You are on page 1of 64

A Sensorless Induction Machine Drive

- An implementation study
A Master’s thesis by Håkan Carlsson and Johan Bergerlind
Department of Energy and Environment
CHALMERS UNIVERSITY OF TECHNOLOGY
Göteborg, Sweden 2006
Möndal, Sweden 2006
Abstract
This Master’s thesis describes a DSP implementation of the Statically Com-
pensated Voltage Model (SCVM) algorithm for sensorless induction machine
control. The algorithm is aimed to be implemented in a 16-bit fixed-point
processor.
In general, low speed operation with voltage model based control systems
is problematic, especially when in generator mode. In generator mode with
the load at nominal torque, the implemented design fails already at 7.5 Hz.
With a lower load, 2 Nm or 54 % of nominal, the drive is able brake it to
almost zero, but the speed accuracy is low. On the other hand, if the motor
operates in motor mode, low speed is less problematic. Experiments show
that this implementation can handle variations in stator resistance from -37
% up to +60 % with stable low speed operation and variations of at least
±60 % at nominal speed. Stable operation for all speeds is possible even
during an increase of rotor resistance by 60 %, but the speed accuracy will
be reduced by approximatly 2 %. The rotor resistance may decrease by 58
% at low speed and by 53 % at nominal speed before the drive destabilizes.
The total leakage inductance can vary at least ±30 % without noticeably
affecting the system.
The thesis project is conducted in association with Aros electronics AB.
Acknowledgments
We would like to thank our supervisor at Aros electronics AB, Ph.D. Mikael
Alatalo, for his unwavering support, help and encouragement. We also ex-
tend our thanks to our examiner, associate professor Torbjörn Thiringer, and
to Ph.D. Andreas Petersson for their help with theoretical questions and al-
ternative approaches to problems. Professor Lennart Harnefors and Ph.D.
Rolf Ottersten has through their work with the SCVM and to some extent
also by helping us directly inspired and motivated our work. The work of
Ph.D. student Oskar Wallmark has also been a sorce of inspiration during
this thesis work.
Last but certainly not least, a warm thanks to all employees at Aros
electronics AB for their kindness and support, making us feel most welcome
during our thesis work.
Contents
1 Introduction 1
1.1 Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Goal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Induction Motor Basics 3
2.1 Mechanical Construction . . . . . . . . . . . . . . . . . . . . . 3
2.2 Mathematical Circuit Models . . . . . . . . . . . . . . . . . . 3
2.3 Principle of Operation . . . . . . . . . . . . . . . . . . . . . . 6
3 Vector Control 8
3.1 Space Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Synchronous Coordinates . . . . . . . . . . . . . . . . . . . . 9
3.3 Motor Control . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4 Sensorless Motor Control . . . . . . . . . . . . . . . . . . . . . 9
4 Sensorless Control Algorithm 12
4.1 Improved Statically Compensated Voltage Model . . . . . . . 12
4.2 Parameter Sensitivity of the SCVM . . . . . . . . . . . . . . . 14
4.3 Known Limitations of the SCVM . . . . . . . . . . . . . . . . 14
5 Simulations 16
5.1 Motor Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
5.2 Model Verification . . . . . . . . . . . . . . . . . . . . . . . . 18
5.3 Control System Model . . . . . . . . . . . . . . . . . . . . . . 18
5.4 The Improved SCVM Model . . . . . . . . . . . . . . . . . . . 20
5.5 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . 21
5.6 New Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.7 Simulation Conclusions . . . . . . . . . . . . . . . . . . . . . . 24
6 Implementation 25
6.1 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
6.1.1 Controller Card . . . . . . . . . . . . . . . . . . . . . . 25
V
A Sensorless Induction Machine Drive.
6.1.2 Motor . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.2 Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.3 Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.4 Implementation Issues . . . . . . . . . . . . . . . . . . . . . . 27
7 Results 28
7.1 Speed Step Response . . . . . . . . . . . . . . . . . . . . . . . 28
7.2 Slow Speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
7.2.1 Slow Speed Limit . . . . . . . . . . . . . . . . . . . . . 28
7.2.2 Reversing . . . . . . . . . . . . . . . . . . . . . . . . . 31
7.3 Parameter Sensitivity . . . . . . . . . . . . . . . . . . . . . . . 36
7.3.1 Deviations in Stator Resistance . . . . . . . . . . . . . 36
7.3.2 Deviations in Rotor Resistance . . . . . . . . . . . . . 42
7.3.3 Deviations in Leakage Inductance . . . . . . . . . . . . 46
8 Conclusions 50
8.1 Goal Fulfilment . . . . . . . . . . . . . . . . . . . . . . . . . . 50
8.1.1 Stable Operation up to Nominal Speed . . . . . . . . . 50
8.1.2 Stable Operation up to Twice the Nominal Torque . . 50
8.1.3 Stable Operation during Acceleration and Decelera-
tion with an Arbitrary Load . . . . . . . . . . . . . . . 50
8.1.4 Stable Operation during Parameter Variations . . . . . 50
8.2 Further Work . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
A Measurement of Motor Parameters 52
A.1 Locked Rotor Test . . . . . . . . . . . . . . . . . . . . . . . . 53
A.2 No-load Test . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
VI
Chapter 1
Introduction
Vector control of induction machines (IM:s) is a well known tool for advanced
control of AC-machines. This method of control requires knowledge of the
rotor flux angle. Traditionally, this has been derived using current measure-
ments and information about the rotor speed, measured by a device mounted
on the shaft of the machine. However, there are several problems involved
with the speed measurement. The measuring device, often an encoder or a
resolver, is expensive and adds another component that may fail in the sys-
tem. The drive will also be sensitive to disturbances and harsh operational
environments. The measuring device also need space for mounting, which
may prove problematic in some applications, as well as communication be-
tween the device and the control system. Careful mounting and alignment
are often necessary, further increasing the production costs. The problems
concerning rotor speed measurement have spurred a development of methods
for sensorless flux angle estimation and thus advanced control of the machine
without mechanical sensors.
1.1 Purpose
This Master’s thesis describes the implementation of the Statically Com-
pensated Voltage Model flux- and speed-sensorless control strategy for an
induction machine. Some basic vector control theory is also covered.
1.2 Goal
The goal is to achieve a sensorless IM drive that enables stable operation
• up to the nominal speed.
• up to twice the nominal torque.
• during acceleration and deceleration with an arbitrary load.
1
A Sensorless Induction Machine Drive.
• during variations in rotor and stator resistance up to 60 % and in the
total leakage inductance up to 30 %.
1.3 Sources
There is much research done on the subject of sensorless control. Numerous
articles and papers has been published. Earlier work at Aros electronics
AB includes a Master’s thesis by Anders Ottergren [1], "Evaluation of speed
observers for an induction machine". This paper is considered in Chapter 4.
Professor Lennart Harnefors has written two very informative compendia
about control of variable speed drives, [2] and [3]. The second is partly
an extended version of the first, covering also control of power electronic
converters. Harnefors presents a promising algorithm called the Statically
Compensated Voltage Model or SCVM. This is the approach we will further
study.
1.4 Structure
The theory of vector control is covered in Chapter 3 to give an introduction to
the subject of field oriented motor drives. The SCVM method for sensorless
control is discussed in Chapter 4. In Chapter 5 the control system and
the mathematical model of the motor is described and implemented using
Matlab Simulink. The results of the different simulations are displayed and
discussed here as well. The main part of this thesis is the implementation,
Chapter 6, and the results here of are described in Chapter 7. The report is
ended with a conclusion part, Chapter 8.
2
Chapter 2
Induction Motor Basics
As a background to motor control, some basic induction motor principles are
explained.
2.1 Mechanical Construction
The fundamental parts of the induction motor are the stator and the rotor.
In the IM, unlike the DC-machine, the rotor conductors are short-circuited
and are, compared to earlier constructions of plain copper wiring, nowadays
mostly made of solid aluminum. In most cases the rotor slots are skewed
which reduces nonlinear effects and in turn harmonic content of the flux,
with less torque pulsations and lower acoustic noise as a result [2]. The rotor
is indirectly propelled by the magnetic field produced by the stator, hence
induction motor. This makes it important for the airgap between the stator
and rotor to be small in order to limit the flux leakage. Larger airgap means
higher losses in the stator windings, giving a lower efficiency.
2.2 Mathematical Circuit Models
A classical way to model the induction motor is the single phase "T model",
see figure 2.1 [4]. It consists of the resistance and inductance of the stator
and rotor. The magnetizing impedance is represented by an inductance.
Sometimes the active power loss in the core is included in the model, but
hardly ever in simulations. If included, it is represented by a resistance, R
c
,
connected in parallel or in series with the magnetizing inductance L
m
. In
this thesis it is excluded, with exception from the parameter calculations in
appendix A.
Another suitable equivalent model is the "inverse-Γ model", see figure
2.2. Contrary to the T model, all leakage inductance is transferred to one
side, the stator side.
There is a direct translation between T and inverse-Γ parameters:
3
A Sensorless Induction Machine Drive.
V
a
R
s
L
ls
L
m
R
r
s
L
lr
Figure 2.1: Single phase T model of an induction motor.
V
a
R
s
L
σ
L
M
R
R
s
Figure 2.2: Single phase inverse-Γ model of an induction motor.
4
A Sensorless Induction Machine Drive.
V
V
a
R
s
L
ls
L
m
R
r
L
lr

r
ψ
r
Figure 2.3: Single phase dynamic T model of an induction motor.
L
σ
= L
s

L
2
m
L
r
(2.1)
L
M
=
L
2
m
L
r
(2.2)
R
R
= R
r
_
L
m
L
r
_
2
(2.3)
ψ
R
= ψ
r
L
m
L
r
(2.4)
where L
s
= L
ls
+L
m
and L
r
= L
lr
+L
m
.
Since they are steady-state set-ups, these models are not feasible for
analysis of the dynamic behavior of the IM. Instead, two models valid for
transient analysis are used [3]. In figure 2.3 and 2.4, the effective rotor resis-
tance dependent of the slip is replaced with a voltage source corresponding
to the back EMF. For a more detailed description of back EMF, see Section
2.3.
V
V
a
R
s
L
σ
L
M
R
R

r
ψ
R
Figure 2.4: Single phase dynamic inverse-Γ model of an induction motor.
5
A Sensorless Induction Machine Drive.
2.3 Principle of Operation
The three-phase induction motor is based upon electro-magnetic interaction
between the stator and rotor. As alternating voltages displaced 120 degrees
to each other are applied to the stator windings, stator currents are formed.
A rotating magnetic field with varying flux density is generated. The mag-
netic flux induce a voltage over the rotor resulting in a current trough the
rotor windings. As there is a magnetic field present, the flowing current will
generate a force acting upon the rotor windings, thus producing an electro-
dynamic torque.
The rotor voltage is induced as long as the stator field is not moving with
the same speed as the rotor. As the rotor accelerates, the induced voltage
and current will drop. As an effect the torque is reduced. If the machine
is not loaded, the rotor practically keeps the same speed as the stator field.
Thus the speed difference, called slip speed or lag, will be small and the
torque almost zero. If the motor is loaded, the rotor speed will be lower and
the torque increases until the load is matched according to equation (2.5).
If the load is driving the rotor faster than the stator field, the voltage over
the rotor change polarity and the produced torque will be in the reverse
direction compared to the stator field movement. This will brake the load
[5].The mechanical speed of the motor is often represented by
ω
r
=
(T
e
−T
L
)
J
e
−Bt/J
(2.5)
where T
e
is the electrodynamic torque, T
L
the applied load, J the moment of
inertia and B the viscous friction. figure 2.5 is a sketch of the AC induction
motor torque and steady-state torque limits due to thermal limitations in
the machine. The maximum speed noted in figure 2.5 is due to mechanical
construction limitations of the rotor.
The rotor current is producing a magnetic field of its own, counteracting
the magnetic field produced by the stator, inducing a voltage called rotor
EMF, Electro-Motive Force. In the dynamic models, this is represented by
the term jω
r
ψ
s
R
. Together with the voltage drop over the rotor, a voltage
called back EMF, E
s
b
, is applied over the magnetizing inductance L
M
in
figure 2.4 [3].
E
s
b
=
_

r

R
R
L
M
_
ψ
s
R
(2.6)
From figure 2.4 the voltage equations for the inverse-Γ model can be
derived:
v
s
s
−R
s
i
s
s
−L
σ
di
s
s
dt
−L
M
di
s
M
dt
= 0 (2.7)

r
ψ
s
R
−R
R
i
s
R
−L
M
di
s
M
dt
= 0 (2.8)
6
A Sensorless Induction Machine Drive.
T
ω
Maximum speed
Field weakening region
IM torque
Steady-state torque limit
Figure 2.5: Comparison between IM-torque curves for a fixed stator fre-
quency and for vector control with steady-state torque limit and maximum
speed.
As the rotor currents are hard to measure, they are eliminated from
equations (2.7) and (2.8) by introducing
i
s
R
= i
s
M
−i
s
s
=
ψ
s
R
L
M
−i
s
s
. (2.9)
The voltage equations may now be written as
L
M
di
s
s
dt
= v
s
s
−R
s
i
s
s


s
R
dt
(2.10)

s
R
dt
= R
s
i
s
s

_
R
R
L
M
−jω
r
_
ψ
s
R
(2.11)
Harnefors defines

s
R
dt
as the flux EMF, E
s
. Comparing equation (2.6)
with (2.11) it may be concluded that the difference at nominal speeds are
small. Hence E
s
is often referred to as back EMF [3].
7
Chapter 3
Vector Control
The idea of vector control is to mathematically represent an AC machine as
a DC machine [2]. Vector control relies on field orientation. In all rotating
machines there is a magnetic field and if the control signals are aligned to
this field, field orientated control is achieved.
3.1 Space Vectors
Space vectors were originally invented to describe spatial flux distributions in
AC machines, hence “space” [2]. Due to the fact that v
a
(t) +v
b
(t) +v
c
(t) = 0
∀ t, it is possible to describe the three-phase system as an equivalent two-
phase system, with two perpendicular axes, with the axes denoted as α and
β. It is also convenient to think of these axes as the real and imaginary
axes in a complex plane. The voltage vector in the complex plane is then
denoted as v
s
(t) = v
α
(t) + jv
β
(t). The superscript “s” in v
s
means that
the space vector is referred to the stator reference frame, meaning that the
vector rotates with the stator speed which is zero. v
α
and v
β
are calculated
using the following transformation matrix
_
v
α
v
β
_
= K
_
2
3

1
3

1
3
0
1

3

1

3
_
. ¸¸ .
T
32
_
¸
_
v
a
(t)
v
b
(t)
v
c
(t)
_
¸
_
(3.1)
were K is a scaling factor:
Peak-value/Amplitude-invariant scaling: K = 1
RMS-value scaling: K = 1/

2
Power-invariant scaling: K =
_
3/2.
The inverse matrix of T
32
, T
23
, is used to get from two-phase back to
three-phase [2]
8
A Sensorless Induction Machine Drive.
_
¸
_
v
a
(t)
v
b
(t)
v
c
(t)
_
¸
_=
1
K
_
¸
_
1 0

1
2

3
2

1
2


3
2
_
¸
_
. ¸¸ .
T
23
_
v
α
v
β
_
. (3.2)
3.2 Synchronous Coordinates
The purpose of vector control is, as mentioned earlier, to control the AC
machine as if it were a DC machine. If this is to be done, the control sig-
nals needs to be DC quantities. This is achieved by the transformation
from two-phase to synchronous reference frame, v = e
−jω
1
t
v
s
were ω
1
is
the synchronous frequency. As the reference frame rotates with the sup-
ply/synchronous frequency, the vector appears not to be moving. This is
called dq-transformation and its relations with α and β leads to the follow-
ing transformation matrix [2]
v = e
−jω
1
t
v
s
= e
−jθ
1
v
s
= (cos θ
1
−j sin θ
1
) (v
α
+jv
β
)
= v
α
cos θ
1
+v
β
sin θ
1
+j (−v
α
sin θ
1
+v
β
cos θ
1
)
⇔v =
_
v
d
v
q
_
=
_
cos θ
1
sin θ
1
−sin θ
1
cos θ
1
_
. ¸¸ .
T
dq

1
)
_
v
α
v
β
_
.
(3.3)
3.3 Motor Control
Vector control implies rotor flux orientation. Rotor flux orientation means
that the rotor flux is aligned with the d-axis. The result is that the q-part
of the rotor flux is zero.
Most significant for applying vector control is the measuring of the sta-
tor currents, i
s
, which are transformed to the equivalent two phase system,
αβ. They are then dq transformed for the control system using the rotor
flux oriented angle, θ
1
[2]. The rotor flux oriented angle is derived from
the measured rotor speed ω
r
together with the calculated slip speed. The
control system also provides the desired dq-frame stator voltage vector, V
s
,
which in turn is αβ- and then two-phase/three-phase -transformed before it
is amplified and fed to the IM, see Figure 3.1.
3.4 Sensorless Motor Control
As mentioned earlier, the traditional way of controlling an IM often involves
a shaft sensor that measures the speed/position of the rotor. This sensor is
costly, for low- and medium-powered drives it often matches the cost of the
9
A Sensorless Induction Machine Drive.
IM
V
s
i
s
θ
r
θ
1
θ
1
αβ
αβ
αβ
αβ
dq
dq
3
3
Figure 3.1: The idea of the vector control method, where the control system
gets a voltage reference signal and then produces a three phase voltage to
the IM. The rotor speed ω
r
is measured and added to a calculated slip
speed, creating the synchronous speed ω
1
. This speed is then used to get the
synchronous angle θ
1
necessary for the transformations from control signals
to a three-phase voltage.
motor itself [2]. Besides the cost of the sensor it is also “one more part” that
may fail. It is also possible to use a flux sensor but the usage of such kind
of device is highly impractical due to probe placing between the stator and
rotor.
Harnefors [2] distinguishes between three different types of sensorless
control cases:
1. Flux-sensored, speed-sensorless control: When using a flux sen-
sor, vector control is no more difficult than controlling a DC motor. An
extra sensor for speed measurement is not needed because the speed
can be calculated using slip subtraction.
2. Flux-sensorless, speed sensored control: The typical situation
in most IM drives. Speed measuring allows design of flux estimators
which give good performance at all speeds.
3. Flux- and speed-sensorless control: Estimates both the flux and
speed which is an attractive solution due to reduced costs and parts. At
nominal speeds, a flux-angle estimate of good accuracy can be obtained
but they tend to fail when load torque is applied at very low speeds.
While for 1. and 2., stable operation is possible at all speeds, there is not
yet proof of a flux- and speed-sensorless control that is able to function
10
A Sensorless Induction Machine Drive.
completely stable at low frequencies while subjected to a load torque. The
flux and speed-sensorless control is the focus of this thesis.
11
Chapter 4
Sensorless Control Algorithm
Several sensorless control strategies has been proposed. Anders Ottergren
evaluated three different speed sensorless observers at Aros electronics AB in
2002: Observer based on Current-Voltage model (OCV), Adaptive Speed Ob-
server (ASO) and Voltage Model with Indirect Field Orientation (VMIFO).
Ottergren concluded that ASO was the preferred method [1].
Since Ottergrens work, the voltage model variants have improved. Lennart
Harnefors describes a sensorless flux estimator called the Statically Compen-
sated Voltage Model (SCVM) [2] first proposed in [6]. This method has been
further developed by Rolf Ottersten [7]. In this thesis the improved SCVM
is studied and implemented.
The estimators may be of two types, Direct Field Orientation (DFO)
and Indirect Field Orientation (IFO). DFO estimate the flux space vector
directly while IFO use the estimated rotor position to get the flux.
4.1 Improved Statically Compensated Voltage Model
The SCVM is based upon the classic "Voltage Model (VM)" flux estimator,
derived from the relation
d
´
ψ
s
R
dt
= E
s
. (4.1)
This implies that if the flux EMF is known, the rotor flux may be found.
Harnefors [2] deducts the SCVM quite thoroughly and this thesis only cover
it briefly.
After transferring the VM from DFO to IFO representation and split-
ting the relation into real and imaginary parts, the following equations are
obtained [7]:
d
´
Ψ
R
dt
= E
d
(4.2)
12
A Sensorless Induction Machine Drive.
´ ω
1
=
E
q
´
Ψ
R
(4.3)
where
E
d
= v
d
−R
s
i
d
+ ´ ω
1
L
σ
i
q
(4.4)
and
E
q
= v
q
−R
s
i
d
− ´ ω
1
L
σ
i
d
. (4.5)
The VM achieves good results at nominal speeds, but low speed estima-
tion is problematic. One common improvement is to replace the integrator
needed to calculate
´
ψ
s
R
with a low-pass integrator. Unfortunately, doing so
adds a static error, equal at all frequencies. The error may be compensated
for if a complex valued gain factor is introduced in the filter. This modified
VM is called the Statically Compensated Voltage Model, or SCVM [2]. The
flux will then be
´
ψ
s
R
=
j ´ ω
1
+λ|´ ω
1
|
j ´ ω
1
(p +λ|´ ω
1
|)
E
s
=
1 −jλsign(´ ω
1
)
p +λ|´ ω
1
|
E
s
(4.6)
where p =
d
dt
and λ is a gain parameter greater than zero. Equation (4.6)
may be transformed to IFO by substituting p →p +j ´ ω
1
and E →E
d
+jE
q
and removing superscript "s". Splitting the result into real and imaginary
parts and solving the imaginary part for ´ ω
1
gives
´
ψ
R
=
E
d
+λsign(´ ω
1
)E
q
p +λ|´ ω
1
|
(4.7)
´ ω
1
=
E
q
−λsign(´ ω
1
)E
d
´
ψ
R
(4.8)
Ottersten [7] then modifies the SCVM by introducing a new coefficient,
µ. By doing so, Ottersten adds another degree of freedom to the model.
´
ψ
R
=
µE
d
+λsign(´ ω
1
)E
q
p +λ|´ ω
1
|
(4.9)
´ ω
1
=
E
q
−λsign(´ ω
1
)E
d
´
ψ
R
(4.10)
13
A Sensorless Induction Machine Drive.
4.2 Parameter Sensitivity of the SCVM
A crucial measure of the performance of a sensorless control algorithm is the
sensitivity to variations in motor parameters. As the motor operates, the
parameters may change substantially. The resistances are mainly affected
by heat development and inductances by the level of magnetization and sat-
uration. As a result, the most likely deviations will be increasing resistance
and inductance. Other deviations may also occur, for example if parameters
are wrongly measured or the motors differs in the production line.
Four parameters are included in the inverse-Γ model; stator resistance
R
s
, rotor resistance R
R
, magnetizing inductance L
M
and total leakage in-
ductance L
σ
. In this project, L
σ
is derived from the T-model parameters
L
s
, L
r
and L
m
.
The SCVM includes three of the inverse-Γ parameters, R
s
, R
R
and L
σ
.
Indirectly the magnetizing inductance will affect L
σ
according to equation
(2.1). It may be rewritten as
L
σ
= L
ls
+L
m
(1 −
1
L
lr
L
m
+ 1
)
. ¸¸ .
→L
lr
, as L
m
→∞
→0, as L
m
→0
. (4.11)
The calculated leakage inductance will hence be between L
ls
and L
ls
+L
lr
.
As long as the magnetizing inductance changes are moderate the effect on
L
σ
will be small.
The critical parameter for the SCVM is the stator resistance [3]. It is
used in the calculations of E
d
and E
q
and impact the calculated flux directly.
The rotor resistance affect the slip calculation, thus affecting the accuracy
of the drive to some extent. The total leakage inductance is a part of the
flux estimation.
4.3 Known Limitations of the SCVM
The main problem region for the SCVM is low speed operation and reversing
the motor when subjected to a larger load. Theoretically, the estimator is
stable for all operating conditions except ω
1
= 0 and reversal with i
q
>
i
d

2
[2]. In practice, the low speed region where the SCVM is unstable is
larger. Reversing with load is a problem mainly when doing so slowly. If
the transition is fast, the probability of the problem developing is less. It is
also dependent upon how accurately the stator resistance is modelled. Rolf
Ottersten [7] suggest two improvements to make the SCVM more stable at
reversal and low speed operation. The stator resistance may be modeled
differently if ω
1
and i
q
have the same sign or not, slightly underestimated
in the first case and overestimated in the second. This requires the stator
14
A Sensorless Induction Machine Drive.
resistance to be fairly well known. Ottersten also suggest a modification of
λ at low speeds, choosing lambda to be
λ =

2|ω
1
|
ω
1,min

p
(
´
ψ
R
−ψ
ref
) (4.12)
where
λ
p
= 3
|i
q
|
i
d
. (4.13)
The success of the latter method depends upon that R
s
is estimated in the
manner described above. This approach is not yet fully tested. It is shown
that with λ
p
= 0, this method is an improvement at lower speeds. w
1,min
is
suggested to be between 5 and 10 % of nominal speed.
In this thesis, Otterstens improvements are not used.
15
Chapter 5
Simulations
Before proceeding with the implementation it is necessary to do some simu-
lations in order to test the control strategy. It is also an opportunity to get
acquainted with different problems that might arise during the implementa-
tion.
5.1 Motor Model
In this section, the mathematical model of the motor and its implementation
in Matlab Simulink is described and constructed. The base for the motor
model is the arbitrary dynamic model of an induction machine [5]:
_
¸
¸
¸
_
v
c
qs
v
c
ds
v
c
qr
v
c
dr
_
¸
¸
¸
_
=
_
¸
¸
¸
_
R
s
+L
s
p ω
c
L
s
L
m
p ω
c
L
m
−ω
c
L
s
R
s
+L
s
p −ω
c
L
m
L
m
p
L
m
p (ω
c
−ω
r
)L
m
R
r
+L
r
p (ω
c
−ω
r
)L
r
−(ω
c
−ω
r
)L
m
L
m
p −(ω
c
−ω
r
)L
r
R
r
+L
r
p
_
¸
¸
¸
_
_
¸
¸
¸
_
i
c
qs
i
c
ds
i
c
qr
i
c
dr
_
¸
¸
¸
_
(5.1)
where p =
d
dt
. The super- and subscript "c" denotes that the parameter
should be replaced by the parameter specific for the reference frame chosen.
If the reference frame is aligned with the rotor flux, the system is called
synchronous and ω
c
is replaced with ω
1
. In this thesis, the motor is mod-
eled with stationary references and ω
c
is replaced with 0. With stationary
references, it is common to say that the model is an αβ-system where α is
aligned to the magnetizing d-axis (q = β, d = α). The rotor windings are
short-circuited, leading to zero rotor voltage (v
βr
= v
αr
= 0), yielding
_
¸
¸
¸
_
v
s
βs
v
s
αs
0
0
_
¸
¸
¸
_
=
_
¸
¸
¸
_
R
s
+L
s
p 0 L
m
p 0
0 R
s
+L
s
p 0 L
m
p
L
m
p −ω
r
L
m
R
r
+L
r
p −ω
r
L
r
ω
r
L
m
L
m
p ω
r
L
r
R
r
+L
r
p
_
¸
¸
¸
_
_
¸
¸
¸
_
i
r
βs
i
r
αs
i
r
βr
i
r
αr
_
¸
¸
¸
_
. (5.2)
16
A Sensorless Induction Machine Drive.
The flux linkages are defined as
λ
βs
= L
s
i
βs
+L
m
i
βr
(5.3)
λ
αs
= L
s
i
αs
+L
m
i
αr
(5.4)
λ
βr
= L
m
i
βs
+L
r
i
βr
(5.5)
λ
αr
= L
m
i
αs
+L
r
i
αr
. (5.6)
Rearranging (5.2) and inserting (5.3) - (5.6) and the differentiation of the
same equations yields the differential equations implemented in the model
of the motor. The goal with the transformations of the equations is to base
the model on rotor flux and stator currents:
di
βs
dt
=
1
L
s

L
2
m
L
r
_
V
βs
−R
s
i
βs

L
m
L
r

βr
dt
_
(5.7)
di
αs
dt
=
1
L
s

L
2
m
L
r
_
V
αs
−R
s
i
αs

L
m
L
r

αr
dt
_
(5.8)

βr
dt
= ω
r
λ
αr

R
r
L
r
λ
βr
+
L
m
L
r
R
r
i
βs
(5.9)

αr
dt
= −ω
r
λ
βr

R
r
L
r
λ
βr
+
L
m
L
r
R
r
i
αs
. (5.10)
Further on, the flux and current index r and s will be omitted. The
electrical torque produced is calculated using i
α
, i
β
, λ
α
and λ
β
.
T
e
=
n
p
2
(i
β
λ
α
−i
α
λ
β
) (5.11)
where n
p
is the pole pair number. The mechanical part of the motor is
represented by
T
e
−T
L
= J

r
dt
+Bω
r
(5.12)
where J is the moment of inertia and B is the viscous friction. Solving
the equation for ω
r
gives
ω
r
=
(T
e
−T
L
)
J
e
−Bt/J
. (5.13)
Equations (5.7) - (5.13) are implemented in the “IM block” in Matlab
Simulink together with the control system and transformation equations
(3.1) - (3.3) as can be seen in figure 5.1. The motor parameters in the
model are measured on the motor used in the implementation later on. The
motor parameters can be found in Appendix A.
17
A Sensorless Induction Machine Drive.
Figure 5.1: Simulink implementation of the motor and its controller using
feedback of rotor speed.
5.2 Model Verification
Before beginning the sensorless implementation the model of the motor in
figure 5.1 needs to be verified. In order to verify the motor model the outputs
are compared to those of the real motor subjected to nominal speed and load
torque, see figure 5.2. The phase voltage supplied to the motor in figure 5.2
is 230·
_
2
3
V (delta connected motor) and the nominal load torque is 3.73
Nm according to equation (5.14).
nominal shaft power
nominal speed
=
1100
2815 ·
60

≈ 3.73Nm (5.14)
As can be seen in figure 5.3 the simulated motor speed is ∼46.9 Hz. This
is very close to the speed of the real motor during nominal conditions. The
currents in figure 5.4 are close to the nominal rms phase-current, 4.41 A, of
the machine. The verification testing indicates that the implemented motor
model is realistic.
5.3 Control System Model
In this thesis, the input to the control system is chosen to be a speed refer-
ence. The control system in figure 5.1 is written as an S-function in Matlab
for a more easy conversion to C-code. The control system in the first, non
sensorless, version consists of a speed controller, two current controllers and
a voltage decoupler.
18
A Sensorless Induction Machine Drive.
Figure 5.2: Verification of the implemented IM model.
0 0.5 1 1.5 2 2.5 3 3.5 4
46
46.5
47
47.5
48
48.5
49
49.5
50
50.5
Rotor speed at verification simulation
S
p
e
e
d

[
H
z
]
Time [sec]
Rotor speed
Synchronous speed
Figure 5.3: With the rotor speed at
∼46.9 Hz, the slip is ∼6 % which is
close to the real nominal slip, thus con-
firming the model.
1.85 1.9 1.95 2 2.05
−6
−4
−2
0
2
4
6
Phase currents at verification simulation
C
u
r
r
e
n
t

[
A
]
Time [sec]
I
a
I
b
I
c
Figure 5.4: The peak-value of the cur-
rents are ∼6 A rendering the rms-
value of 4.2 A, which is close to nomi-
nal current when the motor is cold.
19
A Sensorless Induction Machine Drive.
The PI speed controller derives i

q
from the speed error.
As discussed in Section 3.3, λ

qr
is defined as zero. The λ

dr
reference is
set to a constant value [8]. From (5.9) the setpoint for i
ds
is then
i

ds
=
λ

dr
L
M
(5.15)
Considering equation (5.15), the torque is proportional to i

qs
. With
the reference currents and the measured dq motor currents, the two cur-
rent controllers are implemented. The output from these controllers are fed
to a voltage decoupler, designed from the inverse Γ-model to remove the
cross-coupling between d and q dependency while calculating the voltage
commands to the "dq to abc" transformation block [3].
v

q
= H
q
+ ´ ω
1
L
σ
i

d
−(R
s
+R
R
)i

q
(5.16)
v

d
= H
d
− ´ ω
1
L
σ
i

q
−(R
s
+R
R
)i

d
(5.17)
where H
q
and H
d
is the result from the current controllers.
5.4 The Improved SCVM Model
Lennart Harnefors describes a discrete improved SCVM algorithm that is
implemented in this thesis project [2].
E
d
= v
ref
d
−R
s
i
ref
d
+ ´ ω
1
L
σ
i
ref
q
(5.18)
E
q
= v
ref
q
−R
s
i
ref
q
− ´ ω
1
L
σ
i
ref
d
(5.19)
´
ψ
R
=
´
ψ
R
+T
s
_
µE
d
+λsign(´ ω
1
)E
q
−λ|´ ω
1
|
´
ψ
R
_
(5.20)
´ ω
slip
= R
R
i
ref
q
´
ψ
R
(5.21)
´ ω
r
= ´ ω
r
+T
s
α
e
_
E
q
−λsign(´ ω
1
)E
d
´
ψ
R
− ´ ω
slip
− ´ ω
r
_
(5.22)
Equation (5.18), (5.19) and (5.20) corresponds to equation (4.4), (4.5)
and (4.9), respectively. Equation (5.21) is the standard slip relation. As the
observant reader may notice, the calculation of ´ ω
1
, equation (4.10), is a part
of the equation describing the rotor speed, (5.22). This is not a head on
solution but a filter, making the estimate less sensitive to disturbances. This
is possible as the mechanical system is slower than the control system.
The SCVM parameters is chosen as recommended by Harnefors, µ = −1
and λ =

2. The bandwidth α
e
of the rotor speed filter is set equal to the
current controller bandwidth [2].
20
A Sensorless Induction Machine Drive.
0 0.5 1 1.5 2 2.5 3 3.5 4
−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1
Speed difference Model−Estimate
Time [sec]
S
p
e
e
d

[
H
z
]
0 0.5 1 1.5 2 2.5 3 3.5 4
−5
0
5
10
15
20
25
30
35
40
45
50
Rotor Speed
Time [sec]
S
p
e
e
d

[
H
z
]
2.9 3 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9
600
605
610
615
620
625
630
Rotor Angle
Time [sec]
A
n
g
l
e

[
r
a
d
]
Model
Estimate
Model
Estimate
Figure 5.5: Modelled and estimated rotor data during an acceleration and
retardation test, showing both speed, speed difference and angle when the
motor is unloaded.
5.5 Simulation Results
To get some idea of the performance of the estimator, different scenarios are
simulated.
Figure 5.5 shows some simulated and estimated rotor quantities during
an acceleration and retardation of the motor. Figure 5.6 shows the same
as figure 5.5 but the motor is subjected to a 2 Nm load. In both figures
the top left image displays the speed of the rotor. The bottom left image
shows the total angle of the rotor, and the speed difference between the
estimation and the model are shown to the right. As may be seen in both
figures, the angle difference after the test cycle is less than 1 %. The speed
difference is also small, except for short periods when the speed command
change instantaniously.
Section 4.3 states that a slow reversal of the induction machine may be
problematic. To simulate such a case, a speed governed torque limited load
model is developed. The load speed reference is set to a constant value while
the motor speed command is varied as a saw-tooth signal. Two different
cases are studied, one with step changes in load torque limitation, figure 5.8,
and one with a constant load torque limit and also a deviation in the stator
resistance value, figure 5.9. As seen in figure 5.8, the estimator can handle
almost nominal load with slow ramping around zero excitation frequency.
21
A Sensorless Induction Machine Drive.
0 0.5 1 1.5 2 2.5 3 3.5 4
−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1
Speed difference Model−Estimate with load
Time [sec]
S
p
e
e
d

[
H
z
]
0 0.5 1 1.5 2 2.5 3 3.5 4
−5
0
5
10
15
20
25
30
35
40
45
50
Rotor Speed with load
Time [sec]
S
p
e
e
d

[
H
z
]
Model
Estimate
2.8 3 3.2 3.4 3.6 3.8 4
560
570
580
590
600
610
620
Rotor Angle with load
Time [sec]
A
n
g
l
e

[
r
a
d
]
Model
Estimate
Figure 5.6: Modelled and estimated rotor data during an acceleration and
retardation test, showing both speed, speed difference and angle with 2 Nm
load.
0 0.5 1 1.5 2 2.5 3 3.5 4
0.55
0.6
0.65
0.7
0.75
0.8
Estimated rotor flux
Time [sec]
F
l
u
x

[
W
b
]
0 0.5 1 1.5 2 2.5 3 3.5 4
0.4
0.45
0.5
0.55
0.6
0.65
0.7
0.75
0.8
Estimated rotor flux with load
Time [sec]
F
l
u
x

[
W
b
]
0 0.5 1 1.5 2 2.5 3 3.5 4
0.5
0.55
0.6
0.65
Modelled rotor flux
Time [sec]
F
l
u
x

[
W
b
]
0 0.5 1 1.5 2 2.5 3 3.5 4
0.4
0.45
0.5
0.55
0.6
0.65
Modelled rotor flux with load
Time [sec]
F
l
u
x

[
W
b
]
Figure 5.7: Modelled and estimated rotor flux during an acceleration and
retardation with and without load.
22
A Sensorless Induction Machine Drive.
0 2 4 6 8 10 12
0
2
4
L
o
a
d

t
o
r
q
u
e

[
N
m
]
Time [sec]
0 2 4 6 8 10 12
−10
−8
−6
−4
−2
0
2
4
6
8
10
Slow ramping around zero speed with different loads
R
o
t
o
r

s
p
e
e
d

[
H
z
]
0 2 4 6 8 10 12
0.5
0.52
0.54
0.56
0.58
0.6
0.62
0.64
0.66
0.68
0.7
F
l
u
x

[
W
b
]
Time [sec]
0 2 4 6 8 10 12
0
2
4
6
8
10
C
u
r
r
e
n
t

[
A
]
Time [sec]
Model rotor speed
Estimated rotor speed
Model rotor flux
Estimated rotor flux
i
d
i
q
Figure 5.8: Simulated ramping around zero speed with stepwise increase of
load torque.
The difference in estimated and modelled flux depend upon the different
models used. The motor is implemented in the T model and the control
system parameters are derived from the inverse-Γ model and according to
equation (2.4), the flux differs with the factor
L
m
L
r
. The ability to handle
a deviation of the stator resistance in the motor is not good according to
Figure 5.9. The flux will soon destabilize and the ability to follow the speed
command will be reduced.
5.6 New Parameters
To optimize the usage of the MCU, the control system calculations are best
done with integers. To accomplish this, all parameters in the control system
is recalculated with suitable factors. These factors should be a power of two,
as it is logical for the processor to work with. The factor for resistance is 256
(2
8
), for inductances it is 65536 (2
16
). Integer calculation with small values
will render large errors (if the current is 1.49 A, the system will round it to
1 A). To avoid this, the control system is working with different units. The
unit for current is 10 mA, voltage 10 mV and speed 10 mHz.
23
A Sensorless Induction Machine Drive.
0 5 10 15 20 25 30
−10
−5
0
5
10
15
20
25
Slow ramping around zero speed with constant load and ramped stator resistance
R
o
t
o
r

s
p
e
e
d

[
H
z
]
Model rotor speed
Estimated rotor speed
0 5 10 15 20 25 30
2
2.5
3
S
t
a
t
o
r

r
e
s
i
s
t
a
n
c
e

[

]
Time [sec]
0 5 10 15 20 25 30
0.5
1
1.5
2
2.5
3
F
l
u
x

[
W
b
]
Time [sec]
Model rotor flux
Estimated rotor flux
0 5 10 15 20 25 30
0
2
4
6
8
10
C
u
r
r
e
n
t

[
A
]
Time [sec]
i
d
i
q
Figure 5.9: Simulated ramping around zero with constant load and ramped
stator resistance.
5.7 Simulation Conclusions
The SCVM can handle static load, acceleration and deceleration. It also
seem promising in reversing a speed governed load, at least if the motor
parameters are modelled correctly. The quite high dependency of the stator
resistance may be a problem, but all together it is worth to evaluate this
approach experimentally.
24
Chapter 6
Implementation
A difficult feature of the control system when moving from simulations to
implementation is that the DSP use 16-bit integer representation. This calls
for a careful review of the implemented code. Any part of the computations
exceeding 16-bit (+/- 32768 as most integers used are signed) has to be
revised. It is also good if the code is optimized to use the DSP as effective
as possible. Apart from estimator and control algorithms, there has to be
capacity left for applications.
This chapter describes different parts of the implementation, beginning
with hardware, continuing with software and data acquisition then ending
with implementation issues.
6.1 Hardware
6.1.1 Controller Card
This project uses a hardware setup constructed by Aros electronics AB. The
controller card is a prototype from another project and is to large extent a
standard converter-inverter design.
The board is prepared for using an encoder. In this project, the encoder
is used as a reference for evaluation of the speed estimator. The encoder is
capable of measuring speeds up to 50 Hz.
The communication with the computer during test and programming is
operated by a serial interface. A special interface adapter is used to translate
between the PC and the controller card. It is also equipped with optocouplers
to galvanically isolate the two systems. The earth potential of the controller
card may reach several hundreds of volts compared with the potential of the
grid, thus harming the computer. The optocouplers transfer the signals from
one voltage system to another.
The card has an on-board three phase diode bridge rectifier, a 36MT120
from International Rectifier. It is rated 35 A continuously and 1200 V peak
25
A Sensorless Induction Machine Drive.
DC voltage. DC/DC converters use the main DC voltage to extract the
power levels needed by the control card.
The phase currents are measured for the control system and for the over-
current protection. If a too high current, about 40 A, is registered, a signal
is sent to the processor and the system shuts down. The DC current is also
measured, to if necessary invoke an over-current protection. The DC over-
current trigger is set to 50 A. The control system also has to gain knowledge
about the DC voltage. A voltage divider is connected to a voltage follower
and read by an AD channel on the DSP.
6.1.2 Motor
The test motor is constructed by Bonfiglioli Group in Italy. The rated power
is 1,1 kW and it has two poles. Motor data can be found in Appendix A.
6.2 Software
The software at Aros electronics AB is modular and the same modules appear
in many of the company’s products. One aim with this project is to develop
an estimator that is easy to adapt to present and future projects at Aros.
It should be enough to replace the encoder routine calls with the estimator
ditto, adjust the parameters and the motor drive should be operational.
The estimator is only a small part of the controller software. In this the-
sis, a software package developed at Aros electronics AB is used. The basic
control system already present is not altered. There are some differences be-
tween the simulated and the already implemented control system, the most
important being that the speed controller use reference value filtering. It
controls the speed in two levels, stepping the actual reference value toward
the desired value. Reference value filtering allows a more stable controller.
Another difference is the absence of a voltage decoupler in the control system.
To manipulate the system during test, a special PC-program is used. It
is developed at Aros electronics AB. The software can control and monitor
each variable in the DSP. It also has the possibility to display graphs of up
to three variables at the same time.
6.3 Data Acquisition
There are three possibilities to access the parameters from the processor.
• The control PC-program has a logging feature including a graph tool.
It is possible to log and graph three different signals through the se-
rial control interface. Unfortunately, the sampling time of the system
becomes large and erratic. For system supervising while running the
motor it is adequate, but not for keeping track of transient behavior.
26
A Sensorless Induction Machine Drive.
• Some projects at Aros have used MCU or on-board memory for logging
data which is transferred to the computer as a file after the measure-
ment. The equipment used in this project lacks the memory capacity
needed for such approach.
• The board used is equipped with a Serial Peripheral Interface bus
(SPI). SPI is a loose standard for serial communication. The idea
with the interface in this project is to let the software use pins on the
DSP to put out up to three signals to an external decoding circuit.
The signals can then be studied using an oscilloscope. The sample
rate is decided by how often the SPI code is executed in the DSP.
In this project, the SPI call is located in the switch interrupt, ren-
dering a sample rate equal to the switching frequency. The SPI has
low amplitude resolution, 8 bits. The signal range is 0 to 3.68 V, a
level sensitive to the harsh lab environment. While running the motor,
the switched currents induce an electromagnetic field large enough to
introduce noticeable disturbance when measuring the signals.
As described above, two methods are of interest in this case. The sam-
pling time of the control PC-program is bad. On the other hand, the ampli-
tude resolution of the SPI is narrow. In this thesis, the system is sampled via
the SPI as it is assumed that the transient behavior is of more importance
for evaluation of the SCVM than the amplitude resolution.
6.4 Implementation Issues
Going from simulation to the actual system proved to be troublesome. One
large obstacle were the 16-bit limit for all variables. The calculations need to
be large enough to get good accuracy (see Section 5.6), but sufficiently small
to fit inside the 16-bit representation. Several adjustments had to be done
to accomplish this. Some calculations were also split into several lines to get
a better overview as well as making it easier for the compiler to interpret the
code. There were some pure coding errors that took time to find. Even after
careful review, some issues remained. For a final debug of the system, the
implemented c-code was translated back to Matlab. The code was simulated
bit by bit in the model used to develop the same. The errors were then quite
easily found and corrected.
27
Chapter 7
Results
The implemented design is put through several tests to check the perfor-
mance of the drive. It is of interest to see the operational limits concerning
slow speed, parameter variation and transient behavior as well as speed sta-
bility up to twice the nominal load.
In order to get more information from the measurements, the measured
data are filtered. The filter parameters are created using the cheby2 function
in Matlab with a cutoff frequency of 1500 Hz and stop-band attenuation of
110 dB.
7.1 Speed Step Response
Figures 7.1-7.4 depicts the step response as a speed change from zero to near
nominal speed is applied to the control system. As may be seen, the drive is
capable of handling different loads. The somewhat slower response for the
double nominal load, figure 7.4, is a result from the current limit.
7.2 Slow Speed
The slow speed performance is tested by applying different loads and varying
the speed command. The aim is to find the slow speed limit and the load
limit at slow speed reversal of a speed governed load.
7.2.1 Slow Speed Limit
In the first test, the load is set to -16 Hz and nominal torque, 3.7 Nm. The
control system reference speed is programmed to decrease from 10 Hz to
standstill with steps of 0.5 Hz. Figure 7.5 shows that it is possible to lower
the speed and keep control of the motor, although the accuracy decrease.
In the second test, the reference speed is set to go from -10 Hz to zero.
As may be seen in figure 7.6, the system fails already at 7.5 Hz. The test is
28
A Sensorless Induction Machine Drive.
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
−10
0
10
20
30
40
50
60
Speed command step−change with no load
S
p
e
e
d

[
H
z
]
Time [sec]
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
F
l
u
x

[
W
b
]
Time [sec]
Reference rotor speed
Measured rotor speed
Figure 7.1: Speed step with with no load.
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
−10
0
10
20
30
40
50
60
Speed command step−change with 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
F
l
u
x

[
W
b
]
Time [sec]
Reference rotor speed
Measured rotor speed
Figure 7.2: Speed step with 2 Nm load.
29
A Sensorless Induction Machine Drive.
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
F
l
u
x

[
W
b
]
Time [sec]
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
−10
0
10
20
30
40
50
60
Speed command step−change with nominal load
S
p
e
e
d

[
H
z
]
Time [sec]
Reference rotor speed
Measured rotor speed
Figure 7.3: Speed step with nominal load, 3.7 Nm.
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
F
l
u
x

[
W
b
]
Time [sec]
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
−10
0
10
20
30
40
50
60
Speed command step−change with twice the nominal load
S
p
e
e
d

[
H
z
]
Time [sec]
Reference rotor speed
Measured rotor speed
Figure 7.4: Speed step with double nominal load, 7.5 Nm.
30
A Sensorless Induction Machine Drive.
0 5 10 15 20 25 30 35 40 45 50
0
2
4
6
8
10
12
Decreasing speed with nominal load
S
p
e
e
d

[
H
z
]
Time [sec]
0 5 10 15 20 25 30 35 40 45 50
−2
−1.5
−1
−0.5
0
0.5
1
Difference between reference and measured rotor speed
S
p
e
e
d

[
H
z
]
Time [sec]
Reference rotor speed
Measured rotor speed
Figure 7.5: Dercreasing speed with load and motor operating in the opposite
direction. The load torque limit is set to motor nominal torque, 3.7 Nm.
redone with less load torque, 2 Nm. From the result in figure 7.7 it may be
seen that the accuracy is significantly less compared to the test with positive
movement, but the drive does not destabilize.
7.2.2 Reversing
The purpose is to test consequential slow reversals. As may been seen from
figures 7.8-7.10, the load servo is controlled to a speed of -16 Hz. In all
tests, the rotor is first operated at almost standstill. Keeping zero speed is
difficult, in reality there is a small positive movement of the rotor. In figures
7.8 and 7.9, only the rotor speed, reference speed and estimated rotor speed
is plotted. The drive is quite capable of handling reversal of such loads, but
the transition is less smooth for 2 Nm. For a 3 Nm load, figure 7.10, the
drive loses control but regains it once the speed is increased above zero. To
enable an analysis of the occurring problem, the test is redone with logging
of estimated rotor flux and measured stator currents.
31
A Sensorless Induction Machine Drive.
5 10 15 20 25 30 35
−20
−15
−10
−5
0
5
Decreasing speed with braked nominal load
S
p
e
e
d

[
H
z
]
Time [sec]
5 10 15 20 25 30 35
−10
−8
−6
−4
−2
0
2
4
6
8
10
12
Difference between reference and measured rotor speed
S
p
e
e
d

[
H
z
]
Time [sec]
Reference rotor speed
Measured rotor speed
Figure 7.6: Dercreasing speed with load and motor operating in the same
direction, thus the motor tries to brake the load. The load torque limit is
set to motor nominal torque, 3.7 Nm.
32
A Sensorless Induction Machine Drive.
0 10 20 30 40 50 60 70 80 90
−12
−10
−8
−6
−4
−2
0
2
Decreasing speed with braked 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
0 10 20 30 40 50 60 70 80 90
−3
−2
−1
0
1
2
3
Difference between reference and measured rotor speed
S
p
e
e
d

[
H
z
]
Time [sec]
Reference rotor speed
Measured rotor speed
Figure 7.7: Decreasing speed with load and motor operating in the same
direction, thus the motor tries to brake the load. The load torque limit is
set to 2 Nm.
33
A Sensorless Induction Machine Drive.
10 15 20 25 30 35 40
−15
−10
−5
0
5
Slow ramping around zero speed with 1 Nm load torque
S
p
e
e
d

[
H
z
]
Time [sec]
Estimated rotor speed
Measured rotor speed
Reference rotor speed
Figure 7.8: Slow ramping around zero speed with 1 Nm speed governed load.
10 15 20 25 30 35 40
−15
−10
−5
0
5
Slow ramping around zero speed with 2 Nm load torque
S
p
e
e
d

[
H
z
]
Time [sec]
Estimated rotor speed
Measured rotor speed
Reference rotor speed
Figure 7.9: Slow ramping around zero speed with 2 Nm speed governed load.
34
A Sensorless Induction Machine Drive.
10 15 20 25 30 35 40
−15
−10
−5
0
5
Slow ramping around zero speed with 3 Nm load torque
S
p
e
e
d

[
H
z
]
Time [sec]
Estimated rotor speed
Measured rotor speed
Reference rotor speed
10 15 20 25 30 35 40
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
F
l
u
x

[
W
b
]
Time [sec]
10 15 20 25 30 35 40
−20
−15
−10
−5
0
5
10
15
20
C
u
r
r
e
n
t

[
A
]
Time [sec]
iq
id
Figure 7.10: Slow ramping around zero speed with 3 Nm speed governed
load. The two bottom plots are from a second test.
35
A Sensorless Induction Machine Drive.
7.3 Parameter Sensitivity
To properly test the SCVMs ability to withstand parameter deviations, the
motor parameters would need to be controllable. As this is hard to achieve,
parameter changes are emulated by changing the parameter values in the
implemented code. Doing so, the estimator will be subjected to similar
stresses as if the real motor parameters were to change. The resistance values
are changed to emulate a deviation of ±60 % and the leakage inductance to
emulate ±30 %.
The motor is loaded with 2 Nm, as it seems to be able to handle it but
with some difficulty, see figure 7.9. The aim with these tests is to find out
how the performance at nominal speed and at low speed ramping about zero
is affected by variations of stator and rotor resistance and leakage inductance.
7.3.1 Deviations in Stator Resistance
To test the estimators ability to cope with an increase in stator resistance,
R
s
is lowered in the control system. The test is conducted in two modes, slow
ramping around zero with 2 Nm speed governed load, figure 7.11, and near
the nominal speed with the same load, figure 7.12. The test is not conducted
exactly at nominal speed as the encoder with its interface is unable to handle
more than 50 Hz. Nominal rotor speed is 46.9 Hz, but with overshoot and
other effects, it may be too large. Therefore, the rotor speed setpoint is 45
Hz. Each mode is tested twice, first to get encoder speed, estimated rotor
speed and rotor resistance of the control system. In the second run, the
estimated rotor flux and the measured, dq-transformed, currents are logged.
This is due to the limitations of the SPI interface, see Section 6.3.
As seen in figure 7.11, the stability and accuracy reduce as the control
system stator resistance is lowered. When the control system stator resis-
tance is increased at low speed operation, figure 7.13, the drive fails. To see
more exactly where this phenomena occurs, an even slower ramp is tested,
see figure 7.14. From the graph it is found to happen at approximately
3.25 Ω. This corresponds to a motor stator resistance decrease by 37 %.
At higher speeds, a control system stator resistance change affect the drive
insignificantly, as seen in Figures 7.12 and 7.15.
Especially in the low speed region, the SCVM is sensitive to inaccurately
modeled stator resistance. However, it is not as bad as the simulations in
Section 5.5 indicated. In the high speed region, the impact of the R
s
i
s
part in
the equations (5.18) and (5.19) becomes less due to the increased importance
of ´ ω
1
L
σ
i
s
.
36
A Sensorless Induction Machine Drive.
10 15 20 25 30 35 40 45 50 55 60 65
−10
−8
−6
−4
−2
0
2
4
6
8
Slow ramping around zero speed with decreased control system stator resistance and 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
Estimated rotor speed
Measured rotor speed
10 15 20 25 30 35 40 45 50 55 60 65
1
1.5
2
2.5
S
t
a
t
o
r

r
e
s
i
s
t
a
n
c
e

[

]
Time [sec]
10 15 20 25 30 35 40 45 50 55 60 65
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
F
l
u
x

[
W
b
]
Time [sec]
10 15 20 25 30 35 40 45 50 55 60 65
−6
−4
−2
0
2
4
6
8
10
C
u
r
r
e
n
t

[
A
]
Time [sec]
i
q
i
d
Figure 7.11: Speed, estimated rotor flux and measured currents at ramped
speed around zero as the control system stator resistance is changed. A lower
resistance means emulating an increase of motor stator resistance. The test
is conducted twice, first to log speeds and resistance and secondly to log flux
and currents.
37
A Sensorless Induction Machine Drive.
35 40 45 50 55 60
43
43.5
44
44.5
45
45.5
46
46.5
47
47.5
48
48.5
Decreased control system stator resistance near nominal speed with 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
35 40 45 50 55 60
1
1.5
2
2.5
S
t
a
t
o
r

r
e
s
i
s
t
a
n
c
e

[

]
Time [sec]
35 40 45 50 55 60
0.6
0.62
0.64
0.66
0.68
0.7
0.72
0.74
0.76
0.78
0.8
F
l
u
x

[
W
b
]
Time [sec]
35 40 45 50 55 60
3.5
4
4.5
5
5.5
6
6.5
7
C
u
r
r
e
n
t

[
A
]
Time [sec]
Estimated rotor speed
Measured rotor speed
i
q
i
d
Figure 7.12: Speed, estimated flux and measured currents at near nominal
speed as the control system stator resistance is changed. A lower resistance
means emulating an increase of motor stator resistance. The test is con-
ducted twice, first to log speeds and resistance and secondly to log flux and
currents.
38
A Sensorless Induction Machine Drive.
40 50 60 70 80 90 100 110 120 130
−12
−10
−8
−6
−4
−2
0
2
4
6
Slow ramping around zero speed with increased control system stator resistance and 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
40 50 60 70 80 90 100 110 120 130
2
4
6
S
t
a
t
o
r

r
e
s
i
s
t
a
n
c
e

[

]
Time [sec]
40 50 60 70 80 90 100 110 120 130
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
F
l
u
x

[
W
b
]
Time [sec]
Estimated rotor speed
Measured rotor speed
40 50 60 70 80 90 100 110 120 130
−4
−2
0
2
4
6
8
10
12
14
16
C
u
r
r
e
n
t

[
A
]
Time [sec]
i
q
i
d
Figure 7.13: Speed, estimated flux and measured currents at ramped speed
around zero as the control system stator resistance is increased. The test is
conducted twice, first to log speeds and resistance and secondly to log flux
and currents.
39
A Sensorless Induction Machine Drive.
65 66 67 68 69 70 71 72 73 74 75
−8
−6
−4
−2
0
2
4
6
Slow ramping around zero speed with slowly increasing control system stator resistance and 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
65 66 67 68 69 70 71 72 73 74 75
3
3.1
3.2
3.3
3.4
3.5
S
t
a
t
o
r

r
e
s
i
s
t
a
n
c
e

[

]
Time [sec]
Estimated rotor speed
Measured rotor speed
Figure 7.14: Estimated and measured speed as the control system stator
resistance is slowly increased to see the troublesome resistance level.
40
A Sensorless Induction Machine Drive.
30 40 50 60 70 80 90 100 110 120 130
43
44
45
46
47
48
49
Increased control system stator resistance near nominal speed with 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
Estimated rotor speed
Measured rotor speed
30 40 50 60 70 80 90 100 110 120 130
2
4
6
S
t
a
t
o
r

r
e
s
i
s
t
a
n
c
e

[

]
Time [sec]
30 40 50 60 70 80 90 100 110 120 130
0.56
0.58
0.6
0.62
0.64
0.66
0.68
0.7
0.72
F
l
u
x

[
W
b
]
Time [sec]
30 40 50 60 70 80 90 100 110 120 130
2
3
4
5
6
7
8
C
u
r
r
e
n
t

[
A
]
Time [sec]
i
q
i
d
Figure 7.15: Speed, estimated flux and measured currents at near nominal
speed as the control system stator resistance is first increased, then lowered
to the nominal value. The test is conducted twice, first to log speeds and
resistance and secondly to log flux and currents.
41
A Sensorless Induction Machine Drive.
10 15 20 25 30 35 40 45 50 55 60
1.2
1.4
1.6
1.8
2
2.2
R
o
t
o
r

r
e
s
i
s
t
a
n
c
e

[

]
Time [sec]
15 20 25 30 35 40 45 50 55 60
0.5
0.6
0.7
0.8
0.9
1
F
l
u
x

[
W
b
]
Time [sec]
10 15 20 25 30 35 40 45 50 55 60
−6
−4
−2
0
2
4
6
Slow ramping around zero speed with decreased control system rotor resistance and 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
15 20 25 30 35 40 45 50 55 60
−4
−2
0
2
4
6
8
C
u
r
r
e
n
t

[
A
]
Time [sec]
Estimated rotor speed
Measured rotor speed
i
q
i
d
Figure 7.16: Speed, estimated rotor flux and measured currents at speed
ramped about zero as the control system rotor resistance is changed. A
lower resistance means emulating an increase of motor rotor resistance. The
test is conducted twice, first to log speeds and resistance and secondly to log
flux and currents.
7.3.2 Deviations in Rotor Resistance
The rotor resistance affects the slip calculation. The tests are conducted in
the same way as for the stator resistance. As it may be seen in figure 7.16,
the low speed operation of the SCVM hardly change at all with a decrease
in control system rotor resistance. Figures 7.17 and 7.19 on the other hand
depicts that in nominal or high speed region, the impact on the accuracy
is more significant. As the control system rotor resistance is decrease in
figure 7.17, the measured rotor speed decrease approximately 2 %. If the
motor rotor resistance decrease much, the drive will even destabilize. At low
speed, figure 7.18, the speed oscillation starts when the motor rotor speed is
emulated to have decreased with 58 %. At near nominal speed, figure 7.19,
the drive destabilize at 53 %.
Equation (5.21) models a linear relation between rotor resistance and slip
speed. The slip is small at low speeds.
42
A Sensorless Induction Machine Drive.
35 40 45 50 55 60 65
43
43.5
44
44.5
45
45.5
46
46.5
47
47.5
48
Decreased control system rotor resistance near nominal speed with 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
35 40 45 50 55 60 65
1.2
1.4
1.6
1.8
2
2.2
R
o
t
o
r

r
e
s
i
s
t
a
n
c
e

[

]
Time [sec]
35 40 45 50 55 60 65
0.63
0.64
0.65
0.66
0.67
0.68
0.69
0.7
0.71
0.72
F
l
u
x

[
W
b
]
Time [sec]
35 40 45 50 55 60 65
3
3.5
4
4.5
5
5.5
6
6.5
7
C
u
r
r
e
n
t

[
A
]
Time [sec]
i
q
i
d
Estimated rotor speed
Measured rotor speed
Figure 7.17: Speed, estimated rotor flux and measured currents at near
nominal speed as the control system rotor resistance is changed. A lower
resistance means emulating an increase of motor stator resistance. The test
is conducted twice, first to log speeds and resistance and secondly to log flux
and currents.
43
A Sensorless Induction Machine Drive.
30 40 50 60 70 80 90 100 110 120
2
3
4
5
R
o
t
o
r

r
e
s
i
s
t
a
n
c
e

[

]
Time [sec]
40 50 60 70 80 90 100 110 120 130
0.4
0.45
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
F
l
u
x

[
W
b
]
Time [sec]
40 50 60 70 80 90 100 110 120 130
−4
−2
0
2
4
6
8
10
12
14
16
C
u
r
r
e
n
t

[
A
]
Time [sec]
30 40 50 60 70 80 90 100 110 120
−15
−10
−5
0
5
10
15
20
Slow ramping around zero speed with increased control system rotor resistance and 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
Estimated rotor speed
Measured rotor speed
i
q
i
d
Figure 7.18: Speed, estimated flux and measured currents at ramped speed
around zero as the control system rotor resistance is first increased, then
lowered to the nominal value. The test is conducted twice, first to log speeds
and resistance and secondly to log flux and currents.
44
A Sensorless Induction Machine Drive.
40 50 60 70 80 90 100 110 120
−20
−10
0
10
20
30
40
50
60
Increased control system rotor resistance near nominal speed with 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
40 50 60 70 80 90 100 110 120
2
4
6
R
o
t
o
r

r
e
s
is
t
a
n
c
e

[

]
Time [sec]
40 50 60 70 80 90 100 110 120
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
F
lu
x

[
W
b
]
Time [sec]
40 50 60 70 80 90 100 110 120
−10
−5
0
5
10
15
20
C
u
r
r
e
n
t

[
A
]
Time [sec]
Estimated rotor speed
Measured rotor speed
i
q
i
d
Figure 7.19: Speed, estimated flux and measured currents at near nominal
speed as the control system rotor resistance is first increased, then lowered
to the nominal value. The test is conducted twice, first to log speeds and
resistance and secondly to log flux and currents.
45
A Sensorless Induction Machine Drive.
0 10 20 30 40 50 60
0.01
0.012
0.014
L
e
a
k
a
g
e

i
n
d
u
c
t
a
n
c
e

[
H
]
Time [sec]
0 10 20 30 40 50 60
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
F
l
u
x

[
W
b
]
Time [sec]
0 10 20 30 40 50 60
−4
−2
0
2
4
6
8
10
12
14
16
C
u
r
r
e
n
t

[
A
]
Time [sec]
0 10 20 30 40 50 60
−6
−4
−2
0
2
4
6
Slow ramping around zero speed with decreased control system leakage inductance and 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
Estimated rotor speed
Measured rotor speed
i
q
i
d
Figure 7.20: Speed, estimated rotor flux and measured currents at speed
ramped about zero as the control system leakage inductance is decreased.
A lower inductance means emulating an increase of motor inductance. The
test is conducted twice, first to log speeds and inductance and secondly to
log flux and currents.
7.3.3 Deviations in Leakage Inductance
As may be seen from Figures 7.20-7.23, the drive is quite unaffected as the
total leakage inductance is changed.
46
A Sensorless Induction Machine Drive.
30 35 40 45 50 55 60 65 70
0.01
0.012
0.014
L
e
a
k
a
g
e

i
n
d
u
c
t
a
n
c
e

[
H
]
Time [sec]
10 20 30 40 50 60 70 80 90 100
0.6
0.65
0.7
F
l
u
x

[
W
b
]
Time [sec]
30 35 40 45 50 55 60 65 70
3
4
5
6
7
8
9
10
C
u
r
r
e
n
t

[
A
]
Time [sec]
30 35 40 45 50 55 60 65 70
44.5
45
45.5
46
46.5
47
47.5
48
Decreased control system leakage inductance near nominal speed with 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
i
q
i
d
Estimated rotor speed
Measured rotor speed
Figure 7.21: Speed, estimated rotor flux and measured currents at near
nominal speed as the control system leakage inductance is decreased. A
lower inductance means emulating an increase of motor inductance. The
test is conducted twice, first to log speeds and inductance and secondly to
log flux and currents.
47
A Sensorless Induction Machine Drive.
0 10 20 30 40 50 60 70 80 90 100
0.012
0.014
0.016
0.018
0.02
L
e
a
k
a
g
e

i
n
d
u
c
t
a
n
c
e

[
H
]
Time [sec]
0 10 20 30 40 50 60 70 80 90 100
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
F
l
u
x

[
W
b
]
Time [sec]
0 10 20 30 40 50 60 70 80 90 100
−6
−4
−2
0
2
4
6
8
10
12
14
C
u
r
r
e
n
t

[
A
]
Time [sec]
0 10 20 30 40 50 60 70 80 90 100
−6
−4
−2
0
2
4
6
Slow ramping around zero speed with increased control system leakage inductance and 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
Estimated rotor speed
Measured rotor speed
i
q
i
d
Figure 7.22: Speed, estimated flux and measured currents at ramped speed
around zero as the control system leakage inductance is increased. The test
is conducted twice, first to log speeds and inductance and secondly to log
flux and currents.
48
A Sensorless Induction Machine Drive.
10 20 30 40 50 60 70 80 90 100
0.012
0.014
0.016
0.018
0.02
L
e
a
k
a
g
e

i
n
d
u
c
t
a
n
c
e

[
H
]
Time [sec]
10 20 30 40 50 60 70 80 90 100
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
1.05
1.1
F
l
u
x

[
W
b
]
Time [sec]
10 20 30 40 50 60 70 80 90 100
3
4
5
6
7
8
9
10
C
u
r
r
e
n
t

[
A
]
Time [sec]
10 20 30 40 50 60 70 80 90 100
44.5
45
45.5
46
46.5
47
47.5
48
Increased control system leakage inductance near nominal speed with 2 Nm load
S
p
e
e
d

[
H
z
]
Time [sec]
Estimated rotor speed
Measured rotor speed
i
q
i
d
Figure 7.23: Speed, estimated flux and measured currents at near nominal
speed as the control system leakage inductance is increased. The test is
conducted twice, first to log speeds and inductance and secondly to log flux
and currents.
49
Chapter 8
Conclusions
The implementation of the Statically Compensated Voltage Model was com-
pleted with some success. As shown in Chapter 7, the goals were partly
met.
8.1 Goal Fulfilment
In this section, the results are compared to the set goals.
8.1.1 Stable Operation up to Nominal Speed
The drive works well at nominal speed. Low speed operation is more prob-
lematic. Keeping zero speed is not possible with the implemented design.
At slow operation, the accuracy is reduced. Slow braking is possible with 2
Nm load but not with nominal load, 3.7 Nm, where it is unstable already at
7.5 Hz.
8.1.2 Stable Operation up to Twice the Nominal Torque
It has been shown that the SCVM is capable of handling large loads at
nominal speeds. The limitations are more dependent on the inverter design
and the power source present. At low speed, larger loads are not possible.
8.1.3 Stable Operation during Acceleration and Deceleration
with an Arbitrary Load
Slow reversal with more than moderate load is not possible. The problems
occur in the braking mode.
8.1.4 Stable Operation during Parameter Variations
As the motor parameters are not controllable, changes in motor data are
emulated by changing the control system parameters. In this way, the control
50
A Sensorless Induction Machine Drive.
system is put through similar stress as if the real motor parameters change.
• The drive is sensitive to changes in the value of the stator resistance.
It is stable for variations between -37 % and +60 % in low speed
operation. In the high speed region, the drive is less sensitive to stator
resistance deviations. It can handle a discrepancy of ±60 % before
becoming unstable.
• The rotor resistance may decrease by 58 % at low speed and by 53
% at nominal speed before the drive destabilizes. Increasing the rotor
resistance with 60 % will not affect the stability at any speed but the
nominal speed accuracy will be reduced by approximatly 2 %.
• The implemented design is capable of handling deviations with 30 %
in the total leakage inductance without any noticeable effect on per-
formance at any speed.
8.2 Further Work
To improve the drive, one important contribution would be an on-line param-
eter estimator. A good solution should be easy to understand and implement.
On-line tuning would also minimize the risk of failing units due to scatter in
motor parameters. It may also open the possibility for a stand-alone sensor-
less drive, ready for operation with almost any IM within its power range.
The effects of parameter variations could perhaps be less by implementing
the suggested improvements in Section 4.3.
51
Appendix A
Measurement of Motor
Parameters
Two types of tests were conducted; a locked-rotor test to determine R
r
,
L
r
and L
s
and a no load test to get L
m
. The stator resistance R
s
was
measured directly on the motor terminals. All measurements are performed
on a relatively cold motor.
The motor is produced by Bonfiglioli Group and is marked with the data
found in table A.1. From these data, the approximate nominal flux level is
calculated by
T = i
q
λ
r
(A.1)
T =
P
ω
r
(A.2)
λ
r
=
P
ω
r
i
q
. (A.3)
With the parameters from table A.1, the approximate nominal flux level for
this motor is λ

dr
= 0.605Wb.
The motor-bench at Aros electronics AB is equipped with a Panasonic
MSMA402A1G AC servo motor (serial no: 03090001F) which is used as
load. An Infratek 106A Power Analyzer (serial no: 03083312) is used for
measuring currents, voltage, active power, apparent power and power factor.
V ∆ / Y Hz kW A ∆ / Y rpm cos(φ)
230/400 50 1.1 4.41/2.55 2815 0.81
—/460 60 1.3 —/2.5 3420
Table A.1: Motor data from nameplate.
52
A Sensorless Induction Machine Drive.
I
a
(A) P
tot
(W) S
tot
(VA) Q
tot
(VAr)
L
lr
+L
ls
2
(mH) R
r
(Ω)
1.02 12.63 18.29 13.23 6.75 2
2.01 49.05 71.53 52.06 6.84 2
3.01 110.19 160.07 116.11 6.80 2.01
4.06 203.52 292.97 210.74 6.78 2.07
5.04 310.48 447.57 322.37 6.78 2.02
Table A.2: Results from locked rotor test.
A.1 Locked Rotor Test
With the rotor locked, the slip is equal to one. This gives minimum effective
rotor resistance, thus minimizing the current through the magnetizing induc-
tance, which may be neglected. The motor steady-state T model equivalent
circuit is then reduced to figure A.1. The rotor of the IM was locked by
applying a large braking torque on the servo. The current was increased
in steps and the required voltage noted together with apparent power and
active power. The reactive power is calculated from
Q
tot
=
_
S
2
tot
−P
2
tot
. (A.4)
V
a
R
s
L
ls
L
lr
L
m R
r
Figure A.1: Reduced T-model for locked rotor test.
The rotor and stator inductances L
lr
and L
ls
are assumed equal. The
reactive power is then equal to
Q
phase
= ω(L
lr
+L
ls
)I
2
a
(A.5)
where Q
phase
= Q
tot
/3. The result of the measurements is found in table A.2.
The mean value of these tests yields L
ls
= L
lr
= 6.79 (mH). Using the
measured stator resistance, R
s
= 2.05 (Ω), the rotor resistance, R
r
, can be
calculated using the measured active power
P
phase
= (R
r
+R
s
)I
2
a
. (A.6)
The mean value of R
r
= 2.02 (Ω).
53
A Sensorless Induction Machine Drive.
I
a
(A) P
tot
(W) S
tot
(VA) Q
tot
(VAr) L
m
(mH)
2.83 143.36 1134.3 1125.2 144.6
2.95 153.79 1192.5 1182.5 139.7
2.91 148.20 1170.3 1160.9 140.6
Table A.3: Results from no-load test.
A.2 No-load Test
Nominal voltage is supplied to the motor. The servo motor drives the IM
at synchronous speed to control the slip speed to zero. The effective rotor
resistance becomes large and can be neglected. The equivalent circuit is
seen in figure A.2. Current, active power, apparent power and power factor
=>
0 s
R
R
s
V
a
R
s
L
ls
L
lr
L
m
Figure A.2: Reduced T-model for locked rotor test.
is noted.
The current I
a
splits between the magnetizing inductance L
m
and the
core loss resistance R
c
according to
I
m
= sin(φ)I
a
. (A.7)
To extract the magnetizing inductance L
m
, the following relation is used:
Q
phase
= I
2
a
ω
1
L
l
s +I
2
m
ω
1
L
m
(A.8)
The values from the power analyzer is read three times with aid of a hold
function. The test results can be found in table A.3. The mean value of the
magnetizing inductance is found to be 141.6 mH.
54
Bibliography
[1] A. Ottergren, “Evaluation of speed observers for an induction machine,”
Master’s thesis, Department of Electric Power Engineering, Chalmers
University of Technology, Gothenburg, 2003.
[2] L. Harnefors, Control of Variable-Speed Drives. Västerås, Sweden: De-
partment of Electronics, Mälardalen University, 2002.
[3] ——, Control of Power Electronic Converters and Variable-Speed Drives.
Västerås, Sweden: Department of Electronics, Mälardalen University,
2005.
[4] G. R. Slemon, “Modelling of induction machines for electric drives,” IEEE
Transactions on Industry Applications, vol. 25, no. 6, pp. 1126–1131,
November/December 1989.
[5] R. Krishnan, Electric Motor Drives. Upper Saddle River, New Jersey:
Prentice Hall, 2001.
[6] M.-H. Shin, D.-S. Hyun, S.-B. Cho, and S.-Y. Choe, “An improved stator
flux estimation for speed sensorless stator flux orientation control of in-
duction motors,” IEEE Transactions on Power Electronics, vol. 15, no. 2,
pp. 312–318, March 2000.
[7] R. Ottersten, “On control of back-to-back converters and sensorless induc-
tion machine drives,” Ph.D. dissertation, Department of Electric Power
Engineering, Chalmers University of Technology, Gothenburg, 2002.
[8] N. Mohan, Electric Drives. Minneapolis, Minnesota: MNPERE, 2003.
55

Abstract This Master’s thesis describes a DSP implementation of the Statically Compensated Voltage Model (SCVM) algorithm for sensorless induction machine control. The algorithm is aimed to be implemented in a 16-bit fixed-point processor. In general, low speed operation with voltage model based control systems is problematic, especially when in generator mode. In generator mode with the load at nominal torque, the implemented design fails already at 7.5 Hz. With a lower load, 2 Nm or 54 % of nominal, the drive is able brake it to almost zero, but the speed accuracy is low. On the other hand, if the motor operates in motor mode, low speed is less problematic. Experiments show that this implementation can handle variations in stator resistance from -37 % up to +60 % with stable low speed operation and variations of at least ±60 % at nominal speed. Stable operation for all speeds is possible even during an increase of rotor resistance by 60 %, but the speed accuracy will be reduced by approximatly 2 %. The rotor resistance may decrease by 58 % at low speed and by 53 % at nominal speed before the drive destabilizes. The total leakage inductance can vary at least ±30 % without noticeably affecting the system. The thesis project is conducted in association with Aros electronics AB.

.

Andreas Petersson for their help with theoretical questions and alternative approaches to problems. student Oskar Wallmark has also been a sorce of inspiration during this thesis work. a warm thanks to all employees at Aros electronics AB for their kindness and support. and to Ph. Last but certainly not least. Ph.D. Mikael Alatalo. help and encouragement. Rolf Ottersten has through their work with the SCVM and to some extent also by helping us directly inspired and motivated our work. making us feel most welcome during our thesis work. We also extend our thanks to our examiner. for his unwavering support. associate professor Torbjörn Thiringer.D. .D. Professor Lennart Harnefors and Ph.Acknowledgments We would like to thank our supervisor at Aros electronics AB. The work of Ph.D.

.

. . . . . . . . .1 Mechanical Construction . . . . .4 Structure 1 1 1 2 2 3 3 3 6 8 8 9 9 9 . .6 New Parameters . . . . . . . . . . . 3 Vector Control 3. . . . . . . . . . . . . . . . . . . . . . . . . .4 The Improved SCVM Model 5. . . . . . . . . . .4 Sensorless Motor Control . . . . 25 V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Synchronous Coordinates 3. . . 14 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Induction Motor Basics 2. .Contents 1 Introduction 1. . . . . . . . 3. . . 16 16 18 18 20 21 23 24 . . . . . . . . . 5. . . . . 1. . . . . . . . .2 Parameter Sensitivity of the SCVM . . . 5. . . . . .1 Hardware . . . . . . . . . . . . . . . . . . . . . . .2 Model Verification . . . . 14 5 Simulations 5. . .3 Sources . . . . . . 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Known Limitations of the SCVM . . . . . . . . . . . .1 Controller Card . . . . . . . . . . . . . . . . . . 2.2 Goal . . . . . . . . . . . . . . . . . . . . . 3. . . . .5 Simulation Results . . . . . . . 25 6. . . . . . . . . . .1 Motor Model . . . 6 Implementation 25 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Purpose . . . . . . . . . . . . . . . . . . 4 Sensorless Control Algorithm 12 4. . . 2. . . . . . . . . . . . . . . . . . . . . . . . 1. . . . . . . 12 4. . . . . . . . . . . . . . . . .3 Motor Control . 5. . . . . .7 Simulation Conclusions . .1. . . . . . . . . . . . . . . . . . . . . . .2 Mathematical Circuit Models . . . . . .1 Space Vectors . . . . . . .3 Principle of Operation . . . . .3 Control System Model .1 Improved Statically Compensated Voltage Model . .

A Sensorless Induction Machine Drive.

6.2 6.3 6.4

6.1.2 Motor . . . . . Software . . . . . . . . Data Acquisition . . . Implementation Issues

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

26 26 26 27 28 28 28 28 31 36 36 42 46

7 Results 7.1 Speed Step Response . . . . . . . . . . . 7.2 Slow Speed . . . . . . . . . . . . . . . . 7.2.1 Slow Speed Limit . . . . . . . . . 7.2.2 Reversing . . . . . . . . . . . . . 7.3 Parameter Sensitivity . . . . . . . . . . . 7.3.1 Deviations in Stator Resistance . 7.3.2 Deviations in Rotor Resistance . 7.3.3 Deviations in Leakage Inductance

8 Conclusions 50 8.1 Goal Fulfilment . . . . . . . . . . . . . . . . . . . . . . . . . . 50 8.1.1 Stable Operation up to Nominal Speed . . . . . . . . . 50 8.1.2 Stable Operation up to Twice the Nominal Torque . . 50 8.1.3 Stable Operation during Acceleration and Deceleration with an Arbitrary Load . . . . . . . . . . . . . . . 50 8.1.4 Stable Operation during Parameter Variations . . . . . 50 8.2 Further Work . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 A Measurement of Motor Parameters 52 A.1 Locked Rotor Test . . . . . . . . . . . . . . . . . . . . . . . . 53 A.2 No-load Test . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

VI

Chapter 1

Introduction
Vector control of induction machines (IM:s) is a well known tool for advanced control of AC-machines. This method of control requires knowledge of the rotor flux angle. Traditionally, this has been derived using current measurements and information about the rotor speed, measured by a device mounted on the shaft of the machine. However, there are several problems involved with the speed measurement. The measuring device, often an encoder or a resolver, is expensive and adds another component that may fail in the system. The drive will also be sensitive to disturbances and harsh operational environments. The measuring device also need space for mounting, which may prove problematic in some applications, as well as communication between the device and the control system. Careful mounting and alignment are often necessary, further increasing the production costs. The problems concerning rotor speed measurement have spurred a development of methods for sensorless flux angle estimation and thus advanced control of the machine without mechanical sensors.

1.1

Purpose

This Master’s thesis describes the implementation of the Statically Compensated Voltage Model flux- and speed-sensorless control strategy for an induction machine. Some basic vector control theory is also covered.

1.2

Goal

The goal is to achieve a sensorless IM drive that enables stable operation • up to the nominal speed. • up to twice the nominal torque. • during acceleration and deceleration with an arbitrary load. 1

A Sensorless Induction Machine Drive.

• during variations in rotor and stator resistance up to 60 % and in the total leakage inductance up to 30 %.

1.3

Sources

There is much research done on the subject of sensorless control. Numerous articles and papers has been published. Earlier work at Aros electronics AB includes a Master’s thesis by Anders Ottergren [1], "Evaluation of speed observers for an induction machine". This paper is considered in Chapter 4. Professor Lennart Harnefors has written two very informative compendia about control of variable speed drives, [2] and [3]. The second is partly an extended version of the first, covering also control of power electronic converters. Harnefors presents a promising algorithm called the Statically Compensated Voltage Model or SCVM. This is the approach we will further study.

1.4

Structure

The theory of vector control is covered in Chapter 3 to give an introduction to the subject of field oriented motor drives. The SCVM method for sensorless control is discussed in Chapter 4. In Chapter 5 the control system and the mathematical model of the motor is described and implemented using Matlab Simulink. The results of the different simulations are displayed and discussed here as well. The main part of this thesis is the implementation, Chapter 6, and the results here of are described in Chapter 7. The report is ended with a conclusion part, Chapter 8.

2

Contrary to the T model. Larger airgap means higher losses in the stator windings. some basic induction motor principles are explained. Rc .1 [4]. In this thesis it is excluded. the stator side. 2. There is a direct translation between T and inverse-Γ parameters: 3 .2. the rotor conductors are short-circuited and are. This makes it important for the airgap between the stator and rotor to be small in order to limit the flux leakage. The rotor is indirectly propelled by the magnetic field produced by the stator. The magnetizing impedance is represented by an inductance. nowadays mostly made of solid aluminum. see figure 2. connected in parallel or in series with the magnetizing inductance Lm . with exception from the parameter calculations in appendix A. all leakage inductance is transferred to one side. with less torque pulsations and lower acoustic noise as a result [2]. If included. unlike the DC-machine. 2. In most cases the rotor slots are skewed which reduces nonlinear effects and in turn harmonic content of the flux.Chapter 2 Induction Motor Basics As a background to motor control. hence induction motor. see figure 2. it is represented by a resistance.1 Mechanical Construction The fundamental parts of the induction motor are the stator and the rotor.2 Mathematical Circuit Models A classical way to model the induction motor is the single phase "T model". giving a lower efficiency. In the IM. It consists of the resistance and inductance of the stator and rotor. Another suitable equivalent model is the "inverse-Γ model". Sometimes the active power loss in the core is included in the model. but hardly ever in simulations. compared to earlier constructions of plain copper wiring.

A Sensorless Induction Machine Drive.2: Single phase inverse-Γ model of an induction motor.1: Single phase T model of an induction motor. Rs Lls Llr Va Lm Rr s Figure 2. 4 . Rs Lσ Va LM RR s Figure 2.

L2 m Lr L2 LM = m Lr Lm 2 RR = Rr Lr Lm ψR = ψr Lr Lσ = Ls − (2.3: Single phase dynamic T model of an induction motor.4: Single phase dynamic inverse-Γ model of an induction motor. Since they are steady-state set-ups. these models are not feasible for analysis of the dynamic behavior of the IM. For a more detailed description of back EMF. two models valid for transient analysis are used [3]. 5 .3 and 2.3) (2. Rs Lσ RR Va LM jωr ψR V Figure 2.2) (2. Instead.4. see Section 2. Rs Lls Llr Rr Va Lm jωr ψr V Figure 2. the effective rotor resistance dependent of the slip is replaced with a voltage source corresponding to the back EMF.4) where Ls = Lls + Lm and Lr = Llr + Lm . In figure 2.1) (2.3.A Sensorless Induction Machine Drive.

ωr = RR ψs (2.7) (2.8) jωr ψ s − RR is − LM R R 6 . stator currents are formed. A rotating magnetic field with varying flux density is generated. If the machine is not loaded.5 is a sketch of the AC induction motor torque and steady-state torque limits due to thermal limitations in the machine. As an effect the torque is reduced. the induced voltage and current will drop. is applied over the magnetizing inductance LM in b figure 2. This will brake the load [5]. The rotor voltage is induced as long as the stator field is not moving with the same speed as the rotor. As alternating voltages displaced 120 degrees to each other are applied to the stator windings. As the rotor accelerates. If the load is driving the rotor faster than the stator field.5 is due to mechanical construction limitations of the rotor. If the motor is loaded.A Sensorless Induction Machine Drive. the rotor practically keeps the same speed as the stator field. thus producing an electrodynamic torque.4 [3]. will be small and the torque almost zero. inducing a voltage called rotor EMF. The magnetic flux induce a voltage over the rotor resulting in a current trough the rotor windings. called slip speed or lag.The mechanical speed of the motor is often represented by (Te − TL ) −Bt/J e (2. J the moment of inertia and B the viscous friction. the flowing current will generate a force acting upon the rotor windings.5). this is represented by s the term jωr ψR . TL the applied load. The rotor current is producing a magnetic field of its own. As there is a magnetic field present. figure 2. Electro-Motive Force. a voltage called back EMF. Thus the speed difference. Together with the voltage drop over the rotor. Es .6) R LM From figure 2. The maximum speed noted in figure 2. In the dynamic models. counteracting the magnetic field produced by the stator.5) J where Te is the electrodynamic torque. 2. the voltage over the rotor change polarity and the produced torque will be in the reverse direction compared to the stator field movement. the rotor speed will be lower and the torque increases until the load is matched according to equation (2.4 the voltage equations for the inverse-Γ model can be derived: Es = jωr − b vs − Rs is − Lσ s s dis dis s − LM M = 0 dt dt dis M =0 dt (2.3 Principle of Operation The three-phase induction motor is based upon electro-magnetic interaction between the stator and rotor.

11) it may be concluded that the difference at nominal speeds are small.10) (2.9) The voltage equations may now be written as LM dψ s dis R s = v s − Rs i s − s s dt dt RR − jωr ψ s R LM (2. As the rotor currents are hard to measure.8) by introducing is = is − is = R M s ψs R − is . s LM (2. Comparing equation (2. T IM torque Field weakening region Maximum speed ω Steady-state torque limit Figure 2.A Sensorless Induction Machine Drive. they are eliminated from equations (2.6) with (2.5: Comparison between IM-torque curves for a fixed stator frequency and for vector control with steady-state torque limit and maximum speed. Es .7) and (2.11) dψ s R = Rs i s − s dt dψ s Harnefors defines dtR as the flux EMF. Hence Es is often referred to as back EMF [3]. 7 .

It is also convenient to think of these axes as the real and imaginary axes in a complex plane. with the axes denoted as α and β. Vector control relies on field orientation. it is possible to describe the three-phase system as an equivalent twophase system. is used to get from two-phase back to three-phase [2] 8 . meaning that the vector rotates with the stator speed which is zero. vα and vβ are calculated using the following transformation matrix vα vβ 2 3 1 −3 1 √ 3 =K 0 1 − √3 −1 3  T32 va (t)    vb (t)  vc (t)  (3. The inverse matrix of T32 . hence “space” [2]. 3. The voltage vector in the complex plane is then denoted as vs (t) = vα (t) + jvβ (t). Due to the fact that va (t) + vb (t) + vc (t) = 0 ∀ t.1 Space Vectors Space vectors were originally invented to describe spatial flux distributions in AC machines. T23 . In all rotating machines there is a magnetic field and if the control signals are aligned to this field. The superscript “s” in vs means that the space vector is referred to the stator reference frame.1) were K is a scaling factor: Peak-value/Amplitude-invariant scaling: K = 1 √ RMS-value scaling: K = 1/ 2 Power-invariant scaling: K = 3/2. with two perpendicular axes. field orientated control is achieved.Chapter 3 Vector Control The idea of vector control is to mathematically represent an AC machine as a DC machine [2].

As the reference frame rotates with the supply/synchronous frequency. Vs . the control signals needs to be DC quantities. which are transformed to the equivalent two phase system.A Sensorless Induction Machine Drive. the traditional way of controlling an IM often involves a shaft sensor that measures the speed/position of the rotor.2 Synchronous Coordinates The purpose of vector control is. v = e−jω1 t vs were ω1 is the synchronous frequency. Most significant for applying vector control is the measuring of the stator currents. 3.  1 va (t) 1  1    −2  vb (t)  = K vc (t) −1 2   √ 0 3 2 √ 3 2  − T23  vα  vβ . Rotor flux orientation means that the rotor flux is aligned with the d-axis.3) 3. for low.3 Motor Control Vector control implies rotor flux orientation. If this is to be done.2) 3. the vector appears not to be moving. θ1 [2].1. (3. see Figure 3. αβ. This is called dq-transformation and its relations with α and β leads to the following transformation matrix [2] v = e−jω1 t vs = e−jθ1 vs = (cos θ1 − j sin θ1 ) (vα + jvβ ) = vα cos θ1 + vβ sin θ1 + j (−vα sin θ1 + vβ cos θ1 ) ⇔v = vd vq = cos θ1 sin θ1 − sin θ1 cos θ1 Tdq (θ1 ) vα vβ .and medium-powered drives it often matches the cost of the 9 . as mentioned earlier. to control the AC machine as if it were a DC machine.and then two-phase/three-phase -transformed before it is amplified and fed to the IM. This sensor is costly. is . The control system also provides the desired dq-frame stator voltage vector. This is achieved by the transformation from two-phase to synchronous reference frame. The result is that the q-part of the rotor flux is zero. The rotor flux oriented angle is derived from the measured rotor speed ωr together with the calculated slip speed. They are then dq transformed for the control system using the rotor flux oriented angle. which in turn is αβ. (3.4 Sensorless Motor Control As mentioned earlier.

speed-sensorless control: When using a flux sensor.and speed-sensorless control: Estimates both the flux and speed which is an attractive solution due to reduced costs and parts. It is also possible to use a flux sensor but the usage of such kind of device is highly impractical due to probe placing between the stator and rotor.and speed-sensorless control that is able to function 10 . This speed is then used to get the synchronous angle θ1 necessary for the transformations from control signals to a three-phase voltage. creating the synchronous speed ω1 . Flux-sensorless. speed sensored control: The typical situation in most IM drives. where the control system gets a voltage reference signal and then produces a three phase voltage to the IM. Speed measuring allows design of flux estimators which give good performance at all speeds. vector control is no more difficult than controlling a DC motor. Besides the cost of the sensor it is also “one more part” that may fail. 3. The rotor speed ωr is measured and added to a calculated slip speed. θ1 is dq αβ αβ 3 Vs dq αβ αβ IM 3 θr θ1 Figure 3.A Sensorless Induction Machine Drive. 2. Harnefors [2] distinguishes between three different types of sensorless control cases: 1. Flux-sensored.1: The idea of the vector control method. At nominal speeds. a flux-angle estimate of good accuracy can be obtained but they tend to fail when load torque is applied at very low speeds.. While for 1. Flux. there is not yet proof of a flux. and 2. motor itself [2]. An extra sensor for speed measurement is not needed because the speed can be calculated using slip subtraction. stable operation is possible at all speeds.

The flux and speed-sensorless control is the focus of this thesis. 11 . completely stable at low frequencies while subjected to a load torque.A Sensorless Induction Machine Drive.

Adaptive Speed Observer (ASO) and Voltage Model with Indirect Field Orientation (VMIFO). Since Ottergrens work.2) . derived from the relation dψ R = Es . dt s (4. This method has been further developed by Rolf Ottersten [7].1 Improved Statically Compensated Voltage Model The SCVM is based upon the classic "Voltage Model (VM)" flux estimator. the following equations are obtained [7]: dΨR = Ed dt 12 (4.1) This implies that if the flux EMF is known. The estimators may be of two types. After transferring the VM from DFO to IFO representation and splitting the relation into real and imaginary parts. the voltage model variants have improved. DFO estimate the flux space vector directly while IFO use the estimated rotor position to get the flux. Harnefors [2] deducts the SCVM quite thoroughly and this thesis only cover it briefly. Ottergren concluded that ASO was the preferred method [1]. In this thesis the improved SCVM is studied and implemented. Lennart Harnefors describes a sensorless flux estimator called the Statically Compensated Voltage Model (SCVM) [2] first proposed in [6]. 4. the rotor flux may be found. Anders Ottergren evaluated three different speed sensorless observers at Aros electronics AB in 2002: Observer based on Current-Voltage model (OCV). Direct Field Orientation (DFO) and Indirect Field Orientation (IFO).Chapter 4 Sensorless Control Algorithm Several sensorless control strategies has been proposed.

By doing so. or SCVM [2]. Ottersten adds another degree of freedom to the model. doing so adds a static error. (4. The error may be compensated for if a complex valued gain factor is introduced in the filter. equal at all frequencies.3) Ed = vd − Rs id + ω1 Lσ iq and Eq = vq − Rs id − ω1 Lσ id .4) (4.A Sensorless Induction Machine Drive.5) The VM achieves good results at nominal speeds. but low speed estimation is problematic.8) Ottersten [7] then modifies the SCVM by introducing a new coefficient.9) ω1 = (4. The flux will then be ψR = s j ω1 + λ|ω1 | s 1 − jλsign(ω1 ) s E = E j ω1 (p + λ|ω1 |) p + λ|ω1 | (4.6) may be transformed to IFO by substituting p → p + j ω1 and E → Ed + jEq and removing superscript "s". One common improvement is to replace the integrator s needed to calculate ψ R with a low-pass integrator. ψR = µEd + λsign(ω1 )Eq p + λ|ω1 | Eq − λsign(ω1 )Ed ψR (4.7) ω1 = (4. This modified VM is called the Statically Compensated Voltage Model.10) 13 . Unfortunately. Equation (4.6) d where p = dt and λ is a gain parameter greater than zero. ω1 = where Eq ΨR (4. µ. Splitting the result into real and imaginary parts and solving the imaginary part for ω1 gives ψR = Ed + λsign(ω1 )Eq p + λ|ω1 | Eq − λsign(ω1 )Ed ψR (4.

4. It is also dependent upon how accurately the stator resistance is modelled. 4. Other deviations may also occur. The stator resistance may be modeled differently if ω1 and iq have the same sign or not. The rotor resistance affect the slip calculation.3 Known Limitations of the SCVM The main problem region for the SCVM is low speed operation and reversing the motor when subjected to a larger load. for example if parameters are wrongly measured or the motors differs in the production line. It may be rewritten as Lσ = Lls + Lm (1 − →Llr . Rs . In practice. stator resistance Rs .11) →0. the estimator is stable for all operating conditions except ω1 = 0 and reversal with iq > id √ [2]. +1 (4. The critical parameter for the SCVM is the stator resistance [3]. As a result. Lσ is derived from the T-model parameters Ls . Theoretically.1).A Sensorless Induction Machine Drive. Four parameters are included in the inverse-Γ model. In this project. the most likely deviations will be increasing resistance and inductance. Rolf Ottersten [7] suggest two improvements to make the SCVM more stable at reversal and low speed operation. Reversing with load is a problem mainly when doing so slowly. The resistances are mainly affected by heat development and inductances by the level of magnetization and saturation.2 Parameter Sensitivity of the SCVM A crucial measure of the performance of a sensorless control algorithm is the sensitivity to variations in motor parameters. The SCVM includes three of the inverse-Γ parameters. This requires the stator 14 . the probability of the problem developing is less. Lr and Lm . RR and Lσ . Indirectly the magnetizing inductance will affect Lσ according to equation (2. If the transition is fast. Llr Lm 1 ). the parameters may change substantially. thus affecting the accuracy of the drive to some extent. The total leakage inductance is a part of the flux estimation. the low speed region where the SCVM is unstable is 2 larger. magnetizing inductance LM and total leakage inductance Lσ . slightly underestimated in the first case and overestimated in the second. As long as the magnetizing inductance changes are moderate the effect on Lσ will be small. as Lm →∞ as Lm →0 The calculated leakage inductance will hence be between Lls and Lls + Llr . As the motor operates. It is used in the calculations of Ed and Eq and impact the calculated flux directly. rotor resistance RR .

min where λp = 3 |iq | .min is suggested to be between 5 and 10 % of nominal speed. 15 .13) The success of the latter method depends upon that Rs is estimated in the manner described above. Ottersten also suggest a modification of λ at low speeds.12) λ= ω1. choosing lambda to be √ 2|ω1 | + λp (ψR − ψref ) (4. id (4.A Sensorless Induction Machine Drive. This approach is not yet fully tested. Otterstens improvements are not used. this method is an improvement at lower speeds. In this thesis. It is shown that with λp = 0. resistance to be fairly well known. w1.

(5. The rotor windings are short-circuited.1) d where p = dt .Chapter 5 Simulations Before proceeding with the implementation it is necessary to do some simulations in order to test the control strategy. d = α). the motor is modeled with stationary references and ωc is replaced with 0. leading to zero rotor voltage (vβr = vαr = 0).2)  .1 Motor Model In this section. With stationary references. The super. In this thesis. 5.and subscript "c" denotes that the parameter should be replaced by the parameter specific for the reference frame chosen. If the reference frame is aligned with the rotor flux. yielding      s vβs s vαs 0 0 c vqs c vds c vqr c vdr    ic qs ic ds ic qr ic dr           =    Rs + Ls p 0 Lm p 0 0 Rs + Ls p 0 Lm p Lm p −ωr Lm Rr + Lr p −ωr Lr ωr Lm Lm p ωr Lr Rr + Lr p 16      ir βs ir αs ir βr ir αr     . The base for the motor model is the arbitrary dynamic model of an induction machine [5]:  Rs + Ls p ωc Ls Lm p ωc Lm     −ωc Ls Rs + Ls p −ωc Lm Lm p       =    Lm p (ωc − ωr )Lm Rr + Lr p (ωc − ωr )Lr   −(ωc − ωr )Lm Lm p −(ωc − ωr )Lr Rr + Lr p (5. It is also an opportunity to get acquainted with different problems that might arise during the implementation. the system is called synchronous and ωc is replaced with ω1 . it is common to say that the model is an αβ-system where α is aligned to the magnetizing d-axis (q = β. the mathematical model of the motor and its implementation in Matlab Simulink is described and constructed.

10) dλβr Rr Lm = ωr λαr − λβr + Rr iβs dt Lr Lr Rr Lm dλαr = −ωr λβr − λβr + Rr iαs .9) (5. (5.8) (5.2) and inserting (5.7) (5.(3. The electrical torque produced is calculated using iα .A Sensorless Induction Machine Drive.12) where J is the moment of inertia and B is the viscous friction. 17 .6) Rearranging (5. J (5. Solving the equation for ωr gives ωr = (Te − TL ) −Bt/J e . dt Lr Lr Further on.13) Equations (5.3) as can be seen in figure 5.5) (5.(5.11) where np is the pole pair number.7) . The motor parameters in the model are measured on the motor used in the implementation later on.1. λα and λβ .4) (5. iβ . The motor parameters can be found in Appendix A.(5. The flux linkages are defined as λβs = Ls iβs + Lm iβr λαs = Ls iαs + Lm iαr λβr = Lm iβs + Lr iβr λαr = Lm iαs + Lr iαr . The mechanical part of the motor is represented by Te − T L = J dωr + Bωr dt (5.1) .6) and the differentiation of the same equations yields the differential equations implemented in the model of the motor. Te = np (iβ λα − iα λβ ) 2 (5.13) are implemented in the “IM block” in Matlab Simulink together with the control system and transformation equations (3. The goal with the transformations of the equations is to base the model on rotor flux and stator currents: diβs 1 = 2 dt Ls − Lm Lr 1 diαs = 2 dt Ls − Lm Lr Vβs − Rs iβs − Vαs − Rs iαs − Lm dλβr Lr dt Lm dλαr Lr dt (5. the flux and current index r and s will be omitted.3) (5.3) .

2.A Sensorless Induction Machine Drive. The control system in figure 5.9 Hz. The verification testing indicates that the implemented motor model is realistic.1 is written as an S-function in Matlab for a more easy conversion to C-code.73 3 Nm according to equation (5.1: Simulink implementation of the motor and its controller using feedback of rotor speed. 18 . The control system in the first.1 needs to be verified.41 A. The currents in figure 5. of the machine. 1100 nominal shaft power = 60 ≈ 3. In order to verify the motor model the outputs are compared to those of the real motor subjected to nominal speed and load torque. 5. 4.2 Model Verification Before beginning the sensorless implementation the model of the motor in figure 5. two current controllers and a voltage decoupler.4 are close to the nominal rms phase-current. Figure 5.14). The phase voltage supplied to the motor in figure 5.73N m nominal speed 2815 · 2π (5.14) As can be seen in figure 5. version consists of a speed controller.2 is 230· 2 V (delta connected motor) and the nominal load torque is 3. see figure 5. 5. This is very close to the speed of the real motor during nominal conditions. the input to the control system is chosen to be a speed reference.3 the simulated motor speed is ∼46. non sensorless.3 Control System Model In this thesis.

85 1.5 49 Rotor speed Synchronous speed 4 Phase currents at verification simulation I a Ib Speed [Hz] 48.95 2 2. 19 .5 4 Time [sec] 1.5 6 50 49.2: Verification of the implemented IM model. Figure 5.4: The peak-value of the currents are ∼6 A rendering the rmsvalue of 4. Figure 5. thus confirming the model.5 1 1.05 Time [sec] Figure 5.5 47 Current [A] 2 I c 0 −2 −4 46.3: With the rotor speed at ∼46.A Sensorless Induction Machine Drive.5 3 3.2 A.9 Hz.5 2 2.5 48 47. which is close to nominal current when the motor is cold. the slip is ∼6 % which is close to the real nominal slip. Rotor speed at verification simulation 50.5 46 −6 0 0.9 1.

(5. ∗ vq = Hq + ω1 Lσ i∗ − (Rs + RR )i∗ d q ∗ vd = Hd − ω1 Lσ i∗ − (Rs + RR )i∗ q d (5.3. The output from these controllers are fed to a voltage decoupler. the torque is proportional to i∗ . The λ∗ reference is qr dr set to a constant value [8].21) ψR = ψR + Ts µEd + λsign(ω1 )Eq − λ|ω1 |ψR ωslip = RR ωr = ωr + Ts αe iref q ψR − ωslip − ωr Eq − λsign(ω1 )Ed ψR (5.4 The Improved SCVM Model Lennart Harnefors describes a discrete improved SCVM algorithm that is implemented in this thesis project [2].21) is the standard slip relation.19) and (5.22).A Sensorless Induction Machine Drive. is a part of the equation describing the rotor speed. equation (4.4). λ∗ is defined as zero. The PI speed controller derives i∗ from the speed error. The bandwidth αe of the rotor speed filter is set equal to the current controller bandwidth [2].10). ref Ed = vd − Rs iref + ω1 Lσ iref q d ref Eq = vq − Rs iref − ω1 Lσ iref q d (5. q As discussed in Section 3. µ = −1 √ and λ = 2. From (5. designed from the inverse Γ-model to remove the cross-coupling between d and q dependency while calculating the voltage commands to the "dq to abc" transformation block [3]. (4.20) corresponds to equation (4.16) (5.18) (5.19) (5. This is not a head on solution but a filter.15). 20 . the two current controllers are implemented. making the estimate less sensitive to disturbances. This is possible as the mechanical system is slower than the control system.9) the setpoint for ids is then i∗ = ds λ∗ dr LM (5. Equation (5. (5.9).15) Considering equation (5. As the observant reader may notice.22) Equation (5. the calculation of ω1 .20) (5. respectively. The SCVM parameters is chosen as recommended by Harnefors. 5.5) and (4.18). With qs the reference currents and the measured dq motor currents.17) where Hq and Hd is the result from the current controllers.

5 3. except for short periods when the speed command change instantaniously.5 shows some simulated and estimated rotor quantities during an acceleration and retardation of the motor.5 Simulation Results To get some idea of the performance of the estimator. one with step changes in load torque limitation.5 1 1.5 but the motor is subjected to a 2 Nm load.4 0.9.6 shows the same as figure 5.A Sensorless Induction Machine Drive. 5. figure 5. As seen in figure 5.5: Modelled and estimated rotor data during an acceleration and retardation test.7 3. The load speed reference is set to a constant value while the motor speed command is varied as a saw-tooth signal. a speed governed torque limited load model is developed. The bottom left image shows the total angle of the rotor. showing both speed.5 3 3.8 3. the angle difference after the test cycle is less than 1 %. the estimator can handle almost nominal load with slow ramping around zero excitation frequency.6 615 610 Model Estimate 605 −0. The speed difference is also small.5 3 3.4 3.2 3. To simulate such a case.9 3 3. different scenarios are simulated.8 600 2.1 3. and the speed difference between the estimation and the model are shown to the right. Figure 5.3 states that a slow reversal of the induction machine may be problematic. As may be seen in both figures. and one with a constant load torque limit and also a deviation in the stator resistance value.5 2 2. Section 4. 21 .5 1 1.2 625 −0.3 3.6 1 Speed difference Model−Estimate Speed [Hz] 30 25 20 15 10 5 0.6 3. figure 5.8.5 2 2. Figure 5. Two different cases are studied. In both figures the top left image displays the speed of the rotor.5 4 Time [sec] Time [sec] Figure 5.8. speed difference and angle when the motor is unloaded.9 −1 0 0.2 0 −5 0 0.4 Angle [rad] 620 −0.5 4 Speed [Hz] Time [sec] 0 Rotor Angle 630 −0. Rotor Speed 50 45 40 35 0.8 Model Estimate 0.

speed difference and angle with 2 Nm load.5 1 1. Estimated rotor flux 0.5 1 1.5 2 2.5 4 Speed [Hz] Time [sec] 0 Rotor Angle with load 620 −0.75 0.4 Angle [rad] 600 −0.4 0.5 3 3.6 3.5 4 Time [sec] Modelled rotor flux 0.75 0.6 0.55 0.2 0 −5 0 0.5 4 Flux [Wb] 0 0.5 2 2.5 1 1.5 4 0.2 3.5 1 1.65 0.5 1 1.5 3 3. Rotor Speed with load 50 45 40 35 0.55 0.8 0.5 1 1.7 0.5 2 2.45 0.5 2 2.45 0.5 4 Time [sec] Time [sec] Figure 5.7 0.65 Time [sec] Modelled rotor flux with load 0.5 2 2.65 0.55 0.6 590 580 Model Estimate 570 −0.5 3 3.A Sensorless Induction Machine Drive.6 0.8 3 3.6 Flux [Wb] Flux [Wb] 0.6: Modelled and estimated rotor data during an acceleration and retardation test.8 Model Estimate 0.6 0.55 0.5 0.5 3 3.5 4 Time [sec] Time [sec] Figure 5.4 3.5 0.8 4 −1 0 0. showing both speed.6 1 Speed difference Model−Estimate with load Speed [Hz] 30 25 20 15 10 5 0.65 0 0.4 Estimated rotor flux with load Flux [Wb] 0.7: Modelled and estimated rotor flux during an acceleration and retardation with and without load.4 0 0.5 2 2.8 560 2.8 0. 22 .2 610 −0.5 3 3.5 3 3.5 0 0.

The flux will soon destabilize and the ability to follow the speed command will be reduced. all parameters in the control system is recalculated with suitable factors.A Sensorless Induction Machine Drive. Integer calculation with small values will render large errors (if the current is 1.66 0.9. The difference in estimated and modelled flux depend upon the different models used. The unit for current is 10 mA.64 8 Model rotor flux Estimated rotor flux id 10 i q Flux [Wb] 0.58 0.49 A.6 0. 5. The motor is implemented in the T model and the control system parameters are derived from the inverse-Γ model and according to equation (2.8: Simulated ramping around zero speed with stepwise increase of load torque.68 0. for inductances it is 65536 (216 ). the control system calculations are best done with integers. To avoid this. Slow ramping around zero speed with different loads 10 8 6 Model rotor speed Estimated rotor speed Rotor speed [Hz] 4 2 0 −2 −4 −6 −8 −10 0 2 4 6 8 10 12 Load torque [Nm] 4 2 0 0 2 4 6 8 10 12 Time [sec] 0. To accomplish this.5 0 2 4 6 8 10 12 Current [A] 6 4 2 0 0 2 4 6 8 10 12 Time [sec] Time [sec] Figure 5. the control system is working with different units.62 0.52 0.54 0.56 0. 23 . the system will round it to 1 A). the flux differs with the factor Lm .4). These factors should be a power of two.7 0. voltage 10 mV and speed 10 mHz. The ability to handle Lr a deviation of the stator resistance in the motor is not good according to Figure 5.6 New Parameters To optimize the usage of the MCU. The factor for resistance is 256 (28 ). as it is logical for the processor to work with.

5 4 1 2 0. but all together it is worth to evaluate this approach experimentally. The quite high dependency of the stator resistance may be a problem.7 Simulation Conclusions The SCVM can handle static load.A Sensorless Induction Machine Drive. at least if the motor parameters are modelled correctly. It also seem promising in reversing a speed governed load.5 0 0 5 10 15 20 25 30 Time [sec] Time [sec] Figure 5. 24 . 5. Slow ramping around zero speed with constant load and ramped stator resistance 25 20 15 Model rotor speed Estimated rotor speed Rotor speed [Hz] 10 5 0 −5 −10 0 5 10 15 20 25 30 Stator resistance [Ω] 3 2.5 2 0 5 10 15 20 25 30 Time [sec] 3 Model rotor flux Estimated rotor flux 2.5 8 i 10 i d q Current [A] 0 5 10 15 20 25 30 2 Flux [Wb] 6 1. acceleration and deceleration.9: Simulated ramping around zero with constant load and ramped stator resistance.

This calls for a careful review of the implemented code. The card has an on-board three phase diode bridge rectifier. beginning with hardware. the encoder is used as a reference for evaluation of the speed estimator. thus harming the computer. Any part of the computations exceeding 16-bit (+/. The board is prepared for using an encoder. It is rated 35 A continuously and 1200 V peak 25 . a 36MT120 from International Rectifier. It is also good if the code is optimized to use the DSP as effective as possible. Apart from estimator and control algorithms.1. The earth potential of the controller card may reach several hundreds of volts compared with the potential of the grid. The controller card is a prototype from another project and is to large extent a standard converter-inverter design. there has to be capacity left for applications. This chapter describes different parts of the implementation.1 Hardware Controller Card This project uses a hardware setup constructed by Aros electronics AB. In this project.1 6. continuing with software and data acquisition then ending with implementation issues.32768 as most integers used are signed) has to be revised. The encoder is capable of measuring speeds up to 50 Hz. The optocouplers transfer the signals from one voltage system to another. The communication with the computer during test and programming is operated by a serial interface. 6. It is also equipped with optocouplers to galvanically isolate the two systems.Chapter 6 Implementation A difficult feature of the control system when moving from simulations to implementation is that the DSP use 16-bit integer representation. A special interface adapter is used to translate between the PC and the controller card.

the sampling time of the system becomes large and erratic. For system supervising while running the motor it is adequate. It controls the speed in two levels. but not for keeping track of transient behavior.2 Motor The test motor is constructed by Bonfiglioli Group in Italy.1 kW and it has two poles. A voltage divider is connected to a voltage follower and read by an AD channel on the DSP. It is possible to log and graph three different signals through the serial control interface. 6. To manipulate the system during test. It should be enough to replace the encoder routine calls with the estimator ditto. the most important being that the speed controller use reference value filtering. The DC current is also measured. The control system also has to gain knowledge about the DC voltage. The estimator is only a small part of the controller software. to if necessary invoke an over-current protection. a software package developed at Aros electronics AB is used.A Sensorless Induction Machine Drive. stepping the actual reference value toward the desired value.1. It also has the possibility to display graphs of up to three variables at the same time. Reference value filtering allows a more stable controller. In this thesis. adjust the parameters and the motor drive should be operational. DC/DC converters use the main DC voltage to extract the power levels needed by the control card. a signal is sent to the processor and the system shuts down. The software can control and monitor each variable in the DSP. a special PC-program is used. Another difference is the absence of a voltage decoupler in the control system. If a too high current. The DC overcurrent trigger is set to 50 A.2 Software The software at Aros electronics AB is modular and the same modules appear in many of the company’s products. There are some differences between the simulated and the already implemented control system. One aim with this project is to develop an estimator that is easy to adapt to present and future projects at Aros. Motor data can be found in Appendix A. 26 . about 40 A. 6. The phase currents are measured for the control system and for the overcurrent protection. 6. The rated power is 1. is registered. • The control PC-program has a logging feature including a graph tool. The basic control system already present is not altered. It is developed at Aros electronics AB. Unfortunately. DC voltage.3 Data Acquisition There are three possibilities to access the parameters from the processor.

SPI is a loose standard for serial communication. the implemented c-code was translated back to Matlab. The sampling time of the control PC-program is bad. In this project. Even after careful review. The calculations need to be large enough to get good accuracy (see Section 5. a level sensitive to the harsh lab environment. Several adjustments had to be done to accomplish this. There were some pure coding errors that took time to find. For a final debug of the system.4 Implementation Issues Going from simulation to the actual system proved to be troublesome. but sufficiently small to fit inside the 16-bit representation. two methods are of interest in this case. the amplitude resolution of the SPI is narrow. rendering a sample rate equal to the switching frequency. While running the motor. On the other hand. the SPI call is located in the switch interrupt. The code was simulated bit by bit in the model used to develop the same.6). • The board used is equipped with a Serial Peripheral Interface bus (SPI). Some calculations were also split into several lines to get a better overview as well as making it easier for the compiler to interpret the code. The SPI has low amplitude resolution. The signal range is 0 to 3. the system is sampled via the SPI as it is assumed that the transient behavior is of more importance for evaluation of the SCVM than the amplitude resolution. The errors were then quite easily found and corrected. One large obstacle were the 16-bit limit for all variables. The idea with the interface in this project is to let the software use pins on the DSP to put out up to three signals to an external decoding circuit. • Some projects at Aros have used MCU or on-board memory for logging data which is transferred to the computer as a file after the measurement. As described above. The equipment used in this project lacks the memory capacity needed for such approach. 8 bits. some issues remained. In this thesis. the switched currents induce an electromagnetic field large enough to introduce noticeable disturbance when measuring the signals.68 V. 6. 27 .A Sensorless Induction Machine Drive. The sample rate is decided by how often the SPI code is executed in the DSP. The signals can then be studied using an oscilloscope.

As may be seen in figure 7.7 Nm. is a result from the current limit.2.Chapter 7 Results The implemented design is put through several tests to check the performance of the drive.6. although the accuracy decrease.5 Hz.1 Speed Step Response Figures 7. the system fails already at 7.1-7. In order to get more information from the measurements.4. As may be seen.1 Slow Speed Limit In the first test. The aim is to find the slow speed limit and the load limit at slow speed reversal of a speed governed load.5 shows that it is possible to lower the speed and keep control of the motor.2 Slow Speed The slow speed performance is tested by applying different loads and varying the speed command. the drive is capable of handling different loads. The somewhat slower response for the double nominal load. 7. The test is 28 . The control system reference speed is programmed to decrease from 10 Hz to standstill with steps of 0. 3. parameter variation and transient behavior as well as speed stability up to twice the nominal load. figure 7. Figure 7. the load is set to -16 Hz and nominal torque. 7. the reference speed is set to go from -10 Hz to zero. It is of interest to see the operational limits concerning slow speed. the measured data are filtered. In the second test.4 depicts the step response as a speed change from zero to near nominal speed is applied to the control system. The filter parameters are created using the cheby2 function in Matlab with a cutoff frequency of 1500 Hz and stop-band attenuation of 110 dB. 7.5 Hz.

2 0.2 0.2 1.8 2 Time [sec] 1.2 0.4 0 0. Speed command step−change with no load 60 50 40 Speed [Hz] 30 20 10 Reference rotor speed Measured rotor speed 0 −10 0 0.8 1 1.4 1.3 1.8 2 Time [sec] Figure 7.4 0.4 1.5 0 0.4 1.4 1.6 0.8 1 1.6 1.8 1 1.4 0.2 0.4 0.7 0.6 0.6 1.6 0.2 1.6 1.6 0.8 2 Time [sec] Figure 7.1 1 Flux [Wb] 0.1: Speed step with with no load.2 1.6 1.9 0.4 0.6 0.6 0. 29 .2 1.2 1.3 1.7 0.8 0.8 2 Time [sec] 1.9 0.8 0.2 1.2: Speed step with 2 Nm load. Speed command step−change with 2 Nm load 60 50 40 Speed [Hz] 30 20 10 Reference rotor speed Measured rotor speed 0 −10 0 0.5 0.1 1 Flux [Wb] 0.A Sensorless Induction Machine Drive.8 1 1.

4 1.8 2 Time [sec] Figure 7.8 1 1.2 0.2 0.2 1.5 0.2 0.2 1.9 0.4 1.2 0.8 1 1.6 0.1 1 Flux [Wb] 0.5 Nm.A Sensorless Induction Machine Drive.8 1 Time [sec] 1.4 0.8 2 Time [sec] 1. 30 .7 0.4 0.7 0.4: Speed step with double nominal load.3 1.2 1.3 1.8 0.2 1.1 1 Flux [Wb] 0.4 0.6 1.6 0.6 1.8 1 1. 7.6 0.6 0.8 2 1.4 0 0.4 1. Speed command step−change with twice the nominal load 60 50 40 Speed [Hz] 30 20 10 0 Reference rotor speed Measured rotor speed −10 0 0.9 0.6 0.7 Nm. 3.5 0.3: Speed step with nominal load.4 1.8 2 Time [sec] Figure 7.4 0.2 1.6 1.2 1.4 0 0. Speed command step−change with nominal load 60 50 40 Speed [Hz] 30 20 10 Reference rotor speed Measured rotor speed 0 −10 0 0.6 1.8 0.6 0.

8-7. From the result in figure 7.9.5 −1 −1. but the drive does not destabilize.5 −2 0 5 10 15 20 25 30 35 40 45 50 Time [sec] Figure 7.5: Dercreasing speed with load and motor operating in the opposite direction. Decreasing speed with nominal load 12 Reference rotor speed Measured rotor speed 10 8 Speed [Hz] 6 4 2 0 0 5 10 15 20 25 30 35 40 45 50 Time [sec] Difference between reference and measured rotor speed 1 0.7 it may be seen that the accuracy is significantly less compared to the test with positive movement. but the transition is less smooth for 2 Nm. To enable an analysis of the occurring problem.2. reference speed and estimated rotor speed is plotted. 31 . figure 7. For a 3 Nm load. the rotor is first operated at almost standstill.10.5 0 Speed [Hz] −0. In all tests. only the rotor speed. the test is redone with logging of estimated rotor flux and measured stator currents.2 Reversing The purpose is to test consequential slow reversals.10. the load servo is controlled to a speed of -16 Hz. in reality there is a small positive movement of the rotor. redone with less load torque. Keeping zero speed is difficult. 7. the drive loses control but regains it once the speed is increased above zero. As may been seen from figures 7. The load torque limit is set to motor nominal torque.A Sensorless Induction Machine Drive.7 Nm.8 and 7. In figures 7. 3. 2 Nm. The drive is quite capable of handling reversal of such loads.

A Sensorless Induction Machine Drive.6: Dercreasing speed with load and motor operating in the same direction. Decreasing speed with braked nominal load 5 Reference rotor speed Measured rotor speed 0 Speed [Hz] −5 −10 −15 −20 5 10 15 20 25 30 35 Time [sec] Difference between reference and measured rotor speed 12 10 8 6 4 Speed [Hz] 2 0 −2 −4 −6 −8 −10 5 10 15 20 25 30 35 Time [sec] Figure 7. 3. 32 .7 Nm. The load torque limit is set to motor nominal torque. thus the motor tries to brake the load.

thus the motor tries to brake the load.7: Decreasing speed with load and motor operating in the same direction. Decreasing speed with braked 2 Nm load 2 0 −2 Speed [Hz] −4 −6 −8 −10 Reference rotor speed Measured rotor speed −12 0 10 20 30 40 50 60 70 80 90 Time [sec] Difference between reference and measured rotor speed 3 2 1 Speed [Hz] 0 −1 −2 −3 0 10 20 30 40 50 60 70 80 90 Time [sec] Figure 7.A Sensorless Induction Machine Drive. The load torque limit is set to 2 Nm. 33 .

Slow ramping around zero speed with 2 Nm load torque 5 0 Speed [Hz] −5 −10 −15 Estimated rotor speed Measured rotor speed Reference rotor speed 10 15 20 25 30 35 40 Time [sec] Figure 7.9: Slow ramping around zero speed with 2 Nm speed governed load.8: Slow ramping around zero speed with 1 Nm speed governed load. 34 . Slow ramping around zero speed with 1 Nm load torque 5 0 Speed [Hz] −5 −10 −15 Estimated rotor speed Measured rotor speed Reference rotor speed 10 15 20 25 30 35 40 Time [sec] Figure 7.A Sensorless Induction Machine Drive.

Slow ramping around zero speed with 3 Nm load torque 5 0 Speed [Hz] −5 −10 Estimated rotor speed Measured rotor speed Reference rotor speed −15 10 15 20 25 30 35 40 Time [sec] 1 20 15 0.10: Slow ramping around zero speed with 3 Nm speed governed load.6 0.A Sensorless Induction Machine Drive.8 10 0.5 −10 0. The two bottom plots are from a second test.9 0.3 10 15 20 25 30 35 40 −20 10 15 20 25 30 35 40 Time [sec] Time [sec] Figure 7.4 iq id −15 0.7 Current [A] 5 Flux [Wb] 0 −5 0. 35 .

A Sensorless Induction Machine Drive. see figure 7. slow ramping around zero with 2 Nm speed governed load. the estimator will be subjected to similar stresses as if the real motor parameters were to change. currents are logged. At higher speeds. 7. see Section 6. The test is conducted in two modes.19) becomes less due to the increased importance of ω1 Lσ is .3. This corresponds to a motor stator resistance decrease by 37 %.5 indicated.11. first to get encoder speed. see figure 7. As seen in figure 7. Therefore. As this is hard to achieve. a control system stator resistance change affect the drive insignificantly. the drive fails.12 and 7. it is not as bad as the simulations in Section 5. the SCVM is sensitive to inaccurately modeled stator resistance. 7. When the control system stator resistance is increased at low speed operation. figure 7.13. To see more exactly where this phenomena occurs. the estimated rotor flux and the measured.11. the impact of the Rs is part in the equations (5. In the second run. This is due to the limitations of the SPI interface. it may be too large. as it seems to be able to handle it but with some difficulty.9 Hz.3.1 Deviations in Stator Resistance To test the estimators ability to cope with an increase in stator resistance.18) and (5. The test is not conducted exactly at nominal speed as the encoder with its interface is unable to handle more than 50 Hz. Rs is lowered in the control system. the motor parameters would need to be controllable. parameter changes are emulated by changing the parameter values in the implemented code. However.3 Parameter Sensitivity To properly test the SCVMs ability to withstand parameter deviations. The motor is loaded with 2 Nm. dq-transformed. and near the nominal speed with the same load. but with overshoot and other effects. as seen in Figures 7. figure 7. the stability and accuracy reduce as the control system stator resistance is lowered.14. From the graph it is found to happen at approximately 3. Each mode is tested twice.25 Ω. Nominal rotor speed is 46.9. Doing so. In the high speed region. estimated rotor speed and rotor resistance of the control system. 36 . The resistance values are changed to emulate a deviation of ±60 % and the leakage inductance to emulate ±30 %.12. The aim with these tests is to find out how the performance at nominal speed and at low speed ramping about zero is affected by variations of stator and rotor resistance and leakage inductance. the rotor speed setpoint is 45 Hz. figure 7. an even slower ramp is tested.15. Especially in the low speed region.

8 0.3 10 15 20 25 30 35 40 45 50 55 60 65 Current [A] Flux [Wb] 2 0 −2 −4 −6 10 15 20 25 30 35 40 iq id 45 50 55 60 65 Time [sec] Time [sec] Figure 7.4 0.5 2 1.6 0.11: Speed.A Sensorless Induction Machine Drive.5 0. first to log speeds and resistance and secondly to log flux and currents.1 1 0.9 0.7 0. 37 .5 1 10 15 20 25 30 35 40 10 8 6 4 45 50 55 60 65 Time [sec] 1. estimated rotor flux and measured currents at ramped speed around zero as the control system stator resistance is changed. Slow ramping around zero speed with decreased control system stator resistance and 2 Nm load 8 6 4 2 Speed [Hz] 0 −2 −4 −6 −8 −10 Estimated rotor speed Measured rotor speed 10 15 20 25 30 35 40 45 50 55 60 65 Stator resistance [Ω] Time [sec] 2. The test is conducted twice. A lower resistance means emulating an increase of motor stator resistance.

A lower resistance means emulating an increase of motor stator resistance.5 43 35 40 45 Estimated rotor speed Measured rotor speed 50 55 60 Stator resistance [Ω] Time [sec] 2.5 45 44.5 35 40 45 d 4 0.8 0. Decreased control system stator resistance near nominal speed with 2 Nm load 48.5 0.68 0.6 35 50 55 60 Time [sec] Time [sec] Figure 7.A Sensorless Induction Machine Drive.5 5 4. first to log speeds and resistance and secondly to log flux and currents. 38 .78 6.76 0.66 0.5 47 Speed [Hz] 46.72 0.7 0. estimated flux and measured currents at near nominal speed as the control system stator resistance is changed.12: Speed.5 1 35 0.62 0.64 Current [A] 5.74 6 40 45 50 55 60 Time [sec] 7 Flux [Wb] 0.5 2 1. The test is conducted twice.5 iq i 40 45 50 55 60 3.5 44 43.5 48 47.5 46 45.

Slow ramping around zero speed with increased control system stator resistance and 2 Nm load 6 4 2 Speed [Hz] 0 −2 −4 −6 −8 −10 −12 40 50 60 70 80 90 100 110 Estimated rotor speed Measured rotor speed 120 130 Stator resistance [Ω] Time [sec] 6 4 2 40 50 60 70 80 90 100 110 120 130 Time [sec] 16 1 14 12 1. 39 .9 10 Current [A] Flux [Wb] 0. first to log speeds and resistance and secondly to log flux and currents.5 −2 −4 iq i 40 50 60 70 80 90 d 0.7 0.8 8 6 4 2 0 0.4 40 50 60 70 80 90 100 110 120 130 100 110 120 130 Time [sec] Time [sec] Figure 7.1 0.13: Speed. The test is conducted twice.A Sensorless Induction Machine Drive.6 0. estimated flux and measured currents at ramped speed around zero as the control system stator resistance is increased.

A Sensorless Induction Machine Drive.2 3.1 3 65 66 67 68 69 70 71 72 73 74 75 Time [sec] Figure 7.14: Estimated and measured speed as the control system stator resistance is slowly increased to see the troublesome resistance level.5 66 67 68 69 70 71 72 73 74 75 Time [sec] Stator resistance [Ω] 3.3 3. 40 .4 3. Slow ramping around zero speed with slowly increasing control system stator resistance and 2 Nm load 6 Estimated rotor speed Measured rotor speed 4 2 Speed [Hz] 0 −2 −4 −6 −8 65 3.

6 0.68 6 0.56 30 40 50 60 70 80 90 100 110 120 130 Current [A] Flux [Wb] 0. then lowered to the nominal value. estimated flux and measured currents at near nominal speed as the control system stator resistance is first increased.58 0.A Sensorless Induction Machine Drive. 41 .62 0.72 0.64 0.7 0.15: Speed. The test is conducted twice. first to log speeds and resistance and secondly to log flux and currents.66 5 4 3 2 30 40 50 60 70 80 90 100 110 120 130 Time [sec] Time [sec] Figure 7. Increased control system stator resistance near nominal speed with 2 Nm load 49 Estimated rotor speed Measured rotor speed 48 47 Speed [Hz] 46 45 44 43 30 40 50 60 70 80 90 100 110 120 130 Stator resistance [Ω] Time [sec] 6 4 2 30 40 50 60 70 80 90 8 100 110 120 130 Time [sec] i i 7 q d 0.

the impact on the accuracy is more significant.8 1.7 0 −2 0.2 Deviations in Rotor Resistance The rotor resistance affects the slip calculation. figure 7. Slow ramping around zero speed with decreased control system rotor resistance and 2 Nm load 6 4 Speed [Hz] 2 0 −2 −4 Estimated rotor speed Measured rotor speed 15 20 25 30 35 40 45 50 55 60 −6 10 Rotor resistance [Ω] Time [sec] 2. Figures 7. the drive destabilize at 53 %.3.17. As the control system rotor resistance is decrease in figure 7. At near nominal speed. The test is conducted twice.2 10 15 20 25 30 35 40 8 45 50 55 60 Time [sec] 1 6 0. estimated rotor flux and measured currents at speed ramped about zero as the control system rotor resistance is changed. first to log speeds and resistance and secondly to log flux and currents. As it may be seen in figure 7.6 −4 0.19 on the other hand depicts that in nominal or high speed region.9 4 0. At low speed. 7. If the motor rotor resistance decrease much.8 Current [A] Flux [Wb] 2 0. The tests are conducted in the same way as for the stator resistance.2 2 1.17 and 7. The slip is small at low speeds. the measured rotor speed decrease approximately 2 %.21) models a linear relation between rotor resistance and slip speed.5 15 20 25 30 35 40 45 50 55 60 15 20 i i q d 25 30 35 40 45 50 55 60 Time [sec] Time [sec] Figure 7.16: Speed.16. 42 . figure 7. the low speed operation of the SCVM hardly change at all with a decrease in control system rotor resistance.6 1.4 1.18. the drive will even destabilize. the speed oscillation starts when the motor rotor speed is emulated to have decreased with 58 %.19.A Sensorless Induction Machine Drive. A lower resistance means emulating an increase of motor rotor resistance. Equation (5.

first to log speeds and resistance and secondly to log flux and currents.5 6 5.5 45 44.63 35 4 3.72 0.69 0.5 60 65 Time [sec] Current [A] Flux [Wb] 65 Time [sec] Time [sec] Figure 7.65 0.A Sensorless Induction Machine Drive.71 0.67 0.8 1.7 0.64 0. The test is conducted twice.5 43 35 40 45 50 Estimated rotor speed Measured rotor speed 55 60 65 Rotor resistance [Ω] Time [sec] 2.5 Speed [Hz] 46 45.17: Speed.66 0.68 0. A lower resistance means emulating an increase of motor stator resistance. Decreased control system rotor resistance near nominal speed with 2 Nm load 48 47.5 47 46. estimated rotor flux and measured currents at near nominal speed as the control system rotor resistance is changed.4 1.6 1.5 3 35 i i 40 45 50 55 60 65 40 45 50 55 60 q d 40 45 50 55 7 6. 43 .2 2 1.5 5 4.2 35 0.5 44 43.

45 0.A Sensorless Induction Machine Drive.5 0. Slow ramping around zero speed with increased control system rotor resistance and 2 Nm load 20 Estimated rotor speed Measured rotor speed 15 10 Speed [Hz] 5 0 −5 −10 −15 30 40 50 60 70 80 90 100 110 120 Rotor resistance [Ω] Time [sec] 5 4 3 2 30 40 50 60 70 80 90 100 110 120 Time [sec] 16 iq i d 0. The test is conducted twice.55 0.8 0.18: Speed.6 0. estimated flux and measured currents at ramped speed around zero as the control system rotor resistance is first increased.75 14 12 10 Flux [Wb] 0.85 0. then lowered to the nominal value.65 0. first to log speeds and resistance and secondly to log flux and currents.4 40 50 60 70 80 90 100 110 120 130 Current [A] 8 6 4 2 0 −2 −4 40 50 60 70 80 90 100 110 120 130 Time [sec] Time [sec] Figure 7.7 0. 44 .9 0.

7 i i 15 0.3 0 0.8 0.2 −5 0. first to log speeds and resistance and secondly to log flux and currents. estimated flux and measured currents at near nominal speed as the control system rotor resistance is first increased. Increased control system rotor resistance near nominal speed with 2 Nm load 60 50 40 Speed [Hz] 30 20 10 0 −10 −20 40 Rotor resistance [Ω] 6 4 2 40 50 60 70 80 Time [sec] 20 90 100 110 120 50 60 70 80 Time [sec] 90 100 110 120 Estimated rotor speed Measured rotor speed 0. 45 . then lowered to the nominal value. The test is conducted twice.19: Speed.6 10 0.A Sensorless Induction Machine Drive.1 0 −10 Current [A] Flux [Wb] 5 q d 40 50 60 70 80 Time [sec] 90 100 110 120 40 50 60 70 80 Time [sec] 90 100 110 120 Figure 7.5 0.4 0.

7.6 0. estimated rotor flux and measured currents at speed ramped about zero as the control system leakage inductance is decreased.9 0.3 Deviations in Leakage Inductance As may be seen from Figures 7. first to log speeds and inductance and secondly to log flux and currents.7 0.01 0 1.1 1 0.23.2 16 1.8 0.4 14 12 10 iq i d 10 20 30 40 50 60 Time [sec] Current [A] 0 10 20 30 40 50 60 Flux [Wb] 8 6 4 2 0 −2 −4 0 10 20 30 40 50 60 Time [sec] Time [sec] Figure 7.20: Speed. the drive is quite unaffected as the total leakage inductance is changed.3. A lower inductance means emulating an increase of motor inductance.014 0. The test is conducted twice.20-7.012 0.5 0. 46 .A Sensorless Induction Machine Drive. Slow ramping around zero speed with decreased control system leakage inductance and 2 Nm load 6 4 2 Speed [Hz] 0 −2 −4 Estimated rotor speed Measured rotor speed −6 Leakage inductance [H] 0 10 20 30 40 50 60 Time [sec] 0.

21: Speed.014 0. 47 . The test is conducted twice. A lower inductance means emulating an increase of motor inductance.5 47 Speed [Hz] 46.65 0. estimated rotor flux and measured currents at near nominal speed as the control system leakage inductance is decreased.01 30 35 40 45 50 55 10 iq 9 i d 60 65 70 Time [sec] 8 0.012 0.5 45 Leakage inductance [H] 44.A Sensorless Induction Machine Drive. first to log speeds and inductance and secondly to log flux and currents.5 46 45.7 Current [A] 0. Decreased control system leakage inductance near nominal speed with 2 Nm load 48 Estimated rotor speed Measured rotor speed 47.5 30 35 40 45 50 55 60 65 70 Time [sec] 0.6 10 20 30 40 50 60 70 80 90 100 Flux [Wb] 7 6 5 4 3 30 35 40 45 50 55 60 65 70 Time [sec] Time [sec] Figure 7.

6 −4 0.A Sensorless Induction Machine Drive. first to log speeds and inductance and secondly to log flux and currents.012 0 10 20 30 40 50 60 70 80 90 100 Time [sec] 1.22: Speed.5 −6 0 10 20 30 40 50 60 70 80 90 100 Time [sec] Time [sec] Figure 7. 48 . Slow ramping around zero speed with increased control system leakage inductance and 2 Nm load 6 Estimated rotor speed Measured rotor speed 4 Speed [Hz] 2 0 −2 −4 −6 Leakage inductance [H] 0 10 20 30 40 50 60 70 80 90 100 Time [sec] 0.8 0.7 0. The test is conducted twice.014 0.2 14 1.018 0. estimated flux and measured currents at ramped speed around zero as the control system leakage inductance is increased.02 0.016 0.1 12 10 1 8 0.9 iq i d Current [A] 0 10 20 30 40 50 60 70 80 90 100 Flux [Wb] 6 4 2 0 −2 0.

05 9 1 0.5 47 Speed [Hz] 46.1 1. The test is conducted twice.018 0.8 0.016 0.9 0.7 Current [A] 20 30 40 50 60 70 80 90 100 7 6 5 4 0.5 45 Estimated rotor speed Measured rotor speed 20 30 40 50 60 70 80 90 100 Leakage inductance [H] 44.A Sensorless Induction Machine Drive.95 8 20 30 40 50 60 70 80 90 100 Time [sec] 10 iq i d Flux [Wb] 0.5 10 Time [sec] 0.6 10 3 10 20 30 40 50 60 70 80 90 100 Time [sec] Time [sec] Figure 7.5 46 45.012 10 1. estimated flux and measured currents at near nominal speed as the control system leakage inductance is increased.014 0.85 0. Increased control system leakage inductance near nominal speed with 2 Nm load 48 47.23: Speed. first to log speeds and inductance and secondly to log flux and currents.75 0.02 0. 49 .65 0.

8. the results are compared to the set goals. 3. Keeping zero speed is not possible with the implemented design.4 Stable Operation during Parameter Variations As the motor parameters are not controllable. the goals were partly met. 8. In this way. 8. where it is unstable already at 7. larger loads are not possible.1. changes in motor data are emulated by changing the control system parameters.1. The problems occur in the braking mode. 8. The limitations are more dependent on the inverter design and the power source present.5 Hz.2 Stable Operation up to Twice the Nominal Torque It has been shown that the SCVM is capable of handling large loads at nominal speeds. the control 50 . Slow braking is possible with 2 Nm load but not with nominal load.1.7 Nm.Chapter 8 Conclusions The implementation of the Statically Compensated Voltage Model was completed with some success.1 Stable Operation up to Nominal Speed The drive works well at nominal speed. At slow operation.1 Goal Fulfilment In this section.1. the accuracy is reduced. 8.3 Stable Operation during Acceleration and Deceleration with an Arbitrary Load Slow reversal with more than moderate load is not possible. Low speed operation is more problematic. As shown in Chapter 7. At low speed.

• The drive is sensitive to changes in the value of the stator resistance. On-line tuning would also minimize the risk of failing units due to scatter in motor parameters. In the high speed region. It can handle a discrepancy of ±60 % before becoming unstable. A good solution should be easy to understand and implement. The effects of parameter variations could perhaps be less by implementing the suggested improvements in Section 4. It may also open the possibility for a stand-alone sensorless drive. • The implemented design is capable of handling deviations with 30 % in the total leakage inductance without any noticeable effect on performance at any speed. the drive is less sensitive to stator resistance deviations. 8. It is stable for variations between -37 % and +60 % in low speed operation. • The rotor resistance may decrease by 58 % at low speed and by 53 % at nominal speed before the drive destabilizes.3.A Sensorless Induction Machine Drive.2 Further Work To improve the drive. Increasing the rotor resistance with 60 % will not affect the stability at any speed but the nominal speed accuracy will be reduced by approximatly 2 %. ready for operation with almost any IM within its power range. one important contribution would be an on-line parameter estimator. 51 . system is put through similar stress as if the real motor parameters change.

1. From these data.5 rpm 2815 3420 cos(φ) 0.3) With the parameters from table A.2) λr = (A. ωr iq (A.3 A∆/Y 4. Lr and Ls and a no load test to get Lm . active power. the approximate nominal flux level is calculated by T = iq λr T = P ωr P .41/2. 52 . voltage.81 Table A.Appendix A Measurement of Motor Parameters Two types of tests were conducted. An Infratek 106A Power Analyzer (serial no: 03083312) is used for measuring currents.1: Motor data from nameplate.1 1. the approximate nominal flux level for this motor is λ∗ = 0.1) (A.55 —/2.605Wb. The motor is produced by Bonfiglioli Group and is marked with the data found in table A. apparent power and power factor.1. a locked-rotor test to determine Rr . dr The motor-bench at Aros electronics AB is equipped with a Panasonic MSMA402A1G AC servo motor (serial no: 03090001F) which is used as load. V∆/Y 230/400 —/460 Hz 50 60 kW 1. The stator resistance Rs was measured directly on the motor terminals. All measurements are performed on a relatively cold motor.

Using the measured stator resistance. (A.06 5.01 3.19 203.84 6. The mean value of these tests yields Lls = Llr = 6. thus minimizing the current through the magnetizing inductance.1 Locked Rotor Test With the rotor locked.4) Llr Va Lm Rr Figure A.79 (mH). 53 .05 (Ω). The rotor of the IM was locked by applying a large braking torque on the servo. (A.48 Stot (VA) 18. which may be neglected. The motor steady-state T model equivalent circuit is then reduced to figure A.02 2.29 71. the rotor resistance.05 110.78 6.04 Ptot (W) 12.11 210.80 6.2: Results from locked rotor test.02 Table A.52 310.A Sensorless Induction Machine Drive.53 160. The reactive power is calculated from Qtot = Rs Lls 2 2 Stot − Ptot . Llr +Lls 2 Ia (A) 1.01 2.07 2.57 Qtot (VAr) 13.01 4.37 (mH) 6.97 447.6) The mean value of Rr = 2.2.1: Reduced T-model for locked rotor test.78 Rr (Ω) 2 2 2. Rr . can be calculated using the measured active power 2 Pphase = (Rr + Rs )Ia .23 52. the slip is equal to one.63 49.07 292.02 (Ω).75 6. This gives minimum effective rotor resistance.5) where Qphase = Qtot /3. A.06 116.74 322. The result of the measurements is found in table A. The rotor and stator inductances Llr and Lls are assumed equal. Rs = 2. The current was increased in steps and the required voltage noted together with apparent power and active power. The reactive power is then equal to 2 Qphase = ω(Llr + Lls )Ia (A.1.

6 139.83 2. apparent power and power factor Rs Lls Llr Va Lm s => 0 RR s Figure A. The equivalent circuit is seen in figure A. The effective rotor resistance becomes large and can be neglected.5 1170.2.3 Qtot (VAr) 1125.7 140. 54 .8) The values from the power analyzer is read three times with aid of a hold function.36 153.3.2 1182.5 1160.6 mH. the following relation is used: 2 2 Qphase = Ia ω1 Ll s + Im ω1 Lm (A. Current.2: Reduced T-model for locked rotor test.3 1192.A Sensorless Induction Machine Drive.2 No-load Test Nominal voltage is supplied to the motor. active power. The current Ia splits between the magnetizing inductance Lm and the core loss resistance Rc according to Im = sin(φ)Ia .3: Results from no-load test.9 Lm (mH) 144. The mean value of the magnetizing inductance is found to be 141.7) To extract the magnetizing inductance Lm . Ia (A) 2.79 148.91 Ptot (W) 143. is noted.6 Table A.95 2. A. (A. The test results can be found in table A.20 Stot (VA) 1134. The servo motor drives the IM at synchronous speed to control the slip speed to zero.

” IEEE Transactions on Industry Applications. vol. Krishnan. “An improved stator flux estimation for speed sensorless stator flux orientation control of induction motors. November/December 1989. Department of Electric Power Engineering. Slemon. Upper Saddle River. Electric Motor Drives. Cho. Västerås. 2. 2005. Chalmers University of Technology. and S. D. “On control of back-to-back converters and sensorless induction machine drives.-Y.-H. [3] ——. Control of Power Electronic Converters and Variable-Speed Drives. 6.-B. no. [5] R.Bibliography [1] A. R. Harnefors. [2] L. Västerås. Mälardalen University. Control of Variable-Speed Drives.” Ph. 2003. Sweden: Department of Electronics. March 2000. Gothenburg. 2001. 2002. [7] R. pp. S.” Master’s thesis. Mälardalen University. 2002. 15.-S. 55 . Department of Electric Power Engineering. New Jersey: [6] M. Shin. pp. 25. [4] G. vol. Sweden: Department of Electronics. Electric Drives. “Evaluation of speed observers for an induction machine. Ottersten. Ottergren. Minneapolis. Prentice Hall. Hyun. Mohan. 1126–1131. 312–318.D. Choe. “Modelling of induction machines for electric drives. 2003. dissertation.” IEEE Transactions on Power Electronics. Chalmers University of Technology. Gothenburg. Minnesota: MNPERE. [8] N. no.