Proceedings of the 6th WSEAS International Conference on Power Systems, Lisbon, Portugal, September 22-24, 2006 108

Dynamic Model Identification of Induction Motors using Intelligent

Search Techniques with taking Core Loss into Account

BOONRUANG MARUNGSRI*, NITTAYA MEEBOON and ANANT OONSIVILAI

Electrical and Energy Systems Research Unit, School of Electrical Engineering
Suranaree University of Technology
Nakhon Ratchasima, 30000
THAILAND

Abstract: - Traditionally, dynamic parameters of induction motors can be roughly estimated through
conventional tests (no load test, block rotor test and retardation test) and core loss is neglected in the dynamic
behaviours analysis. Due to the complication of dynamic behaviours of induction motors, inaccuracy of
transient characteristics may obtain when using these dynamic parameters. In order to improving accuracy of
dynamic behaviour analysis, however, the inclusion of core loss in the machine model needs to be re-addressed
and an intelligent approach to estimated dynamic parameters needs to be adopted. In this paper, three of
intelligent search techniques, which are i) Tabu Search (TS), ii) Adaptive Tabu Search (ATS) and iii) Genetic
Algorithm (GA), are employed to demonstrate the effectiveness of intelligent identification compared with the
conventional model with and without core loss parameter(RC). The simulation results from dynamic parameters
including RC obtained by the GA in comparison with the experimental results are convinced the effectiveness
for this aim.

Key-Words: - Induction Motor, Dynamic Model, Intelligent Search, Core Loss, Tabu Search, Adaptive Tabu
Search, Genetic Algorithm.

1 Introduction drives systems. Thus, the accurate parameter

Induction motors are very commonly used in identification of induction motors is challenged for
industrial applications over half a century because those controlled drives system. Many identification
there are cheap, robust efficient and reliable [1]. techniques have been studied and have been
Traditionally, the conventional steady state per- proposed for dynamic parameter identification of
phase equivalent circuit has been used to describing induction motors [6–11]. But there is still lack to
steady-state behaviours of three-phase induction include the effect of core loss in dynamic model.
motors [1–5]. Furthermore, in simple control, However, there has been increasing studied in the
where the accuracy and the precision are not that detuning effect of that the core loss produces [12–
much significant, such a steady-state model seems 20]. Hence, the effect of core loss should be taken
to be moderate for describing and represent the into account and core loss parameter should be re-
dynamic behaviour of induction motors. It is well addressed in dynamic model of induction motors.
known that only the standard steady-state per-phase
equivalent circuit for three-phase induction motors In this paper, the authors have proposed two kinds
has included core loss parameter in a simple and of dynamic model of induction motor. One is the
effective manner. But the generalized theory of dynamic model without core loss taking in to
machines neglects core loss parameter in the account. The other one is with core loss taking into
transient analysis of induction motors. account. And this paper also presents the
effectiveness of intelligent search techniques, Tabu
Presently, however, with the development to Search (TS), Adaptive Tabu Search (ATS) and
complication and high accuracy controlled drive Genetic Algorithm (GA) techniques, for identifying
systems of induction motors, dynamic model from dynamic parameter of induction motors compared
traditional steady-state parameters is not enough to with the conventional techniques (no-load, locked-
represent the exactly dynamic behaviors of rotor and retardation tests). The validity of the
induction motor in high accuracy vector controlled authors aim is verified by stator current response
Proceedings of the 6th WSEAS International Conference on Power Systems, Lisbon, Portugal, September 22-24, 2006 109

and speed response from the simulation and the

vs   Rs +Ls p Lmp  is 
 0  = L p − jω L R +L p − jω L ⋅i 
experimental results. (2)
  m r m r r r r  r
2 Equivalent Circuit and Modelling
of Three Phase Induction Motors Where Ls = Lls + Lm, Lr = Llr + Lm and in the steady
Traditionally, the steady state model of a three phase state p = jωs.
induction motor is represeted by the per phase
equivalent circuit, including RC, where RC represents The direct and quadrature axis model(d-q model)
the equivalent resistance for core loss [2,3]. The based on the space phasor theory is widely used for
steady state per phase equivalent circuit referred to simulation the dynamic behaviour of three-phase
the stator side of a three phase induction motor, induction motors as shown in Fig. 2.
neglecting and including RC, is given in Fig. 1.
iqs
IS jXls jXlr Ir vqs
RS Im

VS
RC jXm Rr/s
ωr
iqr
vqr
a) With RC included

IS jXls Im jXlr Ir
RS Idr Ids

VS jXm Vdr Vds

Rr/s
Fig. 2 d-q Axis Representation of Induction Motor.

The generalized theory of machine neglects core

b) With RC neglected
loss in the transient analysis of induction motors.
Fig. 1 Per Phase Equivalent Circuit of Three Phase From Equation (2) and from Fig. 2, neglecting RC
Induction Motors. and with vdr = vqr = 0, the following equations can
be obtained.
As shown in Fig. 1, Vs is the stator phase voltage; vqs Rs +LSp 0 Lmp 0  iqs
Rs and Rr are the stator and rotor winding v   0 Rs +LSp 0 Lmp  ids (3)
resistances; Xls and Xlr are the stator and rotor  ds =  ⋅ 
leakage reactance. Xm is the magnetizing reactance;    m
0 L p − ω L
r m R +
r r L p −ω r r  i qr 
L
and Rc is the equivalent resistance for core loss.  0   ωL Lmp ωr Lr Rr +Lr p idr 
   r m
From Fig. 1a, with power being supplied only on Where ωr is the rotor speed in electrical radians/sec.
the stator side, the voltages and currents can be To represent RC in the d-q model, Equation (3)
related as: should be modified and be written to include RC in

vs   Rs + Xp
the d-q model as:
Kp  is  vqs  Rs + Xp 0 Kp 0  i qs 
 0  = Kp− jω K R + Yp− jω Y⋅ i  (1) v   0 Rs + Xp 0 Kp  i ds 
   r r r   r
 ds  = ⋅   (4)

Where p = d/dt, K =
R CL m  0   Kp −ωr K Rr + Yp −ωr Y  i qr 
R C + L m p , X = L 1s + K    ωr Y Rr + Yp i dr 
 0   ωr K Kp
and Y = L 1 r + K . The voltage equations are valid under both transient
From Fig. 1b, the relation between the input stator and steady state condition. In order to prove the
voltage and the stator and rotor currents can be authors aim, the state equations in the stationary
derived as: reference frame are chosen. Under transient
Proceedings of the 6th WSEAS International Conference on Power Systems, Lisbon, Portugal, September 22-24, 2006 110

conditions, the equation of motion as functions of Table 1 Induction Motor Parameters from the
stator currents, rotor currents and rotor speed as Conventional Tests.
shown in equation (5) is used to simulate the speed
response compared with the experimental results. Induction Motor Parameters
Rls (Ω) 74.02
dω Rlr (Ω) 62.01
J m dt r = ( Tem − TL ) + B m ω r (5) Rc (Ω) 0.6482
Lls (H) 0.2087
Where Tem = − 23 PL m ( i ds i qr − i qs i dr ) , Llr (H) 0.2087
Lms H) 3.4377
J m = Moment of inertia Bm(N.m.s/rad) 0.0000
B m = Friction coefficient Jm(N.m.s2/rad) 0.0025
TL = Load torque
In this paper, transient condition is starting process 3
of induction motor from standstill to full speed at Experimental Result

no load ( TL = 0 ).
2

Current, A
1

0
3 Experimental Results
A squirrel-cage induction motor, 0.5-kW, 220/380- -1
V, 50-Hz, 3-phase, is used for the conventional -2 Simulation Result
tests as shown in Fig. 4.
-3
0 0.05 0. 1 0.15 0.2 0.25 0.3 0.35 0 .4
Digital Oscilloscope Time, sec
3
Phase
a) Stator Current Responses
Tacho 2000
Variable
Voltage
Transformer Induction Pendulum Control Unit for
Motor Machine Pendulum Machine
1500 Experimental Result
Speed, rpm

1000
Fig. 3 Block Diagram Representing the
Experimental Set. 500 Simulation Result

With the conventional no-load, locked-rotor and 0

retardation tests, the parameters of the induction 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
motor can be obtained. They are put in Table 1. Time, sec

These parameters are used to simulate the stator

b) Speed Responses
current and the speed responses compared with the
results from experimental test. The comparison Fig. 5 Comparison the Dynamic Response between
results are shown in Fig. 5. the Experimental Results and the Simulation from
Parameters obtained by the Conventional Test
As shown in Fig. 5, inaccuracy of dynamic Method.
responses was obtained when using dynamic
parameters from conventional tests. Smaller stator 4 Parameters Identification using
current was obtained in the simulation results
Intelligent Search Techniques
compared with the experimental results. As illustrated in Section 3, the conventional steady
state model can not used to estimate accurately and
precisely dynamic parameters. By this reason,
many different approaches to identify dynamic
parameters of induction motors have been
proposed. In this paper, the authors would like to
Proceedings of the 6th WSEAS International Conference on Power Systems, Lisbon, Portugal, September 22-24, 2006 111

demonstrate the effectiveness of some intelligent Input

search techniques, TS, ATS and GA techniques, to
identify dynamic parameters and also would like to

Adjustment
Parameters
demonstrate the effect of core loss in dynamic Space Phasor Experimental
Model
model for the dynamic response simulation of
induction motors. Due to there exist many of
Simulation
works employing intelligent search techniques. So, Experimental
the details of such techniques are not illustrated in Dynamic

Intelligent Search Engine

Simulation Responses

( TS, ATS and GA )

this paper. However, more detail in [21] and [22] Dynamic
may be useful. Responses

In order to elucidate the effectiveness of intelligent

Objective Function
search techniques, firstly, dynamic response of (Sum of Square Error)
stator current and rotor speed of a tested induction
motor are measured. Secondly, TS, ATS and GA
techniques are selected to develop an algorithm to No Less than Maximum
identify dynamic parameters of such induction Error Allowance
motor based on space phasor model. The procedure
of searching process shows in Fig. 6. For the best Yes

searching results in each search technique, the End

parameters are adjusted to give response best fitting
experimental data. An objective function (the sum
Fig. 6 Intelligent Dynamic Parameter Identification
of squared errors) in equation (6) is the key to
Procedure.
perform the properly searching results.
N
f obj = ∑ ( y measured − y simulated )2 (6) Table 2 Comparison among Obtained Parameters.
i =1
where ymeasured is the measured signal CT Tabu ATS GA
ysimulated is the simulated signal
fobj is the objective function Lls (H) 0.2087 0.07046 0.0713 0.07142
Each searching techniques was used for identify
Llr (H) 0.2087 0.0979 0.1081 0.09858
dynamic parameter of induction motor until obtain
5 sets of the best searching results. The average of
Lms (H) 3.4377 1.00368 0.9463 1.51958
the best searching results for each technique
compared with the conventional test results are Rls (Ω) 74.02 38.41318 37.6793 38.6659
shown in Table 2. Although variation of searching
results were obtained from each searching Rlr (Ω) 62.01 37.32834 37.2042 28.9187
techniques but the same tendency and around ½ of
the conventional test results of the results were Jm 0.0000 0.00378 0.0037 0.00382
obtained. From the average of parameters of
searching techniques given in Table 2, the rotor Bm 0.0025 0.00098 0.0008 0.0016
speed and the stator current responses were
simulated using the space phasor model as And, also, slight deviation of the speed response
described in Section 2. The simulation results were was obtained. While with taking RC into account,
compared with the experimental results. The near the same magnitude of stator current was
effectiveness and the accuracy of each technique obtained from the simulation result compare with
are shown in Fig. 7, Fig. 8 and Fig. 9, respectively. the experimental result. And the speed response
from the simulation results is best fitting with the
In case of parameters obtained by TS technique as experimental results.
shown in Fig. 7, without taking RC in to account,
larger magnitude of stator current before the steady In case of parameters obtained by ATS technique
state condition were obtained from the simulation as shown in Fig. 8, without taking RC in to account,
results compared with the experimental results. larger magnitude of stator current before the steady
Proceedings of the 6th WSEAS International Conference on Power Systems, Lisbon, Portugal, September 22-24, 2006 112

state condition were obtained from the simulation However, the effectiveness of each searching
results compared with the experimental results. techniques can be comparing by the calculation
While with taking RC into account, near the same time. High accuracy and precisely with shortest
magnitude of stator current was obtained. Speed time for calculation are the key for indicate the
responses from the simulation results with and most effective searching technique. As shown in
without taking RC into account are best fitting with Table 3, calculation time of GA technique is shorter
the experimental results. than the other techniques. So, dynamic parameters
identification using GA technique may be the most
In case of parameters obtained by GA technique as effectiveness searching techniques for induction
shown in Fig. 9, without taking RC in to account, motor compare with those the two techniques.
largest magnitude of stator current before the
steady state condition were obtained from the Table 3 Comparison of Calculation Times.
simulation results compared with the other
TS ATS GA
techniques and the experimental results. While with
Calculation
taking RC into account, near the same magnitude of Time (sec)
7595.658 902.57 349.963
stator current was obtained. Speed responses from
the simulation results with taking RC into account is
best fitting with the experimental results than the
simulation results with taking RC into account.

All the simulation results with taking core loss into

account show the accuracy of each searching
techniques for parameter identification of induction
motor.

3 3
Experimental Result Experimental Result
2 2
1 1
Current, A
Current, A

0 0
-1 -1
-2 Simulation Result -2 Simulation Result
-3 -3
0 0.05 0.1 0 .15 0 .2 0.25 0.3 0.35 0.4 0 0.05 0.1 0 .15 0.2 0.25 0.3 0.35 0 .4
Time, sec Time, sec
i) Stator Current Responses i) Stator Current Responses
2000 2000
Experimental Result Experimental Result
1500 1500
Speed, rpm
Speed, rpm

Simulation Result
1000 Simulation Result 1000

500 500

0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 0 .05 0.1 0.15 0 .2 0.25 0.3 0.35 0. 4
Time, sec Time, sec
ii) Speed Responses ii) Speed Responses
a) Without taking RC into account b) With taking RC into account
Fig. 7 Comparison the Dynamic Response between the Experimental Results and the Simulation
Results from Parameters obtained by the Tabu Search based Method.
Proceedings of the 6th WSEAS International Conference on Power Systems, Lisbon, Portugal, September 22-24, 2006 113

3 3

2 Experimental Result 2 Experimental Result

1 1

Current, A
Current, A

0 0

-1 -1

-2 -2
Simulation Result Simulation Result
-3 -3
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 0.05 0 .1 0.15 0.2 0.25 0.3 0.35 0.4
Time, sec Time, sec

i) Stator Current Responses i) Stator Current Responses

2000 2000
Experimental Result Experimental Result

1500 1500

Speed, rpm
Speed, rpm

1000 Simulation Result 1000 Simulation Result

500 500

0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 0.05 0. 1 0.15 0.2 0.25 0.3 0.35 0.4
Time, sec Time, sec

ii) Speed Responses ii) Speed Responses

a) Without taking RC into account b) With taking RC into account
Fig. 8 Comparison the Dynamic Response between the Experimental Results and the Simulation
Results from Parameters obtained by the AdaptiveTabu Search based Method.
3 3
Experimental Result Experimental Result
2 2

1
Current, A

1
Current, A

0 0

-1 -1

-2 -2
Simulation Result Simulation Result
-3 -3
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Time, sec Time, sec
i) Stator Current Responses i) Stator Current Responses
2000 2000
Experimental Result
Experimental Result
1500 1500
Speed, rpm
Speed, rpm

1000 Simulation Result 1000

500 500 Simulation Result

0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Time, sec Time, sec
ii) Speed Responses ii) Speed Responses
b) With taking RC into account b) With taking RC into account

Fig. 9 Comparison the Dynamic Response between the Experimental Results and the Simulation
Results from Parameters obtained by the Genetic Algorithm based Method.
Proceedings of the 6th WSEAS International Conference on Power Systems, Lisbon, Portugal, September 22-24, 2006 114

5 Conclusions Induction Motors”, Proc. of 6th Int. Conf. on

The effectiveness of an intelligent approach to Optimization of Electrical and Electronic
estimate dynamic parameters of induction motors Equipments, Vol. 2, 14-15May1998, pp. 499 –
with and without taking core loss into account 504.
illustrates in this paper. Although, the parameters [10] R. K. Ursem and P. Vadstrup, “Parameter
of induction motor can be roughly estimated Identification of Induction Motors using
through conventional tests (no-load, locked-rotor Differential Evolution”, The 2003 Congress on
and retardation tests) due to the complication of Evolutionary Computation CEC’03, Vol. 2, 8-12
space phasor equations describing dynamic December 2003, pp.790 - 796 .
behaviours of induction motors, they may cause [11] T. Kulworawanichpong, K.-L, Areerak, K.-N,
inaccurate estimation, especially when transient Areerak, P. Pao-la-or, D. PuangDownreong and S.
characteristics are seriously required. Tabu Search Sujitjorn, “Dynamic Parameter Identification of
(TS), Adaptive Tabu Search (ATS) and Genetic Induction Motors using Intelligent Search
Algorithm (GA) techniques are employed to Techniques”, IASTED Int. Conf. on Modelling,
demonstrate this intelligent identification. All the Identification and Control(MIC2005), 16-18
simulation results with taking core loss into account February 2005, pp. 328-332.
show the effectiveness of each technique for [12] Levi, E., Sokola, M., Boglietti, A. and
parameter identification of induction motor. Pastorelli,M., “Iron Loss in Rotor-Flux-Oriented
However, high accuracy and precisely with shortest Induction Machines: Identification, Assessment of
time for calculation are the key for indicate the Detuning, and Compensation”, IEEE Trans. On
most effective searching technique. Due to shortest Power Electronics, Vol. 11, No. 5, September
calculation time, dynamic parameters identification 1996, pp. 698 – 709.
using GA technique may be the most effectiveness [13] Wee, Sung-Don., Shin, Myoung-Ho. and
searching techniques for induction motor compare Hyun, Dong-Seok., “Stator-flux-oriented control of
with Tabu Search (TS) and Adaptive Tabu Search induction motor considering iron loss”, IEEE
(ATS) techniques. Trans. on Industrial Electronics, Vol. 48, No. 3,
June 2001, pp. 602 – 608.
References: [14] Boglietti, A., Ferraris, P. and Lazzari, M.,
[1] A. Husain, Electric Machines, Dhanpat Rai “Induction motor iron losses measurement with a
&Co, 2006. static converter supply using a slotless rotor test
[2] S. J. Chapman, Electric machinery bench”, IEEE Trans. on Magnetics, Vol. 30, No. 6,
fundamentals, Singapore: McGraw-Hill,1991. November 1994, pp. 4599 – 4601.
[3] R. Krishnan, Electric motor drives modelling, [15] Santisteban, J.A. and Stephan, R.M., “Vector
analysis and control, Upper Saddle River, NJ: control methods for induction machines: an
Prentice-Hall, 2001. overview”, IEEE Trans. on Education, Vol. 44, No.
[4] M.G. Say, “Alternating current machines” 2, May 2001, pp. 170 – 175.
Pitman, London, 1983. [16] Boldea, I. and Nasar, S.A., “Unified treatment
[5] P. C. Krause, O. Wasynczuk and S. D. Sudhoff, of core losses and saturation in the orthogonal-axis
“Analysis of Electric Machinery”, IEEE press, model of electric machines”, IEE Proc., Vol 134,
New York, 1995. Pt. B, No. 6, November 1987, pp. 355 – 363.
[6] J. W. Kim and S. W. Kim, “Parameter [17] Jong-Woo Choi, Dae-Woong Chug and
Identification of Induction Motors using Dynamic Seung-Ki SUI, “Implementation of Field Oriented
Encoding Algorithm for Searches (DEAS)”, IEEE Induction Machine Considering lron Losses”, in
Trans. on Energy conversion, Vol. 20, Issue Conf Rec. IEEE IAS’96, 1996, pp. 375-379
1, March 2005, pp.16 – 24. [18] J. lung and K. Nam, “A Vector Control
[7] F. Alonge, F. D'Ippolito, G. Ferrante and Scheme for EV Induction Motors with a Series Iron
F.M.Raimondi, “Parameter identification of Loss Model”, IEEE LE tra., vol. 45, pp. 617-624,
induction motor model using genetic algorithms”, 1998.
IEE Proc. on Control Theory and Applications, [19] T. Mizuno, J. Takayam& T. Ichioka, “
Vol.145, Issue 6, November 1998 , pp. 587 – 593. Decoupling Control Method of Induction Motor
[8] R. Craciun and I. Margineanu, “Method And Taking Stator Core Loss Consideration”, in Conf
Testing Equipment For Parameter Identification Of Rec. IPEC’90, 1990, pp. 69-74
Proceedings of the 6th WSEAS International Conference on Power Systems, Lisbon, Portugal, September 22-24, 2006 115

[20] G. 0. Garcia, J. A. Santisteban and S. D.

Brignone, “Iron Losses Influence on a Field-
Oriented Contmller”, in Conf Rec. IEEE
IECON’94, 1994, pp. 633-638.
[21] D. Puangdownreong, K.-N. Areerak, A.
Srikaew, S. Sujitjorn and P. Totarong, “System
Identification via Adaptive Tabu Search”, IEEE Int.
Conf. on Industrial Technology, Vol. 2, 11-14
December 2002, pp. 915 – 920.
[22] C. Chaiyaratana and A.M.S.;Zalzala, “Recent
Developments in Evolutionary and Genetic
Algorithms: Theory and Applications”, 2nd Int.
Conf. on Genetic Algorithms in Engineering
System: Innovations and Applications, 2-4
September 1997, pp. 270 – 277.