Professional Documents
Culture Documents
DISSERTATION
By
*****
2005
Engineering
UMI Number: 3197781
and more attention has been given to design and development of induction machine
orientation control (FOC). Speed sensorless technology has also been proposed for
decades to overcome the disadvantages of cost and fragility of a mechanical speed sensor.
However, due to the high order, multiple variables and nonlinearity of induction machine
task.
In this research, a sliding mode based flux and speed estimation technique for speed
sensorless control of field oriented induction machine is first investigated. The parameter
sensitivity of the control method is also analyzed. A robust sliding mode speed controller
is also presented, which has the advantage of disturbance rejection and avoiding re-tuning
Then an adaptive sliding mode observer is proposed and the stability is verified by
Lyapunov theory. Two sliding mode current observers are utilized to compensate the
effects of parameter variation on the rotor flux estimation, which make flux estimation
more accurate and insensitive to parameter variation. The convergence of the estimated
ii
Finally, an efficiency optimization method which does not require extra hardware and
minimization and motor loss minimization in the induction motor vector control system is
determined that the motor loss minimization can be achieved by minimizing stator
current in practice.
The simulation and experimental results are presented to demonstrate the potential and
iii
Dedicated to my wife and my parents
iv
ACKNOWLEDGMENTS
Xu, for his academic guidance, his constant help and support of my research. His
supervision has broadened my knowledge in power electronics and drive system. I had
I would like to thank Professor Donald Kasten and Professor Vadim Utkin for being
Rizzoni and Dr. Zheng Zhang for their kindly help during my research work.
I thank all my colleagues of the Power Electronics and Electric Machines (PEEM)
group at The Ohio State University and especially to Dr. Mongkol Konghirun, Dr. Jingbo
liu, Mr. Song Chi, Mr. Jiangang Hu, Dr. Mihai Comanescu, Mr. Reza Esmaili and Ms.
Debosmita Das. We had a very good corporation and had many fruitful discussions
Finally, I would like to express my deepest appreciation to my wife, Yan, who has
been sharing hardships and happiness over years, and my parents and my brothers. Their
Without their constant support none of this would have been possible.
v
VITA
Shenzhen, China
PUBLICATIONS
Research Publication
Jingchuan Li, Longya Xu, Zheng Zhang, “An Adaptive Sliding Mode Observer for
Induction Motor Sensorless Speed Control,” IEEE Trans. Industry Applications, Vol. 41,
No. 4, pp.1039 - 1046, 2005.
Jingchuan Li, Longya Xu, Zheng Zhang, “A New Efficiency Optimization Method on
Vector Control of Induction Motors,” Electric Machines and drives Conference, IEMDC
2005.
Jingchuan Li, Longya Xu, Zheng Zhang, “An Adaptive Sliding Mode Observer for
Induction Motor Sensorless Speed Control,” IEEE Industry Applications Conference,
IAS 2004. Volume 2, pp1329 – 1334, 2004.
vi
Jingchuan Li, Longya Xu, “Investigation of cross-saturation and deep bar effects of
induction motors by augmented d-q modeling method”, IEEE Industry Applications
Conference, IAS 2001, Volume 2, pp 745 – 750, 2001.
Codrin-Gruie Cantemir, Gabriel Ursescu, Jingchuan Li, Chris Hubert, Giorgio Rizzoni,
“An 1800 HP, Street Legal Corvette: An Introduction to the AWD Electrically-Variable
Transmission,” SAE 2005 World Congress.
Codrin G. Cantemir, David Mikesell, Nicholas Dembski, Jingchuan Li, Giorgio Rizzoni,
“Hybrid Electric Refuse Vehicle,” IEEE Vehicular Power and Propulsion – IEEE VPP
2004, Paris.
FIELDS OF STUFY
Studies in:
vii
TABLE OF CONTENTS
Page
Abstract……………………………………………….…………….…..…...ii
Dedication………………………………………………………….……….iv
Acknowledgments……………………………………….……….………….v
Vita……………………………………….……….……………………..….vi
List of Tables………………………………………….…………...………..xi
List of Figures……………………………………………………..……….xii
Chapters
1. Introduction................................................................................................1
viii
3. Sliding Mode Flux Observer for DFO sensorless Control ......................26
Bibliography ................................................................................................120
x
LIST OF TABLES
Table Page
xi
LIST OF FIGURES
Figures Pages
2.4 General block diagram for a field orientation control system .................................... 13
xii
3.12 Estimated and measured current (current iαs ; curve 2: phase current îαs ) ............. 41
3.15 Four quadrate operation (±1200 rpm) (curve1: speed command ω r ; curve 2:
*
3.16 Motor response for step speed change from standstill to 950 rpm ........................... 43
3.20 Speed response with Rr unchage under 0.5pu torque disturbance ............................ 48
3.21 Speed response with Rr decreasing 20% and 0.5pu torque disturbance ................... 48
3.22 Speed response with Rr increasing 20% and 0.5pu torque disturbance .................... 49
3.23 Experimental result with Tr unchage under 2.0 N.m torque disturbance.................. 49
3.24 Experimental result with Tr increase 22% under 2.0 N.m torque disturbance.......... 50
3.25 Experimental result with Tr decrease 22% under 2.0 N.m torque disturbance......... 50
4.4 Speed response at Tl =1.0 Nm (curve1: real speed ωr ; curve 2: estimated speed ω̂r ;
curve 3: flux current id ; curve 4: torque current i q ; curve 5: phase current ia )........ 60
4.5 Speed response at Tl =2.0 Nm (curve1: real speed ωr ; curve 2: estimated speed ω̂r ;
curve 3: flux current id ; curve 4: torque current i q ; curve 5: phase current ia )........ 61
4.6 Speed response with external torque step change Tl =0.5Nm ..................................... 61
xiii
4.7 Speed response with external torque step change Tl =1.0Nm ..................................... 62
4.8 Speed response with external torque step change Tl =2.0 Nm .................................... 62
4.9 Four quadrant operation with trapezoidal speed command (between 300rpm and
1400rpm) ................................................................................................................... 63
4.10 Four quadrant operation with trapezoidal speed command (between ±1200rpm)
.................................................................................................................................... 63
5.7 Speed step response from –0.5pu to 0.5pu (curve 1: speed command ω r * ; curve 2: real
speed ω r ; curve 3: estimated speed ω~r ) .................................................................... 78
~
5.8 Rotor flux estimation. (curve 1: real flux λαr ; curve 2: estimated flux λαr ; curve 3:
~
estimated flux angle θ r ) .............................................................................................. 79
5.9 Trapezoidal speed at ±0.5pu. (curve 1: phase current i a ; curve 2: torque current i q ;
curve 3: estimated speed ω~r ) ..................................................................................... 79
5.10 Transient response to speed step command ±900rpm at no load (curve1: speed
command ω r * ; curve 2: estimated speed ω~r ; curve 3: torque current i q ; curve 4:
phase current ia ) ........................................................................................................ 81
5.11 Real and estimated currents. (curve 1: measured current iαs ; curve 2: observed
~
current iˆαs ; curve 3: observed current iαs ) ................................................................. 82
xiv
5.12 Transient response due to trapezoidal speed command (±900rpm) at no load (curve
1: speed command ω r * ; curve 2: estimated speed ω~r ; curve 3: torque current i q ,;
curve 4: phase current ia ) .......................................................................................... 82
5.13 Transient response due to trapezoidal speed command (±900rpm) at T = 0.5 pu (curve l
5.14 Speed response due to step change command from 360rpm to 1260rpm at
~
T = 0.5 pu . (curve1: real speed; curve 2: estimated speed ω r ; curve 3: torque current
l
5.15 Transient response for step disturbance torque (curve1: real speed ω r ; curve 2:
estimated speed ω~r ; curve 3: torque current i q ) ........................................................ 84
6.5 Motor losses with respect to ids at different load torque ............................................ 96
6.8 Minimum input power point and input power point corresponding to minimum stator
current......................................................................................................................... 99
6.10 Measured stator current vs flux level at different torques ...................................... 100
6.11 The ratio k ip with frequency for different motors ................................................... 101
xv
6.14 Membership for fuzzy logic controller ................................................................... 105
6.22 Current, power and efficiency variation at Tl = 0.2 Nm, n = 900rpm ............................. 112
6.23 Current, power and efficiency variation at Tl = 0.5Nm, n = 900rpm ............................ 113
6.24 Current, power and efficiency variation at Tl = 0.2Nm, n = 1200rpm ......................... 114
6.25 Current, power and efficiency variation at Tl = 0.2Nm, n = 600rpm .......................... 115
6.26 Comparison of efficiency curves at different operation points with and without fuzzy
optimization control. (solid line is with fuzzy optimization, dash line is without fuzzy
optimization) ............................................................................................................ 116
xvi
NOMENCLATURE
• Lm magnetizing inductance
2
L
• σ leakage coefficient, σ = 1− m
Ls Lr
Lr
• Tr rotor time constant, Tr =
Rr
• Bf coefficient of friction
• p differential operator
xvii
• idr , iqr d and q components of the rotor currents
~ ~
• iαs , iβs estimated stator currents by second current observer
~ ~
• λαr , λ βr estimated rotor flux linkages by adaptive sliding mode
observer
• ia phase a current
List of Abbreviations
• SC search controller
xix
CHAPTER 1
1. INTRODUCTION
INTRODUCTION
Induction motors are relatively inexpensive and rugged machines because they can be
built without slip rings or commutators. They are widely used in industry application.
Consequently much attention has been given to induction motor control for starting,
braking, four-quadrant operation, etc. Open loop control of the machine with variable
frequency may provide a satisfactory variable speed drive when the motor has to
operate at steady torque without stringent requirements on speed regulation. When the
drive requirements include fast dynamic response and accurate speed or torque
control, an open loop control is unsatisfactory. Hence it is necessary to operate the motor
in a closed loop mode. The dynamic operation of the induction machine drive system has
an important effect on the overall performance of the system. The control of induction
motors is a challenging problem since it has a nonlinear model, rotor variables are rarely
Several techniques are used to control the induction motor. These schemes can be
classified into two main categories: 1) Scalar control, One of the first ways of controlling
induction machines was the volts/hertz speed control also known as scalar method in
which the machine was excited with constant voltage to frequency ratio in order to
maintain a constant air gap flux and hence provide maximum torque sensitivity. This
method is relatively simple but does not yield satisfactory results for high
1
performance applications. This is due to the fact that in the scalar method, an
inherent coupling exists between the torque and air gap flux, which leads to a
sluggish response of the induction machine. 2) Field oriented control (vector control). To
overcome the limitation of the scalar control method, field oriented methods were
developed. In field oriented control methods the variables are transformed into a
reference frame in which the dynamics behave like dc quantities. The decoupling
control between the flux and torque allows the induction machine to achieve fast
transient response. The field oriented induction machine drive therefore, can be used for
The above traditional control schemes require a speed sensor for closed loop
operation. The speed sensor has several disadvantages from the standpoint view of drive
cost, reliability, and noise immunity. Various speed sensorless approaches have been
proposed in the literature recently. However, due to the high order, multiple variables and
nonlinearity of induction motor dynamics, estimation of the rotor speed and flux
problem for classic vector control is the efficiency improvement. It has been reported that
65% of the electric energy in US is consumed by electric motors [1]. In industrial sector
alone, 76% is consumed by motors and over 90% of these are induction motors.
Induction motors have a high efficiency at rated speed and load. However, at light loads,
iron losses increase, reducing the efficiency considerably. For a 500 hp motor with proper
control, it is reported that the reduction in losses translates into annual savings of $7000
at the energy cost of $0.05/kW.Hr [2] . Obviously, there is a clear motivation for
efficiency improvement.
2
This dissertation is organized as follows. The background and principle of field
existing work up to now about induction motor sensorless control and efficiency
improvement are reviewed. An effective sliding mode flux and speed observer for direct
field orientation control is presented and the parameter sensitivity is analyzed in Chapter
3. The simulation and experimental results are also presented. In Chapter 4, a robust
chattering free sliding mode speed controller is presented and analyzed. The results are
in Chapter5. The stability is derived using Lyapunov theory and the speed estimation is
induction motor control is presented in Chapter 6. This new method has the advantage
that it is insensitive to motor parameters and does not require extra hardware, at the same
time it can improve motor efficiency dramatically especially in light load. The simulation
and experimental results show the effectiveness and validation of the method. The
3
CHAPTER 2
2. BACKGROUD
BACKGROUDAND
ANDLITERATURE
LITERATUREREVIEW
REVIEW
As stated above, induction motor drive system is becoming more and more a
competitive system in many high performance motion drive application. There are three
main major components in an induction motor drive system: an induction motor, a power
electronic device and a controller. The field orientation control (FOC), integrating modern
control theory, power electronic and DSP/micro-processor technology, has made possible
the development of high performance induction motor drive systems. In this chapter, the
basis of field orientation control of induction motors is summarized. The state of art of
speed sensorless control and motor loss minimization is reviewed. The existing problems in
developed in the electric motor is a result of the interaction between current in the
armature and the magnetic filed produce by motor. Independent control of the field and
stator winding determines the magnetic field of the motor, while the current in the rotor
armature winding can be used as a direct means of torque control. In a similar manner to
between the field flux and the armature MMF is emulated in induction machines by
orienting the stator current with respect to the rotor flux so as to attain decoupled
controlled flux and torque. Such controllers are called field orientation controllers (also
A basic understanding of the decoupled flux and torque control resulting from field
orientation can be obtained from the d-q axis model of an induction machine with the
reference axes rotating at synchronous speed. This control is based on projections that
transform a three-phase time and speed dependent system into a two co-ordinate (d- and
q- axis) time invariant system. These projections lead to a structure similar to that of a
DC machine control.
The three-phase voltages, currents and fluxes of induction motors can be analyzed in
terms of complex space vectors [3-6]. With regard to the currents, the space vector can be
defined as follows. Assuming that ia , ib , ic are the instantaneous currents in the stator
v
phases, ia + ib + ic = 0 , the complex stator current vector is is defined by:
v
is = k (ia + αib + α 2ic ) ( 2.1)
2 4
j π j π 2
where α = e 3
, α =e
2 3
represent the spatial operators, k = . The following
3
5
β
b
iβs is
iαs α
where (a, b, c) are the three-phase system axes. This current space vector depicts the three
phase sinusoidal system. The space vector can also be plotted in another reference frame
with only two orthogonal α − β axis. The real part of the space vector is equal to the
instantaneous value of the direct-axis stator current component iαs . The imaginary part is
equal to the quadrature axis stator current component iβs . Thus, the stator current space
vector in the stationary reference frame attached to the stator can be expressed as:
v
is = iαs + jiβs ( 2.2)
The space vectors of other motor quantities (voltages, rotor currents, magnetic fluxes,
etc.) can be defined in the same way as the stator current space vector.
1) Clarke transformation
In symmetrical three-phase machines, the direct and quadrature axis stator currents
(as shown in Figure 2.1) are fictitious two-phase current components. Assuming α -axis
is in the same direction with a-axis, we have following relations with respect to the actual
2
The constant k ( = ) for the non-power-invariant transformation. In this case, the
3
quantities ias and iαs are equal. If it’s assumed that ia + ib + ic = 0 , the quadrature-phase
components can be expressed utilizing only two phases of the three-phase system:
iαs = ias
1 2 (2.4)
i βs = ibs + ics
3 3
The components iαs and iβs , calculated with a Clarke transformation, are attached to
the stator reference frame α − β system. In vector control, all quantities must be
expressed in the same reference frame. The stator reference frame is not suitable for the
control process. The space vector is is rotating at a rate equal to the angular frequency of
the phase currents. The components iαs and iβs change with time and speed. These
components can be transformed from the stator reference frame to the d-q reference
frame rotating at the same speed as the angular frequency of the phase currents. The ids
and iqs components do not then depend on time and speed. If the d-axis is aligned with
7
β
q
is
i sβ d
ψr
isd
isq
θ
i sα α
The components ids and iqs of the current space vector in the d-q reference frame are
The inverse Park transformation from the d-q to the α − β coordinate system is found
The system model defined in the stationary α − β coordinate system attached to the
8
d
vαs = Rs iαs + λαs (2.7)
dt
d
v βs = Rs i βs + λ βs ( 2.8)
dt
d
vαr = Rr iαr + λαr + ωλβr ( 2.9
dt
d
v β r = Rr i β r + λβr − ωλαr ( 2.10)
dt
where
Besides the stationary reference frame, induction motor model can be formulated in a
general d-q reference frame, which rotates at a general speed ωe . The motor model
voltage equations in the general reference frame can be expressed by using the
transformations of the motor quantities from one reference frame to the general reference
frame. The two phase d-q model of an induction machine rotating at the synchronous
speed will help to carry over this decoupled control concept. This model can be described
d
vds = Rs ids + λds − ωe λqs ( 2.15)
dt
d
vqs = Rs iqs + λqs + ωe λds ( 2.16)
dt
9
d
vdr = Rridr + λdr + ( ωe − ωr )λqr ( 2.17)
dt
d
v qr = Rr iqr + λqr − ( ωe − ωr )λdr ( 2.18)
dt
3 PLm
Te = (λds iqs − λqr ids ) ( 2.23)
2 Lr
J B
Te − Tl = pω r + ω r ( 2.24)
P P
This induction motor model is often used in field orientation control (vector control)
algorithms. To achieve this, the reference frames may be aligned with the stator flux-
linkage space vector, the rotor flux-linkage space vector or the magnetizing space vector.
The most popular reference frame is the reference frame attached to the rotor flux
flux and locking the phase of reference system such that the rotor flux is entirely in the d-
axis, resulting in
λqr = 0 ( 2.25)
This expresses the field orientation concept in d-q variables. Assuming the machine is
supplied from a current regulated source so the stator equation can be omitted, the d-q
10
0 = Rr idr + pλdr ( 2.26)
3 PLm
Te = λdr iqs ( 2.29)
2 Lr
Equation (2.29) demonstrates the desired torque control properties in terms of the
current components and the rotor. If the rotor can be kept constant just as it is in the
D.C. machine, then the instantaneous torque control can be achieved by controlling
the current component. From these equations, the following relations can be
obtained:
Lm
iqr = − iqs ( 2.30)
Lr
Lm
λdr = ids ( 2.31)
1 + Tr p
Lm iqs
ωs = ( 2.32)
Tr λdr
the steady state, λdr = Lm ids and idr = 0 . The phasor diagram of the field oriented
11
Lm
iqre = − iqs iqse
Lr q- axis
idse
idqs
r r
λdr = λdqr
d- axis
Equation (2.31) shows that the machine flux can be determined by controlling the current
component ids . Therefore, in the steady state, the constant flux can be obtained by
constant ids . As a result, the torque control can be easily obtained by controlling ids as
seen in (2.29). Equation (2.32) is the most important expression for the practical
implementation of the induction machine in indirect field control which will be discussed
later.
Field orientated controlled machines need two constants as input references: the
torque component (aligned with the q-axis) and the flux component (aligned with d-axis).
Since the field orientation control is simply depended on projections, the control structure
can handle instantaneous electrical quantities. This makes the control accurate in every
working operation and independent of the limited bandwidth mathematical model. The
12
1) The ease of reaching constant reference (torque component and flux component of
2) The ease of applying direct torque control because in the d-q reference frame the
By maintaining the amplitude of the rotor flux at a fixed value we have a linear
relationship between torque and torque component current. We can then control the
torque by controlling the torque component of stator current vector. The general block
diagram of a field orientation control system for an induction motor is shown in Figure
2.4.
*
iq *
PI Uα
iq
* T −1 (θ ) Uβ
* PWM Inverter
id
PI
id
iq
iα
a,b,c ia
T (θ )
id iβ to
α, β ib
Induction
Motor
Figure 2.4 General block diagram for a field orientation control system
Depending on the reference frame transformation used, two types of field orientation
control are mostly used: the rotor flux orientation (RFO) [8-10] and the stator flux
13
orientation (SFO) [11-13]. In the rotor flux orientation vector control, the reference frame
rotates synchronously with the rotor flux, while in the stator flux orientation the reference
frame rotates with the stator flux. In both these reference frames, the dynamics of an
machine. The rotor field orientation control of induction machine can also be classified as
a direct field orientation control [14-17] or an indirect field orientation control [18-20]
depending on how the flux information necessary to perform the reference frame
transformation is obtained.
Knowledge of the instantaneous position of the flux vector, with which the revolving
reference frame is aligned, constitutes the necessary requirement for proper field
so called direct field orientation (DFO). Only the air gap flux can be measured directly. A
simple scheme for estimation of rotor flux vector is based on measurement of air-gap flux
and stator current. The disadvantage of direct measurement method is that a flux sensor is
expensive and needs special installation and maintenance, thus, spoil the ruggedness of
the induction motor. In practice, the rotor flux is usually computed from the stator voltage
and current. This technique requires the knowledge of the stator resistance along with
the leakage and magnetizing inductance. This method is commonly known as the voltage
model observer [21]. The scheme of a direct field orientation is shown in Figure 2.5.
14
iαβs
λdqr θe
Flux
vαβs observer
tg −1 (λβr / λαr )
The stator flux along the α and β axes, in the stationary frame of reference, can
Lr ˆ
λˆαr = ( λαs − Lσ iαs ) (2.35)
Lm
Lr ˆ
λˆ βr = ( λβs − Lσ iβs ) (2.36)
Lm
2
Lm
where Lσ = ( Ls − ) is leakage induction. This method depends on parameters
Lr
such as the stator resistance and the leakage inductance. The study of parameter
sensitivity [22,23] shows that the leakage inductance can significantly effect system
performance such as stability, dynamic response, and utilization of the machine and the
inverter. The major difficulty in this case is the need for three motor parameters. The
inductance parameters are only moderately affected by saturation. There are also problems
with integrating low frequency signals and with the fact that the stator resistance
15
voltage drop becomes dominant at low speed. These limitations preclude use of this
scheme at low speed. However, it is practical over a reasonable speed range and is used in
many implementations.
Indirect field orientation is based on the slip relation as shown in Equation (2.32). The
control algorithm for calculation of the rotor flux angle using IFO control is shown in
the Figure 2.6. This algorithm is based on the assumption that the flux along the q-
Lm iqs
axis is zero which imposes a condition on the command slip that is ω s = , a
Tr λdr
necessary and sufficient condition to guarantee that all the flux are aligned along d-
axis and the flux along q-axis is zero. The angle can be then calculated by adding the
slip angle and the rotor angle. The slip angle includes the necessary and sufficient
i *ds
i *qs
T −1 IM
θe
∫
ωe ωr
1 i * qs
Tr i * ds ωs
The IFOC is an open loop feed forward control in which the slip frequency is fed
forward, guaranteeing the field orientation. This feed forward control is very sensitive to
16
the rotor open circuit time constant Tr. Therefore, T r must be known in order to achieve
a decoupled control of torque and flux by controlling iqs and ids respectively. When T r is
not set correctly the motor will be detuned, and the controller performance will become
are two feedback loops typically used to implement field orientation control and speed
control. The field orientation control is implemented in the inner current loop, the
decoupled control of flux and torque can be obtained by d- and q-axis current
regulator. In the area of controlled electric drives, the drive inertia and load
performance, the speed controller has directly impacts the system performance. It is
desirable to have a drive system that can provide fast dynamic response, a parameter-
insensitive control feature, and rapid recovery from speed drop caused by impact loads. A
speed regulation loop [24-26]. The PI controller offers fair performance in a stable
and robust manner if it is well tuned. The PI controller is usually designed in a linear
region ignoring the saturation-type non-linearity. At some working area, the behavior of
such controller could be satisfied. When the controller is applied to variable speed motor
large overshoot, slow setting time, and, sometimes, even instability. So the parameters of
17
the PI controller should be modified according to the various operating condition of
induction motor. This would add difficulty to the on-line debugging and cause several
gains cannot satisfy the requirements for different speed commands. 2) PI speed
controller lacks the ability to handle detuning problems when parameters of the
machine vary. 3) Tuning PI gains is very time consuming. The limitations of the PI
controller have motivated research into alternative control techniques such as fuzzy logic,
experience in the controller [27-37]. FLC can perform better with high nonlinearities
techniques other than a conventional PI controller. Recent literature has explored the
potential of fuzzy control for machine drive applications It has been shown that a
properly designed direct fuzzy controller can outperform conventional PID controllers.
However, the performance will still degrade when the machine is severely detuned.
Another approach proposed and widely studied is the discontinuous sliding mode
control [38-43]. SLM control offers attractive features such as insensitivity to parameter
variations (as long as the bounds of the parameter variations are known) and
sliding mode in motion control exhibits chattering imposed by the discontinuity of the
control action. The essential of sliding mode control is that the discontinuous feedback
control switches on one or more manifolds in the state space. Ideally, the switching of
18
manifolds. In practice, the frequency is not infinitely high due to the finite switching time
and with effects of un-modeled dynamics. This chattering is quite undesirable for most
applications. Different schemes have been suggested to eliminate the chattering such as
sliding-mode controller with boundary layer [44], fuzzy sliding-mode controller (FSMC)
[45]. Bartolini [46,47] proposes to introduce an integrator into the controller and design a
discontinuous control as the derivative of the actual control signal. The chattering
chattering free sliding mode technique which produces a continuous signal to the system
[49-52].
The approach of speed senseless control of induction motor has been receiving more
and more attention in industry application since it can reduce cost and avoid fragility of a
mechanical speed sensor, and eliminate the difficulty of installing the sensor in some
applications. Different techniques for obtaining the rotor speed, estimating the rotor flux
of an induction machine for sensorless control have been extensively studied in the past
2) Voltage model and current model flux and speed estimations [56-58];
19
5) Extended Kalman filters [65,66];
7) Artificial neural network and artificial intelligence based sensorless control [71,72].
low speed operation. These approaches rely on the motor response to the injection of
relatively high-frequency test signals, which investigate the motor saliency due to
saturation or geometric construction. They need high precision in the measurement and
increase the hardware and software complexity with respect to a standard vector control
scheme. When applied with the high frequency signal injection [55], the method may cause
torque ripples, vibration, and audible noise. Moreover, motors having a low saliency
content do not give an appreciable response, whereas enhancing the saliency requires a
proper machine design, therefore the saliency based technique is machine specific and
unavailability of the signal at low speed and parameter sensitivity. Different schemes to
overcome these problems and to improve the sensorless control have been proposed in the
literature.
The voltage model flux estimations [56,57] have problems at low frequency regions,
because the signal to noise ratio of the stator voltage measurement is very poor, and
voltage drop on the stator resistance is dominant. The voltage model is also sensitive to
the leakage inductance. The current model flux observer is considered to have better
performance at low speeds. Also, its accuracy is relatively unaffected by the leakage
inductance for any operating condition. However, it does not work well at high speed due
20
to rotor resistance variation. To get better performance, it has been suggested to use the
current model observer at low speed and voltage model observer at high speed [56, 57].
To further improve the observer performance, close loop rotor flux observers are proposed
[58].
Model reference adaptive schemes (MRAS) are proposed in [59-61], where one of the
flux estimators acts as a reference model, and the other acts as the adaptive estimator. The
estimation is based on the comparison between the outputs of two estimates, and the
output errors are then used to drive a suitable adaptation mechanism that generates the
estimated speed. These schemes require integration and. To overcome the integration
problem, Peng [60] suggested the use of back-EMF and instantaneous reactive power as
alternative ways to estimate the velocity in the adaptive controller. However, the
performances are still limited by parameter variations and the accurate flux estimation
Adaptive observer based approaches [62,63] can have preferred performance using
the derived adaptive laws with relatively simple computation. However, their robustness
[64], in which only the rotor flux, not the stator current is estimated. The correction is
then applied by using the error between the actual stator voltage vector and an estimate
ones. However, this requires adding voltage sensors to the system, which is not desirable.
Extended Kalman filters have been proposed in [65, 66] as a potential solution for
better flux estimation. Unfortunately, this approach contains some inherent disadvantages
such as computational expense and having no specific design and tuning criteria.
21
Sliding mode has been documented to have the advantages of robustness and
parameters insensitivity [67] and been recognized as the prospective control methodology
for electric motors. Flux observers have been designed [67-69] using the sliding mode
technique for sensorless speed control of induction machines. These algorithms use a
current model flux observer and apply a correction term based on the current estimation
error. The observers require the rotor speed and rotor time constant for the current and
flux estimations. Therefore, an error in the estimated speed or rotor time constant will
affect the current and flux estimations, and thus degrade the observer accuracy.
Other algorithms for speed sensorless vector control, such as artificial neural network
[71] and artificial intelligence (AI) [72], can achieve high performance, but are relatively
An effective sliding mode based flux and speed estimation technique for sensorless
decoupled by the proposed sliding function, which makes the observed rotor flux
independent of rotor speed. But the observed flux calculation is still sensitive to
parameter variation. To overcome this sensitivity and make flux and speed estimation
robust to parameter variations, an adaptive sliding mode flux and speed observer is
proposed. Two sliding mode current observers are used in the method. The effects of
parameter deviation in the rotor flux observer can be alleviated by these two current
sliding mode observers. The stability of the method is proven by Lyapunov theory. An
22
2.3. Efficiency optimization of induction machine control
Induction machines consume most of the world’s electrical energy every year.
Improving efficiency of electrical drives is important not only for energy saving, but also
variable frequency drive to get a best transient response. However, most of the time, the
drive system operates with light loads. In this case the core losses become excessive
causing poor efficiency. To improve the motor efficiency, the flux must be reduced,
A number of methods for efficiency improvement through flux control have been
proposed in the literature. They can be classified into three basic types. The simple pre-
computed flux program as a function of torque is widely used for light load efficiency
improvement. This method, however, yields only a partial improvement in the system
efficiency. The second approach is based on the modeling of the motor and the losses to
maximized) to yield the maximum efficiency. Thus, this method treats the situation
analytically by properly modeling the losses and is called Loss Model Controller
(LMC)[73-75]. The third method is on-line efficiency optimization control on the basis of
search, and has a feedback nature that finds the maximum efficiency and is called Search
The LMC method has the advantage that it is fast, however, the accuracy depends on
correct modeling of the motor drive and the losses. Garcia [73] proposed a simple loss
model consisting of computation of iron loss, rotor and stator losses in function of stator
current and in the frame. For a given speed and torque, the solution of the loss model
23
yields the flux current for which the total loss is minimal. Lorenz and Yang [74] pointed
that major loss saving is possible by considering the system as a whole and employing
dynamic programming to select the operating flux. Kioskeridis and Margaris [75]
calculated the total of iron loss, copper losses, and stray loss and found an optimal flux
level that minimizes the total loss. Thus, LMC method is to develop controllers for
different drive systems by building the loss models and including different applications.
The LMC method consists in the real time computation of losses and corresponding
selection of flux level that results in minimum losses. As the loss computation is based
SC method on the other hand offers optimum efficiency based on the exact
measurement of input power (or DC bus power). Sul and Park [76] proposed a method that
maximizes the efficiency by means of finding optimal slip. The technique can be
considered as an in direct way to minimize the input power. For the vector drive, Kirschen
et al. [77] reduced the flux in small steps to reach to the optimum condition. Kim et al.
[78] adjusted the squared rotor flux according to a minimum power algorithm using
search method. Sousa et al. [79] reduced the reference flux current by minimizing input
DC bus power using fuzzy logic, where the torque pulsation is overcome by using feed
forward pulsating torque compensation. Ta and Hori [80] improved the convergence rate
the upper and lower limit of the flux-producing current before the algorithm starts. Thus,
the object in the SC is to reduce the search time and torque pulsations. Moreover, The SC
24
method does not require the knowledge of machine parameters and completely
LMC and the LMC works on the model and not on the actual drives. In LMC, the loss
simplicity of the method and not requiring extra hardware. However, it is sensitive to
motor parameters which change considerably with temperature and load condition.
Performance of the LMC method deteriorates when parameters change, the online
estimation of the parameters makes the method far more complicated. On the other
hand, SC method measures input power to searches the flux where the motor runs at
condition. However, it does require extra hardware to measure DC bus current and does
not be used in the classical vector control system where additional sensor is not available.
To take the advantages of both LMC and SC, an efficiency optimization method by
minimizing the stator current is presented. This approach does not require extra hardware
minimization and motor losses minimization in the induction motor vector control system
has been investigated. It is pointed that minimum stator current point is very close to
minimum losses point in most cases and the losses minimization can be achieved by
minimizing stator current in practice. A fuzzy logic based search method is simulated and
implemented. Simulation and experimental results are given in the paper to verify the
proposed method.
25
CHAPTER 3
3.SLIDING
SLIDINGMODE
MODEFLUX
FLUXOBSERVER
OBSERVERFOR
FORDFO
DFOSENSORLESS
SENSORLESS
CONTROL
CONTROL
3.1. Introduction
In the past decade, a wide range of nonlinear methods for feedback control, state
estimation, and parameter identification has merged. Among them, sliding mode control
gained wide acceptance because sliding mode method can offer many good properties, such as
insensitivity to parameter variations, external disturbance rejection, and fast dynamic response.
In this chapter, based on the concept of equivalent control of sliding mode, a speed
observation system, which comprises a current observer, a rotor flux observer and a rotor
speed observer, is presented for a direct rotor flux oriented induction motor drive.
reference frames include stationary reference frame, which is fixed to the stator, and
defining stator currents and rotor fluxes as the state variables, we can rewrite the
d
vαs = Rs iαs + λαs ( 3.1)
dt
26
d
v βs = Rs i βs + λ βs (3.2)
dt
d
vαr = Rr iαr + λαr + ωλβr (3.3)
dt
d
v β r = Rr i β r + λβr − ωλαr (3.4)
dt
For squirrel rotor, vαr = 0, v βr = 0 . Eliminating the rotor currents and stator fluxes from
Lm
λα s = λα r + σ L s iα s (3.9)
Lr
L
λ β s = m λ β r + σ Ls iβ s (3.10)
Lr
1
iα r = ( λ α r − L m iα s ) (3.11)
Lr
1
iβ r = ( λ β r − Lm iβ s ) (3.12)
Lr
Substituting Equations (3.9)-(3.12) into Equations (3.1)-(3.4), yield the induction motor
2
d 1 Lm 1 1 Lm 1 L 1
iαs = λαr + ω r λ βr − ( Rs + m )iαs + v (3.13)
dt σLs Lr Tr σLs Lr σLs Lr Tr σLs αs
2
d 1 Lm 1 1 Lm 1 L 1
i βs = λαr − ωr λαr − ( Rs + m )iβs + v (3.14)
dt σLs Lr Tr σLs Lr σLs Lr Tr σLs βs
27
d 1 L
λαr = − λαr − ωr λβr + m iαs (3.15)
dt Tr Tr
d 1 L
λβr = − λβr + ωr λαr + m iβs (3.16)
dt Tr Tr
2
d L 1 L
Let p = , Tr = r , k 2 = , σ =1− m ,
dt Rr σLs Ls Lr
2
k L L L
β = 2 m , k1 = k 2 ( Rs + m ) , k 3 = m ,
Lr LrTr Tr
the above equations become:
β
piαs = λαr + βω r λβr − k1iαs + k 2 vαs (3.17)
Tr
β
piβs = λβr − βω r λαr − k1iβs + k 2 v βs (3.18)
Tr
1
pλαr = − λαr − ωr λβr + k 3iαs (3.19)
Tr
1
pλβr = − λβr + ωr λαr + k 3iβs (3.20)
Tr
pΛ = − AΛ + k 3I (3.22)
1
T ωr
where I = [iαs , iβs ] , Λ = [λαr , λβr ] , A = r
T T
− ω 1
r
T r
It can be seen that the term AΛ appears in both current and flux equations of the
machine. So this model has the advantage that the coupling terms between α and β axes
28
are exactly same and the coupling terms can be replaced by the same sliding function
As noted above, the current and flux observer can be designed by replacing the term
AΛ with sliding function U . The sliding mode current observer is defined as:
ˆ = −U + k ˆI
pΛ (3.26)
3
where
ˆ = [λˆ ,λˆ ]T , U = [U ,U ]T
ˆI = [î ,î ]T , Λ
αs βs αr βr αr βr
and
Two independent sliding functions U αr and U βr are designed for the α and β
axes of the current observer, respectively. It is noted that the sliding functions U αr
29
and U βr are only dependent on the error between the measured and estimated phase
currents. The α and β axes rotor fluxes are only intergration of these sliding
functions and their own currents. So, the designed current and flux observers for
the α and β axes have no coupling between them, making the current and the flux
The stability of the observre can be proved by Lyapunov stability theory. Let us select
1 T
V = sn sn (3.27)
2
where sn = [sαs s βs ]. The Lyapunov function V is positive definite, which satisfies the
V& = sn s&n
T
(3.28)
To satisfy the Lyapunov stability, second condition must satisfy V& < 0 . From Equation
Thus
|Γ|+|Φ|+|Ω|
u 0 > Bslid =
ids + iqs
where Bslid is the boundary of sliding function. When the system reach the sliding
surface s n = 0 , that means the observed currents converge to the actual ones, then the
flux estimation just an integration of sliding mode function without need of other
information related motor parameter or speed. The resulting equivalent control depends
equivalent control is the slow component of real control that can be obtained by using a
low-pass filter.
1 1
U αreq = U α r , U βreq = U βr ( 3.30)
µs + 1 µs + 1
1
ω̂r
U αeqr T̂r λ̂
eq = αr ( 3.31)
1 λ̂β r
U β r − ω̂r
T̂r
measured ones, the rotor flux can be calculated from Equation (3.26)
31
Moreover, the angle of the rotor flux can be calculated by the following equation:
λˆ βr
θˆ = tan −1 ( ) (3.34)
λˆαr
This rotor flux angle is used in the field orientation control. Noted that the rotor fluxes
are estimated only by sliding function and the their own currents. It requires no speed
1 λ̂ β r U αeqr
T̂ = 1 λ̂α r ( 3.35)
r λ̂ r λ̂ β r − λ̂α r U βeqr
ω̂ r
where λˆ r = λˆαr 2 + λˆ βr 2
The rotor speed and rotor time constant can be calculated as:
ω̂r =
1
(λ̂ U
βr
eq
αr − λ̂αrU βeqr ) (3.36)
λ̂r
It is important to notice that the observer structure is decoupled in the sense that the
estimation process for α and β axis fluxes are independent because of the choice of
Figure 3.1 shows the diagram of system block using sliding mode observer for direct
filed orientation control of induction motors. Two motor phase currents are measured.
These measured currents are used for the Clarke transformation module. The outputs of
this projection are designated by iαs and iβs . These two components of the current are the
inputs of the Park transformation that gives the current in the d, q rotating reference
32
*
frame. The ids and iqs components are compared to the references ids (which is the
*
output of speed regulation) and iqs (the torque reference).
Recitifier
ωr * ∆ωr iqs
*
PI Uq
*
*
PWM1~6
Speed Uα
regulator regulator d,q
Space
iqs to
ωr ids
*
* α, β Uβ * Vector 3-phase
Flux PI Ud PWM Inverter
T −1(θ )
weakening regulator
ids
θ
iαs
α, β
to a,b,c ia
d,q to
α, β ib
T (θ )
iβs
Induction
Motor
Speed Sliding Mode
observer Flux Observer
The presented flux and speed observers have been simulated by Matlab for direct
field orientation control. The simulation is based on per unit system. The motor
1 HP 4 poles
Figure 3.2 through 3.6 show the simulation results. The real motor speed and
estimated speed are shown in Figure 3.2 for a step speed command at start up. Figure 3.3
33
shows the real and estimated stator currents. Once the estimated currents converge to real
currents, the rotor fluxes can be calculated from the equivalent control as shown in Figure
3.4, where the estimated rotor fluxes converge to real rotor fluxes. The transient speed
response for trapezoid command is shown in Figure 3.5. Figure 3.5(b) shows the speed
track error and Figure 3.6 shows the real and estimated rotor flux under this condition.
34
Figure 3.4 Real and estimated rotor fluxes
35
Figure 3.6 Real and estimated rotor flux
digital signal processor. HIL evaluation is to use a computer model of the process as the
real target hardware, and on the other hand, the control and estimation algorithm are
implemented in real time. The purpose of HIL is to make evaluation of the proposed
algorithm as closely as possible to those that would be encountered in the real time
equations. The control and estimated algorithms are implemented in 32-bit Q-math
approach, interacting with the motor model rather than the real targeted physical system.
The main advantages of this evaluation are: 1) the control software are implemented
and evaluated in real time and can be debugged very easily in the absence of motor; 2)
The control software can be easily transferred to the real drive system with only minor
changes. Figures 3.7 through 3.10 show the HIL simulation for induction motors at
36
different power rates. (from 1 hp motor to 50hp motor). The results show that the sliding
simulated for different motors. The results are very similar to the results of Matlab
simulation. The results also prove that the sliding mode algorithm is stable and can be
37
Figure 3.9 HIL simulation results for 5hp motor
38
3.7. Experimental results
TMS320x24xx DSP generates six pulse width modulated (PWM) signals, which control
the six power devices in the inverter. Two of the motor phase currents (ia and ib) are
sensed using the inverter leg resistors and measured by the two analog-to-digital
converters (ADCs) in TMS320x24xx. The advantages of this method are the low cost and
referenced to the DC bus common. The measured current of this method is no longer
motor phase current, but half-bridge current. If the low side switch is conducting (through
either the transistor or freewheeling diode) then the current is equal to that motor phase
current. This certainly occurs periodically throughout the PWM cycle, so a reconstruction
circuit including a sample and hold amplifier is required. In addition, the DC bus voltage
is also measured by an ADC channel. This information is used to calculate the three
The experimental setup and connection is illustrated in Figure 3.11. The experimental
results are shown in Figures 3.12 through 3.17. The estimated stator current and
39
measured current are compared in Figure 3.12. It shows these currents are very close to
each other and eventually the estimated one will converge to the measured one. Figure
3.13 shows the estimated rotor flux and flux angle, which is used for the Park and
inverse-Park transformation. The rotor flux trajectory estimated by the sliding mode
observer is shown in Figure 3.14. Figure 3.15 shows the transient response of drive
system to a trapezoidal speed command. The performance of motor speed step change
AC~
DSP board
40
iαs 1A/div
îαs
1A/div
20ms/div
Figure 3.12 Estimated and measured current (current iαs ; curve 2: phase current îαs )
θ̂ r
λˆαr
ia 5A/div
20ms/div
41
λ̂ r
ωr∗
1818rpm/div
ω̂ r
iq
5A/div
ia
1s/div
Figure 3.15 Four quadrate operation (±1200 rpm) (curve1: speed command ω r ; curve 2:
*
42
Speed step response
1200
1000
800
400
200
0
0 1 2 3 4 5
Time (s)
Figure 3.16 Motor response for step speed change from standstill to 950 rpm
3.00
2.00
1.00
Phase current (A)
0.00
0 1 2 3 4 5
-1.00
-2.00
-3.00
Time (s)
43
3.8. Parameter sensitivity analysis
The flux observer model presented above has accurate flux estimation when motor
parameters are exactly known. However, when the motor parameters are changed due to
temperature or different from preset values, the estimated flux and speed will deviate from the
real values. To investigate the influence of parameter variation, we change the coefficients
k1 , k2 , and β in observers. There will be errors ∆k1 , ∆k 2 , ∆β exist if these parameter are
β ∆k1 ˆ ∆k 2
U eq = AΛ + I− V ( 3.39)
β + ∆β β + ∆β β + ∆β
The error is
∆β ∆k1 ˆ ∆k 2
∆U = U eq − AΛ = − AΛ + I− V ( 3.40)
β + ∆β β + ∆β β + ∆β
This error will cause incorrect flux and speed estimation. Figures 3.18 and 3.19 show
the simulation results when k1 and β are change by 20%. It can be seen from the
simulation that the estimated flux is deviated from actual value and the observed speed
fluctuate around real speed. The influence of rotor resistance change is shown in Figures
3.20 - 3.22. In Figure 3.20 the parameters in observers have the same value as the
induction motor model does. In Figures 3.21 and 3.22, the rotor resistance in flux
observers is increased and decreased by 20% separately. As can be easily observed, the
44
changes of rotor resistance will produce substantial speed error in steady state. This
Figure 3.23, the rotor time constant in flux observers is exactly the same as motor actual
value in the experiments. In Figures 3.24 and 3.25, the rotor time constant value in
the system is insensitive to rotor parameter under no load condition. When load increases,
the motor currents increase, the error of equivalent control will increase, thus the speed
error is more dependent on rotor parameter. In low speed, system performance is also
45
(a) λαr real rotor flux, λobs estimated rotor flux
46
(a) λαr real rotor flux, λobs estimated rotor flux
47
Figure 3.20 Speed response with Rr unchage under 0.5pu torque disturbance
Figure 3.21 Speed response with Rr decreasing 20% and 0.5pu torque disturbance
48
Figure 3.22 Speed response with Rr increasing 20% and 0.5pu torque disturbance
700
600
500
Real Speed
Speed (rpm)
300
200
100
0
0 2 4 6 8 10
Time (s)
Figure 3.23 Experimental result with Tr unchage under 2.0 N.m torque disturbance
49
700
600
500
Real Speed
Speed (rpm)
Estimated Speed
400
300
200
100
0
0 2 4 6 8 10
Time (s)
Figure 3.24 Experimental result with Tr increase 22% under 2.0 N.m torque disturbance
800
700
600
500
Speed (rpm)
Real Speed
400
Estimated speed
300
200
100
0
0 2 4 6 8 10
Time (s)
Figure 3.25 Experimental result with Tr decrease 22% under 2.0 N.m torque disturbance
50
3.9. Conclusion
A sliding mode flux and speed observer is presented and implemented. The terms
containing fluxes, which are common in both current and flux equation, are estimated by
a sliding function, which makes d- and q- axis flux equations decoupled in the stationary
frame. The flux estimation is easy to calculate and merely an integration of known terms,
which make the algorithm simple to implement. However, the parameter sensitivity
analysis shows that the equivalent control will detune if the parameters in observer is
incorrect, causing flux and speed estimation incorrect, which is confirmed by simulation
and experiments.
51
CHAPTER 4
4. ROBUST
ROBUST SLIDING
SLIDING MODE
MODE SPEED
SPEED CONTROLLER
CONTROLLER
4.1. Introduction
A high performance speed motor drive should possess good command tracking and
load regulation dynamic responses, and these responses should be insensitive to the
operating condition. The uncertainties usually are composed of plant variations, external
load disturbance, and nonlinear dynamics of the plant. Many researches have been
reported for the robust speed control of an induction motor. The proportional (P),
proportional plus integral (PI), proportional plus integral plus derivative (PID)
conventional controllers are very easy to design and implement. The proposed continuous
sliding mode controller is robust to load changes and system disturbances. Also this
method overcomes the chattering problem which is the main concern when using
discontinuous sliding mode controller. It can prevent the performance degradation and
simulations and experimental results prove that the proposed controller is robust to
external disturbance and can also follow speed command trajectories very well without
52
4.2. Sliding mode controller design
x ∈ ℜn ,u ∈ ℜm , rank (B( x )) = m
The components of the control input and the derivative of are assumed bounded. The
design goal is to find control u such that the motion of the system (4.1) is restricted
S = { x : σ ( x , t ) = 0 }, σ = (σ 1 ,σ 2 ,...,σ m )T (4.2)
where σ i ,( i = 1,2,..., m ) are continuous functions. Based on sliding mode theory, the
behavior of the system (4.1) on manifold (4.2) is determined by the selection of the
1 T
V ( x ,t ) = σ σ (4.3)
2
The state will move toward and reach manifold S and the stability will be proved if
V& ( x , t ) < 0
when σ ≠ 0 . That is
53
x& 2 = f 2 (x1 , x 2 ) + B( x1 , x2 )u( t ) (4.6)
x1 ∈ ℜ n −m , x 2 ∈ ℜ m , u ∈ ℜ m
B is a nonsingular matrix, rank (B) = m . The aim is to drive the state of the system to
manifold defined by
S = { x : ϕ ( t ) − ξ ( x ) = σ ( x ,t ) = 0 } (4.7)
where x is the state vector obtained by augmenting x1 and x2 . ϕ ( t ) is the time dependent
part of the sliding function and contains reference inputs to be applied to the controlled
The stability conditions for selected control must be examined first. This selection
should ensure the stability of the system’s motion in the origin of the subspace, whose
coordinates are distances from the sliding mode manifold. For the selected manifold (4.6)
the first choice is the Lyapunov function in a quadratic form of control error as in Equation
(4.3). The solution σ ( x ,t ) = 0 will he stable if the time derivate of the Lyapunov
V& = −σ T Dσ (4.8)
where D is a positive definite matrix. Thus, the derivative of the Lyapunov function will
be negative definite, and this will ensure stability. From (4.4) and (4.8), we have
σ T ( Dσ + σ& ) = 0 (4.9)
( Dσ + σ& ) = 0 (4.10)
54
σ& = ϕ& − ξ&
(4.11)
= ϕ& − ( G1 f1 + G2 f 2 + G2 Bu )
where
Then
σ& = ϕ& − ( G1 f1 + G2 f 2 + G2 Bu )
(4.14)
= − Dσ
then
u = ( G2 B )−1 ( ϕ& − G1 f1 − G2 f 2 ) + ( G2 B ) −1 Dσ
(4.15)
= ueq + ( G2 B ) −1 Dσ
Replacing the first term on the right by ueq , the above equation becomes
That is
55
Which indicates that the equivalent control is the sum of current control value and
( G2 B )−1σ& . The current value of control input is not available to use. An approximation
where t − = t − ∆, ∆ → +0
u( t ) = u( t − ) + ( G2 B ) −1 ( σ& + Dσ ) |t =t − (4.20)
The term ( G2 B )−1 ( σ& + Dσ ) |t =t − is used in updating a recursive formula for the control
input. On the sliding manifold, u( t − ) becomes the same as the equivalent control.
push σ toward zero, so that (4.10) holds and stability is reached [49].
dθ
= ωr (4.21)
dt
dωr 1
= ( Te − Tl − B f ωr ) (4.22)
dt J
If the induction motor is in field oriented control, for a fixed rotor flux, the motor torque
can be written as
2 PLm
Te = λdr iqs = k t iqs (4.23)
3 Lr
56
where
2 PLm
kt = λdr
3 Lr
dωr 1 k
= ( −Tl − B f ωr ) − t iqs (4.24)
dt J J
Let θ = x1 ,ωr = x 2 , the regular form for motor control can be rearranged as
dx1
= x2 (4.25)
dt
dx2 1 1
= − B f x2 − Tl + Ku (4.26)
dt J J
kt
where K=
J
simulated. Figures 4.1 through 4.3 show the simulation results. The simulation
results of PI controller and the proposed sliding mode controller at induction motor
start-up are compared in Figure 4.1. In the figure, the curves PI-1, PI-2, PI-3
much better transient performance, and even more important, it overcomes the
performance degradation with speed and avoids tedious tuning process. The
where a 0.5 pu step load is applied to the machine between 0.6 and 1 second and
then released. The sliding mode controller rejects this disturbance very well.
57
Figure 4.3 shows the speed tracking simulation results for four-quadrant
PI-1
PI-3
58
Figure 4.3 Speed tracking simulation with a triangle speed command
The proposed chattering free sliding mode controller has been implemented on
the prototype 1 H P induction machine. Figure 4.4 and 4.5 show the motor transient
response to step load changes at 1.0 Nm and 2.0 Nm respectively, where ω r is
current, ia is phase current. The results show that the controller has good dynamic
performance and speed rejection to load change. To further demonstrate the speed
robustness, the test data have been collected through the data acquisition system as
shown in Appendix A.1. Figures 4.6 - 4.8 show the motor speed response and
external step torque applied to the motor at Tl =0.5Nm, 1.0Nm and 2.0Nm
respectively. It can be seen from the results that even at high load torque ( Tl =2.0
Nm) step change, the motor speed change is within 10 rpm, about 2%, which shows
59
the effectiveness of the proposed controller. The speed tracking is also tested for
four-quadrant operation. Figure 4.9 and 4.10 show the different experimental
results (300rpm to 1400rpm and ± 1200rpm) for the operation with trapezoidal
ωr
ω̂ r 909rpm/div
id
iq
ia 5A/div
500ms/div
Figure 4.4 Speed response at Tl =1.0 Nm (curve1: real speed ωr ; curve 2: estimated speed
ω̂r ; curve 3: flux current id ; curve 4: torque current i q ; curve 5: phase current ia )
60
ωr 909rpm/div
ω̂ r
id
iq
ia 5A/div
500ms/div
Figure 4.5 Speed response at Tl =2.0 Nm (curve1: real speed ωr ; curve 2: estimated speed
ω̂r ; curve 3: flux current id ; curve 4: torque current i q ; curve 5: phase current ia )
700 1
Speed
Torque 0.9
600
0.8
500
0.7
0.6
Torque (Nm)
Speed (rpm)
400
0.5
300
0.4
0.3
200
0.2
100
0.1
0 0
1 2 3 4 5 6 7 8 9 10
Time (s)
Figure 4.6 Speed response with external torque step change Tl =0.5Nm
61
Speed response under step torque disturbance (1.0 Nm)
700 2
Speed
Torque
600
1.5
500
Torque (Nm)
Speed (rpm) 400
300
200
0.5
100
0 0
1 3 5 7 9 11
Time( s )
Figure 4.7 Speed response with external torque step change Tl =1.0Nm
700 4
Speed
Torque
3.5
600
3
500
2.5
Torque (Nm)
Speed (rpm)
400
300
1.5
200
1
100
0.5
0 0
0 2 4 6 8 10 12 14 16
Time (s)
Figure 4.8 Speed response with external torque step change Tl =2.0 Nm
62
1800 2
1.5
1500
1200
0.5
Torque (Nm)
Speed (rpm)
900 0
-0.5
600
-1
300
-1.5
0 -2
0 2 4 6 8 10 12 14
Time (s)
Figure 4.9 Four quadrant operation with trapezoidal speed command (between
1500 5
4
1000
3
2
500
Torque (Nm)
1
Speed (rpm)
0 0
0 2 4 6 8 10 -1
-500
-2
-3
-1000
-4
-1500 -5
Time (s)
Figure 4.10 Four quadrant operation with trapezoidal speed command (between
±1200rpm)
63
4.7. Conclusion
A robust and chattering free continues sliding mode controller is presented and
implemented in this Chapter. The controller has been tested for various command
speeds through the simulation and experimental results. These results prove that the
proposed continues sliding mode controller is robust and have good rejection to external
disturbance. The simulation and experimental results show that its dynamic performance
as well as steady state performance is much better than conventional PI controller and
free of re-tuning.
64
CHAPTER 5
5. ADAPTIVE
ADAPTIVESLIDING
SLIDINGMODE
MODEROTOR
ROTORFLUX
FLUXAND
ANDSPEED
SPEED
OBSERVERS
OBSERVERS
5.1. Introduction
In the sensorless speed control of induction motors with direct field orientation, the rotor
flux and speed information are dependent on the observers. However, the exact values of the
parameters that construct the observers are difficult to measure and changeable with respect to
the operating conditions. When the motor parameters are changed and thus different from the
preset values, the estimated flux and speed will deviate from the real values. To make flux and
speed estimation robust to parameter variations, an adaptive sliding mode flux and speed
observer is proposed in the Chapter. Two sliding mode current observers are used in the
proposed method. The effects of parameter deviations on the rotor flux observer can be
alleviated by the interaction of these two current sliding mode observers. The stability of
the method is proven by Lyapunov theory. An adaptive speed estimation is also derived
As defined in Chapter 3, the induction motor model can be expressed in the stationary
frame as:
65
pI = βAΛ − k 1 I + k 2 V ( 5.1)
pΛ = − AΛ + k 3 I ( 5.2)
where
The configuration of the proposed flux and speed estimators is shown in Figure 5.1.
The adaptive sliding mode observer consists of two sliding mode current observers and one
rotor flux observer. The rotor flux observer is based on the current estimation from the
two current observers. The rotor speed observer takes the outputs from the second current
observer and the rotor flux observer as its inputs and generates the estimated rotor speed
as the output. The estimated speed is then fed back to the second current observer for its
adaptation. The estimation of the motor speed is derived from a Lyapunov function,
which guarantees the system convergence and stability. Once the sliding functions of the
current observers reach the sliding surfaces, the rotor flux will converge to the real value
66
~ ~
Current sliding mode observer I U eq1
λαr , λ βr
pIˆ = βU1 − k1Iˆ + k 2 V Rotor Flux observer
~ ~
pΛ = −(U eq1 − L U eq 2 β ) + k3 I
pe I 1 = β (U 1 − AΛ ) − k1e I 1 ( 5.5)
where e I 1 = Iˆ − I
By selecting u 01 large enough, the sliding mode will occur ( s1n = 0 ), and then it
follows that
67
pe I 1 = e I 1 = 0 (5.6)
From the equivalent control concept [67], if the current trajectories reach the sliding
manifold, we have
U eq1 = AΛ (5.7)
Equation (5.7) indicates that the equivalent control equals to the rotor flux multiplied
by the A matrix, which is the common part in (5.2). The rotor flux can be obtained by
The flux estimation is accurate when the motor parameters k1 , k 2 , and β are known.
However, if the parameters in observers are different from the real values, there will be
some errors ∆ k 1 , ∆ k 2 , ∆ β in the coefficients of the observers. Then the estimated flux
and speed will be incorrect. In order to compensate this divergence, a second sliding
The second sliding mode current observer is designed differently from (5.3) as
~ ~ ~
p I = βAΛ − k1 I + k 2 V + U 2 (5.8)
~ ~ ~
Λ = [λαs , λβs ]T , the observed rotor fluxes
~
U α 2 = −u02 sign( sα 2 ) sα 2 = iαs − iαs
, ~
U β 2 = −u02 sign( sβ 2 ) s β 2 = iβs − iβs
68
sn 2 = [sα 2 , s β 2 ] (5.9)
~
where e I 2 = I − I
~
where e Λ = Λ − Λ . The second equivalent control equals to the negative multiplication of
the estimated rotor flux error and the A matrix. It is noticed that the second current
Combining the results from (5.7) and (5.11), the rotor flux observer can be
constructed as
~ ~
pΛ = −( U eq1 − L U eq 2 β ) + k 3 I (5.12)
where L is the observer gain matrix to be decided such that the observer is asymptotically
stable.
From (5.3) and (5.8), the equivalent controls obtained individually by the two current
observers will deviate from their real values if the motor parameters are incorrect.
Consequently the rotor flux estimation based on each individual control will also be
inaccurate. To reduce this deviation on rotor flux estimation, the rotor flux observer is
designed from the combination of two equivalent controls, where the effects of parameter
variations are largely cancelled. From (5.7) and (5.11), the error equation for the rotor
flux is
69
pe Λ = −( U eq1 − L U eq 2 β ) + AΛ + k 3e I 2
(5.13)
= − LAe Λ + k 3e I 2 = − LAe Λ
consider the rotor speed as a variable parameter, the error equation of flux observer is
0 ∆ω ~ −ω
where ∆A = , ∆ω = ω
0
r r
− ∆ω
V = e Λ e Λ + ∆ω 2 / λ
T
(5.15)
where λ is a positive constant. We know that V is positive definite. The time derivative of
V becomes
T T ~
pV = −e Λ ( LA + A T LT )e Λ − e Λ L∆ AΛ
~ d
− Λ T ∆ALT e Λ + 2∆ωr ω~r / λ
dt
T
U eq2 ( A −1 ) T (5.16)
T ~
= −e Λ ( LA + A L )e Λ −
T T
L∆ AΛ
β
−1
~ ( A )U eq2 d
− Λ T ∆ALT + 2∆ωr ω~r / λ
β dt
above becomes:
70
T
pV = −e Λ (LA + A T LT )e Λ
~ ~ (5.17)
γU eq2 T ∆AΛ − γΛ T ∆AU eq2 d
− + 2∆ω r ω~ r / λ
β dt
Let the second term equal to the third term in (5.17), we can find the following adaptive
d ~ γλ ~ ~
ωr = (U β 2λαr − U α 2 λβr ) (5.18)
dt β
where U eq2 = [U α 2 , U β 2 ]T
In practice, the speed can be found by the following proportional and integral adaptive
scheme:
( ~ ~
) ( ~ ~
ω~r = K P U β 2 λαr − U α 2 λβr + K I ∫ U β 2 λαr − U α 2 λβr ) (5.19)
Since the second term equal to the third term in (5.17), the time derivative of V
becomes
T
pV = −e Λ ( LA + A T LT )e Λ
1 2
2γ 2 + ω r 0 ( 5.20)
T Tr
= −e Λ 2 eΛ < 0
1 2
0 2γ 2 + ω r
Tr
It is apparent that (5.20) is negative definite. From Lyapunov stability theory, the flux
observer is asymptotically stable, guaranteeing the observed flux to converge to the real
rotor flux.
71
5.5. Simulation results
To evaluate the proposed algorithm for the rotor flux and speed estimation, computer
feasibility, the estimation and control algorithm are evaluated by HIL (hardware-in-the-
loop) testing. A 1 HP induction motor was used in the simulation and also in the
experiments.
Figures 5.2 and 5.3 show the induction motor response to a step speed command of
±0.5pu (±900rpm) where the motor parameters are exactly known. The actual machine
model is used to calculate the current, flux and speed of the motor. The observer model as
described above is used to estimate the rotor flux and speed. Figure 5.2 shows the speed
command, real speed, estimated speed and the speed estimation error. Figure 5.3 shows
the real and estimated rotor flux and the flux estimation error. It can be seen that the
estimated speed and flux converge to the real values very quickly.
72
Figure 5.2 Real and estimated speed at a step speed command.
73
Figure 5.4 Real and estimated rotor flux.
To study the effects of parameter variation on the speed and flux observers, the
parameters in the observers are changed on purpose in the simulation. Figure 5.5 shows
the simulation results when the coefficient k 1 in the observers is changed by 20% from
its actual value, where the flux λαr _ obs and speed ω robs are estimated by the proposed
method, and λαr _ obs1 and ω robs1 are estimated by the previous method using only one
current sliding mode observer as in Chapter 3. It is noticed that even k 1 is incorrect, the
estimated rotor flux and speed by the new observer still converge to the real values, but in
previous model, there is an offset in the rotor flux estimation and fluctuation in the rotor
speed estimation. The dc offset of flux estimation by previous method is caused by the
incorrect equivalent control U eq1 . If k 1 changes, the equivalent control U eq1 will
74
detune. The integration of incorrect U eq1 causes dc offset on the flux estimation.
Whereas in the new flux observer, this dc offset is cancelled by using two current
observers. The effects of coefficient β variation on the flux and speed estimation are
shown in Figure 5.6. We can also observe obvious fluctuations in speed estimation. There
is still an error on the rotor flux estimation by the proposed method as shown in Figure
5.6(a), but the new method eliminates the dc offset caused by the parameter variation,
75
(a) λαr : real rotor flux, λαr _ obs : estimated by the proposed method, λαr _ obs 1 : estimated by
previous method.
(b) ω r : real rotor speed, ω robs : estimated by the proposed method, ω robs1 : estimated by
previous method.
Figure 5.5 Coefficient k1 in the observer is increased by 20% .
76
(a) λαr : real rotor flux, λαr _ obs : estimated by the proposed method, λαr _ obs1 : estimated by
previous method.
(b) ω r : real rotor flux, ω robs : estimated by the proposed method, ω robs1 : estimated by
previous method.
Figure 5.6 Coefficient β in the observer is increased by 20%.
77
5.5.2. HIL Evaluation results by TI 2812 DSP
The results evaluated by HIL are shown in Figures 5.7 through 5.9. Figure 5.7 shows
the motor step response to a speed command at ±0.5pu (±900rpm). Figure 5.8 shows the
real and estimated rotor flux and the estimated flux angle. Figure 5.9 shows the motor
response to a trapezoidal speed command. The results show that the method can be
Figure 5.7 Speed step response from –0.5pu to 0.5pu (curve 1: speed command ω r * ; curve
2: real speed ω r ; curve 3: estimated speed ω~r )
78
λαr
~
λαr
~
θr
100ms/div
~
Figure 5.8 Rotor flux estimation. (curve 1: real flux λαr ; curve 2: estimated flux λαr ;
~
curve 3: estimated flux angle θ r )
0.6pu/div
1.2126pu/div
Figure 5.9 Trapezoidal speed at ±0.5pu. (curve 1: phase current i a ; curve 2: torque
current i q ; curve 3: estimated speed ω~r )
79
5.6. Experimental results
induction motor drive system was set up. The setup consists of a 1 HP induction motor, a
power drive board and a DSP controller board. The external load is imposed by a
hysteresis dynamometer. The experimental setup is shown in Appendix A.1. The control
The test was first performed on the motor in four-quadrant operations. Figure 5.10
shows the motor response to a commanded step change speed at ±900rpm. Figure 5.11
shows the measured current and two sliding mode observer currents. It is seen that the
sliding mode functions enforce the two observed currents to the measured ones very
closely. Once these two observer currents converge to the measured ones, the estimated
rotor flux converges to the real rotor flux. The motor responses to a trapezoidal speed
command when the motor runs at no load are shown in Figure 5.12. To further
T = 0.5 pu is applied when the motor runs at the same trapezoidal speed command as in
l
Figure 5.12. The waveform of speed command ω r * , estimated speed ω~r , torque current
iq , and phase current ia are shown in Figure 5.13. The estimated rotor speed response to a
80
step change of command from 360rpm to 1260rpm with a load torque of T = 0.5 pu is
l
shown in Figure 5.14. To investigate the speed robustness, a step disturbance torque
( T = 0.5 pu) is applied and then removed at motor speed n=900rpm. Figure 5.15 shows the
l
estimated rotor speed response and the torque current response. As evidenced by the
testing results, the induction motor drive functions very well by the proposed algorithm.
1091rpm/div
5A/div
Figure 5.10 Transient response to speed step command ±900rpm at no load (curve1:
speed command ω r * ; curve 2: estimated speed ω~r ; curve 3: torque current i q ; curve 4:
phase current ia )
81
iαs 3A/div
iˆα s 3A/div
~ 3A/div
iα s
50ms/div
Figure 5.11 Real and estimated currents. (curve 1: measured current iαs ; curve 2:
~
observed current iˆαs ; curve 3: observed current iαs )
2182rpm/div
2182rpm/div
5A/div
82
1091rpm/div
5A/div
1091rpm/div
1091rpm/div
5A/div
Figure 5.14 Speed response due to step change command from 360rpm to 1260rpm at
~
T = 0.5 pu . (curve1: real speed; curve 2: estimated speed ω r ; curve 3: torque current i q ;
l
83
1091rpm/div
1091rpm/div
1.5A/div
Figure 5.15 Transient response for step disturbance torque (curve1: real speed ω r ;
curve 2: estimated speed ω~r ; curve 3: torque current i q )
5.7. Conclusion
An adaptive sliding mode observer for sensorless speed control of induction motor is
presented in this Chapter. The proposed algorithm consists of two current observers and
one rotor flux observer. The two sliding mode current observers are utilized to
compensate the effects of parameter variations on the rotor flux estimation. When the
motor parameters are deviated from initial value by temperature or operation conditions,
the errors of two equivalent controls from current observers will be largely cancelled,
which make the flux estimation more accurate and insensitive to parameter variations.
Although additional sliding mode current observer is used, the complexity of the method
is not increased too much. The stability and convergence of the estimated flux to real
rotor flux are proved by the Lyapunov stability theory. Digital simulation and
84
experiments have been performed in the paper. The effectiveness of the approach is
85
CHAPTER 6
6. EFFICIENCY
EFFICIENCYOPTIMIZATION
OPTIMIZATIONON
ONVECTOR
VECTORCONTROL
CONTROLOF
OF
INDUCTIONMOTORS
INDUCTION MOTORS
6.1. Introduction
Electrical machines consume most of the world’s electrical energy every year.
Improving efficiency of electrical drives is important not only for energy saving, but also
for environmental protection. In an induction motor drive system, to get a best transient
response the induction motor normally operates at rated flux. However, when the drive
system operates with light loads, the core losses become excessive, causing drive system
poor efficiency. To improve the motor efficiency, the flux must be decreased, obtaining a
banlance between the copper and iron losses. This phenomenon can be illustrated in
Figure 6.1. T1 and T2 are motor torque-speed curves at different frequencies f1, f2, where
T1 is at rated frequency and voltage, T2 is at reduced flux level because the applied
frequency is increased (f2>f1) and voltage is reduced. η1 and η 2 are efficiency curves
corresponding to T1 and T2 respectively. The operating point ‘a’ in the Figure 6.1 can be
achieved either by curve T1 or T2, but the efficiencies are quit different. It is seen that the
efficiency (point ‘c’) corresponding to curve T2, which nearly reaches maximum point, is
much higher than the efficiency (point ‘b’) at which the motor operates at rated flux level.
86
In the vector control drive system, the flux optimum strategy is to find the maximum
A number of methods for efficiency improvement through flux control have been
proposed in the literature. They can be classified into two main categories. The first
category is called Loss Model Controller (LMC) [73-75]. This method is based on the
loss model of the induction motor. The flux level is selected according to the computation
of minimum motor losses. The second method is Search Controller (SC) [29-33]. This
method searches the maximum motor efficiency by measuring the input power or DC bus
power.
The LMC method has the advantage that it is simple and fast. However, the accuracy
depends on correct modeling of the motor drive and the losses. On the other hand, SC
method measures input power or DC bus power to searches the flux where the motor runs
An efficiency optimization method by minimizing the stator current which does not
stator current method has been used in induction motor scalar control by I. Kioskeridis
[75]. In this Chapter, the relationship between the stator current minimization and the
pointed that minimum stator current point is very close to minimum loss point in most
cases and the loss minimization can be achieved by minimizing stator current in practice.
A fuzzy logic based search method is simulated and implemented. Simulation and
experimental results are given in the paper to verify the proposed method.
The fuzzy logic control is based on fuzzy logic or fuzzy inference system that is able
mapping of a given input data set into an output data set. A block diagram of a fuzzy
control system is shown in Figure 6.2. The fuzzy controller is composed of the
following four elements: fuzzifier, rule base, inference engine, and defuzzifier.
88
Fuzzy controller
Crisp output
Inference u(t) Output
Defuzzifier
Crisp input r(t) y(t)
Fuzzifier
engine
Process
Rule Base
6.2.1. Fuzzifier
A fuzzifier maps crisp numbers into fuzzy sets and converts controller inputs into
information that the inference engine can easily use to activate and apply rules. It is
needed in order to activate rules which are in terms of linguistic variables. The
fuzzifier includes two parts: choice of membership function and choice of scaling
factor.
A fuzzy variable has values that are expressed by the natural language. This
example, the stator current of a motor can be defined by the qualifying linguistic
defines how the values of a fuzzy variable in a certain region are mapped to a membership
value between 0 and 1. The fuzzy sets can have more subdivisions such as Zero, Very small,
Medium small, Medium large, Very large for a more precise description of the fuzzy
variable.
89
For example, in Figure 6.3, if the current is below 20A, it belongs completely to the set
Small, that is, the MF value is 1; whereas for 35A, it is in the set Small by 25% (MF = 0.25)
and to the set Medium by 50% (MF = 0.5). At current 40A, it belongs completely to the set
Medium (MF = 1) and not in the set Small and Big (MF = 0). If the current is above 70A,
it belongs completely to the set Big (MF = 1), where MF = 0 for Small and Medium.
trapezoidal, or Gaussian membership function which are shown in Figure 6.4 (a),
(b), and (c). The simplest and most commonly used membership function is the
0 10 20 30 40 50 60 70
Current (A)
90
(a) Triangular (b) Trapezoidal
(c) Gaussian
A rule base (a set of If-Then rules), which contains a fuzzy logic quantification
of the expert’s linguistic description of how to achieve good control. Once the rules
have been established, a fuzzy logic system can be viewed as a mapping from inputs
to outputs.
Rules may be provided by experts or can be extracted from numerical data. The
As it can be seen, the premises (which are sometimes called “antecedents”) are
associated with the fuzzy controller inputs and are on the left-hand-side of the rules.
The consequents (sometimes called “actions”) are associated with the fuzzy
controller outputs and are on the right-hand-side of the rules. Notice that each
91
premise (or consequent) can be composed of the conjunction of several terms, e.g., ‘IF t l is
very cold AND v 1 is quite low, THEN voltage u must be very big.’ This one rule
1) Linguistic variables versus numerical values of a variable (e.g., very cold versus
-5°C);
terms associated with it, ranging from extremely hot to extremely cold), which
Using the above approach, we could write down rules for all possible cases. In practical,
since only a finite number of linguistic variables and linguistic values are specified, the
Inference engine, also called fuzzy inference, which emulates the expert’s
decision making in interpreting and applying knowledge about how best to control
the plant. It handles the way in which rules are combined. Just as humans use many
there are many different fuzzy logic inferential procedures. Only a very small number of
them are actually being used in engineering applications of fuzzy logic system.
6.2.4. Defuzzifier
The defuzzifier maps output sets into crisp numbers. It converts the conclusions
of the inference mechanism into actual inputs for the process. The most common
92
used defuzzification method is Center of Gravity (COG). COG defuzzification
the COG method, the crisp output of is obtained by using the center of gravity, in
which the crisp or variable is taken to be the geometric center of the output fuzzy
δ crisp =
∑ b ∫ µ( i )
i i
(6.1)
∑ ∫ µ( i )
i
where bi is the center of membership function, ∫µ i denote the area under membership
function.
In order to design a fuzzy logic based algorithm, the following steps need to be
performed.
logic application. Get all the information from the operator of the plant to be
control led.
necessary.
3) Definition of membership functions. Formulate the fuzzy sets and select the
are needed. If a variable requires more precision near steady state, use more
93
4) Selection of inference method. Formulate the rule table.
6) If the mathematical plant model is available, simulate the system with the fuzzy
controller. Iterate the fuzzy sets and rule table until the performance is
optimized.
7) Implement the control in real time and further iterate to improve performance.
The d-q model for a three-phase squirrel cage induction motor in the synchronous
v ds = R s i ds + p λ ds − ω e λ qs
v qs = R s i qs + p λ qs + ω e λ ds
(6.2)
0 = R r i dr + p λ dr + (ω e − ω r ) λ qr
0 = R r i qr + p λ qr − (ω e − ω r ) λ dr
The total losses of an induction motor consist of stator and rotor copper losses Pcu, core
losses Pfe and mechanical losses Pm. In the steady state, the stator and rotor copper losses
are given by
3 2 2 3 2 2
Pcu = R s (i ds + i qs ) + R r (i dr + i qr ) (6.3)
2 2
The core losses include eddy current losses and hysteresis losses.
3
( k h ω e λ m + k eω e λ m )
2 2 2
Pfe = (6.4)
2
where k h , k e are the hysteresis and eddy current loss coefficient. As a reasonable
When the motor is running under the rotor flux field orientation, we have relationship
idr = 0 (6.6)
and
Lm
i qr = − i qs ( 6.7)
Lr
in steady state. From (6.1) to (6.4), we can get the expression for total motor losses:
3 2
( 2 2
Ploss = {R s i ds + i qs + R r i dr + i qr +
2
)2
( )
2 2 (6.8)
Lσr Lm
(k h ω e + k e ω e )[ Lm i ds + i qs ]}+ k m ω r
2 2 2 2 2
2
Lr
2
3 p Lm
where K = .
2 2 Lr
where
A = R s + ( k h ω e + k eω e ) L m ,
2 2
2 2 2
Lm Lσr Lm
+ ( k h ω e + k eω e )
2
B = R s + Rr 2 2
.
Lr Lr
95
Equation (6.10) gives an expression of induction motor losses in a vector control
system. It shows the relationship of motor losses with rotor flux ( ids ), motor torque and
speed. It is obvious that if speed and torque are constant, the loss just is a function of ids .
Figure 6.5 shows the curves of total losses with respect to ids at different load torque
(speed is 900rpm). It is seen from Figure 6.5 that if speed is fixed, then the d-axis current
ids corresponding to minimum loss point is different at different load torque. So we can
find the ids value that corresponds to minimum loss at each operating point.
Figure 6.5 Motor losses with respect to ids at different load torque
96
6.4. Comparison of minimum losses point and minimum stator current
point
As state above, assuming the motor parameters are constant and independent on rotor
flux, it can be seen from the loss expression (6.10) that at each operating point ( ω r and
Te fixed), the motor loss is just a function of d-axis current component ids . In general,
minimum loss corresponds to minimum input power if output power is constant. Figure
6.6 shows the curves of input power with respect to ids (output power is fixed). This plot
*
gives us the appropriate ids corresponding to the point of the optimum efficiency. To
investigate the relationship between stator current minimization and motor loss
minimization, the stator current variation with respect to ids is plotted in Figure 6.7. It is
interesting to note that the minimum stator current points are very close to minimum loss
(or input power) points at each load condition. To illustrate this more explicitly, the
minimum input power points and the input power points corresponding to minimum
stator currents are plotted in same figure as shown in Figure 6.8. This phenomenon is also
demonstrated by experimental results. Figures 6.9 and 6.10 show the experimental input
power and stator current with motor flux level (v/f value).
97
Figure 6.6 Input power with respect to ids
98
Figure 6.8 Minimum input power point and input power point corresponding to minimum
stator current
99
Figure 6.10 Measured stator current vs flux level at different torques
To further investigate the relationship between minimum losses and minimum stator
expression (6.10) with respect to ids , the magnetizing current i * ds _ p corresponding to the
B Te
i * ds _ p = 4 (6.11)
A K
2
2 2 2 T 1
i s = i ds + i qs = i ds + e2 2 (6.12)
K i ds
By setting the derivative of (6.12) with respect to ids to zero, it yields that
Te
i * ds _ i = (6.13)
K
100
i * ds _ i 4 A
k ip = = (6.14)
i * ds _ p B
The plot of the ratio k ip with respect to frequency for different motors is shown in
Figure 6.11. It can be seen that the ratio k ip is dependent on motor parameters and
operating frequency. The value of k ip increases as the frequency increases. But the value
is still in the range from 0.8 to 1.4. As illustrated in Figures 6.6 and 6.9, the motor input
power (or loss) curves are quit flat around the minimum loss points. This means that even
the minimum stator current points are somewhat away from the minimum loss points, the
motor losses by finding minimum stator current are still very close to minimum losses.
Therefore, the minimum stator current can be used to minimize the motor losses in
practice.
50hp
10hp
1hp
101
6.5. Fuzzy controller for efficiency optimization
Several fuzzy logic based loss minimum algorithms have been reported previously
[78,79]. The advantage of minimizing stator current to minimize motor losses is that no
explained in Figure 6.12. The program searches the minimum stator current by adjusting
the magnetizing current. If the magnetizing current is decreased, then the rotor flux is
reduced, causing a corresponding increase in the torque current to keep the developed
torque constant. As the rotor flux is decreased, the iron loss and copper loss decreases at the
same time, resulting in a decrease of stator current. After the flux level reaches to some
level, the iron loss will continue reduce while the copper loss will increase. However, the
total system loss and stator current will still decrease. Since the minimum stator current
point is very close to minimum losses point, the search continues until the system settles
down at the minimum stator current point. Any excursion beyond the minimum point will
102
n
T
Based on the principle above, the fuzzy controller is designed as in Figure 6.13. The
fuzzy controller uses stator peak current change ∆is and d-axis current change ∆ids as its
103
The membership functions for fuzzy controller are shown in Figure 6.14. All
membership functions are triangular for simplicity. The fuzzy output is calculated using
The rule base for fuzzy control is given in Table 6.1. The basic idea is that if the last
control action indicated a decrease of stator current, the search proceeds in the same
direction. In case the last control action resulted in an increase of stator current, the
search direction is reversed. For example, IF ∆is = NS AND ∆ids = N , THEN ∆ids _ ref = NS .
This rule means that IF the stator current increment ∆is is negative small (NS) and the
last d-axis current ∆ids is negative (N), THEN the new excitation current increment
104
NB NS ZE PS PB
-1 -0.5 0 0.5 1
∆ is
N P
-1 0 1 ∆ i ds
NB NS ZE PS PB
105
Table 1 Rule base
∆ids N P
PB PB NB
PS PS NS
ZE ZE ZE
NS NS PS
NB NB PB
P = Positive
N = Negative
PB = Positive Big
PS = Positive Small
ZE = Zero
NS = Negative Small
NB = Negative Big
106
The disadvantage of this efficiency control is that the transient response is relatively
slow. To overcome this, for any change in load or speed command, the fuzzy efficiency
controller is turned off and the controller’s attention is directed to the system
The system is first simulated by MATLAB. The simulation results are shown in
Figures 6.15 through 6.17. Figure 6.15 shows the search process of d-axis current
command change and q-axis current variation at a load toque of Tl = 0.1 pu. Figures 6.16
and 6.17 show that the stator current and motor losses decrease until the controller
reaches steady state (almost minimum point). The core losses and copper losses are also
shown in the process. The results show that the motor loss is greatly reduced by the
proposed method.
Fuzzy begins
ids _ ref
iqs
107
Fuzzy begins
is
Fuzzy begins
Ptotal losses
Pfe
Pcu
induction motor drive system was set up. The external load is imposed by a hysteresis
fixed-point DSP.
108
The test was implemented at different motor speeds and load torques. Figures 6.18
through 6.26 show the experimental results. The fuzzy logic search processes at 900rpm
with load torque Tl = 0.2 Nm and Tl = 0.5 Nm are shown in Figures 6.18 and 6.19
respectively, where is is stator current magnitude, ids _ ref d-is axis current command,
iqs q-axis torque current, and ia phase current. Figure 6.20 and 6.21 show the test
results at 600rpm and 1200rpm respectively. It can be seen from the results that at light
load, the stator current is greatly reduced, which will cause motor efficiency increase.
Figures 22 - 25 show the input power and motor efficiency variation with stator current
reduction during fuzzy search. The results are shown at Tl = 0.2 Nm, n = 900rpm ,
Tl = 0.5Nm, n = 900rpm , Tl = 0.2 Nm, n = 1200rpm , and Tl = 0.2 Nm, n = 600rpm respectively.
After fuzzy logic is switched on, the stator current is deceased from 1.25A to 0.65A in
Figure 6.22, while the input power is deceased from 62W to 39W and efficiency is
increased from 30% to 52%. In Figure 6.23, when the load is increased, the input power
is reduced from 89W to 78W. At the same time the efficiency is increased from 49% to
61%. The same results are expected in Figures 6.24 and 6.25 for motor speed at 1200rpm
and 600rpm. The experimental results show the motor efficiency is greatly improved at
light load. The comparison of efficiency curves at different operation points with and
109
Fuzzy begins
is
ids _ ref
iqs
ia 2A/div
500ms/div
Fuzzy begins
is
ids _ ref
iqs
ia 2A/div
500ms/div
110
is
Fuzzy begins
ids _ ref
iqs
ia 2A/div
500ms/div
Fuzzy begins
is
ids _ ref
iqs
ia 2A/div
500ms/div
111
1.6
0.8
0.4
0
6 7 8 9 10 11 12
Time (s)
120
Power (W)
80
40
0
6 7 8 9 10 11 12
Time (s)
0.8
Efficiency
0.6
0.4
0.2
0
6 7 8 9 10 11 12
Time (s)
(c) Efficiency
Figure 6.22 Current, power and efficiency variation at Tl = 0.2 Nm, n = 900rpm
112
1.6
0.8
0.4
0
4 6 8 10 12 14
Time (s)
160
Input power (W)
120
80
40
0
4 6 8 10 12 14
Time (s)
0.8
Efficiency
0.6
0.4
0.2
0
4 6 8 10 12 14
Time (s)
(c) Efficiency
113
1.6
1.2
0.4
0
0 2 4 6 8 10 12 14
Time (s)
80
40
0
0 2 4 6 8 10 12 14
Time (s)
0.8
Efficiency
0.6
0.4
0.2
0
0 2 4 6 8 10 12 14
Time (s)
(c) Efficiency
Figure 6.24 Current, power and efficiency variation at Tl = 0.2Nm, n = 1200rpm
114
1.6
1.2
0.4
0
0 5 10 15 20
Time (s)
60
Input power (W)
40
20
0
0 5 10 15 20
Time (s)
0.4
Efficiency
0.2
0
0 5 10 15 20
Time (s)
(c) Efficiency
115
0.8
1200rpm
900rpm
0.7
0.6
0.5
600rpm
Efficiency
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5
Torque (Nm)
Figure 6.26 Comparison of efficiency curves at different operation points with and
without fuzzy optimization control. (solid line is with fuzzy optimization, dash line is
without fuzzy optimization)
6.8. Conclusion
An efficiency optimization method which does not require extra hardware and
minimization and motor losses minimization in the induction motor vector control system
pointed that the motor loss minimization can be achieved by minimizing stator current in
practice. The simulation and experimental results demonstrate the effectiveness of the
approach.
116
CHAPTER 7
7. SUMMARY
SUMMARY AND
AND FUTURE
FUTURE WORK
WORK
7.1. Summary
applications. Open loop control systems maintain the stator v/f ratio at a predetermined
level to establish the desired machine flux. The ratio is satisfied only at low or moderate
dynamic requirements. Field orientation technology can provide high performance control
selected flux and allowing the induction motor to emulate a separately excited dc machine.
The speed sensorless control and loss minimization of induction drive have gained more
and more attention because the fragile speed sensor and energy crisis, which are also the
There are two key issues related to a direct field oriented drive system: flux estimation
and speed estimation. In this research, a flux and speed observer using the sliding
Chapter 5 and the stability is verified by Lyapunov theory. A continuous sliding mode
approach.
of sliding mode method. It shows that the deviation of motor parameters will
cause error in the equivalent control of sliding mode, and the system
parameter variation on the rotor flux estimation, which make flux estimation
operating condition and tuning the gains is a very time consuming job. To
achieve better performance and avoid tuning, a continuous sliding mode speed
118
• An efficiency optimization method is presented which has the advantage of no
logic based search method is simulated and implemented using TI 2812 DSP.
The simulation and experimental results show that this approach has greatly
Simulations and experimental results for the flux estimation, speed estimation
and loss minimization show the great promise of the methods proposed in this
dissertation. However, the robustness of the proposed adaptive sliding mode observer
will be further investigated in the future work. To reduce the search time and torque
pulsations, the fuzzy logic search method will be improved in practical implementation.
119
BIBLIOGRAPHY
8. BIBLIOGRAPHY
[1] J. Moreira, T. Lipo and V. Blasko, "Simple efficiency maximizer for an adjustable
frequency induction motor drive," IEEE Transactions on Industry Applications,
Vol 27, No. 5, pg. 940-945, Sep/Oct 1991.
[2] T.C. Huang, M.A. El-Sharkawi, “Induction motor efficiency maximizer using
multi-layer fuzzy control,” Proceeding of International Conference on Intelligent
Systems Applications to Power Systems 28, pp.109-113, 1996.
[6] I. Boldea and S. A. Nassar, Vector Control of AC Drives, CRC Press, 1992.
[7] D. W. Novotny and T.A. Lipo, Vector Control and Dynamics of AC Drives,
Oxford University Press Inc., New York, 1996.
[8] F. Blaschke, “The pricinple of field orientation as applied to the new transvector
close-loop controlsystem for rotating-filed machines,” Siemens Review, 1972,
pp217-223.
120
[12] X. Xu and D. W. Novotny, " Implementation of direct stator flux orientation
control on a versatile DSP based system", IEEE Transactions on Industry Ap-
plications, Vol. 27, No. 4, pp. 694-709, July./Aug. 1991.
[13] X. Xu, R. De Doncker D. W. Novotny, "A stator flux oriented induction machine
drive system", IEEE PESC Conference Record , Vol. 2, pp. 870-876, 1988.
[14] T. A. Lipo, "Flux sensing and control of static ac drives by the use of flux coils",
IEEE Transactions on Magnetics, Vol. MAG-13, pp. 1403-1408, Sept. 1977.
[15] A. B. Plunkett, "Direct flux and torque regulation in a PWM inverter induction
motor motor drive", IEEE Transactions on Industry Applications, Vol. IA-13, NO.
2, pp.139-146, 1977.
[18] A. Kawamura, R. Hoft, "An analysis of induction motor for field oriented or vector
control," IEEE Power Electronics Specialists Conference, pp. 91-100, 1983.
[20] S. Ogasawara, H. Akagi, A. Nabae, "The generalized theory of indirect vector control
for ac machine", IEEE Transactions on Industrial Electronics, Vol. 35, pp.
470-478, 1988.
[22] P. L. Jansen and R. D. Lorenz, "A physical insightful approach to the design and
accuracy assessment of flux observers for field oriented induction machine drives"
IEEE Transactions on Industry Applications, Vol. 30, No. 1, pp. 101110, Jan./Feb.
1994.
[23] P. L. Jansen and R. D. Lorenz, D. W. Novotny, " Observer-based direct field ori-
entation: Analysis and comparison of alternative methods" , IEEE Transactions on
Industry Applications, Vol. 30, No. 4, pp. 945-953, July./Aug. 1994
121
[24] Y. Ping, D. Vrancic, and R. Hanus, “ Anti-windup, bumpless, and conditioned
transfer techniques for PID controllers,” IEEE Trans. Contr. Syst. Technol., vol.
16, pp. 48-57, July 1996.
[27] G. El-Saady, A. M. Sharaf, et al., "A high performance induction motor drive
system using fuzzy logic controller", Electromechanical Conference Proceeding,
Vol. 3, pp. 1058-1061, 1994
[29] B. Heber, L. Xu and Y. Tang, " Fuzzy logic enhanced speed control of an
indirect field-oriented induction machine drive", IEEE Transactions on Power
Electronics, Vol. 12, No. 5, September 1997
[30] L. Zhen and L. Xu, "On-line fuzzy tuning of indirect field oriented induction
machine drives", in Proceedings of IEEE APEC, San Jose, California, March
1996.
[31] F. Cheng, and S. Yeh , "Application of fuzzy logic in the speed control of AC servo
systems and an intelligent inverter" IEEE Trans. on Energy Conversion, Vol. 8, No.
2, September 1993.
[32] J. Bu, L. Xu , "A high performance full fuzzy controller for induction machine
drives" IEEE APEC 98 Conference, March 1998.
[33] Y. Tang, L. Xu , " Fuzzy Logic application for intelligent control of a variable
speed drive.", Conference Record of IEEE PES winter Meeting, 1994.
[34] S. Mir, D. Zinger and M. Elbuluk, "Fuzzy controller for inverter fed induction ma-
chines" IEEE Trans. on Industry Applications, Vol. 30, No. 1, January-February
1994.
[35] E. Cerruto, A. Consoli, A. Raciti and A. Testa, "Adaptive fuzzy control of high
performance motion systems" Proc. of the International Conference on Motion
Control pp.88-94, 1992.
122
[36] M. Ta-cao and H. Le-huy, "Model reference adaptive fuzzy controller and fuzzy
estimator for high performance induction motor drives" IEEE Trans. IAS Annual
Meeting, pp. 380-387, 1996.
[37] L. Zhen, L. Xu, and Y. Tang , "Fuzzy learning enhanced speed control of an
indirect field oriented induction machine drive. " IEEE Trans. on Control System
Technology, Vol. 8, No. 2,pp. 270-278, March 2000.
[40] K. K. Shyu and H. J. Shieh, "A new switching surface sliding mode speed
control for induction motor drive systems" IEEE Trans. on Power
Electronics, Vol. 11, No. 4, pp. 660-667, 1996
[41] C. C. Chan and H. Q. Wang, "New scheme of sliding mode control for high
performance induction motor drives" IEE Proc. on Electronic Power
Applications, Vol. 143, pp. 177-185, 1996.
[44] J. J. Slotine and W. Li, Applied Nonlinear Control. Englewood Cliffs, NJ: Prentice-
Hall, 1991.
[45] F. Y. Hsu and L. C. Fu, “Nonlinear control of robot manipulators using adaptive fuzzy
sliding mode control,” Proc. IEEE/RSI Int. Conf. Intelligent Robots Systems, vol. 1,
pp. 156-161, 1995
[46] G. Bartolini, P. Pydynowski, “An Improved, Chattering Free, V.S.C. Scheme for
Uncertain Dynamical Systems,” IEEE Transactions on Automatic Control, Vol. 41,
No. 8, August 1996.
[53] M. Schroedl, “Sensorless control of AC machines at low speed and standstill based
on the “INFORM” method,” in Conf. Rec. IEEE-IAS Annu. Meeting, vol. 1, 1996,
pp. 270–277.
[56] C. Lascu, I. Boldea, and F. Blaabjerg, “A modified direct torque control for
induction motor sensorless drive”, IEEE Trans. Ind. Appl., vol. 36, no. 1, pp. 122-
130, January/February 2000.
[57] H. Tajima, Y. Hori, “Speed sensorless field oritation control of the induction
machine”, IEEE Trans. Ind. Appl., vol. 29, pp. 175-180, 1993.
124
[59] C. Schauder, “Adaptive speed identification for vector control of induction motors
without rotational transducers,” IEEE Trans. Ind Applicat., vol. 28, pp. 1054-
1061,1992.
[60] F.-Z. Peng and T. Fukao, “Robust speed identification for speed-sensorless vector
control of induction motors,” IEEE Trans. Ind. Applicat., vol. 30, pp. 1234–1240,
Sept./Oct. 1994.
[61] L. Zhen and L. Xu, “A mutual MRAS identification scheme for position sensorless
field orientation control of induction machines,” in Proc. IEEE-IAS Annu. Meeting
Conf., Orlando, FL, Oct. 1995, pp. 159–165.
[62] H. Kubota, K. Matsuse, and T. Nakano, “DSP- based speed adaptive flux observer
of induction motor,” IEEE Trans. Ind Applicat., vol. 29, pp. 344-348,1993
[64] L. Harnefors, “Design and analysis of general rotor-flux oriented vector control
systems,” IEEE Trans. Ind. Electron., vol. 48, pp. 383–389, Apr. 2001
[65] Y.-R. Kim, S.-K. Sul, and M.-H. Park, “speed sensorless vector control of an
induction motor using an extended kalman filter,” IEEE Trans. Ind Applicat., vol.
30, pp. 1225-1233,1994.
[67] V. I. Utkin, “Sliding mode control design principles and applications to electric
drives,” IEEE Trans. Ind. Electron., vol. 40, pp. 23–36, Feb. 1993.
[69] Derdiyok, A., Guven, M.K., Rehman, H., Inanc, N., Longya Xu, “Design and
implementation of a new sliding-mode observer for speed-sensorless control of
induction machine,” IEEE Transactions on Industrial Electronics, No. 5, pp1177-
1182, 2002.
[70] H. Rehman, A. Derdiyok, M. K. Guven, and L. Xu, “A new current model flux
observer for wide speed range sensorless control of an induction machine,” IEEE
Transactions on Industrial Electronics, No. 6, pp1041-1048, 2002.
125
[71] P. Mehrotra, J. E. Quaico, and R. Venkatesan, “Speed estimation of induction motor
using artificial neural networks,” Proc IEEE IECON’96, 1996, pp. 881-886.
[76] S. K. Sul and M. H. Park, “A novel technique for optimal efficiency control of a
current-source inverter-fed induction motor,” IEEE Trans. Power Electron., vol. 3,
pp. 192–199, Apr. 1988.
[78] G. K. Kim, I. J. Ha, and M. S. Ko, “Control of induction motors for both high
dynamic performance and high power efficiency,” IEEE Trans. Ind. Electron., vol.
39, pp. 323–333, Aug. 1992.
126
APPENDIX A EXPERIMENTAL
9. APPENDIX SETUP
A EXPERIMENTAL SETUP
induction motor drive system was set up. The setup consists of a induction motor, a
power drive board and a DSP controller board. The experimental setup is shown in
Figure A.1
Power analyzer
Motor AC
Drive board
under Power
test
Dynamometer
controller DSP board
`
Figure A.2. The dynamometer controller DSP 6500 (Figure A.3 (lower one)) can provide
technology. Precise torque loading can be provided independent of shaft speed. The
motor input power and efficiency are measured through MAGTROL 6530 power
127
Figure A.2 Hysteresis dynamometer
128