1126

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 25, NO. 6, NOVEMBERIDECEMBER 1989

Modelling of Induction Machines for Electric

Drives
GORDON R. SLEMON, FELLOW,
IEEE

Abstract-Most analyses of induction machine drives make use of a

conventional linear machine model, either in the form of self and mutual
inductances of the familiar T-form of equivalent circuit. This model is
actually more complex than necessary for analysis in the linear regime.
On the ofher hand, it is inadequate for use when the machine is saturated.
This paper presents several linear and nonlinear models suitable for
transient and steady-state analysis. The models are presented in equivalent circuit form to preserve the identity of non-linear parameters. A note
on modelling for time harmonics is also included.

rotating sinusoidal distribution. For example, the space vector

of the stator currents is derived from the phase currents as

where CY =
It is assumed that the neutral connection is
open. The inverse transformation is of the form

[i, ib iCblf=Re[l

CY

CY~]~Z.

INTRODUCTION

N RECENT YEARS, many papers have been published on

variable speed drives using induction machines. A majority
of these papers contain contributions to the power converter
and control system technologies. In their analyses, most of
these papers have used conventional representations of the
machines such as those based on Park's equations [ l ] or the
familiar T-form of linear equivalent circuit [ 2 ] .
Such models are adequate for many situations. They are,
however, overly simplified for proper analysis of some
operating conditions such as occur with magnetic saturation.
In addition, they are more complex than is necessary for most
linear machine analysis.
This paper examines some of the issues of adequate machine
modelling and attempts to provide a firmer basis for selection
of an appropriate model for use in transient or steady-state
analyses of induction machine drives. In doing so it draws on a
number of specialized contributions that have been scattered
through prior literature on the subject.
SPACE
VECTOR
NOTATION
FOR INDUCTION
MACHINES

(2)

The same transformations apply for flux linkages and phase

voltages.
For a magnetically linear machine, the stator and rotor flux
linkages are given by

[t]

Ls Lm
[Lm L,]

[;I

(3)

In most drives, the induction machines are of the squirrelcage type where the rotor variables are not available for
measurement. It is conventional to transform the actual rotor
variables h,! and i,! to a stator reference frame using a turns
ratio n and the rotor position 0 relative to the stator such that

[21 [ 4 [
=eip

(4)

where /3 = \ W O dt and oois the rotor speed in electrical radls.

In drives with solid-state power converters, the stator
voltages and currents are frequently of complex waveform. It
is usually most convenient to use a machine model developed
for a stator reference frame so that the terminal variables of the
machine and the power converter may be easily related. This
model in equation form is

In an induction machine, the polyphase stator windings are

usually distributed so that the magnetomotive force may be
assumed to be sinusoidally distributed in space around the air
= RsZ +pXs
gap periphery. If, in addition, the machine can be assumed to
be magnetically linear, the flux density will also be sinusoi0 = R,i, +ph, -jooh,
dally distributed. For such machines, the space vector notation
and concepts introduced by Ratz and Kovacs [3] are particu- where p = d/dt. All transient models developed in the paper
may, however, be readily modified for any other reference
larly useful.
In this approach, all variables are represented by polar frame.
vectors indicating the magnitude and angular position of the
T MODEL

- -

Paper IF'CSD 89-14, approved by the Electric Machines Committee of the

IEEE Industry Applications Society for presentation at the 1988 Industry
Applications Society Annual Meeting, Pittsburgh, PA, October 2-7. Manuscript released for publication April 17, 1989.
The author is with the Department of Electrical Engineering, University of
Toronto, Toronto, ON, Canada, M5S 1A4.
IEEE Log Number 8931466.

The information of (3), (3,and (6) may be expressed in the

T-form of transient equivalent circuit shown in Fig. 1. The
complete transient analysis can be carried out in complex
number form. For those who prefer to work in real numbers
only, the variables may be expanded as = uds + j u q s , =
ids jiqs, etc. Separation of the real and imaginary parts then

OO93-9994/89/11OO-1126$01.OO

0 1989 IEEE

1127

SLEMON: MODELLING OF INDUCTION MACHINES FOR ELECTRIC DRIVES

Fig. 1. T-form of transient equivalent circuit. L,, = L, - L,;

L,, = L , - L,.

Fig. 2.

1 7
rq3
t

gives the usual but more complex and repetitive set of four
differential equations:

[;I=[

R,+pLs
0
PLm
0
0
R,+pL,
0
PLm
PLm
wdrn Rr+PLr U d r
-woLm
PLm
-woLr Rr+pLr

*[[I.

T-form of steady-state equivalent circuit.

LL

Rs

jooTR

y3GTi
(a)

(7)

In the steady state, the analysis is simplified by resolving

each variable into a component at fundamental frequency w,
plus harmonics. The fundamental stator voltage vector is then
of the form

josh,

LM

R~ 0
0 ,
s

josk.q

(b)
Fig. 3. F equivalent circuit models. (a) Transient circuit. (b) Steady-state
circuit.

choose a value such that the inductances L, and L , in ( 3 ) are

equal. This would give a new set of rotor variables related to
those of ( 3 ) by

and may be expressed as a phasor

[k] [ [i]

(9)

Equations (5) and (6) may then be stated for the steady-state as

V,= RJ, +j w s Xs

*R

RR

(10)

where the rotor frequency wr = w, - w o . This formulation

provides the familiar T-form steady-state equivalent circuit of
Fig. 2.
A recurring problem in using the T models of Figs. 1 and 2
arises from the inaccessibility of the rotor variables for
measurement. Assuming constancy of parameters, two values
of complex impedance or four parameters can be measured on
the stator typically by operating at two values of speed such as
00 = U , and 00 = 0. This is not sufficient to give the five
parameters of these equivalent circuits. Frequently this is
addressed by arbitrarily making LIS= L,,.
LINEAR
r MODEL
The T-form circuit models of Figs. 1 and 2 are actually
more complex than necessary. They can be transformed into
simpler models with no loss of information or accuracy [4][61.
Since actual rotor variables are generally not required, the
turns ratio n in (4) may be chosen arbitrarily. Suppose we

l:Y]

where y = L,/L, . Combining this with ( 5 ) and (6) gives the

transient equivalent circuit of Fig. 3(a) and the steady-state
equivalent circuit of Fig. 3(b). This configuration has been
denoted as the r form from the structure of its two
inductances. The parameters of this circuit are related to those
of Fig. 1 by

LM= y L ,

= L,

L~=yLis+yLi~

RR = y 2 R r .

(1 3 )

(14)
(15)

These parameters may by derived directly from the usual no

load and standstill measurements on the machine.
This l7 model is particularly well suited for analysis of
drives, which attempt to keep the stator flux linkage As
constant by applying a stator voltage that is proportional to
frequency (possibly with a small correction for stator resistance drop. For a P-pole machine, the air gap torque is

3P
T = - Im
4

(XF;)

3~

KsZRRwr

--

N.m.

4 R:+(w,LL)

in the steady state.

(17)

1128

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 2 5 , NO. 6, NOVEMBERIDECEMBER 1989

where
a,=/30+

or dt.

(25)

The rotor position Po is obtained from an angular position

sensor, and the rotor frequency U, is obtained from (23) for the
required torque.
INCLUSION
OF MAGNETIC
NONLINEARITY

(b)
Fig. 4. Inverse F equivalent circuit models. (a) Transient circuit. (b) Steadystate circuit.

The I model is nearly the same as the approximate

equivalent circuit, which has frequently been presented in
traditional literature [2] in which the branch L , in Fig. 1 is
moved out to the stator terminals.

r MODEL

INVERSE

Suppose the turns ratio n is (4)is chosen such that the

inductances L , and L, in (3) are equal. The new set of rotor
variables is then given by

[5][
=

3 [I

l /yI

ER

1,

where y = L,/L,. The result is the inverse r form of

transient and steady-state equivalent circuits shown in Fig. 4 in
which
L ; = L,, y L,,
(19)

L&=yL,

(20)

R;=yI2R,.

(21)

The circuit models of Fig. 4 are particularly appropriate for

the understanding and analysis of vector control systems. In
these, the stator current space vector is controlled in such a
way as to maiEtain a constant magnitude of the e t o r flux
linkage vector A; through the magnetizing current i&, while
providing a rotor current vector 7; in space quadrature with ?h
to produce the required torque:
3P

T=-

Im

( x i ?;*)

Thus, the stator current space vector for a given required

torque is

- - i- + ? /
j =

4T

In the linear T-form equivalent circuits of Figs. 1 and 2, the

stator and rotor leakage inductances are influenced principally
by the shapes and dimensions of the stator and rotor slots. The
inductance L, is inversely proportional to the air gap
reluctance and is most influenced by the length of the air gap.
In truly linear modelling, the iron of both the stator and the
rotor is assumed to be of negligible reluctance.
In practical design, it would generally be inefficient to
operate at maximum flux density levels, which would produce
negligible iron reluctance. A considerable increase in torque
output can be achieved for a given machine frame by operating
at flux densities in the stator and rotor for which the iron
reluctances are significant in comparison with the air gap
reluctance.
Most publications that consider the effects of iron saturation
assume that it can be included by appropriate adjustment of the
inductance L , in Figs. 1 and 2 and in (7). It is frequently
assumed that the leakage inductances are constant although
saturation in these is sometimes included for very high values
of stator and rotor current [8], [9].
The association of magnetic nonlinearity with the parameter
L , is based on an assumption that the saturation is predominantly dependent on the radially directed air gap flux. It is
often, and incorrectly, assumed that the leakage fluxes have a
negligible effect on the state of saturation in the rotor and
stator iron because these fluxes are mainly circumferentially
directed across the tooth tips [2]. It has in fact been shown that
the flux densities in the teeth and yoke of the stator are
dependent primarily on the total stator flux linkage, which in
turn results from the radially directed air gap flux plus the
angular rate of change of the circumferentially directed stator
leakage flux [ 5 ] , [lo]. Similarly, the flux density in the rotor
iron depends primarily on the net rotor flux linkage, which is
derived from the air gap flux minus the effect of the rotor
leakage flux.
The additional magnetizing current required by the stator
teeth and yoke is mainly dependent on the stator flux linkage A,
and is structurally best modelled as a nonlinear inductance L,i
as shown in the circuit of Fig. 5. A typical A, - isi relation
defining L,i is shown in Fig. 6.
A small part of the stator leakage inductance derives from
the leakage flux in the end windings and within the stator
conductors. This part should properly be placed in series with
the stator resistance R, and thus outside the Lsi inductance.
The effect of this refinement is usually too small to be
measurable [ 101.
Saturation in the rotor teeth and core is similarly dependent
on the rotor flux linkage A, and may be represented by the
nonlinear inductance Lri, connected as shown in Fig. 5.

1129

SLEMON: MODELLING OF INDUCTION MACHINES FOR ELECTRIC DRIVES

RR

3,

Fig. 5 .

T-form equivalent circuit including elements to represent stator and

rotor saturation.
A
/

(a)

,/

v,

iwsL

Fig. 7.

Nonlinear

r models.

(b)
(a) Transient circuit. (b) Steady-state circuit.

/
/

/.
/

--

id

Fig. 6 . Typical stator saturation characteristic (for comparison, the dashed

line shows L , = &Jim).
j@o?R

It will be noted that the inductance L , remains linear since it

continues to represent the air gap reluctance.
The nonlinear reluctance of the stator teeth and yoke cause
the stator flux density to be nonsinusoidally distributed in
space. However, only the fundamental space component for
this flux density produces a net flux linkage in the stator
winding because of the assumed sinusoidal distribution of that
winding. Thus, sinusoidal currents in the stator windings still
produce sinusoidally induced phase voltages in spite of the
magnetic nonlinearity. For any instantaneous values of winding currents, the flux linkages may still be represented by their
fundamental space vectors in spite of the nonsinusoidal
distribution of the flux density.

r MODEL
NONLINEAR
The relative importance of the two nonlinear inductance
elements of Fig. 5 depends on the design of the machine and
on its loading condition. Frequently, the machine is designed
for equal stator and rotor flux density at no load. With constant
stator flux linkage (i.e., constant V/Hz), the rotor iron will
experience its greatest flux density at no load. The rotor flux
density will progressively decrease as the stator and rotor
currents are increased.
In practice, it is found that the effect of rotor saturation can
usually be ignored for most machines and for most loading
conditions, leaving the effective saturation to be represented
with adequate accuracy by LSialone. This is particularly true
for full load and overload conditions [ 101.
The saturable stator inductance L,; may readily be incorporated into the r models of Fig. 3 since it is directly in parallel
with the inductance L M .For simplicity, the parallel combination is still denoted as LM (nonlinear) as shown in the circuit
models of Fig. 7.

Fig. 8. Nonlinear inverse r models.

In the conventional T-form equivalent circuits (Figs. 1 and

2), saturation is often associated with the central inductance
L,. Such models predict that a machine operated with
constant stator flux linkage will experience reduced saturation
effect as the load is increased. In contrast, the models of Fig. 7
show that the saturation effect will be essentially constant for a
machine operated with h, constant.
The equivalent circuit of Fig. 7(b) shows that the maximum
steady-state torque obtainable from the induction machine is
limited by the maximum permissible value of the stator flux
to a value from (17), of
linkage

xs,

- 3Pq
T=-N.m.
8 LL
NONLINEAR
INVERSE

r MODEL

The effect of stator saturation can be included in the inverse

r circuits of Fig. 4 by connection of the nonlinear inductance
Ls; across the voltage produced by the time rate of change of
stator flux linkage, as shown in Fig. 8. For low values of stator
flux linkage this added element may be ignored.
INFLUENCE
OF NONLINEARITY
ON VECTOR
CONTROL
In vector control, the objective is to maintain a rotor flux
linkage space vector that is constant in magnitude and rotates
at a controlled angular velocity U, while regulating the rotor
current to obtain the demand torque.
When the induction machine can be considered linear the
inverse r model of Fig. 4(a) is particularly appropriate. The
stator current to produce instantaneous torque T with constant
rotor flux linkage
is given in (24) and (25). A control

1130

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 25, NO. 6 , NOVEMBERIDECEMBER 1989

system that implements these equations is valid as long as the

stator flux linkage X, is such as to produce negligible stator
saturation.
With constant rotor flux linkage, the stator flux linkage will
increase with increase in load. For larger values of stator flux
linkage, the stator current space vector of (24) must be
augmented by the addition of a space vector isi. From Fig. 8,

-I , =%-++i .
Lsi

The nonlinear model of Fig. 7(a) may also be used to

design a vector control system to give instantaneous torque
response in the nonlinear regime. Ignoring the eddy current
effect, the required stator current space vector is obtained
from the relations

and

where LMis equal to its linear value in parallel with L,i from
Fig. 6 .

NONLINEARITY
IN LEAKAGE
INDUCTANCE
The leakage inductance in the equivalent circuit models is
dependent mainly on the stator and rotor slot shapes, particularly in the region near the air gap. If significant nonlinearity
in leakage inductance occurs, it is dominantly caused by @e
circumferentially directed flux established in the tooth faces
and tooth tips. Such nonlinearity normally occurs only with
very high current in the windings for which the current space
vectors in stator and rotor are nearly equal in magnitude and
oppositely directed. The effect of this nonlinearity can be
included with adequate accuracy for most situations by making
the leakage inductances LL and L t in the r and inverse r
circuits nonlinear functions of the stator current vector

z.

CORELoss EFFECTS
The core loss in the stator is a nonlinear function of the
stator frequency and the flux densities in the stator teeth and
yoke. It can be shown [IO] that these flux densities can be
associated with the total stator flux linkage As. The core loss
may therefore be represented by a nonlinear resistance R,,
connected as shown in Fig. 7. Over the range for which the
core loss can be assumed to be proportional to flux density
squared and frequency squared, the resistance R, will be a
constant.
While core losses occur in the rotor iron, the fundamental
space component contributes directly to torque production,

and its effect is included in the equivalent circuit as a minor

reduction in the equivalent rotor resistance RR.
For most analyses of transient and steady-state behavior, the
core losses may be safely be ignored. They are, however,
important in efficiency determination and in the study of
particular phenomena such as self excitation.

HARMONIC
MODELS
Inverter supplies usually impose either voltage or current
harmonics on the induction motor. For steady-state analysis it
is desirable to have equivalent circuit models from which the
impedance to each of the harmonics can be determined. In the
magnetically linear regime, the impedance at each pertinent
harmonic frequency can be obtained using the models of either
Fig. 3(b) or Fig. 4(b), using appropriate values for w, and w, in
each instance and correcting the rotor resistance (and possibly
also the leakage inductance) for the influence of skin effect.
In the magnetically nonlinear regime where the stator iron
may be significantly saturated, it has frequently been assumed
that the leakage inductance dominates and that the harmonic
impedance is only negligibly affected by saturation of the
magnetization inductance.
It has been shown, however, that this assumption can lead to
gross overprediction of the harmonic impedance [ 111. If the
effect of the inverter is such as to keep the magnitude A, of the
stator flux linkage essentially constant while modulating the
angular velocity w, at harmonic frequency, the appropriate
value for the nonlinear magnetizing inductance LM is A,/iM,
and the harmonic inductance will be approximately this value
in parallel with the leakage inductance LL. If, however, the
effect of the inverter is to modulate the amplitude of the stator
flux linkage at harmonic frequency with little harmonic effect
on its angular velocity, the appropriate value of the magnetizing inductance is

dXS
LMh=-.
diM

(32)

For a highly saturated machine, this value may be smaller than

the leakage inductance and thus may have a major effect on the
harmonic impedance.

CONCLUSION
This paper presents equivalent circuit models for induction
machines for transient and steady-state operation. The circuits, designated as r or inverse r form, are simpler than the
conventional T-form of circuit. Their parameters are readily
determined from terminal measurements. The major effects of
magnetic nonlinearity are included in the models in a way that
is more accurate than that usually obtained with the conventional T-form circuit.

REFERENCES
R. H. Park, Two-reaction theory of synchronous machines, AZEE
Trans., vol. 48, Part I, p. 716, 1929, and vol. 52, Part 11, p. 352, 1933.
P. L. Alger, Induction Machines. New York: Gordon and Breach,
1970.
K. P. Kovacs, Transient Processes in A C Machines. Budapest
Akademia Kiado, 1980.
G. R. Slemon, Magnetoelectric Drives- Transducers, Transformers
and Machines. New York: Wiley, 1966.
G. R . Slemon, Circuit models for polyphase machines, Electric

SLEMON: MODELLING OF INDUCTION MACHINES FOR ELECTRIC DRlVES

Machines and Power Systems, vol. 8 , no. 415, pp. 369-379, Nov./
Dec. 1983.
S. Yamamura, A C Motors for High-Performance Applications.
New York: Marcel Dekker, 1986.
F. Khater, R. D. Lorenz, D. W. Novotny, and K. Tang, The selection
of flux level in field oriented induction machine controllers with
consideration of magnetic saturation effects, in Con!. Rec. 1986
IEEE Industry Applications Society Annu. Meeting, Part I, pp.
124- 131.
T. A. Lip0 and A. Consoli, Modelling and simulation of induction
motors with saturable leakage reactances, IEEE Trans. Ind. Appl.,
vol. IA-20, no. I., pp. 180-189, Jan./Feb. 1984.
G. J. Rogers and D. Shimohammadi, Induction machine modeling for
electromagnetic transient program, IEEE Power Engineering Society,
Winter Meeting, 87 WM 213-2, pp. 622-628, Dec. 1987.
G. R. Slemon, Analytical models for saturated synchronous machines, IEEE Trans. Power App. Syst., vol. PAS190, no. 2, pp.
409-417, Mar./Apr. 1971.
G. R. Slemon and E. A. Ismailov, An analysis of the harmonic
impedance of a saturated induction machine, IEEE Trans. Power
App. Syst., vol. PAS/99, no. 4, pp. 1663-1669, July/Aug. 1980.

1131
Gordon R. Slemon received B.A.Sc. and M.A.Sc.
degrees from the University of Toronto, Toronto,
ON, Canada, the D.I.C. of the Imperial College of
Science and Technology, London, and the Ph.D.
and D.Sc. degrees from the University of London.
Following employment with Ontario Hydro and
Atomic Energy of Canada, he taught at Nova Scotia
Technical College prior to his appointment at the
Universitv of Toronto in 1955. He served as Head
of the Electrical Engineering Department at Toronto
from 1966-1976 and as Dean of the Faculty of
Applied Science and Engineering from 1979-1986 He is currently a
Professor of Electrical Engineering at the University
Dr Slemon is a Fellow of the Institution of Electrical Engineers, the
Institute of Electrical and Electronics Engineers, the Engineering Institute of
Canada, and the Canadian Academy of Engineering. In 1965 he received the
Western Electric Award for excellence in teaching. He has acted as consultant
to many corporations and utilities. He was president of the consulting firm of
Electrical kngineering Consociates Ltd from 1976-1979 He is c q e n t l y
Charman of the Board of the University of Toronto Innovations Foundation