Intel Core 2 Duo, 3 GHz personal
computer with 2 GB RAM.
6.1 Parameters selection for bio-inspired algorithms
To compare the performance of GA, PSO and BFO algorithms, the population size and
number of generations are set as 20 and 200, respectively, for all the three algorithms. In
GA, the cross over probability and mutation rate are taken as 0.8 and 0.05, respectively,
as these values found to give the best results.
Specications Motor 1 Motor 2
Capacity (hp) 5 40
Voltage (V) 400 400
Current (A) 8 45
Frequency (Hz) 50 50
No. of poles 4 4
Full-load slip 0.07 0.09
Full-load torque (Nm) 25 190
Full-load efciency (%) 88 90
Table I.
Specications
of the test motors
Bio-inspired
optimization
algorithms
539
Selection of the acceleration constants and inertia weight factor considerably affect the
performance of the PSO algorithm. The acceleration constants, C
1
and C
2
are set as 2.
These values are experimentally determined as the best choice and also commonly used.
The concept of time varying inertial weight is implemented and W is varied from 0.9 to
0.4 using equation (14).
For the implementation of BFO algorithm, several parameters such as
population size (S), maximum number of chemotactic steps (N
c
), maximum
number of swims (N
s
), maximum number of reproduction steps (N
re
), maximum
number of elimination-dispersal events (N
ed
), depth of attractant (d
attract
), width of
attractant (v
attract
), height of repellent (h
repellant
) and width of repellent (v
repellant
) are
initialized.
If the parameters S, N
c
, N
s
, N
re
and N
ed
are too small, the algorithm may converge
prematurely; however, larger values of the parameters increase the computational
complexity and also affect the solution quality. A general biologically inspired
thumb-of-rule for choosing the parameters of BFO is: N
c
. N
re
. N
ed
(Das et al., 2008).
The BFO algorithm is applied repetitively by considering different values of BFO key
parameters such as N
c
, N
s
, N
re
and N
ed
. In the implementation, N
c
, N
s
, N
re
and N
ed
are
selected as 10, 5, 4 and 2, respectively. To choose the parameters of swarming, the
algorithm is run for different values of d
attract
, v
attract
, h
repellant
and v
repellant
. It was
found that these values are chosen as 0.1, 0.2, 0.1 and 10, respectively, give the fastest
convergence.
6.2 Case studies
6.2.1 Case study 1 (single-cage rotor model). GA, PSO and BFO algorithms are used to
determine the optimal equivalent circuit parameters, R
1
, R
c
, X
1
, X
m
, X
2
and R
2
of the
single-cage model. The equivalent circuit parameters obtained from various methods
are presented in Table II for the two motors. The efcacy of the proposed methods of
determining the motor parameters are appraised by calculating the starting, maximum
and full-load torques through its equivalent circuit parameters and comparing them
with the corresponding manufacturer data. The percentage error in the determined
values is dened as:
e%
X
d
2X
mf
X
m:f
100 18
Tables III and IV summarize the torque values and errors obtained from classical, GA,
PSO and BFO methods. For both the motors, the single-cage rotor model produces
less errors in maximum torque and full-load torque, and large errors in starting torque.
Motor 1 Motor 2
Parameters IEEE Std. 112-F GA PSO BFO IEEE Std. 112-F GA PSO BFO
R
1
2.67 2.69 2.63 2.09 0.0153 0.0188 0.011 0.02
R
2
7.4 6.01 5.87 5.844 0.417 0.4396 0.435 0.438
X
1
, X
2
14.82 14.8 13.42 15.45 0.592 0.609 0.622 0.6
R
c
1,245 1,268 1,307 1,238 190 213.68 220.4 205
X
m
310 297.48 244.26 206 10.5 11.35 9.83 10.54
Table II.
Summary of single-cage
rotor model parameters
for motors 1 and 2
COMPEL
31,2
540
I
E
E
E
S
t
d
.
1
1
2
-
F
G
A
P
S
O
B
F
O
T
o
r
q
u
e
M
a
n
u
f
a
c
t
u
r
e
r
d
a
t
a
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
T
s
t
1
5
1
5
.
8
9
5
.
9
3
1
7
.
0
7
1
3
.
8
1
6
.
3
9
9
.
2
7
1
6
.
0
5
7
.
0
T
m
a
x
4
2
3
7
.
1
2
2
1
1
.
6
2
4
2
.
2
2
0
.
5
2
4
4
1
.
4
2
2
1
.
3
8
4
1
.
1
1
2
2
.
1
2
T
f
u
l
l
2
5
2
2
.
7
2
9
.
2
2
7
.
2
1
8
.
8
4
2
6
.
4
2
5
.
6
8
2
3
.
9
2
2
4
.
3
2
Table III.
Comparison of
bio-inspired algorithms
and IEEE Std. 112-F
method results with
manufacturer data for
single-cage rotor model of
motor 1
Bio-inspired
optimization
algorithms
541
I
E
E
E
S
t
d
.
1
1
2
-
F
G
A
P
S
O
B
F
O
T
o
r
q
u
e
M
a
n
u
f
a
c
t
u
r
e
r
d
a
t
a
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
T
s
t
2
6
0
2
5
1
.
3
5
2
3
.
3
3
2
3
7
.
9
8
2
8
.
4
7
2
3
9
.
2
2
8
.
0
7
2
4
2
.
8
4
2
6
.
6
T
m
a
x
3
7
0
3
9
1
.
0
9
5
.
7
3
6
9
.
3
7
2
0
.
1
7
3
6
7
.
6
3
2
0
.
6
4
3
7
1
.
2
2
0
.
3
3
T
f
u
l
l
1
9
0
1
8
4
.
5
9
2
2
.
8
5
1
8
4
.
1
2
3
.
1
1
1
8
4
.
2
6
2
3
.
0
2
1
9
3
.
4
5
1
.
8
2
Table IV.
Comparison of
bio-inspired algorithms
and IEEE Std. 112-F
method results with
manufacturer data for
single-cage rotor model of
motor 2
COMPEL
31,2
542
Furthermore, the BFO algorithm yields lesser error than those produced by GA and
PSO methods.
6.2.2 Case study 2 (double-cage rotor model). Case study 2 deals with the parameter
determination of double-cage rotor model. Here, GA, PSO and BFO are used to
determine the seven optimal equivalent circuit parameters, R
1
, X
1
, X
m
, R
21
, R
22
, X
21
and
X
22
of the double-cage rotor model. Table V presents the equivalent circuit parameters
achieved by the bio-inspired algorithms. The torque values and errors of various
methods are given in Tables VI and VII. It can be seen that the magnitudes of torque
errors obtained by the BFO method is smaller than those using the GA and PSO
methods. On the whole, it is clear that the parameter determination of double-cage rotor
model by BFO method gave better results. The torque versus slip characteristics
obtained fromthe various methods for the test motors are shown in Figures 3 and 4. It is
obvious that the torque-slip characteristic of the motors obtained by the double-cage
model parameters of the BFO algorithm shows good agreement with the manufacturer
data for the entire slip range. Thus, the double-cage rotor model can be used as an
accurate model in the parameter determination problem over a wide speed range.
6.3 Comparison of bio-inspired algorithms
6.3.1 Convergence characteristics. To test the convergence behavior of the bio-inspired
algorithms, the convergence test is carried out by employing same number of function
evaluations. The convergence characteristics of the different algorithms are shown in
Figure 5 from which it is evident that the BFO and PSO algorithms converge quickly
than the GA algorithm.
6.3.2 Robustness. To verify the robustness of the bio-inspired algorithms, 20
independent runs are executed. The standard deviation of objective values obtained by
each algorithm is given in Table VIII. It can be observed that the standard deviation
produced by BFO is the least as compared to other algorithms, emphasizing the
robustness of the method.
6.3.3 Computational efciency. The computational efciencies of all the algorithms
are compared based on the average central processing unit (CPU) time to converge the
solution. The CPUtime taken by each algorithmis depicted in Table IX. FromTable IX,
it is clear that the average convergence time taken by the BFO is minimum.
7. Conclusion and future work
Bio-inspired optimization algorithms GA, PSOand BFOwith adaptive chemotactic step
have been developed in this paper for determining the equivalent circuit parameters of
Motor 1 Motor 2
Parameters GA PSO BFO GA PSO BFO
R
1
1.93 2.05 1.94 0.0234 0.0228 0.0244
X
1
13.72 15.65 15.93 0.588 0.597 0.59
X
m
205.04 332.23 386 10.15 12.515 12.18
R
21
0.436 0.842 0.72 0.276 0.218 0.23
R
22
5.678 5.86 5.94 0.462 0.449 0.437
X
21
53.2 48.34 68.3 1.162 1.23 1.06
X
22
15.07 14.65 16.93 0.726 0.674 0.681
Table V.
Summary of double-cage
rotor model parameters
for motors 1 and 2
Bio-inspired
optimization
algorithms
543
I
E
E
E
S
t
d
.
1
1
2
-
F
G
A
P
S
O
B
F
O
T
o
r
q
u
e
M
a
n
u
f
a
c
t
u
r
e
r
d
a
t
a
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
T
s
t
1
5
1
5
.
8
9
5
.
9
3
1
5
.
7
2
4
.
8
1
5
.
6
8
4
.
5
3
1
5
.
4
2
.
6
7
T
m
a
x
4
2
3
7
.
1
2
2
1
1
.
6
2
4
1
.
4
5
2
1
.
3
1
4
1
.
6
3
2
0
.
8
8
4
1
.
6
5
2
0
.
8
3
3
T
f
u
l
l
2
5
2
2
.
7
2
9
.
2
2
6
.
8
6
7
.
4
4
2
3
.
7
7
2
4
.
9
2
2
5
.
5
2
2
.
0
8
Table VI.
Comparison of
bio-inspired algorithms
and IEEE Std. 112-F
method results with
manufacturer data for
double-cage rotor model
of motor 1
COMPEL
31,2
544
I
E
E
E
S
t
d
.
1
1
2
-
F
G
A
P
S
O
B
F
O
T
o
r
q
u
e
M
a
n
u
f
a
c
t
u
r
e
r
d
a
t
a
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
D
e
t
e
r
m
i
n
e
d
d
a
t
a
E
r
r
o
r
(
%
)
T
s
t
2
6
0
2
5
1
.
3
5
2
3
.
3
3
2
5
1
.
2
1
2
3
.
3
8
2
5
2
.
4
3
2
2
.
9
1
2
5
6
.
2
2
1
.
4
6
T
m
a
x
3
7
0
3
9
1
.
0
9
5
.
7
3
7
8
.
0
6
2
.
1
8
3
8
0
.
0
3
2
.
7
1
3
8
1
.
7
3
3
.
1
7
T
f
u
l
l
1
9
0
1
8
4
.
5
9
2
2
.
8
5
1
8
2
.
4
7
2
3
.
9
6
1
9
5
.
2
6
2
.
7
7
1
8
9
.
8
5
2
0
.
0
8
Table VII.
Comparison of
bio-inspired algorithms
and IEEE Std. 112-F
method results with
manufacturer data for
double-cage rotor model
of motor 2
Bio-inspired
optimization
algorithms
545
three-phase induction motor frommanufacturer data. The bio-inspired algorithms have
been tested on two test motors. Comparing with the IEEE Std. 112-F method, the
bio-inspired algorithms-based parameter determination of double-cage model provides
lesser deviation between the determined and the manufacturer data. Besides, BFO
algorithmoutperforms GAand PSOmethods in terms of robustness and computational
efciency. Most signicantly, the equivalent circuit parameters of the motor are
determined only from the manufacturer data. The parameter determination methods
described in this paper are faster and less intrusive than the IEEE Std. 112-F method,
and also constitute a promising tool in analyzing the stalling and/or reacceleration
process of a loaded motor following a fault or during voltage sag condition as well as in
system-level studies. The mathematical model improvement of single-cage rotor and
the parameter determination of dynamic model of an induction motor using the
biologically inspired algorithms will be considered in the future work.
Figure 3.
Torque versus slip
characteristics of 5-hp
motor obtained from IEEE
Std. 112-F, GA, PSO and
BFO methods
50
40
30
T
o
r
q
u
e
(
N
m
)
20
10
0
0 0.2 0.4 0.6
Slip
0.8 1
Manufacturer data
IEEE Std. 112 method
GA method for single cage rotor model
PSO method for single cage rotor model
PSO method for double cage rotor model
BFO method for double cage rotor model
BFO method for single cage rotor model
GA method for double cage rotor model
Figure 4.
Torque versus slip
characteristics of
40-hp motor obtained
from IEEE Std. 112-F,
GA, PSO and BFO
methods
500
400
300
200
100
0
T
o
r
q
u
e
(
N
m
)
0 0.2 0.4 0.6
Slip
0.8 1
Manufacturer data
IEEE Std. 112 method
GA method for single cage rotor model
PSO method for single cage rotor model
PSO method for double cage rotor model
BFO method for double cage rotor model
BFO method for single cage rotor model
GA method for double cage rotor model
COMPEL
31,2
546
References
Abdelhadi, B., Benoudjit, A. and Nait Said, N. (2004), Identication of induction machine
parameters using a new adaptive genetic algorithm, Electric Power Components and
Systems, Vol. 32 No. 8, pp. 767-84.
Alonge, F., Dippolito, F., Ferrante, G. and Raimondi, F.M. (1998), Parameter identication of
induction motor model using genetic algorithms, IEE Proceedings Control Theory
& Applications, Vol. 145 No. 6, pp. 587-93.
Ansuj, S., Shokooh, F. and Schinzinger, R. (1989), Parameter estimation for induction machines
based on sensitivity analysis, IEEE Transactions on Industry Applications, Vol. 25 No. 6,
pp. 1035-40.
Attaianese, C., Nardi, V. and Tomasso, G. (2002), Self-commissioning of induction motors fed by
VSI, Electric Power Components and Systems, Vol. 30 No. 1, pp. 1-9.
Awadallah, M.A. (2008), Parameter estimation of induction machines from nameplate data
using particle swarm optimization and genetic algorithm techniques, Electric Power
Components and Systems, Vol. 36 No. 8, pp. 801-14.
Motor 1 Motor 2
Methods Case 1 Case 2 Case 1 Case 2
GA 0.0198 0.0132 0.0014 0.0078
PSO 0.0016 0.0013 0.0020 0.0015
BFO 0.0010 0.0008 0.0003 0.00025
Table VIII.
Comparison of standard
deviation results
Methods Case 1 Case 2
GA 2.8 3.1
PSO 2.54 2.87
BFO 2.36 2.64
Table IX.
Comparison of
computation time in
seconds of bio-inspired
algorithms for motor 1
Figure 5.
Convergence
characteristics of
bio-inspired algorithms
for motor 1
0.64
GA method
PSO method
BFO method
0.54
0.44
0.34
0.24
0.14
0.04
0.06
0 200 400 600
Number of Evaluation
O
b
j
e
c
t
i
v
e
v
a
l
u
e
800 1,000
Bio-inspired
optimization
algorithms
547
Bae, D. (1997), Determination of induction motor parameters by using neural network based on
FEM results, IEEE Transactions on Magnetics, Vol. 33 No. 2, pp. 1924-7.
Belmans, R., Verdyck, D., Geysen, W., Findlay, R.D., Szabados, B., Spenser, S. and Lie, S. (1992),
Magnetic eld analysis in squirrel cage induction motors, IEEE Transactions on
Magnetics, Vol. 28 No. 2, pp. 1367-70.
Bishop, R.R. and Richards, G.G. (1990), Identifying induction machine parameters using a
genetic optimization algorithm, Proceedings of IEEE Southeastcon-90, New Orleans, LA,
USA, April 01-04, Vol. 2, pp. 476-9.
Chaturvedi, K.T., Pandit, M. and Srivastava, L. (2009), Particle swarm optimization with time
varying acceleration coefcients for non-convex economic power dispatch, Electrical
Power and Energy Systems, Vol. 31 No. 6, pp. 249-57.
Chung, Y.D. and Tsu, P.H. (2010), A particle swarm optimization for solving joint pricing and
lot-sizing problem with uctuating demand and unit purchasing cost, Computers and
Mathematics with Applications, Vol. 60 No. 7, pp. 1895-907.
Corcoles, F., Pedra, J., Salichs, M. and Sainz, L. (2002), Analysis of the induction machine
parameter identication, IEEE Transactions on Energy Conversion, Vol. 17 No. 2,
pp. 183-90.
Das, T.K., Venayagamoorthy, G.K. and Aliyu, U.O. (2008), Bio-inspired algorithms for the
design of multiple optimal power system stabilizers: SPPSO and BFA, IEEE
Transactions on Industrial Applications, Vol. 44 No. 5, pp. 1445-57.
De Kock, J.A., Van der Merwe, F.S. and Vermeulen, H.J. (1994), Induction machine parameter
estimation through an output error technique, IEEE Transactions on Energy Conversion,
Vol. 9 No. 1, pp. 69-76.
Dolinar, D., Stumberger, G. and Grear, B. (1998), Calculation of the linear induction motor model
parameters using nite elements, IEEE Transactions on Magnetics, Vol. 34 No. 5,
pp. 3640-3.
Eberhart, R.C. and Shi, Y. (2000), Comparing inertia weights and constriction factors in particle
swarm optimization, Proceedings of the IEEE International Congress Evolutionary
Computation, La Jolla, CA, USA, Vol. 1, pp. 84-8.
Elbeltagi, E., Hegazy, T. and Grierson, D. (2005), Comparison among ve evolutionary-based
optimization algorithms, Advanced Engineering Informatics, Vol. 19 No. 1, pp. 43-53.
Filho, E.B.S., Lima, A.M.N. and Jacobina, C.B. (1991), Parameter estimation for induction
machines via non-linear least squares method, Proceedings of the International
Conference on Industrial Electronics, Control and Instrumentation IECON 91., Kobe,
Japan, October 28-November 1, pp. 639-43.
Goldberg, D.E. (1989), Genetic Algorithms in Search Optimization and Machine Learning,
Addison-Wesley, Reading, MA.
Huang, K.S., Kent, W., Wu, Q.H. and Turner, D.R. (2001), Parameter identication for induction
motors using genetic algorithm with improved mathematical model, Electric Power
Components and Systems, Vol. 29 No. 3, pp. 247-58.
Hung, R. and Dommel, H.W. (1996), Synchronous machine models for simulation of induction
motor transients, IEEE Transactions on Power Systems, Vol. 11 No. 2, pp. 833-8.
IEEE Power Engineering Society (2004), IEEE standard test procedure for polyphase induction
motors and generators, IEEE Std. 112, IEEE Power Engineering Society, Austin, Tx.
Johnson, B.K. and Willis, J.R. (1991), Tailoring induction motor analytical models to t known
motor performance characteristics and satisfy particular study needs, IEEE Transactions
on Power Systems, Vol. 6 No. 3, pp. 965-9.
COMPEL
31,2
548
Ju, P., Handschin, E., Wei, Z.N. and Schluecking, U. (1996), Sequential parameter estimation of a
simplied induction motor load model, IEEE Transactions on Energy Conversion, Vol. 11
No. 1, pp. 319-24.
Kennedy, J. and Eberhart, R. (1995), Particle swarm optimization, Proceedings of the IEEE
Conference on Neural Networks ICNN95, Perth, Australia, Vol. 4, pp. 1942-8.
Lindenmeyer, D., Dommel, H.W., Moshref, A. and Kundur, P. (2001), An induction motor
parameter estimation method, Electrical Power and Energy Systems, Vol. 23 No. 4,
pp. 251-62.
Michael, T.W. and Ronald, G.H. (1995), Identication and control of induction machines using
articial neural networks, IEEE Transactions on Industry Applications, Vol. 31 No. 3,
pp. 612-9.
Mishra, S. (2005), A hybrid least square-fuzzy bacterial foraging strategy for harmonic
estimation, IEEE Transactions on Evolutionary Computation, Vol. 9 No. 1, pp. 61-73.
Moon, S.I. and Keyhani, A. (1994), Estimation of induction motor parameters from standstill
time-domain data, IEEE Transactions on Industry Applications, Vol. 30 No. 6, pp. 1609-15.
Nangsue, P., Pillay, P. and Conry, S. (1999), Evolutionary algorithms for induction motor
parameter determination, IEEE Transactions on Energy Conversion, Vol. 14 No. 3,
pp. 447-53.
Nikranajbar, A., Ebrahimi, M.K. and Wood, A.S. (2010), Parameter identication of a cage
induction motor using particle swarm optimization, Journal of Systems and Control
Engineering, Vol. 224 No. 5, pp. 479-91.
Nollan, R., Pillay, P. and Haque, T. (1994), Application of genetic algorithms to motor parameter
determination, Proceedings of 1994 IEEE-IAS Conference, Denvar, CO, USA, October
2-6, pp. 47-54.
Orlowska Kowalska, T., Lis, J. and Szabat, K. (2006), Identication of the induction motor
parameters using soft computing methods, COMPEL: International Journal of
Computation and Mathematics in Electrical and Electronics Engineering, Vol. 25 No. 1,
pp. 181-92.
Panigrahi, B.K. and Pandi, V.R. (2008), Bacterial foraging optimization: Nelder-Mead hybrid
algorithm for economic load dispatch, IET Generation, Transmission and Distribution,
Vol. 2 No. 4, pp. 556-65.
Passino, K.M. (2002), Biomimicry of bacterial foraging for distributed optimization and control,
IEEE Control Systems Magazine, Vol. 22 No. 3, pp. 52-67.
Pedra, J. and Corcoles, F. (2004), Estimation of induction motor double-cage model parameters
from manufacturer data, IEEE Transactions on Energy Conversion, Vol. 19 No. 2,
pp. 310-7.
Pedra, J. and Sainz, L. (2006), Parameter estimation of squirrel-cage induction motors without
torque measurements, IEE Proceedings Electric Power Applications, Vol. 153 No. 2,
pp. 263-9.
Pillay, P., Nollan, R. and Haque, T. (1997), Application of genetic algorithms to motor parameter
determination for transient torque calculations, IEEE Transactions on Industry
Applications, Vol. 33 No. 5, pp. 1273-82.
Rahimpour, E., Rashtchi, V. and Pesaran, M. (2007), Parameter identication of deep-bar
induction motors using genetic algorithm, Electrical Engineering, Vol. 89, pp. 547-52.
Sadeghierad, M., Darabi, A., Lesani, H. and Monsef, H. (2010), Optimal design of the generator of
micro turbine using genetic algorithm and PSO, International Journal of Electrical Power
& Energy Systems, Vol. 32 No. 7, pp. 804-8.
Bio-inspired
optimization
algorithms
549
Sakthivel, V.P., Bhuvaneswari, R. and Subramanian, S. (2010a), Design optimization of
three-phase energy efcient induction motor using adaptive bacterial foraging algorithm,
COMPEL: International Journal of Computation and Mathematics in Electrical and
Electronics Engineering, Vol. 29 No. 3, pp. 699-726.
Sakthivel, V.P., Bhuvaneswari, R. and Subramanian, S. (2010b), Multi-objective parameter
estimation of induction motor using particle swarm optimization, Engineering
Applications of Articial Intelligence: The International Journal of Intelligent Real-Time
Automation, Vol. 23 No. 3, pp. 302-12.
Sakthivel, V.P., Bhuvaneswari, R. and Subramanian, S. (2010c), Nonintrusive efciency
estimation method for energy auditing and management of in-service induction motor
using bacterial foraging algorithm, IET Electric Power Applications, Vol. 4 No. 8,
pp. 579-90.
Say, M.G. (1983), Alternating Current Machines, 4th ed., Pitman, Totowa, NJ.
Stephan, J., Bodson, M. and Chiasson, J. (1994), Real time estimation of induction motor
parameters, IEEE Transactions on Industrial Applications, Vol. 30 No. 3, pp. 746-59.
Tang, W.J., Li, M.S., He, S., Wu, Q.H. and Saunders, J.R. (2006), Optimal power ow with
dynamic loads using bacterial foraging algorithm, Proceedings of the International
Conference on Power System Technology (POWERCON), Chongqing, China, October
22-26, pp. 1-5.
Toliyat, H.A., Levi, E. and Raina, M. (2003), A review of RFO induction motor parameter
estimation techniques, IEEETransactions on Energy Conversion, Vol. 18 No. 3, pp. 271-83.
Ursem, R.K. and Vadstrup, P. (2003), Parameter identication of induction motors using
differential evolution, The 2993 Congress on Evolutionary Computation CEC 03, Vol. 2,
pp. 790-6.
Vahedi, H., Hosseini, S.H. and Noroozian, R. (2010), Bacterial foraging algorithm for security
constrained optimal power ow, Proceedings of the 7th International Conference on the
European Energy Market, Madrid, Spain, June 23-25, pp. 1-6.
Willis, J.R., Brock, G.J. and Edmonds, J.S. (1989), Derivation of induction motor models from
standstill frequency response tests, IEEE Transactions on Energy Conversion, Vol. 4
No. 4, pp. 608-15.
Appendix. Pseudo-code for BFO algorithm
1: Initialize S, N, N
c
, N
s
, N
re
, N
ed
, r
ed
, d
attract
, h
repellant
, v
attract
, v
repellant
, X
min
and X
max
2: Initialize X
i
randomly for i 1, 2, . . . , S
3: Initialize C(i) for i 1, 2, . . . ,S
4: Set the loops counters j, k and l to 0
5: //Elimination-Dispersal loop:
6: while l # N
ed
do
7: l l 1
8: //Reproduction loop:
9: while k # N
re
do
10: k k 1
11: //Chemotaxis loop:
12: while j # N
c
do
13: j j 1
14: for each bacterium i 1, 2, . . . , S do
15: Compute F(i, j, k, l)
16: Let F (i, j, k, l) F(i, j, k, l) F
cc
(X
g
, X
i
)
COMPEL
31,2
550
17: Let F
last
F (i, j, k, l)
18: //Tumble:
19: Generate a N-dimensional random vector D
m
(i), i 1, 2, . . . , N on [21, 1]
20: //Move:
21: Let X
i
j 1; k; l X
i
j; k; l Ci
Di
D
T
iDi
p
22: Compute F (i, j 1, k, l) with X
i
( j 1, k, l)
23: //Swim:
24: Let m 0
25: while m< N
s
do
26: Let m m 1
27: if F (i, j 1, k, l) < F
last
then
28: Let F
last
F (i, j 1, k, l)
29: Let X
i
j 1; k; l X
i
j 1; k; l Ci
Di
D
T
iDi
p
30: Use this X
i
( j 1, k, l) to compute new F (i, j 1, k, l)
31: else
32: m N
s
33: end if
34: end while
35: end for
36: end while
37: Compute for each bacterium i, for given k and l:
F
i
health
N
c
1
j1
Fi; j; k; l
38: Eliminate S
r
fraction of bacteria with highest F
health
and split the other bacteria into two at
their locations.
39: end while
40: For each bacterium, with probability r
ed
eliminate the bacterium and create a new one at a
random position.
41: end while
Corresponding author
V.P. Sakthivel can be contacted at: vp.sakthivel@yahoo.com
Bio-inspired
optimization
algorithms
551
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints