Simulink Model of an Induction Machine

This document describes the defining equations for a dynamic induction machine
model to be implemented in Simulink. There are a number of research publica-
tions on this topic however they often lack key pieces of information which the
author either decided were so self-evident they need not be included or just ne-
glected due to their familiarity with the subject matter. Something that may be
about to happen again.
In terms of research into the design and control of electrical machines the process
described here is “classical” (i.e. more than a few years old). In practice a
number of companies known to the author use a process similar to this for several
types of machine drive but with additions to accommodate effects particular
to their application. For example in high speed work the frictional heating of
the rotor due to wind-age and aerodynamic losses may be included as these
are often a function of speed. Saturation of the machine iron may be included
by parameterising the leakage and mutual inductances as a function of flux or
current. Copper and Eddy current losses are relatively easily added and thereafter
the thermal effects that these create may be included as well. The final form of the
simulation and its scope is limited only by technical necessity, the human resource
available to perform the development work and the financial cost the developer
is prepared to incur to possess the simulation. The main cost of developing a
simulation for a machine is the validation stage(s) which require physical space,
test equipment and human resources to gather the validation data. By comparison
the programming of the model including the HR costs of the programming and
the cost of the software is often a minor component of a project.

Figure 1: Per phase equivalent circuit of an induction machine.

The three phase induction machine per phase equivalent model which is com-
monly taught to undergraduates is shown in fig. 1. A discursive introduction to
machines and control may be found in [1]. Some common questions that under-
graduates might face involve referring the rotor to the stator side and deriving
the torque equation, the pull out torque, the starting torque and sketching a

1
Bear Instruments Ltd August 2018

graph of the torque/power speed curve. This can be done without the necessity
of teaching any field oriented control material. The difficulty with this approach
is that it works well for setting exams but not so well for practical control of
machines or for doing simulations of the machines as if they were systems. It is
not common to perform control calculations in the abc reference frame.
In the case where one is performing magneto-static finite element simulations to
interrogate the machine design, the control aspects of the system are often not
of great concern and working in abc is usually acceptable. it is common to inject
sine wave currents into the phase windings to develop the flux density plots and
thereafter derive the torque speed and efficiency and ultimately the torque per
amp or per unit volume etc. for a given rotor and stator geometry, magnetic
material and winding configuration etc.
Relatively recently it has become popular to link the control and power electronics
simulations with FE simulation ‘in the loop’ i.e. one clock tick of the control and
power electronics simulation in Simulink gives rise to a set of voltages on the
stator windings which are passed to the FE program. The FE program performs
one clock tick (often of the same length or a number of ticks that make up
the same length as the control and power electronics simulation) and solves the
magneto-static equations to provide the current that will flow in the machine
windings as a result of the applied voltage. These currents are used by the
control and power electronics simulations to calculate, for example, the I R drops
across the power switches (among many other things) and this gives rise to the
applied voltage on the windings in the next time-step. This coupled simulation
capability exists in a number of proprietary packages including Flux, Ansys and
COMSOL. Although, COMSOL’s mechanism of interacting with MATLAB is not
as easily amenable to the per clock tick coupling described here as for example
Flux’s method, nevertheless, it is possible. Others such as Vector Field and
JMAG have something similar. In principle such a coupling should be possible
between OpenModelica and Elmer or OpenFOAM, but I am unaware of any
existing implementation.
From a control systems perspective it is preferable to mode most machines as
if they are DC machines. When there are steady state quantities representing
energy transfer in the DC machine the currents and voltages that represent that
energy on the electrical side are constant i.e. not time varying (like sine waves).
This is desirable because it makes the calculations easier for us to interpret, and
the control optimisation less challenging. More importantly it is potentially less
computationally demanding as well, remembering that in an industrial drive the
embedded system should be as inexpensive as possible but still able to complete
the task effectively.
To facilitate this approach to modelling the AC machines as DC machines the

2
Bear Instruments Ltd August 2018

work of Clark and Park is used either in the Clarke Park transform or the dq
transform. This transform converts the three phase currents of the stationary abc
frame to the stationary α, β reference frame and then to the rotating d-q reference
frame. If a refresher is needed then there is an excellent video at https://www.
youtube.com/watch?v=vdeVVTltr1M (not one of mine and I’m not affiliated with
the author). The transformation may be performed easily in Simulink using a
built in block. It is not necessary to describe the transform mathematics here, it
may be found in textbooks including [2–4]. A discussion of PWM techniques can
be found in [5] and [6].

Figure 2: d-q axis decomposition of an induction machine.

The d-q transformed equivalent circuit is shown in Fig. 2 where the nomenclature
is given in table 1 and the value column provides representative parameter values
for a 1.1 kW machine used in a light industrial machine tool.

Stator series resistance The stator series resistance is a representation of the

finite DC resistance of the stator winding.

Stator leakage inductance The stator leakage inductance represents the mag-
netic flux surrounding the stator windings (coils), due to the current passing

3
Bear Instruments Ltd August 2018

Parameter Description Value Units

Rs Stator winding series resistance 6 Ω
Lls Stator winding leakage inductance. 30 mH
Lm Mutual inductance coupling stator and rotor 500 mH
Llr Rotor winding leakage inductance 30 mH
Rr Rotor winding resistance 6 Ω
Power Rated average power of the machine 1.1 kW
Voltage Rated phase to phase voltage of the machine 415 V
Current Rated phase current of the machine 2.77 A
Frequency Rated frequency 50 Hz
Speed Rated speed 1415 rpm
Poles 4 poles (2 pole pairs)
J Rotor inertia 0.02 kg m2
Nm
D Total Friction/Drag 0.003 rad s−1

Table 1: Nomenclature and representative values for the d-q decomposition of a

1.1 kW induction machine.

through the stator windings, which does not cut the rotor bars/windings and
therefore does not have the opportunity to induce a current in the rotor. This
inductance represents the portion of the stator flux which does not manage to
couple the stator to the rotor. It is sometimes said to be due to the ‘end wind-
ing(s)’. Winding schemes with shorter end windings tend to suffer lower leakage
inductances. The advantages and disadvantages of particular winding schemes
is an active area of machine design research. For example, in SR machines [7].
However, in commercial applications the cost of applying the winding to a ma-
chine is usually the most important factor unless there are special design criteria
that must be met. Accepting a slightly longer end winding while having an easy
i.e. automated mechanism of applying the winding to the machine is a common
industrial trade-off.

Rotor series resistance The rotor series resistance is a representation of the

finite DC resistance of the rotor winding or bars. The rotor is often a squir-
rel cage design having bars rather than windings but may be wound in some
circumstances.

Rotor leakage inductance The rotor leakage inductance represents the mag-
netic flux surrounding the rotor bars or windings, due to the current induced in
the rotor by the stator magnetic field, which does not cut the stator coils. This
is effectively a representation of the part of the rotor flux which does not link the
rotor to the stator.

4
Bear Instruments Ltd August 2018

Stator back EMF The d axis stator back EMF is proportional to the syn-
chronous speed and the q axis flux. The current gives rise to a magnetic flux
and the rate of change of that magnetic flux gives rise to a voltage across the
winding in which the current flows the back EMF will act to oppose the change
in current (Lenz’s law). If the current is sinusoidal the voltage that develops will
be co-sinusoidal (derivative of sine is cosine) cosine may be thought of as a sine
wave with a phase shift of π/4. Since the d and q axes are perpendicular to
each-other (shifted π/4 radians) the EMF felt on the d axis is due to the q axis
flux and similarly the EMF felt on the q axis is proportional to the d axis flux.

Rotor back EMF The d axis rotor back EMF is proportional to the difference
between the synchronous speed and the rotor speed (i.e. the “slip”) and the rotor
flux generated by the q axis. The difference in the synchronous and rotor speed
is the rate at which the stator field appears to be moving past us if we’re sitting
on the rotor bars looking radially outwards.
Figure 2 also notes the voltage across the stator windings (Lls + Lm ) and rotor
windings (Llr + Lm ) is given by the rate of change of flux cutting those windings.
Kirchhoff’s laws may be used to develop a set of loop equations for the stator and
rotor. Later these equations will be combined with flux equations to form a self
consistent set of equations that Simulink (or Modelica) can solve. The mechanical
system will also be added to the simulation. Starting with the d axis stator loop
and recalling that the voltage across an inductor is given by the product of the
rate of change of current and the inductance of the coil (v = L di/dt),

d ids d
vds − ids Rs + ωs Φqs − Lls − Lm (ids + idr ) = 0 (1)
dt dt
The relationship between flux and current (flux is the product of inductance and
current) can be used to simplify (1), although this can be obtained by using the
voltages marked by the rate of change of flux in fig. 2 as well
d Φds
vds − ids Rs + ωs Φqs − =0 (2)
dt
d Φds
transposing (2) to make dt
the subject,

d Φds
= vds − ids Rs − ωs Φqs (3)
dt
The expressions involving differential equations are usually cast in the form of
derivative = everything else, because that is the way the Simulink blocks will
be arranged. After the equation is formed the 1/s block is used to integrate
the differential equation. The differential equations are coupled and the solver

5
Bear Instruments Ltd August 2018

that Simulink calls will solve them simultaneously. Well, presuming the solution
converges. If the simulation doesn’t converge it’s quite likely that one of the
equations is either derived incorrectly or has not been transferred into Simulink
correctly. For the q axis,
d Φqs
= vqs − iqs Rs + ωs Φds (4)
dt
The rotor loops can be similarly treated, starting with the d axis,
d Φdr
= −idr Rr − (ωs − ωr ) Φqr (5)
dt
and for the q axis,
d Φqr
= −iqr Rr + (ωs − ωr ) Φdr (6)
dt
Four expressions link the d and q axis rotor and stator flux starting with the q
axis stator flux,
Φsq = (iqs + iqr ) Lm + iqs Lls (7)
for the d axis stator flux,

Φsd = (ids + idr ) Lm + ids Lls (8)

for the q axis rotor flux,

Φrq = (iqs + iqr ) Lm + iqr Llr (9)

for the d axis rotor flux,

Φrd = (ids + idr ) Lm + idr Llr (10)

In this case the flux equations may be written in terms of the d and q axis cur-
rents. This is necessary because the simulation will proceed by a voltage being
impressed on the stator winding. This voltage will be used to calculate the rate
of change of flux, which in turn will be integrated to obtain that flux. The flux
will then be used to calculate the current. The current will be fed back into a
controlled current source which exists between (i.e. in parallel with) the termi-
nals across which the voltage was originally impressed. It’s possible to formulate
another approach in which a current is impressed and the resulting EMF on the
stator is calculated (along with all the mechanical outputs) however this requires
the system of equations to be cast in terms of calculating derivatives. Taking
derivatives in numerical sampled data systems is potentially risky because it is a
noisy process which can promote instability (i.e. lack of convergence). Integra-
tion or numerical quadrature is by comparison a low noise and generally stable
approach. There are some circuit reasons to use an integration approach as well.

6
Bear Instruments Ltd August 2018

When the simulation is set up with some power electronics the machine model
will take the calculated current independent of the other effects in the system.
If a back EMF approach was used the converter supply voltage would have the
machine terminal voltage subtracted from it and the machine current would be
whatever flowed through the series resistance of the power semiconductor device.
If the power device model has no series resistance and the link voltage was a
perfect voltage source there will be an over-constrained matrix in the simulation
as two voltage sources will have been connected in parallel which is undesirable.
Keeping in mind all the foregoing, if the converter design under consideration
is a current source converter or an impedance source converter it’s not impossi-
ble, it may even be desirable, to cast the machine model as being a controlled
voltage device. However, in voltage source inverter applications, which are much
more common, presenting the machine model to the power electronics model as
a controlled current device is certainly preferable. This should be clearer when
considered in association with the video demonstrating the Simulink code (see
https://youtu.be/wM72tarF_to).
The q axis stator flux expression (7) becomes,
Φqs − Lm iqr
iqs = (11)
Lm + Lls
Equation (8) is transposed to provide the d axis stator current,
Φds − Lm idr
ids = (12)
Lm + Lls
Two more expressions are require and can be derived trivially from (9) and (10)
to provide the d and q axis rotor currents in terms of the d and q axis rotor
flux. Looking at (11) and (12) it is clear that the solution of these two equations
and the two which are not written out must be simultaneous as they all depend
on each-other. This is similar to the differential equations for flux, they are
interdependent and the solver must solve them all simultaneously.
The electro-mechanical torque can be expressed as,
3P
Te = (iqs ((ids + idr ) Lm ) + ids ((iqs + iqr ) Lm )) (13)
22
The electro-mechanical torque could be expressed in terms of flux more simply
but the Simulink model that this document relates to uses the form presented
here. Since the integral of the time derivative of flux is calculated prior to the
current the Simulink blocks flow in a more aesthetic way if the current is used to
develop the torque. The rate of change of speed is given by,
dωr Te − D ωr − TL
= (14)
dt J

7
Bear Instruments Ltd August 2018

where ωr is the rotor angular velocity in rad s−1 , Te is the electro-mechanical

torque in Nm, D is the frictional drag in Nm/(rad s−1 ), TL is the load torque,
Nm (which when positive opposes the electro-mechanical torque) and J is the
total mechanical inertia of the rotating parts in kg m2 . The position of the rotor,
in radians, is the integral of the rotor angular velocity,
Z
θ = ωr dt. (15)

The implementation in Simulink is presented in a video at https://youtu.be/

wM72tarF_to

About the Author

Dr. James Green runs Bear Instruments Ltd, an electronics
engineering consultancy specialising in analogue, mixed signal
and power electronics systems design. James was lecturer
in electrical machines and controls in the Department of
Electronic and Electrical Engineering at the University
of Sheffield from 2013 – 2018 and was Royal Academy
of Engineering Industrial Teaching Fellow from 2017 –
2018. He gained his PhD in 2012 in semiconductor device
characterisation and has since worked on academic research
and industrial design problems in, among other fields,
microwave curing of composite materials and automotive
battery testing and modelling. www.bearinstruments.co.uk

8
Bear Instruments Ltd August 2018

References
[1] A. Hughes, Electric Motors and Drives Fundamentals, Types and Applica-
tions, third edition ed., 2006.
[2] B. K. Bose, Power Electronics and Motor Drives: Advances and Trends. El-
sevier, 2006.
[3] N. Mohan, Advanced Electric Drives. Wiley, 2014.
[4] F. Giri, AC Electric Motors Control. Wiley, 2013.
[5] N. Mohan, First Course on Power Electronics and Drives. MNPERE, 2003.
[6] E. C. dos Santos Jr. and E. R. C. da Silva, Advanced Power Electronic Con-
verters, ser. IEEE Press Series on Power Engineering, M. E. El-Hawary, Ed.,
2015.
[7] X. Y. Ma, G. J. Li, G. W. Jewell, and Z. Q. Zhu, “Recent development of re-
luctance machines with different winding configurations, excitation methods,
and machine structures,” CES Transactions on Electrical Machines and Sys-
tems, vol. 2, no. 1, pp. 82–92, March 2018. doi: 10.23919/TEMS.2018.8326454