6th International Conference on Electrical Engineering 11- 13 October 2010

On-Line Losses Minimization of Induction Motor

Vector Control using Real Coded Genetic Algorithms
Z. Rouabah*, B. Abdelhadi* , F. Anayi* * , F.Zidani*
*Department of Electrical Engineering, Batna University, Algeria,
**School of Engineering, Cardiff University, Cardiff, UK.
Email: zinebroua@ieee.org, abdelhadi3b@yahoo.com, Anayi@cardiff.ac.uk, fati_zidani@lycos.com
Abstract This paper deals with on-line real coded
genetic algorithms optimization based on vector control
for loss minimization operation of induction motor.
Losses minimization is obtained by adjusting the
magnetizing current component with respect to the
torque current component to give minimum total copper
and iron losses. Some simulation results are obtained by
using both online real code genetic algorithm
optimization and minimal function which use the
derivative of the total losses of the motor. Those results
are compared with the field-oriented control method
without optimization. Results show that the efficiency of
induction motor can be improved under different
operational conditions especially in light load region.
Keywords Loss minimization, Efficiency, Field-Oriented
Control (FOC), Induction Motor (IM), RealCoded Genetic Algorithm (RCGA).

I. INTRODUCTION
Induction motors (IM) are imposing themselves as a
reliable and more economical choice in a large range
of applications. In industry they present an important
factor of control especially for autonomous electrical
traction; however the motor drive losses minimization
is important for two reasons: economical saving and
reduction of environmental pollution. The very
extensive use of induction motor implies that if losses
in IM drives can be reduced by just a few percent, it
will have a major impact on the total electrical energy
consumption [2-5], then each 1% improvement in
motor efficiency could result in savings of over $l
Billion per year in energy costs, 6-10 million tons
(5.4-9.1 million tonnes) less per year of combusted
coal and approximately 15-20 million tons (13.6-18.1
million tonnes) less carbon dioxide released into the
atmosphere [1].
In high dynamic performances, control schemes
used in industrial applications like vector control and
direct torque control, the flux is usually maintained
constant equal to its nominal value; in this situation
the induction motor runs efficiently around the
nominal operating point [2- 3]. When the load is
reduced considerably, the losses are greatly increased
and the electrical energy consumption is then highly
affected [3-6].
The key to solve this problem and to reduce the IM
losses is to obtain an optimum balance between
copper and iron losses.

Sixth International Conference on Electrical Engineering CEE 2010

This balance can be obtained by controlling the IM

magnetic flux.
Many approaches have been developed in order to
obtain a highly efficient IM drives. It is one of the
most attractive and active subjects in the field of
motion control. The techniques allowing efficiency
improvement can be divided into two categories. The
first category is so called loss-model based
approach; which consists of computing losses by
using the machine model and selecting a flux level
that minimizes theses losses. The second category is
the power measure based approach, also known as
search controllers (SCs),in which the flux is decreased
until the electrical input power settles down to the
lowest value for a given torque and speed [4-7, 9].
This task can be carried out by FOC scheme with
appropriate algorithm as described in the present
work. The used approach is based on real coded
genetic algorithm (RCGA) that is applied to minimize
IM losses in order to evaluate the optimal
magnetizing current, thus maximizing efficiency.
Note that the motor model is including core losses
which uses Razeks and Mendes series model.
Simulation results are compared with those obtained
by using the minimal function [15].
II. MOTOR LOSS MODEL
A. Induction Motor Model
When selecting the stator currents (isd, isq) and the
rotor fluxes (rd, rq) as state variables, the d-q model
for a three phase IM in the synchronous frame can be
written as [15]:

Vsd =R's isd +sLs

R
disd M dFrd
M
fs
+ -w sL i + Frd -w F (1)
s L rq
dt Lr dt s s sq Lr
r

Vsq =Rs' isq +sLs

disq M dFrq
Rfs
M
+ -wssLs isd + Frq -ws Frd (2)
dt Lr dt
Lr
Lr
Rr + R fr
+
L
r

dF
Frd -wrF + rd (3)
rq

dt

Rr + R fr

M
V = R'fr - isq +
rq
Lr
Tr

Where:

dF
Frq - wrFrd + rq (4)

dt

R s' = R

V = R'fr - i
rd
Tr sd

+ R

fs

+1

437

'
fr

= R

The mechanical losses can be neglected and the

motor torque can be expressed as:

sr
Lr
and
, Tr =
sr +1
Rr

fr

Losses of Induction Motor

The total losses of an IM consist of stator and rotor
copper losses, core losses Pfe and mechanical losses
Pm. In the steady state, the stator and rotor copper
losses are defined as follows:
2
Pjs = Rs | is2 | = Rs ( i2sd + isq
)

ir

Rr

Ploss =

(5)

Rr

(1 + s r ) 2

[ ( i m r - i s d ) 2 + i 2sq ]

sLs

III. LOSS MINIMIZING STRATEGY

Control Block Diagram

A simplified block diagram of the proposed
optimization procedure is depicted in Fig. 2. It
consists of loss minimization controller and it is
implemented in classical rotor flux oriented control
(FOC). In this scheme, two phase currents and the
rotor speed are measured in order to calculate the
electromagnetic torque and the magnetizing current
(i) which enables us to express the total motor losses.

Rs

k h w e F 2 + k e w e2 F 2 (7a)

The coefficients of hysteresis and eddy current

losses may be expressed as kh and ke respectively
which can be determined from standard no-load test
data [11].
They can also be written as from the equivalent circuit
of the IM in steady state:
Is

R
T
r
e

+ Rs +
(11)
2
p(1-s ) Li

1
+
s
s
m
(
)

Rs + Rfs im2

The procedure of loss minimization consists of

optimizing the value of the magnetizing current (i)
for a given load torque so as to minimize the total
motor losses in steady state operation.

(6)

The core losses including eddy current and

hysteresis losses are given by:

P fe =

Ir

Vs

d(q)PI

SpeedPI

Rr/g
w

Rfs /1+sr)

usqref

Isqref

wref

I
(1-s)Ls

(10)

Substituting (5), (6) and (7b) into (9) we obtain the

total losses:

P jr =

3 M
p
F r i sq
2 Lr

Te =

-M
.
M

usdref

IM

Inverter

Iopt
Isq
GA

3/2

Te

optim

Fig.1. Equivalent circuit of the IM

fe

= I

2
m

fs

(1 + s

)2

Where :
R fs = A f s + B f s2

(7c)
Where, A and B are constants and fs is stator current
frequency
As a reasonable approximation, the mechanical losses
are dependent on the rotor speed.

Pm = k m w r2

(8)

Where km is the mechanical loss coefficient.

As the stator currents isd and isq are regulated and the
motor is controlled to be field oriented to the rotor
flux, according to the following relation:

= (- M

L r ) i sq and i d r = 0
In steady state, the operating losses of the machine
can be expressed from (1) to (4) as follows:
iq r

P loss = Pjs + Pjr + Pfe + Pm

Fig.2. Block diagram of the Optimization Control System

(7b)

(9)

Sixth International Conference on Electrical Engineering CEE 2010

The operating structure of RGA block represented

on figure 2 is illustrated by the flow chart of Fig.3.
B. Improvement Efficiency Strategy
The philosophy of the controller efficiency
optimization aims to lets motor working at variable
flux which makes efficient conversion of the electrical
power within the motor drive more possible. This can
be achieved by finding the optimal value of the
magnetizing current that satisfies the criterion below:
Pl o s s

/ im

(12)
Then the solution of equation (12) returns the
optimal magnetizing current (flux) which improves
the motor efficiency.
1/ 2
i

= k

opt

(13)

438

The equation (13) is used to obtain the solution

which we call in this paper Fmin method
Initial Population

Tem
From the motor

Fitness
Losses Evaluation
for each flux value

Selection Probability calculating

Criteria Satisfied
(Generation=Max
Generation)

Application of the
Yes best chromosome
(Iopt) to the motor

To the
motor

No
Generation
+1

st
Selection to create 1 generation

Cross Over
Mutate

Classification of
Obtained Solution

Fig. 3. Efficiency optimization Flowchart by RCGA

Where:
k o p t = A / [ n p (1 - s ) L s ]1 / 2

A = [[(Rs + Rr ) / (sr +1)2 +sr Rfs / (1+sr )]/ (Rs + Rfs )]1/2
As one can see this method can minimize losses in
steady state only.
IV.

RCGA OPTIMIZATION PROCEDURE

The genetic algorithm was introduced by J. Holland

during the 1960s. It is known as a stochastic
searching algorithm inspired by principle of the
natural evolution of species and capable to solving
non- smooth, non-continuous and non- differentiable
problems for parallel computation to find global or
near global optimal solutions, [10-12].
This optimization procedure consists of searching
the optimum magnetizing current (flux) value for a
given load torque by relying on genetic algorithm.
The latter is defined as stochastic optimization
technique based on the genetic natural evolution
mechanism of creative beings, [7-8]. Such algorithm
is found to be a powerful computational tool in
seeking optimums and is considered as the most upto-date product of artificial intelligence techniques
that emulate the mechanics of natural selection and
genetics. It explores, with coding parameter set, the
workspace by means of mechanism of reproduction,
with the target of optimizing the process selection.
This mechanism comprises selection, crossover and
mutation operations, [10-11, 13].

Sixth International Conference on Electrical Engineering CEE 2010

For this application, the real coded genetic

algorithms are used; they have many advantages in
numerical optimization over binary encoding.
Efficiency of the RGA is increased as there is no need
to convert chromosomes to phenotypes before each
fitness evaluation, less memory is required and there
is no loss in precision by the conversion between
binary and real values [14].
The application of this approach requires the
introduction of an objective function which evaluates
how good the fitted values of the magnetizing current
are. From this function, a fitness that controls the
reproduction process is derived.
The criterion to select the best individuals for
reproduction is the objective (fitness) function. By
proceeding in this way, the objective function adopted
for this problem is the IM total losses given by
equation (11). Each generation is subjected to the
crossover and mutation mechanisms. The crossover
consists in selecting two parent chromosomes
randomly in order to generate two new individuals.
The mutation follows crossover and is used to insure
that useful genetic data that could have been
disregarded during crossover is not permanently lost.
The procedure of real coded genetic algorithm is
outlined as follows:
1. Initial generation: the RCGA begins by randomly
N individuals inside certain range and forms initial
generation.
2. Fitness evaluation: every individuals fitness is
calculated according to the fitness function (eq. (11)).
3. Reproduction: parent individuals are sorted from
big fitness function value to small one, and
excellent individuals from the headmost are
directly passed to offspring generation and the rest
is put in matching pool.
4. Crossover and Mutation: Those operations are
finished in matching pool, and the generated
individuals are sent to offspring generation.
5. Iteration: the RGA runs iteratively repeating the
procedures 2-4 until population convergence
condition is met or the maximum number of
iteration is reached.
V.

SIMULATION RESULTS

The simulation of efficiency optimization of IM

using genetic algorithms is performed by using
Matlab/Simulink software; the results of different
cases are given below.
The nameplate-rated characteristics of the used
induction machine are shown in table.1 in the
appendix.
The flux must be reduced according to the
efficiency improvement strategy. Its obvious on
figures 5 and 7 that the flux is decreasing when we
switch from the case without optimization (FOC) to
the case with optimization (RCGA). Furthermore, one
can note that the using of RCGA provide the best
optimal value of the flux. This decreasing flux has a
439

w=157rad/s ; Tl=7Nm

1
RCGA

Fmin

FOC

Efficiency (pu)

0.8

w=157rad/s; Tl=0.25T

1.5

F lu xe s (w b )

direct influence on the efficiency which is improved.

At each operating point, the optimal efficiency is
reached when using the optimization algorithm as we
can see on figures 4 and 6. A definite improvement in
efficiency in the transient region is noted as well.

FOC
switchingfrom
Fmin
to RCGA

0.5
switchingfrom
FOC
toFmin

Fmin

RCGA

0.6
0.4

-0.5
0

3
Time (s)

Fig. 7 .Motor fluxes for light load

0.2

Tl=1. 75T

1 60

3
Time(s)

1 40

6
S pe e d (rad / s)

0
0

Fig. 4. Motor efficiency for nominal Load

w=157 rad/s; Tl=7Nm

1.5

1 00
80
60
40

1
FOC

Fmin

0.5

0.5

1 .5
Tim e (s)

2.5

Fig.8. Speed of the motor

Figures 9, 10 and 11 show and confirm that there is

a great efficiency improvement when using the
RCGA optimization algorithm compared with the non
optimization case (FOC), and we can see that the two
optimization methods give very close results at light
loads as seen on the zooms . Also we can note that the
motor works at its optimal efficiency at different load
levels when the optimization algorithm is introduced
in the command.
For the case of low speed as seen on Fig.9, the
motor efficiency is improved only over the light load
region when using real coded
genetic algorithms
w = 5 0 ra d /s

Fird
Firq

-0.5
0

3
Time (s)

Fig. 5. Motor fluxes for nominal load

w=157rad/s; Tl=0.25T

1
0.8
0.6

0.8

Switch from
Fmin to
RCGA

0.7

0.4

E ffic ie n c y( p u )

E ffic ie n c y (p u )

Fmi n
FOC
GA' s

20

RCGA

0.2
Switch from
FOC toFmin

0
0

0.6
0.5
0.4
0.3
w=
5 0a
r /d s

0.5

3
Time (s)

Fig.6. Motor efficiency for light load

Sixth International Conference on Electrical Engineering CEE 2010

0.4

0.1
0

Figure 8 shows the motor speed when using different

methods (FOC, Fmin and RCGA); we can see the
speed is not affected in all cases.

Z oo m

0.2

E f ci ei ncy ( pu)

F lu x e s (w b )

1 20

0.3
0.2
0.1
00.04 0.0 6 0.0 8

0.5

.01 0.12 0. 14 0. 16 0. 18
Loa d(pu)

Lo ad ( pu )

0.2

RCG A
Fm in
FO C
1 .5

Fig. 9 Motor efficiency for low speed

440

Table II. RATING AND PARAMETERS OF INDUCTION M OTOR

Power
P
1.1
kW
Number of pairs of poles
p
2
Stator resistance
Rs
8
W
Rotor resistance
Rr
3.1
W
Total leakage factor

0.12
Mutual inductance
M
0.443 H
Stator and rotor self inductance
Ls = Lr 0.47
H
Inertia
J
0.06
SI
Viscous friction coefficient
f
0.00
SI
Rated Load torque
TL
7
N.m

VIII.

Fig.10 Motor efficiency for high speed

w=157 rad/s

0.8
0.7

Efficiency (pu)

0.6
0.5
0.4
0.3

w=157 r ad/s

0.7
0.6
E f i cie ncy (pu)

0.2
Zoom

0.1

FOC
Fmi n
RCGA

0.5

FOC
Fmin
RCGA

0.4
0.3
0.2
0.1

0
0

0.5

0
0.04 0.06 0.08

0.1

0.1 2 0. 14 0. 16 0. 18
Lo ad(pu)

0.2

1.5

Load (pu)

Fig. 11 Motor efficiency for nominal speed

VI.

CONCLUSION

In this paper an on-line efficiency optimization

algorithm based on real-coded genetic algorithm is
proposed to be implemented for induction motor
drives. The advantages of using real coded genetic
algorithms have been highlighted. Extensive
simulation results of proposed and conventional FOC
methods are investigated and compared and the
superiority of the former is confirmed.
The yielding simulation results show immediate
improvement in efficiency when the on-line RCGA
algorithm is applied with the IM slightly loaded that
usually suffers of low efficiency values.
VII.

APPENDIXE

Table I.
GENETIC ALGORITHMS PARAMETERS
Parameters
Population size
Crossoer probability
Mutate probability
Maximum generation number

Values
20
0.75
0.01
20

Sixth International Conference on Electrical Engineering CEE 2010

REFERENCES

[1] L. Ramesh and all, Efficiency Optimization of

Induction Motor Using a Fuzzy Logic Based Optimum
Flux Search Controller, IEEE International
Conference on Power Electronics Drives and Energy
Systems (PEDES06), 12-15 Dec. 2006, pp.1-6.
[2] Branko Blanusa and all, An Improved Search Based
Algorithm for Efficiency Optimization in the Induction
Motor Drives, XLII Konferencija- za ETRAN, HercyNovi, 2003.
[3] D. S. Kirischen, D. W. Novoty and T. A. Lipo,
Optimal Efficiency Control of an Induction Motor
Drive, IEEE Trans. On Energy Conversion,
Vol. EC-2, N 1, March 1987, pp. 70-76.
[4] S. Ghozzi and all, Energy Optimization of Induction
Motor Drives, IEEE International Conference on
Industrial Technology (ICIT), 8-10 Dec. 2004, Vol. 2,
pp. 602-610.
[5] B. K. Bose, Power Electronics and AC Drives, Prentice
Hall, USA, 2002.
[6] E. Poirier and all, Loss Minimization Control of
Induction Motor Drives Based on Genetic Algorithms,
Electric Machines and Drives, 2001 IEEE International
Conference, 2001, pp. 475-478.
[7] Jinchuan. Li and all, A new Optimization Method on
Vector Control of Induction Motors, Electric
Machines and Drives, 2005 IEEE International
Conference, 15-18 May 2005, pp. 1995-2001.
[8] G. B. Sheble and K.Briting, Refined Genetic
Algorithm- Economic Dispatch Example, IEEE Trans.
On Power Systems, Vol. 10 N 1, Feb 1995, pp. 117-124.
[9] D. E. Golberg, Genetic Algorithms in Search,
Optimization and Machine Learning Reading, MA:
Addisonn-Wisley, USA, 1989.
[10] J.M.D. Murphy and U.B.Honsinger, Efficiency
Optimization of Inverter-fed Induction Motor Drives,
Conf. Rec.1982 IEEE Ind. Appl Soc., pp. 544-552.
[11] J.M.Renders, Genetic Algorithms and Neural Network,
Herms, Paris, 1995.
[12] K. L. She and all., Speed Estimation of an Induction
Motor Drive Using an Optimized Extended Kalman
Filter, IEEE Transaction on Industrial Electronics,
Vol. 49, N 1, Feb 2002, pp.124-133.
[13] J. R. Leigh, Control Theory, Second Edition, Institution
of Electrical Engineers, London, 2004.
[14] F. Zidani and all, Fuzzy Efficient-Optimization
Controller for Induction Motor, IEEE Power
Engineering Review, October2000, pp. 43-44.
[15] Z. Rouabah and all, Optimal Efficiency of a Fuzzy
Controller in a Field Oriented Control Induction Motor
Drive, 4th International Conference on Electrical
Engineering CEE06, Batna, Algeria, 07 - 08 Nov.
2006, pp. 135-138.

441