This document is downloaded from DR‑NTU (https://dr.ntu.edu.

sg)
Nanyang Technological University, Singapore.

Performance improvement of direct torque

controlled permanent magnet synchronous
motor drives

Yang, Hejin

2019

Yang, H. (2019). Performance improvement of direct torque controlled permanent magnet

synchronous motor drives. Doctoral thesis, Nanyang Technological University, Singapore.

https://hdl.handle.net/10356/104238

https://doi.org/10.32657/10356/104238

Downloaded on 06 Jun 2024 17:51:07 SGT

Performance Improvement of Direct Torque Controlled
Permanent Magnet Synchronous Motor Drives

YANG HEJIN

SCHOOL OF ELECTRICAL & ELECTRONIC ENGINEERING

2019
Performance Improvement of Direct Torque Controlled
Permanent Magnet Synchronous Motor Drives

Yang Hejin

School of Electrical and Electronic Engineering

A thesis submitted to the Nanyang Technological University

in partial fulfillment of the requirement for the degree of
Doctor of Philosophy

2019
ACKNOWLEDGEMENT

Acknowledgement

After 7 years of part-time journey, this thesis finally comes to completion. It’s
impossible to finish without any of the people that I would like to thank here. Firstly
of all, I would like to express my sincere gratitude to my supervisor, Professor Wang
Youyi for bringing me as your Ph.D. student, giving me valuable advice and guidance.
You always have faith in me even sometimes I myself want to give up. Your critical
support at the most difficult time makes me get through this.

I also express my sincere appreciation to Dr. Zhang Xinan for being my co-supervisor.
We fight together to get things done despite so many obstacles and difficulties. I won’t
forget your help and a friend in need is a friend indeed.

I would like to take this opportunity to thank EDB and NTU for giving me this
opportunity for doing Ph.D. research. I am deeply grateful to all my employers and
colleagues throughout my Ph.D. career: Peteris Global Ltd, Hope Technik Pte Ltd,
and Dyson Operations Pte Ltd. Thank you for your financial support, lab facilities,
and time support for allowing me to finish the research. I would like to express my
special thanks to my line manager, Dr. Clare Loke, for your great support, trust, and
encouragement.

Finally, I express my deepest love and appreciation to my beloved wife Mr. Ma Yeli
and my lovely daughter Yang Yizhou. The thesis work was carried out mostly in the
evenings and weekends. I am truly sorry for not being able to spend those time with
you and thank you so much for your understanding, support and encourage
throughout these 7 years. It’s you that offers me strength and gives me the reason to
continue.

i
TABLE OF CONTENTS

Table of Contents

ACKNOWLEDGEMENT ........................................................................................ I

TABLE OF CONTENTS ........................................................................................ II

ABSTRACT ............................................................................................................ VI

LIST OF FIGURES ............................................................................................ VIII

LIST OF TABLES ................................................................................................ XII

LIST OF SYMBOLS........................................................................................... XIII

LIST OF ACRONYMS ....................................................................................... XIV

CHAPTER 1 INTRODUCTION EQU ATI ON C HAPTER (NEX T) SECTION 1

................................................................ 1

1.1 Background .................................................................................................... 1

1.1.1 PMSM Control and Applications .................................................................. 1

1.1.2 Flywheel Energy Storage System ................................................................. 5

1.2 Scope of this thesis......................................................................................... 6

1.3 Objective and Contributions .......................................................................... 7

1.4 Organization of the Thesis ............................................................................. 9

CHAPTER 2 FUNDAMENTALS AND LITERATURE REVIEW ............... 11

2.1 Introduction ...................................................................................................11

2.2 PMSM Reference Frames and Model ...........................................................11

2.2.1 Model in the stator reference frame ............................................................ 13

2.2.2 PMSM in 𝒅𝒒 reference frame ..................................................................... 14

2.2.3 PMSM in 𝒙𝒚 reference frame ..................................................................... 15

ii
TABLE OF CONTENTS

2.3 Dual airgap AFPM flywheel motor model ................................................... 16

2.4 Improvements in DTC methods ................................................................... 22

2.5 Field Weakening Methods............................................................................ 24

2.6 Stator Flux Observer .................................................................................... 26

2.7 Summary ...................................................................................................... 27

CHAPTER 3 SENSOR-LESS FLUX AND TORQUE OBSERVER STUDY

FOR THE DIRECT TORQUE AND FLUX CONTROLLED SURFACE
MOUNTED PERMANENT MAGNET FLYWHEEL MOTOR ....................... 29

3.1 Introduction .................................................................................................. 29

3.2 Sensor-less Direct Torque and Flux Control ................................................ 30

3.3 Stator flux observer ...................................................................................... 32

3.3.1 Stator Flux Observers for Evaluation Study ............................................... 33

3.3.2 Explanation with characteristic function..................................................... 36

3.4 Simulation Study .......................................................................................... 40

3.5 Conclusions .................................................................................................. 43

CHAPTER 4 A NEW FIELD WEAKENING CONTROL APPROACH FOR

A DIRECT TORQUE AND FLUX CONTROLLED SURFACE PERMANENT
MAGNET MOTOR ............................................................................................... 45

4.1 Introduction .................................................................................................. 45

4.2 MTPA and Field Weakening Trajectories for SPMSM ................................ 47

4.3.1 Maximum Torque Per Ampere Trajectory .................................................. 47

4.2.2 Field Weakening Trajectory ..................................................................... 48

4.3 Proposed Field Weakening Algorithm ......................................................... 49

4.4 Simulation Study .......................................................................................... 53

iii
TABLE OF CONTENTS

4.5 Conclusion ................................................................................................... 60

CHAPTER 5 DEADBEAT PREDICTIVE TORQUE CONTROL WITH

SPACE VECTOR MODULATION...................................................................... 61

5.1 Introduction .................................................................................................. 61

5.2 Basic Principle of Torque Prediction ........................................................... 61

5.3 Simplified deadbeat PTC with SVM ........................................................... 63

5.4 Simulation Results ....................................................................................... 66

5.5 Experiment Results ...................................................................................... 69

5.6 Conclusion ................................................................................................... 73

CHAPTER 6 DUTY CYCLE BASED MODEL PREDICTIVE TORQUE

CONTROL FOR PMSM ....................................................................................... 74

6.1 Introduction .................................................................................................. 74

6.2 Model and Conventional FS-MPTC ............................................................ 75

6.3 Proposed Duty Ratio Controller and Improvement of Flux Regulation ...... 78

6.4 Simulation Results ....................................................................................... 80

6.5 Conclusion ................................................................................................... 85

CHAPTER 7 CONCLUSION AND FUTURE WORKS EQUATION CHAP TER (N EXT) SEC TI ON 1

........................ 86

7.1 Conclusion ................................................................................................... 86

7.2 Future Works ................................................................................................ 95

APPENDIX A DERIVATION OF (5.7) EQUATI ON CHAP TER 1 SEC TION 1

........................................................ 97

APPENDIX B SIMULATION MODELS ....................................................... 101

B.1 Top Level Overall Model shared across Chapter 3 to Chapter 6 .................... 101

iv
TABLE OF CONTENTS

B.2 Speed Control and Field Weakening Model .................................................... 102

B.3 DTFC Model for Chapter 3 and Chapter 4 ..................................................... 103

B.4 MPTC Model for Chapter 5 and Chapter 6 ..................................................... 104

B.5 DB-SVM-PTC Model for Chapter 5 ............................................................... 105

B.6 DutyCycle-PTC Model for Chapter 6 ............................................................. 106

B.7 Stator Flux, Torque, Speed and Angle Observers for Chapter 3 to Chapter 6 107

B.8 Advanced FieldWeakening ...............................................................................116

APPENDIX C CLARKE AND PARK TRANSFORMS EQU ATI ON CHA PTER (NEX T) SECTION 1

......................... 117

AUTHOR’S PUBLICATIONS ............................................................................ 119

BIBLIOGRAPHY ................................................................................................ 120

v
SUMMARY

Abstract

Permanent Magnet Synchronous Motors (PMSM) have been gaining ever-increasing

popularity in many applications thanks to their advantages such as long life, high
torque and power density, excellent efficiency, etc. The applications of PMSM could
be found in almost every industry such as home appliances, robotics, flywheels,
industrial automation, electric vehicles, so on and so forth. Various control methods
have been proposed for PMSM to attain satisfying performance with different
application needs. For example, V/f control with stabilization loops could be used to
control fan-type applications where dynamic response is not a must but efficiency
and quietness matter. Field Oriented Control (FOC) is the dominant PMSM control
method that achieve good dynamic response, smoothness and energy efficiency.
Direct Torque Control (DTC) excels FOC in simpler control structure and supreme
dynamic response but suffers from variable switching frequency and severe torque
ripples. These defects generally lead to poor torque and flux regulation, high current
harmonics and acoustic noise.

This thesis studies the performance improvement of direct torque controlled PMSM
drives with more emphasis on the Surface-mounted PMSM(SPMSM), which is
widely used due to its cost-effectiveness. Several sub-topics are covered in this thesis.
In the first place, torque ripple reduction is investigated. Emerging techniques to
address the torque ripple problem of DTC can be classified into three groups: (1)
Model Predictive Torque Control (MPTC) which select an optimal voltage vector(VV)
and apply it on the next full control cycle (one VV); (2) Duty cycle based MPTC
which finds an optimal active voltage vector and zero voltage vector (two VVs) for
the next control cycle and (3), Space Vector Modulation (SVM) MPTC which
essentially uses two active vectors and one zero vectors for the next control cycle
(three VVs). All the methods belong to the Finite Set MPC category. This thesis
contributes an improved duty cycle based MPTC which finds the optimal duty cycle
by a PI controller. It obtains excellent torque and flux regulation with fixed switching
frequency.

In the second place, to further enhance the performance of PMSM drives, predictive

vi
SUMMARY

deadbeat control is proposed in this thesis. Compared to the existing work, this
method is implemented in the stationary reference frame and has a simplified nature.
It produces very smooth torque and flux responses using SVM. Moreover, the usage
of prediction technique compensates for the one cycle digital delay and thus, is
desirable for a digital control system.

Thirdly, fast and stable speed control is generally preferred in a wide speed range for
high-speed PMSM drives, such as the flywheel storage system. Field Weakening (FW)
techniques are generally used to extend the speed range to allow extra energy storage.
Most of the existing FW methods for DTFC based drives have constrained constant
power speed range (CPSR) due to the limited usage of DC-link voltage. Moreover,
many of these FW approaches are heavily dependent on machine parameters. This
can lead to degraded performance of drives or even control instability under some
circumstances. To extend the CPSR and simultaneously enhance the parameter
robustness of FW operation, this thesis proposes a new robust FW algorithm for
DTFC based SPMSM drives. The proposed algorithm achieves extended CPSR and
reduced parameter dependency.

It’s worth mentioning that the previously described methods are based on using
encoder as speed and rotor position feedback. Various sensor-less observers have
been proposed in the literature to save the cost of such sensors and to improve the
drives’ reliability. As direct torque control inherently requires flux observers, it is
straightforward to extend the flux observers into position and speed feedback. This
thesis studied three stator flux observers and studied their performance and
compensator gain setting method. The best performance dq frame flux observer is
selected and being applied for all the DTC methods discussed in this thesis.

vii
LIST OF TABLES

List of Figures

Figure 1.1 Classification of electrical machine family [1] ........................................ 2

Figure 1.2 Classification of classic PMSM control methods [12] ............................. 3

Figure 1.3 Main components of the dual-airgap AFPM machine in a simplified

FESS structure [8] .............................................................................................. 6

Figure 1.4 Picture of prototype dual-airgap AFPM machine [8].(a). the rotor and
stator assembly. (b). rotor magnet assembly. ..................................................... 7

Figure 1.5 Structure of this thesis ............................................................................ 10

Figure 2.1 (a).definition of space vector and (b) commonly used reference frame . 12

Figure 2.2 Illustration of the dual-airgap AFPM machine principle ........................ 17

Figure 2.3 Three-phase (a,b,c) and rotating (d,q) windings of a 2-pole AFPM
machine ............................................................................................................ 17

Figure 3.1 Structure of Sensor-less DTFC Control.................................................. 30

Figure 3.2 Stator Flux PI Controller using Vx .......................................................... 32

Figure 3.3 Torque PI Controller using Vx ............................................................... 32

Figure 3.4 Structure of stator flux observer 1 .......................................................... 34

Figure 3.5 Structure of stator flux observer 2 .......................................................... 35

Figure 3.6 Stator Flux Observer 3............................................................................ 35

Figure 3.7 Redrawn stator flux observer showing voltage model and current model
.......................................................................................................................... 37

Figure 3.8 Bode plot of characteristic function for (a) voltage model F(s) and (b)
current model 1-F(s) for both method 1 and method 2 .................................... 39

viii
LIST OF TABLES

Figure 3.9 Simulation of SPMSM Sensor-less DTFC Torque Step with Flux
Observer 1. (a) Speed, Flux and Torque; (b) Error of Speed, Flux and Torque 42

Figure 3.10 Simulation of SPMSM Sensor-less DTFC Torque Step with Flux
Observer 2. (a) Speed, Flux and Torque; (b) Error of Speed, Flux and Torque42

Figure 3.11 Simulation of SPMSM Sensor-less DTFC Torque Step with Flux
Observer 3. (a) Speed, Flux and Torque; (b) Error of Speed, Flux and Torque 43

Figure 4.1 Block diagram of the proposed reference flux adjustment scheme ........ 50

Figure 4.2 Block diagram of the proposed torque limit adjustment scheme ........... 51

Figure 4.3 Block diagram of DTFC incorporating the proposed field-weakening and
torque limit adjustment algorithm.................................................................... 52

Figure 4.4 Conventional FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 4000rpm using (4.5) and (4.7) ............................................... 53

Figure 4.5 Conventional FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 4000rpm with the conventional FW algorithm using (4.5) and
(4.6) .................................................................................................................. 54

Figure 4.6 Proposed FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 4000rpm using (4.10) and (4.8) ............................................. 55

Figure 4.7 Conventional FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 500rpm with the conventional FW using (4.5) and (4.6) ...... 56

Figure 4.8 Proposed FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 5500rpm using (4.10) and (4.8) ............................................. 56

Figure 4.9 Conventional FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 4000rpm with +20% detuning in 𝐿𝑠 ...................................... 57

Figure 4.10 Proposed FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 4000rpm with +20% detuning in 𝐿𝑠 ..................................... 57

ix
LIST OF TABLES

Figure 4.11 Conventional FW algorithm: Simulation of SPMSM step speed

response from 600rpm to 4000rpm with -20% detuning in 𝑅𝑠 ....................... 59

Figure 4.12 Proposed FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 4000rpm with -20% detuning in 𝑅𝑠 ...................................... 59

Figure 5.1 Block diagram of the proposed simplified deadbeat PTC with SVM .... 61

Figure 5.2 Flow chart of predictive torque control. ................................................. 63

Figure 5.3 Reference torque and flux amplitude calculation. .................................. 63

Figure 5.4 Possible voltage vectors and voltage limit imposed by inverter............. 65

Figure 5.5 Reversed torque response of PMSM: (a) torque (up) and zoom in
(below) of the classical MPTC (b) torque (up) and zoomed in (below) of the
proposed method. ............................................................................................. 66

Figure 5.6 Reversed speed response of PMSM: (a) classical MPTC speed and
torque response (b) proposed MPTC speed and torque response .................... 67

Figure 5.7 Steady-state stator flux and current response of the classical MPTC
under 0.3Nm load at 1200 rpm speed: (a) subplot 1: amplitude of stator flux
subplot 2: phase-a current; (b) frequency spectrum and THD of stator current
.......................................................................................................................... 68

Figure 5.8 Steady-state stator flux and current response of the proposed method
under 0.3Nm load at 1200 rpm speed: (a) subplot 1: amplitude of stator flux,
subplot 2: phase-a current; (b) frequency spectrum and THD of stator current.
.......................................................................................................................... 69

Figure 5.9 Experiment platform and the flywheel motor. (a) photo of the overall
experiment setup; (b) system function block diagram ..................................... 70

Figure 5.10 Experiment Result of Classic MPTC: step from 600rpm to 1200rpm. (a)
reference and actual speed (b)torque response (c) stator flux.......................... 71

x
LIST OF TABLES

Figure 5.11 Experiment Result of Proposed MPTC: step from 600rpm to 1200rpm.
(a) reference and actual speed (b)torque response (c) stator flux .................... 72

Figure 5.12 Experiment Result: Steady-state motor current at 1200rpm (a) classic
MPTC; (b) proposed MPTC. ........................................................................... 73

Figure 6.1 Block diagram of the proposed model predictive torque control. .......... 75

Figure 6.2 Proposed duty cycle control method. ..................................................... 79

Figure 6.3 Steady-state response of conventional MPTC(a) and proposed MPTC (b)
at 300rpm with 3.2Nm load. ............................................................................ 82

Figure 6.4 Current harmonics of conventional MPTC(a) and proposed MPTC (b) at
300rpm with 3.2Nm load. ................................................................................ 82

Figure 6.5 Performance of step speed test: (a) speed, torque, and flux responses of
PMSM controlled by classical MPTC, (b) speed, torque and flux responses of
PMSM controlled by the proposed method ..................................................... 83

Figure 6.6 Performance of torque reversal test: (a) speed, torque, and flux responses
of PMSM controlled by classical MPTC, (b) speed, torque and flux responses
of PMSM controlled by the proposed method. ................................................ 83

Figure 6.7 Illustration of flux de-regulation problem of dead-beat based duty cycle
control in (a) and excellent flux regulation of PI-based duty cycle control in
(b). reference in red and actual in blue. ........................................................... 84

Figure 7.1 Torque Regulation Ripple for Four Compared Methods under 1Nm Load
and Reference................................................................................................... 88

Figure 7.2 Field Weakening Operating Region........................................................ 90

Figure 7.3 Voltage at maximum and minimum stator flux point ............................. 91

Figure 7.4 Code Generation Report Screenshot for MPTC ..................................... 95

xi
LIST OF TABLES

List of Tables

Table 2.1 The dual airgap AFPM machine design parameters................................. 19

Table 2.2 Comparison of DTC and MPTC Methods ............................................... 24

Table 3.1 SPMSM parameters for simulation .......................................................... 40

Table 3.2 Control parameter settings for simulation ................................................ 41

Table 5.1 Distribution of Active VV in SVM .......................................................... 65

Table 6.1 PMSM parameter for simulation study .................................................... 81

Table 7.1 Mean and Standard Deviation of Four Compared Methods .................... 89

Table 7.2 Speed Range Extension Numerical Case Study (No Load) ..................... 93

Table 7.3 Computation resource report for the four compared methods ................. 95

xii
LIST OF SYMBOLS

List of Symbols

𝒗𝒔 Stator voltage vector with 𝒗𝒔 = [𝑢𝛼 ; 𝑢𝛽 ] or [𝑢𝑑 ; 𝑢𝑞 ] or [𝑢𝑥 ; 𝑢𝑦 ]

depends on the reference frame, same for other vectors 𝒊𝒔 . 𝝀𝒔 and 𝝀𝒓
𝒊𝒔 Stator current vector
𝝀𝒔 Stator flux vector
𝝀𝒓 Rotor flux vector
𝜆𝑠 Magnitude of stator flux vector
|𝜆𝑠 | Same as λs, magnitude of stator flux vector
𝜆𝑟 Magnitude of the rotor flux vector
𝜆𝑝𝑚 Same as λr, magnitude of rotor flux vector
𝜆𝑓 In chapter 3, the flux linkage caused by virtual field winding
In other chapters: Same as λr, magnitude of rotor flux vector
Rs Motor phase resistance
Ld, Lq d and q axis inductance
Ls Motor phase winding self-inductance
𝜃𝑟𝑒 , 𝜃𝑒 Rotor d-axis electrical angle
𝜔𝑟𝑒 , 𝜔𝑒 Rotor d-axis electrical radius speed
Te Electrical torque generated by motor
TL Load torque
J Moment of inertial
B Viscous coefficient
P Number of pole pairs
̂ Estimated variable
δ Torque angle

xiii
LIST OF ACRONYMS

List of Acronyms

AC Alternative Current
AFPM Axial Flux Permanent Magnet
BLDC Brushless Direct Current
CB Current model Based
CPSR Constant Power and Speed Range
DB Deadbeat
DC Direct Current
DTC Direct Torque Control
DTFC Direct Torque and Flux Control
EMF Electro-Motive Force
FESS Flywheel Energy Storage System
FOC Field Oriented Control
FS-MPTC Finite Set Model Predictive Torque Control
FW Field Weakening or Flux Weakening
IPMSM Interior or Inserted PMSM
MMF Magneto-Motive Force
MPTC Model Predictive Torque Control
MTPA Maximum Torque Per Ampere
MTPF Maximum Torque Per Flux
PM Permanent Magnet
PMSM Permanent Magnet Synchronous Machine or Motor
PTC Predictive Torque Control
PWM Pulse Width Modulation
SPMSM Surface-mounted PMSM
SVM Space Vector Modulation
SVPWM Space Vector Pulse Width Modulation
THD Total Harmonic Distortion
V/f Voltage over frequency
VB Voltage model Based

xiv
CHAPTER 1 Introduction

CHAPTER 1 Introduction

1.1 Background

1.1.1 PMSM Control and Applications

Features and Classification of PMSM

Permanent Magnet Synchronous Motors (PMSM) belongs to the AC synchronous

motor in the AC machine family, as shown in Figure 1.1. It possesses many
advantages as compared with other types of machines. Firstly, if compared with DC
motor, life can last for tens of years as there’s no brush and commutator short life
problem. Secondly, as compared with the winded field machine and AC induction
machine, the rotor filed is provided by a permanent magnet, thus the efficiency is
much higher. In addition, as there’s no field winding, the size can be much more
compact, making it high in power density, suitable for those applications where size
and weight are a concern. Thirdly, the PMSM could be made in a variety forms, such
as Surface-mounted PMSM(SPMSM) or Interior or Inserted PMSM (IPMSM), radial
flux type or axial direction type, linear type or rotational type, inside rotor or external
rotor, etc. making it very flexible and suits a large variety of applications. Lastly, if
compared with BLDC, the only difference is that the BLDC features trapezoidal
shape back-EMF while the PMSM produces sinusoidal back-EMF. Thus, the torque
ripple of PMSM is much smaller under sinusoidal voltage/current control methods
such as FOC.

1
CHAPTER 1 Introduction

Electric
Machine

DC Machine AC Machine

Induction Variable
Synchronous
Machine Reluctance

Squirrel Wound Wound Permanent Synchronus Switched

Stepper
Cage Rotor Field Magnet Reluctance Reluctance

BLDC PMSM

SPMSM IPMSM

Figure 1.1 Classification of electrical machine family [1]

PMSM Application Overview

Due to these excellent advantages of PMSM, applications of PMSM could be found

in a wide range of applications. For example, in industrial servo drive applications[2,
3], PMSM servo drives are found to drive industrial robot, machine tools, that require
precision position/speed control, as well as rigid torque-speed curve to reject external
disturbance. In transportation, electric vehicles[4] and ship propulsion[5] favor
PMSMs due to the high efficiency and high power and torque density; In energy
generation and storage applications, PMSM can be used for wind turbines[6, 7],
flywheels[8-11], etc.

In recent years, in BLDC dominated applications such as wash machines,

compressors in air-conditioners and fridges, fans and purifiers, there’s a trend to shape
the back-EMF into sinusoidal one, making them a PMSM. This is driven by ever
demanding requirement for supreme product efficiency, smooth torque, and low
acoustics noise.

PMSM Control Overview

The control of PMSM shares a lot of similarity with other types of AC machines,
especially the AC induction machine. In fact, the three typical control methods, ie.
The V/f scalar control, the field-oriented control (FOC) and the direct torque control

2
CHAPTER 1 Introduction

(DTC), are first been applied on the AC induction machine, and subsequently being
transplanted to the PMSM machine. Figure 1.2 gives an overview of the classical ac
machine methods based on [12] with updates on the DTC part.

PMSM
Control

Scalar Self-Control FOC DTC

Figure 1.2 Classification of classic PMSM control methods [12]

Scalar control, or V/f control, is an open-loop control method that is widely used by
induction machine drives. The typical applications are those do not require fast
dynamic responses, such as air-conditioners, conveyor belts, ventilating system, etc.
The control of PMSM using V/f control is attracting some research interests due to
the ever-increasing demand for motor efficiency. However, V/f for PMSM is
inherently unstable and the machine can easily lose synchronization in the high-speed
region or under abrupt load change [13-15]. In addition, since the motor current is
not directly controlled to achieve maximum torque per ampere (MTPA), the
efficiency is not guaranteed to be optimum. A substantial amount of research work
has been carried out the efficiency and stability of PMSM drives using V/f control[13,
14, 16-19].

Self-control relies on an absolute position sensor (such as hall sensor or absolute

encoder) to measure the rotor flux position, and then do phase commutation based on
the position angle. It is more widely known as the six-step commutation control in
BLDC control literature[12]. Since PMSM only differs BLDC by the shape of the
back-emf, this method can be readily applied to PMSM. However, it is better to output
a sinusoidal voltage/current so that the torque and acoustics noise can be reduced.
The disadvantages include, an absolute position sensor is required, and the phase
advance angle needs to be tuned. [20, 21 ] .

Field Oriented Control (FOC) is currently the most widely adopted control method
for AC drives. It was first introduced in the 1970s by F. Blaschke [22] and K. Hasse

3
CHAPTER 1 Introduction

[23] to control induction machines. The general idea of FOC is to decompose the
motor current into torque producing component (on q-axis) and flux-producing
component (on d-axis), mimicking the dc motor control principle. That is, the torque
and flux is controlled indirectly by Id and Iq current vector components. The core of
the FOC scheme are coordinate transformation blocks which allow calculation of
field-oriented current components by using αβ/dq transformation (Park
transformation), and reference voltage vector 𝑉𝑠𝛼 𝑉𝑠𝛽 by using dq/αβ transformation
(inverse Park transformation) [23]. To know the field orientation, either position
sensors such as encoders or resolves, or sensor-less observers, are required to know
the rotor angle and perform Park and inverse Park transform.

In the mid-1980s, an alternative high-performance control method, known as Direct

Torque and flux Control (DTC or DTFC, but in this thesis, DTFC refers to DTC with
SVM), or Direct Self-Control (DSC) was introduced by Takahashi and Noguchi [24]
and Depenbrock [25] for induction machines. In the late 1990s, the DTC has been
extended to use with the PMSM machine in the stator reference frame, as reported by
C. French [26] and L. Zhong [27]. The idea of DTC is to directly control the flux and
torque in the stationary reference frame, without using coordinate transformation.
Two hysteresis controllers are used to control the estimated torque and flux within a
tolerance band of the reference torque and flux. There’s no PWM modulation. One
of the 7 vectors of the inverter is selected to control the increase/decrease of the torque
and flux, depends on which sector the voltage vector is traveling in.

Compared with FOC, the classic DTC method possesses several advantages: 1) very
fast torque and flux response, 2) simple structure, no coordinates transform and PI
controllers 3), no precise rotor information is required, thus it is inherently sensor-
less; 4), less parameter dependency. [28]

On the other hand, the classic DTC suffers several drawbacks, such as (1), large
torque and flux ripple, (2) variable switching frequency. These defects generally lead
to poor torque and flux regulation, high current harmonics and acoustic noise. In
recent years, numerous research efforts have been done in the literature to overcome
these limitations for AC drives in general and good results have been achieved, such

4
CHAPTER 1 Introduction

as Space Vector Modulation based DTC (SVM-DTC) [29, 30] (i.e. DTFC), Predictive
Torque Control (PTC)[31, 32] or model predictive direct torque control (MPTC) [33].
These can be considered as improved DTC and a more detailed review will be given
in the literature review part in Section 2.4.

1.1.2 Flywheel Energy Storage System

One of the typical applications of PMSMs is the Flywheel Energy Storage System
(FESS). FESS is an ancient energy storage system where the kinetic energy is stored
in a rotating flywheel. The energy being stored is proportional to the inertia of the
flywheel and to the square of rotation speed, that is, E = 12J𝜔2 . A modern FESS
consists of a flywheel, motor/generator, power electronics, and mechanical
supporting system. As it usually spins very fast to store more energy in a smaller size,
the flywheel is enclosed in a strong vacuum or helium chamber for both safety and
loss reduction purposes.

The core component of FESS is the electric motor/generator that used to drive the
flywheel. Various different electric machine types can be used, for example, induction
machine (IM), Switched Reluctance Machine (SRM) and Synchronous Reluctance
Machine (SynRM), Brushless DC Motor (BLDC), Permanent Magnet Synchronous
Machine (PMSM), Homopolar machine (HM), etc. Among them, PMSM is the most
common choice, for its wide speed range, high efficiency and power density, smooth
operation, as well as versatile configurations are possible. For example, the flux can
be radial direction as most normal PMSMs do, or it can also be axial flux, which can
make the machine much compact, offering even higher efficiency, higher power and
torque density than the radial flux counterparts.

The typical control requirements for FESS includes:

• Fast charge and discharge speed. FESS is typically used for a short period of
energy storage, similar to super-capacitors. The discharge speed must be fast
enough when needed. And it should also be capable of fast charging to absorb
transient energy from load.
• Low torque ripple. Excessive torque ripple would cause vibration and

5
CHAPTER 1 Introduction

acoustics noise which jeopardizes the system life and the usage environment.
• Robust and reliable. As FESS features dozens of years of service life and low
maintenance, the controller must be robust and reliable to match these features.
• Wide speed range. This means more energy can be stored going for higher
speed, and higher discharge depth going for lower speed.

1.2 Scope of this thesis

This thesis focus on the PMSM control for energy storage applications, and the DTC
and its improvements will be of the interest of this thesis. A dual airgap Axial Flux
PM (AFPM) machine prototype for FESS was designed and constructed in the lab by
[34]. Figure 1.3 illustrates the main components of the machine and the pictures of
the machine assembly can be found in Figure 1.4. This machine will be used as an
evaluation platform for validating various control algorithms. This machine can be
treated as two surface-mounted PMSMs connected in series and can be driven by a
single power electronics inverter. It is also cable of controlling the axial displacement
for reducing the bearing supporting force. A more detailed introduction and modeling
of this machine will be given in Chapter 3.

vacuum pump vacuum and safety

chamber
mechanical
supporting
system

upper stator
& windings
magnet
upper
airgap upper rotor
rotor+rim
lower rotor
lower
airgap magnet
lower stator
& windings

base

mechanical
rotary encoder supporting
system

Figure 1.3 Main components of the dual-airgap AFPM machine in a simplified

FESS structure [8]

6
CHAPTER 1 Introduction

magnet with skew magnet with skew

angle = 0 angle = 18.7 deg.

(a) (b)

Figure 1.4 Picture of prototype dual-airgap AFPM machine [8].(a). the rotor and
stator assembly. (b). rotor magnet assembly.

It is worth highlighting that, although the evaluation platform is based on this special
machine, the proposed methods for the torque and flux control could be applied to
any other types of SPMSM as well.

1.3 Objective and Contributions

Given that the FESS requires fast dynamic response as well as high reliability, and
that the direct torque control offers such characteristics, it is motivated to explore the
application of improved direct torque control for the flywheel energy storage
applications. In addition, as the dual airgap AFPM flywheel motor is unique, it would
be interesting to test the motor with various control strategies, especially the proposed
control methods. The objective of this thesis could be summarized as follows:

(1) One major drawback of DTC is large torque ripple, especially if the switching
frequency is low. However, for large power flywheel applications, it is always
desirable to keep the switching frequency low so that the power loss could be
lower. Excessive torque ripple also means poor efficiency of energy storage,
as it produces vibration and acoustics, and increases the mechanical loss and
stress. This study aims to reduce the torque ripple on the family of the DTC
control method and its variants.

7
CHAPTER 1 Introduction

(2) For flywheel applications, it is always important that it could have a wider
speed range. Wide speed range would mean more energy can be stored on the
flywheel. Field weakening is usually adopted to extend the speed range of PM
machines at given dc bus voltage. This study aims to propose an improved
field weakening strategy in conjunction with DTFC.
(3) It is desirable for a flywheel system to last for decades with no or minimal
maintenance. Reliability is another essential feature of a flywheel. PM
machine usually requires rotor position feedback to do commutation and
speed calculation. However, having position feedback sensors not only
increases the cost but also compromises the reliability when the sensor is
faulty. This study aims to study the performance of the observer-based sensor-
less method and its application to the dual airgap AFPM machine.

With all the aforementioned objectives, investigations are carried out to explore better
control performance for the flywheel motor under study. The dual airgap AFPM
flywheel is employed to study the torque and flux control, leaving the axial levitation
control untouched. The contributions of this study can be summarized as follows:

(1) This study proposes a simplified deadbeat model predictive direct torque
control strategy for PMSM drives with reduced torque ripple and constant
switching frequency. Different from most of the existing approaches, the
proposed method calculates the reference voltage vector in the stationary
reference frame based on the deadbeat technique. Moreover, space vector
modulation is directly employed to synthesize the reference voltage vector
which eliminates the complexity of voltage vector selection in conventional
PTC.
(2) This thesis presents a duty cycle based finite-set model predictive torque
control for PMSM drives. The proposed method contributes to significantly
reduce torque ripples experienced in conventional FS-MPTC by using two
voltage vectors(VVs) in one sampling cycle. The duty ratio active VV is
determined based on the PI controller, which equivalently adjusts the
weighting between torque and flux control in FS-MPTC.
(3) This thesis explores the direct torque and flux control (DTFC) with SVM for

8
CHAPTER 1 Introduction

PMSM drives. To ensure a wide speed range, field weakening (FW) methods
for DTFC are generally used. This thesis proposes a new robust FW algorithm
for DTFC based SPMSM drives. The proposed algorithm achieves an
extended speed region and reduces parameter dependency.
(4) This thesis conducts further study on the dual airgap AFPM flywheel motor
and demonstrates that various above-mentioned control methods could be
applied to that motor.

1.4 Organization of the Thesis

There are seven chapters and three appendices, which have been organized as follows:

Chapter 1 briefly introduces the background of the study. The objectives and major
contributions are also stated.

Chapter 2 presents the most often used PMSM reference frames and the
corresponding models in those frames. The AFPM for the FESS model is also
introduced in this chapter. A fundamental literature review related to the research
topics of this thesis is also given in this chapter.

Chapter 3 introduces DTFC for SPMSM control. As the feedback for torque and flux
are key to DTFC, stator flux observers are evaluated together with DTFC. All the
subsequent chapters will be also based on the same evaluated stator flux observer.

Chapter 4 extends the space vector modulation based direct torque and flux wheel
control (SVM-DTFC) to field weakening region. A new field weakening control
approach is proposed to expend the constant power speed range. This chapter first
introduces conventional field weakening operation and MPTA trajectories for
SPMSM. Then the proposed SVM-DTFC based field weakening method is given. To
verify the validity of the proposed algorithm, simulation results are provided.

Chapter 5 presents performance improvement work over another SVM based

predictive torque control (DB-SVM-PTC) strategy. This chapter first gives an
overview of the proposed strategy and the proposed system block diagram, followed

9
CHAPTER 1 Introduction

by a detailed derivation of the voltage vectors with a 2-step ahead prediction on

torque and flux. Then the proposed strategy is first being validated through Simulink
simulation using the dual airgap AFPM machine parameters, and then by experiment
using the actual dual airgap AFPM flywheel motor.

Chapter 6 works on performance improvement using finite set model predictive

torque control (FS-MPTC). A duty cycle based FS-MPTC is presented in this chapter.
This chapter first introduces the mathematic models being used and the conventional
FS-MPTC, followed by the proposed duty cycle FS-MPTC control algorithm.
Comparisons on both steady-state and dynamic performance between classic FS-
MPTC and duty FS-MPTC are carried out by simulation. In addition, a comparison
between the proposed duty cycle FS-MPTC and a deadbeat-based duty-cycle PTC
has also been done to show that duty cycle MPTC does not have the flux-deregulation
problem.

Chapter 7 summarizes the key conclusion of this thesis. Meanwhile, some potential
research topics related to this thesis are recommended for future works.

The structure of this thesis is illustrated by the block diagram as shown in Figure 1.5.

Figure 1.5 Structure of this thesis

10
CHAPTER 2 Fundamentals and Literature Review

CHAPTER 2 Fundamentals and Literature Review

2.1 Introduction

This chapter first briefly introduced the fundamentals of PMSM control in Section
2.1, including space vector concept, reference frames, PMSM models in three most
often used reference frames: αβ frame, dq frame, and xy frame. Then in Section 2.3,
the dual airgap flywheel motor model was presented and one can find that the model
is equivalent to an SPMSM. In Section 2.4, a brief review of the improvement of
DTC control for PMSM, including SVM-DTC, DB-DTC, Finite Control Set Model
Predictive Torque control (FCS-MPTC, or simply MPTC), duty cycle MPTC, etc.
Review of Flux Weakening methods are introduced in Section 2.5. Finally, Section
2.6 reviewed the stator flux observer which is essential for direct torque and flux
control.

2.2 PMSM Reference Frames and Model

For 3-phase AC electric machine modeling and control, a 3-phase variable, such as
voltage, is often expressed in a combined vector form 𝒗𝒔 given by [1]:

2
𝒗𝒔 = (𝑣 + 𝐚𝑣𝑏 + 𝐚𝟐 𝑣𝑐 ) (2.1)
3 𝑎
2𝜋
where 𝐚 = 𝑒 𝑗 3 . 𝑣𝑎 𝑣𝑏 and 𝑣𝑐 instantaneous scalar value in phase a,b,c axis as
shown in Figure 2.1(a). Same is true for the current vector 𝒊𝒔 , stator flux linkage
vector 𝝀𝒔 etc. The combined vector is more convenient expressed in two-axis
2𝜋 2𝜋
coordinate αβ frame by expanding 𝑒 𝑗 3 = cos ( ) + 𝑗𝑠𝑖𝑛(2𝜋
3
), and this conversion
3

is often called Clarke Transform.

𝑣𝛼 = 𝑣𝑎
(2.2)
𝑣𝛽 = (𝑣𝑎 + 2𝑣𝑏 )/√3

The term 2/3 is to make sure the combined vector has the same amplitude as a phase-

11
CHAPTER 2 Fundamentals and Literature Review

amplitude.

(a) (b)

Figure 2.1 (a).definition of space vector and (b) commonly used reference frame

Using the vector definition method, the voltage vector 𝒗𝒔 , current vector 𝒊𝒔 , rotor flux
vector 𝝀𝒓 and stator flux vector 𝝀𝒔 in the αβ frame are illustrated in Figure 2.1(b).
The αβ frame is a static frame, with α axis aligned with phase a. Viewing from αβ
frame, all the vectors are rotating. As PMSM is synchronous motor, the vectors are
rotating at same (average) speed 𝜔𝑒 .

In addition to the αβ frame, there are two most often used rotating reference frames:
dq reference frame and xy reference frame (or 𝑓𝜏 frame in some literature, for
example [35] ). The dq frame is called the rotor flux reference frame, with the d-axis
aligned with rotor permanent flux linkage. The xy frame is called the stator flux
reference frame, with the x-axis aligned with stator flux linkage. The angle between
αβ frame and dq frame is called rotor electrical angle θ𝑒 . The angel between dq and
xy frame is called torque angle δ.

It is more convenient to do FOC current control in dq frame and to do DTFC (i.e.

DTC) in xy frame. To convert a vector from αβ frame to dq frame, just need to

12
CHAPTER 2 Fundamentals and Literature Review

multiply the vector with 𝑒 −𝑗𝜃𝑒 . Still using the voltage vector as an example, after
multiplication, it becomes:

𝑣𝑑 𝑐𝑜𝑠𝜃𝑒 𝑠𝑖𝑛𝜃𝑒 𝑣𝛼
[𝑣 ] = [ ][ ] (2.3)
𝑞 −𝑠𝑖𝑛𝜃𝑒 𝑐𝑜𝑠𝜃𝑒 𝑣𝛽

This process is also known as Park Transform.

To further transform a vector from dq frame to xy frame, simply multiply with 𝑒 −𝑗𝛿 :

𝑣𝑥 𝑐𝑜𝑠𝛿 𝑠𝑖𝑛𝛿 𝑣𝑑
[𝑣 ] = [ ][ ] (2.4)
𝑦 −𝑠𝑖𝑛𝛿 𝑐𝑜𝑠𝛿 𝑣𝑞

The more detailed description of various transforms could be found in Appendix C

and also in some good reference books, such as [1] and [36].

2.2.1 Model in the stator reference frame

The electrical model of a PMSM in the stator reference frame (abc or αβ frame) can
be expressed as:

𝑑𝝀𝒔
𝒗𝒔 = 𝑅𝑠 𝒊𝒔 + (2.5)
𝑑𝑡

𝝀𝒔 = 𝝀𝒂 + 𝝀𝒓 (2.6)

3
𝑇𝑒 = 𝑃(𝝀𝒔 × 𝒊𝒔 ) (2.7)
2

The cross product is calculated by augmenting the two-component vector into a three-
dimensional vector using a zero in the third row. For example, in the dq frame, 𝝀𝒔
and 𝒊𝒔 becomes [𝜆𝑠𝛼 𝜆𝑠𝛽 0]𝑇 and [𝑖𝑠𝛼 𝑖𝑠𝛽 0]𝑇 respectively. Then the result
𝝀𝒔 × 𝒊𝒔 takes the third item from the resulting vector [0 0 𝜆𝑠𝛼 𝑖𝑠𝛽 − 𝜆𝑠𝛽 𝑖𝑠𝛼 ]𝑇 [36].

Note that the flux linkage is a sum of the armature flux linkage 𝝀𝒂 = 𝐿𝑠 𝒊𝒔 and flux
linkage caused by permanent on the rotor 𝝀𝒓 = 𝜆𝑝𝑚 𝑒 𝑗𝜃𝑒 .

13
CHAPTER 2 Fundamentals and Literature Review

So, if expressed in αβ component, the SPMSM model becomes:

𝑑𝜆𝑠𝛼
𝑣𝑠𝛼 = 𝑅𝑠 𝑖𝑠𝛼 +
𝑑𝑡
(2.8)
𝑑𝜆𝑠𝛽
𝑣𝑠𝛽 = 𝑅𝑠 𝑖𝑠𝛽 +
𝑑𝑡

𝜆𝑠𝛼 = 𝐿𝑠 𝑖𝑠𝛼 + 𝜆𝑝𝑚 𝑐𝑜𝑠𝜃𝑒

(2.9)
𝜆𝑠𝛽 = 𝐿𝑠 𝑖𝑠𝛽 + 𝜆𝑝𝑚 𝑠𝑖𝑛𝜃𝑒

3
𝑇𝑒 = 𝑃(𝜆𝑠𝛼 𝑖𝑠𝛽 − 𝜆𝑠𝛽 𝑖𝑠𝛼 ) (2.10)
2

And the mechanical model is given by:

𝑑𝜔𝑚
𝑇𝑒 = 𝐽 + 𝐵𝜔𝑚 + 𝑇𝐿 (2.11)
𝑑𝑡

where 𝑇𝐿 is load torque, J is the moment of inertial, B is viscous coefficient, 𝜔𝑚 is

mechanical angular speed. The mechanical speed relates to rotor electrical speed by
a number of pole pairs P:

𝜔𝑚 = 𝑃𝜔𝑒 (2.12)

Note that the mechanical model is universal regardless of the reference frame. So
the mechanical equation is neglected in other reference frames.

2.2.2 PMSM in 𝒅𝒒 reference frame

The model for both IPMSM and SPMSM in dq reference frame is:

14
CHAPTER 2 Fundamentals and Literature Review

𝑑𝑖𝑑
𝑣𝑑 = 𝑅𝑠 𝑖𝑑 + 𝐿𝑑 − 𝜔𝑒 𝐿𝑞 𝑖𝑞 (2.13)
𝑑𝑡

𝑑𝑖𝑞
𝑣𝑞 = 𝑅𝑠 𝑖𝑞 + 𝐿𝑞 + 𝜔𝑒 𝐿𝑞 𝑖𝑞 + 𝜔𝑒 𝜆𝑝𝑚 (2.14)
𝑑𝑡

𝜆𝑠𝑑 = 𝐿𝑑 𝑖𝑑 + 𝜆𝑝𝑚
(2.15)
𝜆𝑠𝑞 = 𝐿𝑞 𝑖𝑞

3
𝑇𝑒 = 𝑃(𝜆𝑠𝑑 𝑖𝑞 − 𝜆𝑠𝑞 𝑖𝑑 ) (2.16)
2

3
𝑇𝑒 = 𝑃(𝜆𝑝𝑚 𝑖𝑞 + (𝐿𝑑 − 𝐿𝑞 )𝑖𝑑 𝑖𝑞 ) (2.17)
2

When applying for SPMSM, just need to replace 𝐿𝑑 and 𝐿𝑞 with 𝐿𝑠 , since
𝐿𝑑 = 𝐿𝑞 = 𝐿𝑠 . Note that for simplicity, the dq components of 𝒗𝒔 are just expressed
as 𝑣𝑑 and 𝑣𝑞 , instead of 𝑣𝑠𝑑 and 𝑣𝑠𝑞 . Same for 𝑖𝑑 and 𝑖𝑞 .

2.2.3 PMSM in 𝒙𝒚 reference frame

The model for both IPMSM and SPMSM in xy reference frame is given by [28]:

𝑑𝜆𝑠
𝑉𝑥 = 𝑅𝑠 𝑖𝑥 + (2.18)
𝑑𝑡

𝑑𝛿
𝑉𝑦 = 𝑅𝑠 𝑖𝑦 + 𝜔𝑒 𝜆𝑠 + 𝜆 (2.19)
𝑑𝑡 𝑠

3𝑃
𝑇𝑒 = 𝜆 𝜆 𝑠𝑖𝑛𝛿 (2.20)
2𝐿𝑠 𝑠 𝑝𝑚

15
CHAPTER 2 Fundamentals and Literature Review

where 𝛿 is torque angle. 𝜆𝑠 is the amplitude of stator flux linkage vector 𝝀𝒔 and 𝜆𝑝𝑚
is the amplitude of the permanent magnet flux linkage vector 𝝀𝒓 . This model is widely
used in DTC control of PMSM.

2.3 Dual airgap AFPM flywheel motor model

The design and model of this dual airgap AFPM flywheel motor have been introduced
in [8, 11, 37] and in the thesis [34]. For the reader’s convenience, the model is
rederived in this section.

The dual airgap AFPM flywheel machine consists of two halves, namely the upper
half and the lower half. Each upper or lower half has its own stator and rotor. The
upper and lower windings can be connected in series so that only one 3-phase power
converter is required. The principle of the dual airgap AFPM machine is illustrated in
Figure 2.2. The upper half winding and rotor permanent magnet produces an
electromagnetic torque T1 and the lower half winding and rotor permanent magnet
produces an electromagnetic torque T2. The total torque produces by the machine will
be a sum of the two: Te = T1 + T2 .

In this AFPM machine, the torque is used to control the rotation of the flywheel. The
torque equation will be given after the electrical model is derived.

16
CHAPTER 2 Fundamentals and Literature Review

power converter

upper stator
& windings
A B C

airgap (g1)

Z
rotor+rim
T1 (flywheel)
z=0
O

T2
airgap (g2)

A’ B’ C’
lower stator
& windings

Figure 2.2 Illustration of the dual-airgap AFPM machine principle

ia 
d - axis

iq
No
rth d
id
q - axis
q'  A - axis
q

uth
d' So
ib ic

Figure 2.3 Three-phase (a,b,c) and rotating (d,q) windings of a 2-pole AFPM
machine

As the permanent magnets are mounted on the surface of the disk-type rotor, this
machine can be analyzed as a non-salient PM machine. The model of the dual-airgap
AFPM machine is firstly obtained based on the coordinate axes of a 2-pole AFPM
machine, and then be extended to a 2P-pole machine. The three-phase windings and
dq-winding layout of the 2-pole AFPM machine are defined in Figure 2.3. The

17
CHAPTER 2 Fundamentals and Literature Review

rotating d- and q-axis windings are in quadrature of each other and their magnetic
axes are aligned with the d- and q-axis of the rotor. The d-axis of the rotor is chosen
to be aligned with the center lines of the magnets and the q-axis is between the
magnets.

For either of the two halves of this machine, assuming the permanent magnets on the
rotor are represented by the magnetomotive force of 𝑁𝑓 𝑖𝑓 and ignoring the leakage
inductance, the flux matrix of the machine half can be expressed as:

𝐿𝑓𝑓 𝑀𝑓𝑎 𝑀𝑓𝑏 𝑀𝑓𝑐 𝑖𝑓

𝑀𝑎𝑓 𝐿𝑎𝑠 𝑀𝑎𝑏 𝑀𝑎𝑐 𝑖
λ = (𝜆𝑓 𝜆𝑎 𝜆𝑏 𝜆𝑐 ) 𝑇 = ( 𝑎) (2.21)
𝑀𝑏𝑓 𝑀𝑏𝑎 𝐿𝑏𝑠 𝑀𝑏𝑐 𝑖𝑏
𝑀𝑐𝑓 𝑀𝑐𝑎 𝑀𝑐𝑏 𝐿𝑐𝑠 𝑖
( ) 𝑐

where:

𝜇0 𝜋(𝑅𝑜2 − 𝑅𝑖2 )𝑁𝑠2

𝐿𝑎𝑠 = 𝐿𝑏𝑠 = 𝐿𝑐𝑠 = 𝐿𝑠 =
8𝑃2 𝑔
𝜇0 𝜋(𝑅𝑜2 − 𝑅𝑖2 )𝑁𝑓2
𝐿𝑓𝑓 = 𝐿𝑓 =
8𝑃2 𝑔
𝑀𝑎𝑓 = 𝑀𝑓𝑎 = 𝐿𝑚 𝑐𝑜𝑠𝜃
𝑀𝑏𝑓 = 𝑀𝑓𝑏 = 𝐿𝑚 cos (𝜃 − 2𝜋⁄3)
𝑀𝑏𝑓 = 𝑀𝑓𝑏 = 𝐿𝑚 cos (𝜃 + 2𝜋⁄3)

And the PM to winding mutual inductance is given by:

𝜇0 𝜋(𝑅𝑜2 − 𝑅𝑖2 )𝑁𝑠 𝑁𝑓

𝐿𝑚 = (2.22)
8𝑃2 𝑔
𝑁𝑠 = 𝑘𝑤1 𝑁1 (2.23)

𝑁1 is the number of turns in each phase of the stator windings and 𝑘𝑤1 is the winding
factor.

The magnetomotive force 𝑁𝑓 𝑖𝑓 is given by:

18
CHAPTER 2 Fundamentals and Literature Review

𝐵𝑟 𝑙𝑚
𝑁𝑓 𝑖𝑓 = 2𝑃 (2.24)
𝜇𝑟 𝜇 0

where 𝐵𝑟 and 𝜇𝑟 are the remanent flux density and relative permeability of the PM
respectively, 𝑙𝑚 denotes the magnet length and 𝑃 is the number of pole pairs. 𝜇0 is
the magnetic permeability of the air with the value of 4π × 10−7 H/m. The design
parameters of this machine are listed in Table 2.1.

By using the Park’s transform from abc frame to dq frame, the flux linkage λ =
(𝜆𝑓 𝜆𝑎 𝜆𝑏 𝜆𝑐 )𝑇 in (3.1) can be transformed into:

3
𝜆𝑓 𝐿𝑓 𝐿
2 𝑚
0 𝑖𝑓
3
[𝜆𝑑 ] = (𝐿𝑚 𝐿
2 𝑠
0 ) [𝑖𝑑 ] (2.25)
𝜆𝑞 0 0 3
𝐿𝑠 𝑖𝑞
2

Table 2.1 The dual airgap AFPM machine design parameters

Item & Symbol Value & Unit

Number of pole pair, P 2
Equivalent upper air-gap g10 2.877 mm
Equivalent lower air-gap g20 2.763 mm
Magnet type NdFeB
Remanent flux density, Br 1.25T
Stator outer diameter Ro 45mm
Stator inner diameter, Ri 26mm
PM length of upper rotor, lm1 2.0 mm
PM length of lower rotor, lm2 1.87 mm
Number of slots 24
Number of slots per pole per phase 2
Number of phases 3
Skew angle 18.7 degree
Flywheel rim diameter, 12.4 cm
Number of stator winding turns N1 and N2 208 turns each

19
CHAPTER 2 Fundamentals and Literature Review

Ignoring the leakage inductance, the direct axis inductance 𝐿𝑑 and the quadrature
axis inductance 𝐿𝑞 are obtained as:

3 3𝜇0 𝜋(𝑅𝑜2 − 𝑅𝑖2 )𝑁𝑠2

𝐿𝑑 = 𝐿𝑞 = 𝐿𝑠 = (2.26)
2 16𝑃2 𝑔

The amplitude of the flux induced by the permanent magnets in the stator phase
is：

𝜇0 𝜋(𝑅𝑜2 − 𝑅𝑖2 )𝑁𝑠 𝑁𝑓

𝜆𝑝𝑚 = 𝐿𝑚 𝑖𝑓 = 𝑖𝑓 (2.27)
8𝑃2 𝑔

Finally, in the d-q reference frame, the stator flux linkage can be expressed as:

𝜆𝑑 = 𝐿𝑑 𝑖𝑑 + 𝜆𝑝𝑚
(2.28)
𝜆𝑞 = 𝐿𝑞 𝑖𝑞

which is the same as the ordinary radial flux PM machine model in the d-q frame.

The electromagnetic torque can be expressed as:

3
𝑇𝑒 = 𝑃(𝜆𝑑 𝑖𝑞 − 𝜆𝑞 𝑖𝑑 ) (2.29)
2

As 𝐿𝑑 = 𝐿𝑞 , the torque equation can be further simplified as:

3
𝑇𝑒 = 𝑃𝜆𝑝𝑚 𝑖𝑞 (2.30)
2

The voltage equation can be derived as:

20
CHAPTER 2 Fundamentals and Literature Review

𝑑𝑖𝑑
𝑣𝑑 = 𝑅𝑠 𝑖𝑑 + 𝐿𝑑 − 𝜔𝑒 𝐿𝑞 𝑖𝑞
𝑑𝑡 (2.31)
𝑑𝑖𝑞
𝑣𝑞 = 𝑅𝑠 𝑖𝑞 + 𝐿𝑞 + 𝜔𝑒 𝐿𝑑 𝑖𝑑 + 𝜔𝑒 𝜆𝑝𝑚
𝑑𝑡

The axial force is derived by using the co-energy method and is given by:

1 𝜆𝑞 ) 𝑇 ]
𝜕𝑊𝑐𝑜 𝜕 [2 (𝑖𝑓 𝑖𝑑 𝑖𝑞 )(𝜆𝑓 𝜆𝑑
𝐹=− =−
𝜕𝑔 𝜕𝑔
(2.32)
𝜇0 𝜋(𝑅𝑜2 − 𝑅𝑖2 ) 2 5 3
= 2 2
[(𝑁𝑓 𝑖𝑓 ) + 𝑁𝑠 𝑁𝑓 𝑖𝑑 𝑖𝑓 + 𝑁𝑠 2 (𝑖𝑑 2 + 𝑖𝑞 2 )]
16𝑃 𝑔 2 2

Note that in all the above equations the effective air gap is represented by 𝑔. It should
be replaced by 𝑔1 = 𝑔10 − 𝑧 and 𝑔2 = 𝑔20 + 𝑧 for the upper half and lower half
respectively. 𝑔10 and 𝑔20 are design parameters representing the effective airgap
length at the equilibrium point. Also, note that z is much smaller than 𝑔10 and 𝑔20
(um scale versus mm scale).

The force and torque produced by the individual half of the machine can be calculated
by substituting the corresponding parameters for the half-machine with subscript 1
and 2 respectively.

As the machine is controlled by connecting the upper and lower winding in series,
the current will be same ( 𝑖𝑑1 = 𝑖𝑑2 , 𝑖𝑞1 = 𝑖𝑞2 ), the following equations can be
established:

𝑅𝑠12 = 𝑅𝑠1 + 𝑅𝑠2 (2.33)

𝐿𝑠12 = 𝐿𝑠1 + 𝐿𝑠2 (2.34)

𝐿𝑑12 = 𝐿𝑑1 + 𝐿𝑑2 (2.35)

21
CHAPTER 2 Fundamentals and Literature Review

𝐿𝑞12 = 𝐿𝑞1 + 𝐿𝑞2 (2.36)

𝜆𝑝𝑚12 = 𝜆𝑝𝑚1 + 𝜆𝑝𝑚2 (2.37)

3
𝑇𝑒12 = 𝑇1 + 𝑇2 = 𝑃𝜆 𝑖 (2.38)
2 𝑝𝑚12 𝑞

𝐹12 = 𝐹1 + (−𝐹2 ) = 𝐹1 − 𝐹2 (2.39)

Then, the machine can be considered as a single SPMSM, and as the torque is
independent of 𝑖𝑑 current, the rotation of the machine can be controlled the same way
as SPMSM. In the rest part of this thesis, if not stated explicitly, 𝑇𝑒 is to represent the
total electrical torque of full machine(instead of using𝑇𝑒12), and same for the machine
parameters, 𝑅𝑠 , 𝐿𝑠 , 𝐿𝑑 , 𝐿𝑞 , 𝜆𝑝𝑚 , all refer to full machine-level parameters.

2.4 Improvements in DTC methods

Direct torque control (DTC) has been widely used in modern electrical drives due to
its simplicity and fast dynamic response [27, 38]. Nonetheless, classical DTC is
associated with two problems known as high torque ripples and variable switching
frequency. These defects generally lead to poor torque and flux regulation, high
current harmonics and a certain level of acoustic noise at low speeds [39-41]]. A
substantial amount of research work has been carried out to overcome the afore-
mentioned two problems. To alleviate or overcome these issues, many innovative
research proposals have been put forth with varying degrees of success. Notably, the
duty ratio control (D-DTC) [42], predictive torque control (P-DTC) [43], constant
switching frequency-based DTC (CSF-DTC) [44] and the space vector modulation
based DTC (SVM-DTC) [45-47] or DTFC are the most commonly used methods.

Finite Set-Model Predictive Toque Control (FS-MPTC, or simply MPTC) is

considered a significant improvement over the classic DTC as it can predict the future
steps and dynamic select the optimal voltage vector to satisfy certain cost function.

22
CHAPTER 2 Fundamentals and Literature Review

At the same time, it maintains the advantages of fast response and less rotor position
dependency. It also overcomes the one cycle delay in digital control through
prediction [48, 49]. The basic idea of these approaches is to select one optimal VV to
minimize the torque and flux errors in the next sampling cycle. To realize online
optimization, the effects of different VVs are iteratively calculated and compared.
This implies high computational burden since the prediction and cost function
evaluation must be done for all seven VVs, including six active vectors and one zero
vector, for a typical two-level DC/AC inverter fed drives. Although efforts are made
to reduce the computational load of the controller by using selective VVs in MPTC
[50, 51]. Iterative calculation is still unavoidable because of the existence of several
selective VVs in each sampling cycle. Furthermore, for MPTC, the inverter switching
frequency is still variable and the mitigation of torque and flux ripples is limited by
the application of only one VV in an entire sampling cycle [42].

In order to reduce torque and flux ripples whilst maintaining constant switching
frequency, researchers proposed space vector modulation (SVM) based deadbeat
control for AC machine drives [29, 46, 52, 53]. These methods are implemented in
the rotary d-q reference frame which requires coordinate transformation and
decoupling of the cross-coupling terms. This indicates increased algorithm
complexity. Moreover, the majority of them did not employ predictive torque control
(PTC) [29, 46, 53]. As a consequence, the time delay caused by digital control can
not be compensated and the improvement of drive performance is limited. Recently,
a deadbeat MPTC approach with discrete SVM is proposed for PMSM drives [52].
In this approach, the deadbeat technique is employed to determine the reference VV
in the rotary d-q frame. Then, three discretized virtual VVs that are adjacent to the
reference VV are selected as candidates and are iteratively calculated in the online
optimization process. By numerically comparing their corresponding cost functions,
one optimal virtual VV is chosen and subsequently synthesized by using SVM.
Despite its advantages, this approach only approximates the reference VV by using
discretized virtual VVs. Consequently, its contribution to torque and flux ripple
reduction heavily depends on the accuracy of such approximations, which can be
unsatisfying when all three virtual VVs are not very close to the reference VV. Also,
its computational burden is still high because three VVs have to be evaluated

23
CHAPTER 2 Fundamentals and Literature Review

iteratively in online optimization and coordination transformation is required by

deadbeat control.

As the SVM based methods make uses of 3 VVs, the torque ripple as significantly
reduced. However, for high power applications, it is desirable to keep the voltage
vector low in order to reduce switching loss. To achieve a balance between reducing
torque ripple and switching loss, recently, the duty-cycle based MPTC is proposed in
the literature. Compared with conventional MPTC methods, one active VV and one
zero VV is used within one sampling period. The time ratio between the period for
active VV and the sampling time is called duty ratio. As it is sampling-based, it also
resolved the variable switching frequency problem of conventional MPTC methods.
The method to find the duty ratio varies from one to another.

Table 2.2 summarizes the features and advantages of the above-reviewed methods.

Table 2.2 Comparison of DTC and MPTC Methods

Methods Classic DTC; DTFC; Duty MPTC

Classic MPTC Deadbeat-SVM
MPTC
Voltage Vector (VV) 1 VV 3 VVs 2 VVs
Torque Ripple High Low Medium
Switching Freq Variable Fixed Fixed

2.5 Field Weakening Methods

Field weakening (or Flux Weakening, both are abbreviated as FW) is one of the basic
functions that modern ac drive should have [54]. The purpose of field weakening is
to extend the motor speed beyond base speed until available inverter voltage and
current limit is reached. As many applications do not require maximum torque at high
speed, a trade-off between torque and speed can be made, as long as the maximum
available torque satisfies the application need. This is very useful to ensure wide
speed operation of electric drives.

The discussion and derivation of FW trajectories are based on the steady-state model

24
CHAPTER 2 Fundamentals and Literature Review

of PMSM, most conveniently in the dq frame. The max voltage Vmax is usually limited
by dc-link voltage and PWM modulation methods, while the max current Imax is
limited by either inverter max current or motor max current, whichever is lower. Both
refer to amplitude. Hence in dq frame, the constraints are given by:

𝑣𝑑2 + 𝑣𝑞2 ≤ 𝑣𝑚𝑎𝑥

2
(2.40)
𝑖𝑑2 + 𝑖𝑞2 ≤ 𝑖𝑚𝑎𝑥
2 (2.41)

The steady-state PMSM model is obtained by letting di/dt = 0 from the dq reference
frame model. Then the voltage limit constraint defines an ellipse (for IPMSM) or a
circle (SPMSM) with the size inverse proportional to motor speed. The detailed
analytical expression for the ellipse or circle can be found in [1]. In addition, the base
speed at which FW begins and the maximum speed all depends on machine
parameters. For example, for an SPMSM, with 𝜆𝑝𝑚 ≥ 𝐿𝑠 𝑖𝑚𝑎𝑥 (called finite speed
system, which is the case for most SPMSM), after neglecting the stator resistance,
the voltage constraint circle, the base, and maximum speed is given by[1]:

2
𝜆𝑝𝑚 2 𝑣𝑚𝑎𝑥
𝑖𝑞2 + (𝑖𝑑 + ) ≤ 2 2 (2.42)
𝐿𝑠 𝜔𝑒 𝐿𝑠
𝑉𝑚𝑎𝑥 (2.43)
𝜔𝑏𝑎𝑠𝑒 =
√𝜆2𝑝𝑚 + (𝐿𝑠 𝑖𝑚𝑎𝑥 )2
𝑉𝑚𝑎𝑥 (2.44)
𝜔𝑚𝑎𝑥 =
𝜆𝑝𝑚 − 𝐿𝑠 𝑖𝑚𝑎𝑥

In general, there are various existing field-weakening solutions and they can be
roughly divided into feedforward compensation, feedback compensation and a
combination of the two [1, 54, 55]. The feedforward compensation adjusts the torque
limit based on the FW curve, it is an open-loop method and highly depends on the
machine parameter. The feedback compensation method usually employs a PI
controller to adjust the id reference and iq limit. The input of the PI controller is the
difference between the max voltage and the current regulator output voltage
amplitude. This method is robust to parameter variations, but the tuning of the PI
controller is by trial and error. The dynamic performance might also be affected.

As the field weakening is expressed in terms of current id and iq, it is straight forward

25
CHAPTER 2 Fundamentals and Literature Review

to be applied with FOC control methods. The majority of the FW method discussed
in the literature belongs to this group, for example, [54, 56, 57]. For DTC drives, the
control variable is torque and stator flux. The current limiting cannot be directly
imposed. Instead, the current limiting is achieved indirectly by imposing torque limit.
In addition, as the DTC for PMSM is usually done in stator flux frame (xy frame), the
current and voltage limit is expressed with xy components, making it very different
from FOC controlled drives. For example, the voltage constraints, torque constraints,
and stator flux constraints are given by [35]:

2
|𝑉| = √𝑉𝑥2 + 𝑉𝑦2 ≤ 𝑉𝑚𝑎𝑥 (2.45)

2
𝐼𝑦 ≤ √𝐼𝑚𝑎𝑥 − 𝐼𝑥2 (2.46)
3 (2.47)
𝑇𝑒 ≤ 𝑃𝜆 √𝐼 2 − 𝐼𝑥2
2 𝑠 𝑚𝑎𝑥
2
√𝑉𝑚𝑎𝑥 − (𝑅𝑠 𝐼𝑥 )2 − 𝑅𝑠 𝐼𝑥 (2.48)
𝜆𝑠 ≤
𝜔𝑒

Similar work of FW with DTC of PMSM could be found in [8, 35, 58]. A comparison
between FW using FOC and DTC was conducted by [59]. General speaking, FW
using DTC in xy frame has several advantages, such as robust to parameter variations,
simplicity of calculation and more stable in control [59].

2.6 Stator Flux Observer

The motor stator flux observer not only estimates speed and angle for Park
transformation and speed closed-loop control but also provides estimated torque and
flux which are essential for direct torque and flux control. It is well known that
voltage model-based flux observers are more accurate at high speed, while current
model-based flux observers are advantages at low speed. Hence a large number of
state-of-art drives utilize closed-loop flux-observers which combines the voltage
model and current model for a wide speed range of operation.

The combined model was initially introduced for induction machine drives [60] and
employed extensively in recent works on PMSM drives [61],[62],[63],[64],[8, 35, 65,

26
CHAPTER 2 Fundamentals and Literature Review

66], synchronous reluctance drives [67] and induction machine as well [68],[69].

Observer accuracy and dynamic response are of paramount importance to DTC drives.
The parameter sensitivity can be analyzed by using FRF as shown in [60]. [68]
summarized various combination method and unified them using the so-called
observer characteristics functions. It is shown in [68] that the conventional Gopinath-
style model possesses a non-linear flux trajectory when transiting between CB and
VB. Improvement of making them linear are conducted in [68] and [35], at the cost
of increased complexity.

The combined flux observers have significantly improved the sensitivity to dc offset,
motor parameter variations, sensor noise, inverter nonlinearities, etc. At a very low-
speed range, the stator flux observer provides less accurate stator flux and torque
estimation due to motor parameter variations. [65] developed a (disturbance input
decoupling) DID stator flux observer which uses disturbance information from a
current observer to significantly reduce flux estimation error due to magnetic
saturation and PM flux linkage variations for IPMSMs. Stator resistance is the most
sensitive parameter at low speeds. [64] incorporated a novel and simple stator
resistance estimator into the sensor-less drive to compensate for the effects of stator
resistance variation for IPMSM and [8] presented a similar for an SPMSM drive with
an even simpler observer structure.

2.7 Summary

In this chapter, the PMSM model in three most often used reference frames are
introduced: stationary 𝛼𝛽 frame, rotor flux dq frame, and stator flux xy frame. This
thesis is based on these models and reference frames. Specifically, Chapter 3 will
introduce the DTFC which is based on frame xy, but the stator flux observers will be
based on stationary 𝛼𝛽 frame. Chapter 4 will continue the DTFC with Field
Weakening which will still be based on the xy frame. Chapter 5 and Chapter 6 are all
based on the stationary reference frame, demonstrating the simplicity of the MPTC
methods.

A literature review pertaining to the research topic of this thesis is also conducted in

27
CHAPTER 2 Fundamentals and Literature Review

this chapter. SVM-DTFC, Deadbeat PTC, and FS-MPTC are the main improvements
to the DTC and this thesis made contribute based on these methods. The FW method
in the id_iq plane which is convenient for FOC needs to be transformed into the flux-
torque plane for the DTC based methods. The analysis method for combined flux
observer using characteristic function is utilized by this thesis in Chapter 3.2.

28
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

CHAPTER 3 Sensor-less Flux and Torque Observer

Study for the Direct Torque and Flux Controlled Surface
Mounted Permanent Magnet Flywheel Motor

3.1 Introduction

In Chapter 2, the special dual airgap AFPM flywheel motor has been modeled to be
an equivalent surface-mounted permanent motor (SPMSM) for the speed and torque
control. As an energy storage device, the flywheel FESS is required to provide high
burst power with fast dynamic response to meet the stringent demands of future smart
grids where the penetration of renewable generation is high. Furthermore, high
reliability and robustness are necessary to ensure the proper function of FESS. In
practice, all these requirements of FESS are directly imposed on the drives of its
SPMSM.

In recent years, numerous control approaches have been proposed to enhance the
performance of SPMSM drives. Unfortunately, many of them are based on field-
oriented control (FOC) [10, 70, 71], which heavily relies on the knowledge of the
rotor position. In contrast, direct torque control (DTC) offers fast dynamic response
but much less dependency on rotor position and machine parameters. Direct torque
and flux control (DTFC) is one of the most accepted variants of DTC. It employs
space vector modulation (SVM) to produce very smooth torque and flux responses
while maintaining the main features of classical DTC. The overall block diagram of
DTFC is illustrated in Figure 3.1.

The structure of DTFC is very similar to that of FOC. It uses two PI regulators to
regulate torque and flux separately in the xy reference frame. The torque reference
comes from the external speed control output and the flux reference is from the MPTA
curve. The regulator output voltage Vx and Vy are converted into 𝛼𝛽 frame and
modulated by an SVPWM. It effectively reduced the torque ripple and making the
switching frequency constant than conventional DTC.

29
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

Figure 3.1 Structure of Sensor-less DTFC Control

For DTC and all its variants, accurate torque and flux feedback is critical for closed-
loop control. In addition, the rotor position and speed feedback can be readily
extracted from the observed rotor flux. The stator flux, torque, speed and angle
observer in Figure 3.1 takes output voltage and measured current as input, and outputs
all the four feedbacks. For simplicity, the observer is just called a stator flux observer.

This chapter focus on the stator flux observer evaluation. The observer evaluated in
this chapter will also be used in the remaining chapters of this thesis. Three stator flux
observers are studied together with DTFC control. The first is the pure voltage
integration method. The second is the combined voltage model and current model
only in αβ reference frame. The third is the combined voltage model and the current
model in both αβ and dq frame. Through comparative study, it is seen that the third
method yields the best performance and is being selected for the stator flux observer
for the subsequent chapters.

3.2 Sensor-less Direct Torque and Flux Control

The PMSM model in the xy frame in CHAPTER 2 is the basis of the DTFC control.
The equations (2.18)-(2.20) are listed here for convenience.

𝑑𝜆𝑠
𝑉𝑥 = 𝑅𝑠 𝑖𝑥 + (3.1)
𝑑𝑡
𝑑𝛿
𝑉𝑦 = 𝑅𝑠 𝑖𝑦 + 𝜔𝑒 𝜆𝑠 + 𝜆 (3.2)
𝑑𝑡 𝑠

30
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

3𝑃
𝑇𝑒 = 𝜆 𝜆 𝑠𝑖𝑛𝛿 (3.3)
2𝐿𝑠 𝑠 𝑝𝑚

From (3.1), neglecting the resistor voltage drop, the stator flux could be directly
regulated by x-axis voltage Vx, as given by

𝑑𝜆𝑠
𝑉𝑥 ≈ (3.4)
𝑑𝑡

Differentiating (3.3) with time, and assumes that the stator flux amplitude is being
tightly regulated, the change of torque is proportional to the torque angle change as
shown in (3.5):

𝑑𝑇𝑒 3𝑃 𝑑𝛿
= 𝜆𝑠 𝜆𝑝𝑚 𝑐𝑜𝑠𝛿 (3.5)
𝑑𝑡 2𝐿𝑠 𝑑𝑡

The assumption is usually true because DTC gives preference to the flux control.

Neglecting the resistance drop in (3.2) and integrate (3.5) into (3.2), the y-component
voltage becomes:

1 𝑑𝑇𝑒
𝑉𝑦 ≈ + 𝜔𝑒 𝜆𝑠 (3.6)
𝐾 𝑑𝑡

3𝑃
where 𝐾 = 2𝐿 𝜆𝑝𝑚 cos 𝛿 .
𝑠

From the above analysis, the stator flux and torque could be regulated by two PI
regulators separately as shown in Figure 3.2 and Figure 3.3 respectively. 𝑅𝑠 𝑖𝑥 and
𝑅𝑠 𝑖𝑦 could be treated as a disturbance and can be compensated by the PI regulator.
𝜔𝑒 𝜆𝑠 can also be treated as disturbance, it could be added in as feedforward term for
faster response.

The output of the two PI controllers Vx and Vy are then transformed back to the 𝛼𝛽
reference frame by (4.5) and modulated by SVPWM.

31
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

Figure 3.2 Stator Flux PI Controller using Vx

Figure 3.3 Torque PI Controller using Vx

𝑉𝛼 𝑐𝑜𝑠𝜃𝑠 −𝑠𝑖𝑛𝜃𝑠 𝑉𝑥
[𝑉 ] = [ ][ ] (3.7)
𝛽 𝑠𝑖𝑛𝜃𝑠 𝑐𝑜𝑠𝜃𝑠 𝑉𝑦

In (3.7), the stator flux angle could be calculated by the estimated stator flux observer:

𝜆̂𝑠𝛽
θ̂𝑠 = 𝑡𝑎𝑛−1 ( ) (3.8)
𝜆̂𝑠𝛼
The stator flux observer is then being covered by the next section.

3.3 Stator flux observer

Stator flux, angle, electrical torque, and rotor speeds are the key feedbacks for Direct
Torque Control and motor drive speed control. If stator flux is estimated, the rotor
flux, torque, speed, and rotor angle can be calculated from it together with the motor
current and parameters, as given by (3.9) to (3.14).

32
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

3
𝑇̂𝑒 = 𝑃(𝜆̂𝑠𝛼 𝑖𝑠𝛽 − 𝜆̂𝑠𝛽 𝑖𝑠𝛼 ) (3.9)
2
2 2
𝜆̂𝑠 = √𝜆̂𝑠𝛼 + 𝜆̂𝑠𝛽 (3.10)

𝜆̂𝑟𝛼 = 𝜆̂𝑠𝛼 − 𝐿𝑠 𝑖𝑠𝛼 (3.11)

𝜆̂𝑟𝛽 = 𝜆̂𝑠𝛽 − 𝐿𝑠 𝑖𝑠𝛽 (3.12)
𝜆 ̂
𝜃̂𝑟𝑒 = 𝑡𝑎𝑛−1 (̂𝑟𝛽 ) (3.13)
𝜆𝑟𝛼

𝑑
𝜔
̂𝑟𝑒 = 𝜃̂ (3.14)
𝑑𝑡 𝑟𝑒

Now the key is to estimate the stator flux. Various stator flux observers are available
in the literature. For this thesis study, the three most common stator flux observes are
selected for evaluation and they are introduced below.

3.3.1 Stator Flux Observers for Evaluation Study

A. Voltage Model (Observer 1)

From the SPMSM model in (2.5), the stator flux vector can be calculated by
integration using (3.15), The voltage integration based method in (3.15) is known as
voltage model.

𝝀̂𝒔 = ∫(𝒗𝒔 − 𝑅𝑠 𝒊𝒔 )𝑑𝑡 + 𝜆̂𝑠0 (3.15)

where 𝜆̂𝑠0 is the initial stator flux position (i.e. initial rotor flux position as the initial
current is zero).

In practical implementation, low pass filter with a very low cut off frequency 𝜔𝑐 is
usually employed to avoid pure integrator saturation, as given by (3.16).

𝜔𝑐
𝝀̂𝒔 = (𝒖𝒔 − 𝑅𝑠 𝒊𝒔 ) + 𝜆̂𝑠0 (3.16)
𝑠 + 𝜔𝑐

The resultant block diagram for the voltage model is depicted in Figure 3.4:

33
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

Figure 3.4 Structure of stator flux observer 1

B. Voltage Model in αβ frame with PI compensator (Observer 2)

The pure voltage model suffers from integration drifting problem due to measurement
noise and DC-offset, the original stator flux estimator in (3.15) is added with a
compensation voltage Vcomp:

𝝀̂𝒔 = ∫(𝒗𝒔 − 𝑅𝑠 𝒊𝒔 + 𝑽𝒄𝒐𝒎𝒑 )𝑑𝑡 + 𝜆̂𝑠0 (3.17)

Vcomp is used to compensate for various errors in the stator flux estimation, such as
inverter non-linearity, dead time, stator resistance small variations, integration DC-
offset, and measurement noise.

𝐾𝑖
𝑽𝒄𝒐𝒎𝒑 = (𝐾𝑝 + ) (𝒊𝒔 − 𝒊̂𝒔 ) (3.18)
𝑠

The current again is estimated from the estimated stator flux using (3.19)

𝒊̂𝒔 = (𝝀̂𝒔 − 𝝀̂𝒓 )/𝐿𝑠 (3.19)

where the rotor flux estimation is reconstructed from the estimated rotor angle 𝜃̂𝑟𝑒 :

cos (𝜃̂𝑟𝑒 )
𝝀̂𝒓 = 𝜆𝑝𝑚 [ ] (3.20)
sin (𝜃̂𝑟𝑒 )

The structure of this observer is depicted in Figure 3.5, from which it is seen that only
motor input voltage 𝒗𝜶𝜷 and output current 𝒊𝜶𝜷 in the 𝛼𝛽 frame is required. There’s

34
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

no involvement in speed adaptation and coordinates transformation. It’s a very simple

structure.

It’s worth pointing out that, at very low speed, the stator voltage is very small and the
accuracy of the stator flux estimation is limited by the stator resistance variation,
inverter nonlinearities, and measurement noise. An Rs estimation method together
with this observer is introduced in [18]. For the application of the flywheel, the
machine usually operates not at very low speed, thus the resistance change is not
considered in this thesis.

Figure 3.5 Structure of stator flux observer 2

C. Voltage and Current Model in dq frame with PI compensator (Observer 3)

This type of observer can be found in [72] and [28], it is also known as Gopinath flux
observer. The block diagram for this observer can be seen in Figure 3.6.

Figure 3.6 Stator Flux Observer 3

35
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

Comparing Figure 3.5 with Figure 3.6, it is seen that the two observers share a similar
structure, the only difference is that the current model in the left side is calculated in
the dq frame, based on the dq frame flux model in (2.15). Or equivalently, the current
observer is derived from the stator flux model in dq frame as shown in (3.21):

𝒊̂𝒔 = T(𝜃𝑟𝑒 )−1 (T(𝜃𝑒 )𝝀̂𝒔_𝜶𝜷 − 𝝀̂𝒓_𝒅𝒒 )/𝐿𝑑𝑞 (3.21)

where T(𝜃𝑒 ) and T(𝜃𝑟𝑒 )−1 are Park and inverse-Park transformation as given in
Appendix C, and 𝝀̂𝒓_𝒅𝒒 is the vector [𝜆𝑝𝑚 ;0] in dq frame.

The transformation and inverse transformation lead to slightly higher complexity than
the second observer which is only in a stationary frame. However, this transformation
makes both the stator flux and observed angle being involved in the current model of
stator flux, whereas the second observer only used the observed angle.

3.3.2 Explanation with characteristic function

The stator flux observer 3 is a combination of both voltage model which is good at
high speed and a current model that is good at low speed. The two models compensate
each other and able to achieve a wide range of operations. There is numerous
theoretical analysis in the literature for observer 3, such as in [61] and [60]. This
thesis took a slightly different approach. Since observer 3 and observer 2 share a
similar structure, observer 2 is used for the theoretical analysis. Observer 2 could also
be viewed in terms of combined observer theory.

If (3.20) is treated as 𝝀̂𝒓_𝒄𝒎 , the diagram in Figure 3.5 could be redrawn as shown in
Figure 3.7.

The stator flux could be expressed as (3.22):

1
𝝀̂𝒔 = 𝝀̂𝒔_𝒗𝒎 + 𝑽𝒄𝒐𝒎 (3.22)
𝑠

And the compensation voltage is written as (3.23):

36
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

𝑘𝑝 𝑠 + 𝑘𝑖
𝑽𝒄𝒐𝒎 = (𝒊𝒔 − 𝒊̂𝒔 ) (3.23)
𝑠

Figure 3.7 Redrawn stator flux observer showing voltage model and current model

In which the estimated current is derived from:

𝒊̂ 𝒔 = (𝝀̂𝒔 − 𝝀̂𝒓_𝒄𝒎 )/𝐿𝑠 (3.24)

̂
𝝀̂𝒓_𝒄𝒎 = 𝜆𝑝𝑚 𝒆𝒋𝜃𝑒 (3.25)

Substitute (3.24) and (3.25) into (3.23), 𝑽𝒄𝒐𝒎 becomes

𝑘𝑝 𝑠 + 𝑘𝑖 𝑳𝒔 𝒊𝒔 − (𝝀̂𝒔 − 𝝀̂𝒓𝒄𝒎 )
𝑽𝒄𝒐𝒎 = ∙ (3.26)
𝑠 𝐿𝑠

Let 𝝀̂𝒔𝒄𝒎 = 𝝀̂𝒓𝒄𝒎 + 𝐿𝑠 𝒊𝒔 (3.26) can be rewritten as:

𝑘𝑝 𝑠 + 𝑘𝑖 𝝀̂𝒔_𝒄𝒎 − 𝝀̂𝒔
𝑽𝒄𝒐𝒎 = ∙ (3.27)
𝑠 𝐿𝑠

Finally, substituting (3.27) into (3.23), and with some arrangement, it turns into a
combined voltage model and current model in (3.28):

37
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

1
𝑠2 (𝑘𝑝 𝑠 + 𝑘𝑖 )
̂𝝀𝒔 = ̂𝝀𝒔_𝒗𝒎 + 𝐿𝑠
𝝀̂𝒔_𝒄𝒎 (3.28)
2 1 2 1
𝑠 + 𝐿 (𝑘𝑝 𝑠 + 𝑘𝑖 ) 𝑠 + 𝐿 (𝑘𝑝 𝑠 + 𝑘𝑖 )
𝑠 𝑠

𝝀̂𝒔 = 𝐹(𝑠)𝝀̂𝒔𝒗𝒎 + (1 − 𝐹(𝑠))𝝀̂𝒔_𝒄𝒎 (3.29)

The function F(s) is called characteristics equation, given by:

𝑠2 𝑠2
𝐹(𝑠) = = 2 (3.30)
1
𝑠 2 + 𝐿 (𝑘𝑝 𝑠 + 𝑘𝑖 ) 𝑠 + 𝐾𝑝 𝑠 + 𝐾𝑖
𝑠

Where 𝐾𝑝 and 𝐾𝑖 are related to the PI parameters according to (3.31):

𝑘𝑝
𝐾𝑝 =
𝐿𝑠
(3.31)
𝑘𝑖
𝐾𝑖 =
𝐿𝑠

To set the 𝐾𝑝 and 𝐾𝑖 parameters in the characteristic function, one possible way is to
define the transition regions, as given by:

𝐾𝑝 = 𝜔1 + 𝜔2
(3.32)
𝐾𝑖 = 𝜔1 𝜔2

where 𝜔1 < 𝜔2 are motor speed. 𝜔1 and 𝜔2 determines the transition region
between the current model and voltage model. Current model dominates when speed
below 𝜔1 and voltage model dominates when speed above 𝜔2 . For this method, it
can achieve good results if set 𝜔1 ≤ 0.1 and set 𝜔2 to a value that the voltage mode
model works reliably.

Another possible way to set the PI parameters is based on the following:

𝐾𝑝 = 2𝜉𝜔0 (3.33)

38
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

𝐾𝑖 = 𝜔0 2

where 𝜉 is damping ratio and 𝜔0 is crossover frequency. According to control theory,

it’s a good practice to set 𝜉=1. The crossover frequency should be carefully selected
according to the specific applications.

For example, if the transition region-based method is called method 1, and the
crossover frequency-based method is called method 2, a bode plot for comparison
can be plotted for voltage mode characteristic function as shown in Figure 3.5(a), and
for current mode characteristic function as shown in Figure 3.5(b), respectively.

(a)

(b)

Figure 3.8 Bode plot of characteristic function for (a) voltage model F(s) and (b)
current model 1-F(s) for both method 1 and method 2

The bode plot is based on the following values:

• Method 1: 𝜔1 = 0.1 rad/s, 𝜔2 = 10 rad/s

39
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

• Method 2: 𝜉 = 1 and 𝜔0 = 10 rad/s

Based on the above value, it is seen that the transition region-based method yields
better results in terms of earlier entering pure voltage mode.

3.4 Simulation Study

The DTFC drive model for the SPMSM control system is built-in Matlab/Simulink
environment as shown in Appendix B. All the three stator flux observers are included
in the model. Simulation step length is set at 0.5 us, while the control loop and PWM
switching time is set at 10 kHz. The machine parameters listed in Table 3.1 are used
for simulation. The observer sampling time is the same as the torque and flux control
loop. A low pass filter with a 2ms time constant is used for speed filtering.

Table 3.1 SPMSM parameters for simulation

Rated torque, T 2.85 Nm

Magnet flux linkage, λpm 0.1905 V⸱s/rad
Phase resistance, Rs 4.67 Ω
dq-axes inductance, Ld = Lq 26.8 mH
Number of pole pairs, P 2
DC bus voltage 150 V
Rated phase current 5A
No-load speed 1950 rpm
Inertial J 1e-4 kg⸱m2
Viscous coefficient B 1e-4 Nm⸱s/rad

To compare the performance of stator flux observers 1 to 3, the exact same simulation
conditions are used. The Flux and Torque PI controllers for the DTFC are listed in
Table 3.3. They are designed according to the closed-loop response from Figure 3.2
and Figure 3.3, with tighter control for the Flux. The Observer 1 only has a single
cut-off frequency 𝜔𝑐 = 0.02, which cannot be too high. Observer 2 and 3 have the

40
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

same PI parameters. The values are roughly derived from 𝜔1 = 0.05 and 𝜔2 =37 as
given by (3.32) and (3.33).

The simulations use step torque response. A 0.2 Nm load torque and reference torque
are applied to the model from the beginning and the reference torque steps from 0.2
Nm to 0.3 Nm at 0.1 s. This torque drives the flywheel motor from zero speed to max
speed which is around 1920rpm in the 0.2Nm load condition. The estimated speed
from stator flux observers is compared with the actual speed (or idea speed) given by
the model output.

Table 3.2 Control parameter settings for simulation

Observers Observer Flux Controller PI Torque Controller

Parameters Parameters PI parameters
Observer 1 𝜔𝑐 = 0.02
Observer 2 𝑘𝑝_𝑜𝑏𝑠 = 1
𝑘𝑝_𝜆 = 460 𝑘𝑝_𝑇 = 80
𝑘𝑖_𝑜𝑏𝑠 = 0.05
𝑘𝑖_𝜆 = 6 𝑘𝑖_𝑇 = 2
Observer 3 𝑘𝑝_𝑜𝑏𝑠 = 1
𝑘𝑖_𝑜𝑏𝑠 = 0.05

Figure 3.9 presents the simulation results using Observer 1. It is seen that there are
large ripples on all of the torque, flux, and speed, especially when the rpm goes higher.
Note that when the rpm reaches base speed around 1920rpm, there are large
distortions on flux and torque, this is normal phenonium and can also be observed on
other observers.

The simulation results using Observer 2 and Observer 3 are presented in Figure 3.10
and Figure 3.11, respectively. Comparing the first subplot of Figure 3.10 (a) and
Figure 3.11 (a), it is seen that Observer 3 provides a straighter ramp up and reaches
the base speed faster. Comparing the flux and flux error, it is seen that Observer 3
performs much better than Observer 2 as well.

Although Observer 2 is not giving as good result as Observer 3, it still performs better
than Observer 1, which can be observed by comparing Figure 3.10 with Figure 3.9.

41
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

Speed (rpm)
Flux Linkage(Vs)
Torque (Nm)

Time (s) Time (s)

(a) (b)

Figure 3.9 Simulation of SPMSM Sensor-less DTFC Torque Step with Flux
Observer 1. (a) Speed, Flux and Torque; (b) Error of Speed, Flux and Torque
Speed (rpm)
Flux Linkage(Vs)
Torque (Nm)

Time (s) Time (s)

(a) (b)
Figure 3.10 Simulation of SPMSM Sensor-less DTFC Torque Step with Flux
Observer 2. (a) Speed, Flux and Torque; (b) Error of Speed, Flux and Torque

42
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

Figure 3.10 and Figure 3.11 also demonstrate excellent torque and flux tracking
capability for the DTFC control below base speed. As explained above, the large
oscillation toward the end of the simulation is due to the motor speed reaches base
speed and torque and flux regulation are limited. They will be addressed by the Field
Weakening method in Chapter 4.

Through simulation, this thesis concludes that Observer 3 is the best performer and
thus is selected for all the subsequent chapters, despite it’s slightly more complex
than the other two Observers.
Speed (rpm)
Flux Linkage(Vs)
Torque (Nm)

Time (s) Time (s)

(a) (b)

Figure 3.11 Simulation of SPMSM Sensor-less DTFC Torque Step with Flux
Observer 3. (a) Speed, Flux and Torque; (b) Error of Speed, Flux and Torque

3.5 Conclusions

This chapter introduced the direct torque and flux control (DTFC) for the flywheel
SPMSM. As torque and flux observers are key to the success of DTFC, three different
stator flux observers are reviewed and evaluated. Through a simulation study, it is
seen that stator flux observer 3 – the combined voltage and current model in dq frame

43
CHAPTER 3 Sensor-less Flux and Torque Observer Study for the Direct Torque and
Flux Controlled Surface Mounted Permanent Magnet Flywheel Motor

with PI compensator gives the best performance.

In the next chapter, an improved field weakening method with DTFC of PMSM will
be proposed. The simulation is built upon the dual airgap AFPM machine model and
the stator flux observer 3 will be used by the rest of the chapters.

44
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

CHAPTER 4 A new Field Weakening Control approach

for a Direct Torque and Flux Controlled Surface Permanent
Magnet Motor

4.1 Introduction

In Chapter 3, the sensor-less direct torque and flux control (DTFC) of SPMSM
flywheel was introduced. This chapter is still based on the sensor-less DTFC control,
with the focus moving to the flux weakening region.

To guarantee excellent performance of DTFC in high-speed region, the field-

weakening (FW) operation is indispensable. In theory, FW operation starts when the
motor speed exceeds its base speed, before which the maximum torque per ampere
(MTPA) trajectory dominates to minimize the copper loss [6]. When the motor speed
is high, change-over from MTPA to FW operation is necessary since the back
electromotive force (emf) of the motor goes up with the increase of speed. This
indicates that a smaller portion of DC-link voltage is available for torque regulation.
Plenty of research papers have been published to investigate the FW operations of
direct torque controlled electric machines [73-79]. Unfortunately, many of them focus
on the field-weakening operation of direct torque controlled induction motors and
thus are not readily applicable to the PMSMs [73, 77, 79]. In [74], a new FW scheme
is proposed to extend the virtual signal injection control of IPMSM into the high-
speed region. Excellent dynamic response and good parameter robustness are attained
by the proposed scheme. Nonetheless, the method is designed for field-oriented
controlled IPMSM drives. Hence, it is not suitable for DTC based electrical drives. A
linear FW control of IPMSM is initialized in [75] to provide both torque and current
regulation in a wide speed range. It simplifies the calculation of working points for
FW operation to a large extent while obtaining a smooth speed transition in FW areas.
Nevertheless, detailed machine model and d,q-axes currents are used in the method,
leading to high parameter and rotor angular position dependency. In [78], a novel
torque control method of IPMSM in the FW region is proposed to realize improved

45
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

DC-link voltage utilization. It makes use of the voltage angle control with d-axis
current feedback. Due to the employment of rotor reference frame coordinate
transformation, voltage vector selection scheme and parameter dependent
voltage/current vector calculations, this method is associated with high computational
complexity and reduced parameter robustness.

The FW schemes for direct torque and flux controlled PMSM drives in [45] and [47]
compute the base speed and FW trajectory are purely based on the machine model.
Consequently, they are heavily dependent on machine parameters. Recently, a self-
learning direct flux vector control algorithm is proposed for PMSM in both constant
torque and FW regions [76]. It is reported to be insensitive to machine parameter
variations, producing smooth torque response and achieving a good dynamic
response. Despite its advantageous features, this method needs a rotor reference
frame coordinate transformation and a complex self-learning controller. This
increases its dependence on the knowledge of accurate rotor angular position as well
as imposing a higher computational burden.

To extend the Constant Power and Speed Range (CPSR) and simultaneously enhance
the parameter robustness of DTFC based PMSM drives, this thesis proposes a new
FW algorithm that dynamically adjusts reference flux and torque values, using
SPMSM for case study. The proposed algorithm employs a simple PI controller and
thus features at very low computational complexity. Moreover, the entire control
system incorporating the proposed FW algorithm is implemented in the stationary
reference frame, which eliminates coordinate transformation. This reduces the rotor
position dependency of DTFC. The proposed methods could be extended to IPMSM
as well.

The rest of this chapter is organized as follows: The trajectories of MTPA and
conventional FW of SPMSM are given in Section 4.2. Then, the proposed DTFC
based FW algorithm is presented in Section 4.3. To verify the validity of the proposed
algorithm, simulation results are provided in Section 4.4. Finally, a conclusion is
drawn in Section 4.5.

46
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

4.2 MTPA and Field Weakening Trajectories for SPMSM

In high-performance AC machine drives, two trajectories are commonly used to

achieve high efficiency in a wide speed range. They are known as the MTPA and FW
trajectories.

4.3.1 Maximum Torque Per Ampere Trajectory

For SPMSM drives, the MPTA is simply id=0. However, for DTFC, the control
variable is torque and flux, which requires the MTPA trajectory to be expressed as
𝜆𝑠 − 𝑇𝑒 curve.

By defining 𝛽 as the angle between the stator current vector 𝒊𝒔 and the q-axis of rotor
reference frame, dq-axes currents can be expressed as (4.1), where 𝐼𝑠 represents the
amplitude of stator current. In addition, the electromagnetic torque of a PMSM is
given in (4.2) with notation 𝑃 denoting the number of pole pairs.

𝑖𝑑 = −𝐼𝑠 sin 𝛽
(4.1)
𝑖𝑞 = 𝐼𝑠 cos 𝛽
3
𝑇𝑒 = 𝑃𝜆 𝑖 (4.2)
2 𝑝𝑚 𝑞
3
𝑇𝑒 = 𝑃𝜆𝑝𝑚 𝐼𝑠 cos 𝛽 (4.3)
2

Substituting (4.1) into (4.2), the motor torque can be re-derived as (4.3) with respect
to 𝐼𝑠 . From (4.3), it is observable that the maximum torque can be achieved by
controlling 𝐼𝑠 and 𝛽 in an appropriate way. If the derivative of (4.3) is made equal to
zero, the condition for MTPA can be developed. This condition is shown in (4.4). In
principle, the MTPA trajectory can be realized if the reference trajectory of the flux
controller is following the relation in (4.4). The transformation of relation (4.4) to the
torque-flux plane is usually implemented numerically by using a look-up-table that
provides point-to-point mapping of torque and stator flux amplitude.

2
2𝐿𝑠 𝑇𝑟
𝜆𝑠 = √ ( ) + 𝜆2𝑝𝑚 (4.4)
3𝑃𝜆𝑝𝑚

47
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

where Tr is torque reference which is the output of the speed PI controller.

4.2.2 Field Weakening Trajectory

When the rotor speed of SPMSM goes above base speed, the MTPA trajectory will
not be traversed. Consequently, the FW trajectory must be pursued to avoid the
demand for reference phase voltage that exceeds the maximum inverter limit. It is
pointed out that the FW trajectory can be briefly divided into two regions, i.e. Region
I and Region II dominated by the current limit and voltage limit conditions,
respectively.

• Region I: In this region, the transition from MTPA to FW begins when the
demanded torque starts exceeding the torque limit defined by the maximum
available DC-link voltage. This transition is autonomously and smoothly
realized by comparing the reference stator flux governed by (4.5) with its
counterpart determined by (4.4) and selecting the minimum value. The notation
𝑉𝑚𝑎𝑥 represents the maximum phase voltage of SPMSM confined by the
available DC-link voltage.

𝑉𝑚𝑎𝑥
|𝜆∗𝑠 | = (4.5)
𝜔𝑟𝑒

In general, when the SVM of DTFC operates in linear modulation region, 𝑉𝑚𝑎𝑥
is given by (4.6); whereas it is increased to (4.7) in the average sense when the
SVM of DTFC operates in the six-step over-modulation region, which is pushing
the utilization of DC-link voltage to its upper limit. Here, 𝐼𝑚𝑎𝑥 denotes the
maximum phase current. By comparing (4.6) and (4.7), it is obvious that the
latter increases the utilization rate of DC-link voltage, resulting in a larger
average driving torque and subsequently a higher speed limit for SPMSM. Hence,
(4.7) is desirable from the perspectives of producing faster dynamic responses
and enhanced energy storage capacity for FESS. It needs to be pointed out that,
in order to produce an average phase voltage of (4.7), equation (4.8) is usually
used. The detailed reason will be explained in the next section.

48
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

𝑉𝑑𝑐
𝑉𝑚𝑎𝑥 = − 𝐼𝑚𝑎𝑥 𝑅𝑠 (4.6)
√3
2𝑉𝑑𝑐
𝑉𝑚𝑎𝑥 = − 𝐼𝑚𝑎𝑥 𝑅𝑠 (4.7)
𝜋
2
𝑉𝑚𝑎𝑥 = 𝑉𝑑𝑐 − 𝐼𝑚𝑎𝑥 𝑅𝑠 (4.8)
3

• Region II: When the SPMSM reaches very high speed, it enters the deep FW in
Region II. In this region, a new torque limit in (4.9) is imposed by the maximum
allowable motor/inverter current. Physically, by steering the torque along its
limit in (4.9), the torque angle is tracking its limit of 90𝑜 for SPMSM. If
parameter variation occurs, such as the change in value of 𝐿𝑠 caused by magnetic
saturation. It can result in an over-estimated torque limit. In this case, instability
of drives can be triggered because the actual torque is demanded to track a
reference value that is beyond its physical limit in Region II. To ensure the
stability of SPMSM drives in the high-speed region, which is of vital importance
for FESS, the improvement of parameter robustness of FW is necessary.

2
3 𝜆2𝑠 − 𝜆2𝑝𝑚 − 𝐿2𝑠 𝐼𝑚𝑎𝑥
2
𝑇𝑒(𝑙𝑖𝑚𝑖𝑡) = 𝑃𝜆𝑝𝑚 √𝐼𝑚𝑎𝑥
2 −( ) (4.9)
2 2𝜆𝑝𝑚 𝐿𝑠

From the description above, maximized usage of DC-link voltage and parameter
robustness of FW are two main concerns for FESS. The realization of these two goals
can contribute to significantly improve the performance of FESS. This motivates the
proposal of the new FW algorithm in the next section.

4.3 Proposed Field Weakening Algorithm

Based on the previous discussion, the robustness of conventional DTFC based FW is

unsatisfying. To overcome this issue, a new FW algorithm is proposed in this section.
The proposed FW algorithm contributes to deliver the maximum allowable speed for
SPMSM in FESS and control robustness improvement via two simple PI controllers.
They can be defined as the reference flux adjustment PI regulator and the torque limit
adjustment PI regulator. Their mechanisms are explained in sub-section A and B.

49
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

A. Reference Flux Adjustment Scheme

As stated above, the utilization of DC-link voltage must be maximized to produce the
maximum phase voltage. In this manner, the highest possible speed, as well as the
fastest dynamic response, can be obtained in SPMSM drives. Nonetheless, the direct
substitution of (4.7) into (4.5) will cause control failure since (4.7) is only achievable
in the average sense instead of the instantaneous sense. To be specific, (4.7) cannot
be substituted into (4.5) for instantaneous reference flux computation. Instead, the
value of 𝑉𝑚𝑎𝑥 needs to be dynamically adjusted to make it approximately equivalent
to (4.7) on average. Such an adjustment is realized by the proposed algorithm that is
conceptually shown in Figure 4.1. It is visible that equation (4.8) is used in both the
input of adjustment PI controller and the base of flux adjustment in (4.10).

Figure 4.1 Block diagram of the proposed reference flux adjustment scheme

Physically, (4.8) represents the maximum transient phase voltage that is achievable
only at the six vertices of the hexagonal voltage limit in voltage vector space.
Therefore, it can be used as the base of reference flux computation. Decrement (i.e.
negative ∆|𝜆∗𝑠 |) will be made on this base when the maximum available transient
phase voltage is exceeded as shown in (4.10) and in the right part of Figure 4.1.
Similarly, an increment will be added to this base when more phase voltage can be
used. Hence, the final reference stator flux |𝜆∗𝑠 | 𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 is dynamically adjusted. To
find out the actual maximum allowable phase voltage, the comparison of (4.8) and
the output of the PI torque regulator, i.e. 𝑣𝑦 , is performed as shown in the left part of
Figure 4.1. This comparison checks if the torque PI controller output 𝑣𝑦 saturates
caused by an unrealizably high reference flux. Then, based on the level of saturation,
the reference flux will be rapidly reduced until torque control is de-saturated. This is

50
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

the point that the maximum allowable phase voltage is used.

𝑉𝑚𝑎𝑥
|𝜆∗𝑠 | 𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 = + ∆|𝜆∗𝑠 | (4.10)
𝜔𝑟𝑒

B. Torque Limit Adjustment Scheme

It is mentioned in Section 4.2 that the torque limit of Region II in FW operation is

sensitive to machine parameter variations. Therefore, to enhance its robustness
against parameter changes, an auto-adjustment algorithm for torque limit 𝑇𝑒(𝑙𝑖𝑚𝑖𝑡) is
proposed as shown in Figure 4.2.

Figure 4.2 Block diagram of the proposed torque limit adjustment scheme

The fundamental logic of such an adjustment scheme is to properly reduce the torque
limit 𝑇𝑒(𝑙𝑖𝑚𝑖𝑡) when it exceeds the corresponding physical limit caused by variations
in machine parameters. Over-estimated 𝑇𝑒(𝑙𝑖𝑚𝑖𝑡) is compensated by torque angle 𝛿
control in Figure 4.2. In principle, by keeping 𝛿 ≤ 90𝑜 , the stability of FW in Region
II is guaranteed. Moreover, 𝛿 = 𝜃𝑠 − 𝜃̂𝑟𝑒 (𝜃𝑠 and 𝜃̂𝑟𝑒 denotes the stator flux angle
and estimated rotor electrical angle) is independent of machine parameters. Thus, the
error 90𝑜 − 𝛿 can be used directly to compensate for inaccuracy of 𝑇𝑒(𝑙𝑖𝑚𝑖𝑡) . This is
done through the usage of Figure 4.2 and (4.11).

𝑇𝑒(𝑙𝑖𝑚𝑖𝑡)_𝑎𝑑𝑗 = 𝑇𝑒(𝑙𝑖𝑚𝑖𝑡) + Δ𝑇𝑒(𝑙𝑖𝑚𝑖𝑡) (4.11)

By incorporating the proposed FW algorithm with torque limit adjustment into

DTFC, the overall control system block diagram can be obtained as shown in Figure
4.3.

51
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled SPMSM

Figure 4.3 Block diagram of DTFC incorporating the proposed field-weakening and torque limit adjustment algorithm.

52
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

4.4 Simulation Study

To verify the effectiveness of the proposed robust DTFC based FW algorithm,

simulations are carried out in MATLAB/SIMULINK environment as presented in
Appendix B. The parameters of dual airgap AFPM SPMSM in Chapter 3 are used in
the simulations as shown in Table 3.1. Noticeably, the sampling frequency of the
controller is set to be 10 kHz. Simulation results are presented and compared with
their counterparts obtained using the conventional DTFC based FW.

Firstly, simulations of step speed response under both the DTFC with conventional
FW approach and the proposed FW algorithm are carried. They are shown in Figure
4.4, Figure 4.4 for the conventional approach and in Figure 4.5 for the proposed
approach.

4000
Speed (rpm)

Speed Cmd
3000 Speed Estimated
2000
1000
0
0.2 0.4 0.6 0.8 1
0.25
Flux Linkage
(Vs)

0.2

0.15
0.2 0.4 0.6 0.8 1
2
Torque (Nm)

0.2 0.4 0.6 0.8 1

Time (s)

Figure 4.4 Conventional FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 4000rpm using (4.5) and (4.7)

Figure 4.4 presents the situation of acceleration failure because of the substitution of
(4.7) into (4.5) in conventional FW. The direct employment of (4.7) results in a
reference flux that is too high for FW operation. Subsequently, improper FW

53
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

trajectory is utilized and the acceleration fails. Instead, when (4.6) is substituted into
(4.5) in conventional FW, the applied voltage vectors are confined in linear
modulation range of SVM. This gives a conservative FW trajectory, but the
acceleration from 600rpm to 4000rpm is successful as depicted by subplot 1 of Figure
4.4.

4000
Speed (rpm)

3000
2000
0.14s
1000
0
0.2 0.4 0.6 0.8 1

0.2
Flux Linkage
(Vs)

0.1

0
0.2 0.4 0.6 0.8 1
2
Torque (Nm)

0
0.2 0.4 0.6 0.8 1
Time (s)

Figure 4.5 Conventional FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 4000rpm with the conventional FW algorithm using (4.5) and (4.6)

It is noteworthy that, despite the success in acceleration, the speed response is not
smooth. This is caused by the fluctuations in flux and torque responses as illustrated
in subplots 2 and 3 of Figure 4.4, respectively. The fundamental reason for such a
phenomenon is the lower average phase voltage governed by (4.6), which produces
lower flux and torque development capability. This gives fluctuations in flux and
torque when DTFC is tuned to provide fast acceleration responses demanding large
average torque. Although the fine-tuning of DTFC can eliminate these fluctuations,
it will simultaneously slow down the acceleration of SPMSM. In comparison, the
proposed method obtains faster acceleration response with smooth flux and torque
responses under the same condition as shown in Figure 4.6. By comparing subplot 1
of Figure 4.4 and Figure 4.6, it is observed that the acceleration is nearly 20% faster
using the proposed algorithm. Furthermore, from subplot 3 in both of the figures, it

54
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

is seen that the average torque output of the proposed algorithm is larger and
smoother.

4000
Speed (rpm) 3000
2000
0.11s
1000
0.2 0.4 0.6 0.8 1

0.2
Flux Linkage
(Vs)

0.1

0
0.2 0.4 0.6 0.8 1
2
Torque (Nm)

0
0.2 0.4 0.6 0.8 1
Time (s)

Figure 4.6 Proposed FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 4000rpm using (4.10) and (4.8)

Secondly, owing to its capability of applying the maximum phase voltage, the
proposed algorithm is able to deliver the maximum shaft power and subsequently
drives the SPMSM to reach the highest speed. This is demonstrated in Figure 4.7 and
Figure 4.8. By comparing the subplot 1 of these two figures, it is clearly seen that the
maximum speed obtained by the conventional DTFC based FW is below 5000rpm.
However, the proposed FW algorithm can drive the SPMSM to reach 5500rpm,
demonstrating its superior performance. Additionally, the speed response of the
proposed FW is faster because of its higher average output torque seen from the
comparison of subplot 3 in Figure 4.7and Figure 4.8.

55
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

6000

Speed (rpm)
5000
4000
3000
2000
1000
0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

Flux Linkage
0.2

(Vs)
0.1

0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
2
Torque (Nm)

0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Time (s)

Figure 4.7 Conventional FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 500rpm with the conventional FW using (4.5) and (4.6)

6000
Speed (rpm)

5000
4000
3000
2000
1000
0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

0.2
Flux Linkage
(Vs)

0.1

0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
2
Torque (Nm)

0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Time (s)

Figure 4.8 Proposed FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 5500rpm using (4.10) and (4.8)

56
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

4000

Speed (rpm)
3000
2000
1000
0
0.5 1 1.5

Flux Linkage
0.2

(Vs) 0.1

0
0.5 1 1.5
2
Torque (Nm)

0
0.5 1 1.5
Time (s)

Figure 4.9 Conventional FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 4000rpm with +20% detuning in 𝐿𝑠

4000
Speed (rpm)

3000
2000
1000
0
0.5 1 1.5
Flux Linkage

0.2
(Vs)

0.1

0
0.5 1 1.5
2
Torque (Nm)

0
0.5 1 1.5
Time (s)

Figure 4.10 Proposed FW algorithm: Simulation of SPMSM step speed response from
600rpm to 4000rpm with +20% detuning in 𝐿𝑠

57
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

Another advantage of the proposed FW algorithm is its strong robustness against

machine parameter variations. This is verified through the robustness tests shown in
Figure 4.9 to Figure 4.12. On one hand, in Figure 4.9 and Figure 4.10, step speed test
from 600rpm to 4000rpm is carried out with +20% detuning in 𝐿𝑠 . It simulates the
effect of 𝐿𝑠 decrease caused by magnetic saturation, which results in detuning in the
value of 𝐿𝑠 used in the control algorithm . By comparing subplot 1 of the two figures,
it is seen that speed control fails in the conventional FW with detuned 𝐿𝑠 . The speed
of SPMSM is clamped and oscillates around 3000rpm due to the fact that + detuning
in 𝐿𝑠 leads to an over-estimated 𝑇𝑒(𝑙𝑖𝑚𝑖𝑡) , which is above the actual torque production
capability. As a result, the torque control collapses as shown in subplot 3 of Figure
4.9. This is the reason for its speed control failure. In contrast, the proposed FW
properly reduces the over-estimated 𝑇𝑒(𝑙𝑖𝑚𝑖𝑡) back to its actual physical limit through
the torque limit adjustment scheme. Consequently, successful torque and speed
control are maintained as depicted by subplots 3 and 1 in Figure 4.10.

On the other hand, step speed test from 600rpm to 4000rpm is carried out with -20%
detuning in 𝑅𝑠 is also carried out. The results are illustrated in Figure 4.11 and Figure
4.12. This test illustrates the situation of stator resistance 𝑅𝑠 increase caused by
temperature rising in practical FESS applications. Since the increase in 𝑅𝑠 is not
detected by the controller, the value of 𝑅𝑠 used in the algorithm is detuned. By
examining the speed, flux and torque responses in subplots 1, 2 and 3 of Figure 4.11,
it is observed that the SPMSM runs out of control with conventional FW due to the
detuning in 𝑅𝑠 . In addition, the speed estimation using the stator flux observer in
Chapter 3 is also greatly deteriorated when the entire drives system loses control. This
is catastrophic for the FESS. Unlikely, the excellent drives performance is maintained
by the proposed FW with the same -20% detuning in 𝑅𝑠 as illustrated by Figure 4.12.
This is achieved by the auto-adjustment scheme for reference flux computation.

58
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

6000

Speed (rpm)
4000

2,000

0
0.4 0.9 1.4 1.9 2.4

0.2

Flux Linkage
(Vs)
0.1

0
0.4 0.9 1.4 1.9 2.4
2
Torque (Nm)

0.4 0.9 1.4 1.9 2.4

Time (s)

Figure 4.11 Conventional FW algorithm: Simulation of SPMSM step speed

response from 600rpm to 4000rpm with -20% detuning in 𝑅𝑠

6000
Speed (rpm)

4000

2000

0
0.4 0.9 1.4 1.9 2.4

0.2
Flux Linkage
(Vs)

0.1

0
0.4 0.9 1.4 1.9 2.4
2
Torque (Nm)

0.4 0.9 1.4 1.9 2.4

Time (s)

Figure 4.12 Proposed FW algorithm: Simulation of SPMSM step speed response

from 600rpm to 4000rpm with -20% detuning in 𝑅𝑠

59
CHAPTER 4 A New Field Weakening Control Approach for a DTFC Controlled
SPMSM

4.5 Conclusion

This chapter proposes a new field-weakening algorithm for direct torque and flux
controlled (DTFC) surface-mounted permanent magnet synchronous machine drives.
It contributes to enhancing the parameter robustness of field-weakening operations
in high-speed drives. Moreover, it effectively extends the constant power speed range
of SPMSM drives, which is very important for a flywheel-based energy storage
system. Smooth torque and flux responses, as well as constant switching frequency,
are obtained due to the employment of space vector modulation in DTFC.

In the subsequent chapters, predictive torque and flux control methods will be
discussed, with a focus on torque ripple reduction. The FW methods proposed in this
chapter could be potentially extended into the predictive control based methods since
they all use flux and torque as a control reference.

60
CHAPTER 5 Deadbeat Predictive Torque Control with Space Vector Modulation

CHAPTER 5 Deadbeat Predictive Torque Control with

Space Vector Modulation

5.1 Introduction

In this chapter, in order to further diminish the ripples of torque and flux and
simultaneously reduce the computational burden while maintaining a constant
switching frequency, this chapter proposes a simplified PTC strategy for PMSM
drives employing SVM. In the proposed strategy, the reference VV is firstly
determined by the deadbeat technique in the stationary reference frame. Then, SVM
is directly utilized to synthesize the desired reference VV. Consequently, no iterative
calculation is needed and the reference VV is analytically synthesized rather than
being approximated. Due to the fact that one cycle implementation delay [80] is
compensated by PTC and three VVs are used in each sampling cycle, excellent torque
and flux regulation can be achieved. The block diagram of the proposed control
strategy is depicted in Figure 5.1.

Figure 5.1 Block diagram of the proposed simplified deadbeat PTC with SVM

5.2 Basic Principle of Torque Prediction

The mathematical model in 𝛼𝛽 frame (2.5) to (2.7) in Chapter 2 can be rewritten as:

61
CHAPTER 5 Deadbeat Predictive Torque Control with Space Vector Modulation

𝑑𝒊𝒔
𝐿𝑞 = −𝑅𝑠 𝒊𝒔 + 𝒋𝜔𝑟𝑒 (𝝀𝒔 − 𝐿𝑞 𝒊𝒔 ) + 𝒗𝒔 (5.1)
𝑑𝑡
𝑑𝝀𝒔
= 𝒗𝒔 − 𝑅𝑠 𝒊𝒔 (5.2)
𝑑𝑡
3
𝑇𝑒 = 𝑃(𝝀𝒔 × 𝒊𝒔 ) (5.3)
2

It is noted that both the stator flux and the electromagnetic torque can be observed
from measured stator currents and applied voltages. To avoid integral saturation
caused by sensor imperfections, low pass filters can be used to replace the integrator
in (5.2) [81, 82]. The design of flux and torque observer has been studied in Chapter
3 hence not repeated here.

By discretizing (5.1)-(5.3) via forward Euler method and re-writing them in scalar
forms, equations (5.4)-(5.6) can be derived.

𝑅𝑠
1− 𝑇 −𝜔𝑟𝑒 𝑇𝑠
𝑖𝑠𝛼 (𝑘 + 1) 𝐿𝑞 𝑠 𝑖𝑠𝛼 (𝑘)
[ ]= [ ]
𝑖𝑠𝛽 (𝑘 + 1) 𝑅𝑠 𝑖𝑠𝛽 (𝑘)
𝜔𝑟𝑒 𝑇𝑠 1− 𝑇
[ 𝐿𝑞 𝑠 ] (5.4)
𝜔𝑟𝑒
0 𝑇
𝐿𝑞 𝑠 𝜆𝑠𝛼 (𝑘) 𝑇𝑠 𝑣𝑠𝛼 (𝑘)
+ 𝜔𝑟𝑒 [ ]+ [ ]
𝜆𝑠𝛽 (𝑘) 𝐿𝑞 𝑣𝑠𝛽 (𝑘)
− 𝑇𝑠 0
[ 𝐿𝑞 ]
𝜆𝑠𝛼 (𝑘 + 1) 𝜆𝑠𝛼 (𝑘) 𝑣𝑠𝛼 (𝑘) 𝑖𝑠𝛼 (𝑘)
[ ]=[ ] + 𝑇𝑠 ([ ] − 𝑅𝑠 [ ]) (5.5)
𝜆𝑠𝛽 (𝑘 + 1) 𝜆𝑠𝛽 (𝑘) 𝑣𝑠𝛽 (𝑘) 𝑖𝑠𝛽 (𝑘)
3
𝑇𝑒 (𝑘 + 1) = 𝑃 (𝜆𝑠𝛼 (𝑘 + 1)𝑖𝑠𝛽 (𝑘 + 1) − 𝜆𝑠𝛽 (𝑘 + 1)𝑖𝑠𝛼 (𝑘 + 1)) (5.6)
2

where notations 𝑘 and 𝑇𝑠 represent the 𝑘 -th sampling instant and the controller
sampling period, respectively.

In real-time digital control, the presence of one-cycle delay usually leads to increased
torque and flux ripples [42]. To compensate for this effect, the torque and flux should
be predicted for one more step, i.e. the prediction of 𝝀𝒔 (𝑘 + 2) and 𝑇𝑒 (𝑘 + 2) are
needed. This can be done by following the flow chart in Figure 5.2.

62
CHAPTER 5 Deadbeat Predictive Torque Control with Space Vector Modulation

Figure 5.2 Flow chart of predictive torque control.

5.3 Simplified deadbeat PTC with SVM

In PTC, the reference torque 𝑇𝑒∗ and flux amplitude |𝜆∗𝑠 | are generally determined by
speed regulator and the maximum torque per ampere trajectory, respectively, as
illustrated by the block diagram in Figure 5.3.

Figure 5.3 Reference torque and flux amplitude calculation.

Thus, based on the relations 𝑇𝑒∗ = 𝑇𝑒 (𝑘 + 2) and |𝜆∗𝑠 | = √𝜆2𝑠𝛼 (𝑘 + 2) + 𝜆2𝑠𝛽 (𝑘 + 2)

imposed by deadbeat technique, equations (5.7) and (5.8) can be derived (see
Appendix A for detailed derivation).

𝑣𝑠𝛽 (𝑘 + 1) = 𝜇1 + 𝜇2 𝑣𝑠𝛼 (𝑘 + 1) (5.7)

2
|𝜆∗𝑠 |2 = [𝑀1 + 𝑣𝑠𝛼 (𝑘 + 1)]2 + [𝑀2 + 𝑣𝑠𝛽 (𝑘 + 1)] (5.8)

where 𝜇1 , 𝜇2 are constants given by

63
CHAPTER 5 Deadbeat Predictive Torque Control with Space Vector Modulation

2𝑇𝑒∗ 𝑀2
− 𝑀1 𝑁2 + 𝑀2 𝑁1 𝐿𝑞 − 𝑁2
𝜇1 = 3𝑃 , 𝜇2 =
𝑀1 𝑀1
− 𝑁
𝐿𝑞 1 𝐿𝑞 − 𝑁1

with 𝑀1 , 𝑀2 , 𝑁1 , 𝑁2 expressed as

𝑀1 = 𝜆𝑠𝛼 (𝑘 + 1) − 𝑇𝑠 𝑅𝑠 𝑖𝑠𝛼 (𝑘 + 1)

𝑀2 = 𝜆𝑠𝛽 (𝑘 + 1) − 𝑇𝑠 𝑅𝑠 𝑖𝑠𝛽 (𝑘 + 1)

𝑇𝑠 𝑅𝑠 𝜔𝑟𝑒 𝑇𝑠
𝑁1 = (1 − ) 𝑖𝑠𝛼 (𝑘 + 1) − (𝜔𝑟𝑒 𝑇𝑠 )𝑖𝑠𝛽 (𝑘 + 1) + 𝜆 (𝑘 + 1)
𝐿𝑞 𝐿𝑞 𝑠𝛽

𝑇𝑠 𝑅𝑠 𝜔𝑟𝑒 𝑇𝑠
𝑁2 = (𝜔𝑟𝑒 𝑇𝑠 )𝑖𝑠𝛼 (𝑘 + 1) + (1 − ) 𝑖𝑠𝛽 (𝑘 + 1) − 𝜆 (𝑘 + 1)
𝐿𝑞 𝐿𝑞 𝑠𝛼

Substituting (5.7) into (5.8) and re-arranging the terms, the voltage vector governing
equation can be derived as

(1 + 𝜇22 )𝑣𝑠𝛼
2 (𝑘
+ 1) + 2(𝑀1 + 𝑀2 𝜇2 + 𝜇1 𝜇2 )𝑣𝑠𝛼 (𝑘 + 1)
(5.9)
+ (𝜇12 + 2𝑀2 𝜇1 + 𝑀12 + 𝑀22 − |𝜆∗𝑠 |2 ) = 0

Obviously, there are two solutions for 𝑣𝑠𝛽 (𝑘 + 1). It is noted that the smaller solution
is generally chosen since the larger solution usually results in a voltage vector that
exceeds the voltage limit of DC/AC inverter as illustrated by Figure 5.4. After
determining the value of 𝑣𝑠𝛽 (𝑘 + 1) , 𝑣𝑠𝛼 (𝑘 + 1) can be calculated via (5.7).
Consequently, the desired reference VV is determined. To synthesize this reference
VV, SVM is employed in this chapter. In SVM, two adjacent active VVs are selected
𝑣 (𝑘+1)
based on the angular position 𝜃𝑣 = tan−1 ( 𝑠𝛽(𝑘+1)) as elaborated in Table 5.1. The
𝑣 𝑠𝛼

time intervals associated with the two active VVs (𝑡1 and 𝑡2 ) and zero VV (𝑡0 ) are
calculated by (5.10)-(5.12), where 𝜃𝑣∗ is computed in Table I and modulation index 𝑚
is given in (5.13).

64
CHAPTER 5 Deadbeat Predictive Torque Control with Space Vector Modulation

𝜋
𝑠𝑖𝑛 ( 3 − 𝜃𝑣∗ )
𝑡1 = 𝑇𝑠 𝑚 𝜋 (5.10)
𝑠𝑖𝑛 3

sin 𝜃𝑣∗
𝑡2 = 𝑇𝑠 𝑚 𝜋 (5.11)
sin 3

𝑡0 = 𝑇𝑠 − 𝑡1 − 𝑡2 (5.12)

2 (𝑘 + 1) + 𝑣 2 (𝑘 + 1)
√𝑣𝑠𝛼 𝑠𝛽
𝑚= (5.13)
2
3 𝑉𝑑𝑐

Figure 5.4 Possible voltage vectors and voltage limit imposed by inverter.

Table 5.1 Distribution of Active VV in SVM

𝜃𝑣 𝜋 𝜋 2𝜋 2𝜋 2𝜋 2𝜋 𝜋 𝜋
(0, ) ( , ) ( , 𝜋) (−𝜋, − ) (− ,− ) (− , 0)
3 3 3 3 3 3 3 3
Sector 1 2 3 4 5 6
Active V1(100) V2(110) V3(010) V4(011) V5(001) V6(101)
VVs V2(110) V3(010) V4(011) V5(001) V6(101) V1(100)
𝜃𝑣∗ 𝜃𝑣 𝜋 2𝜋 𝜃𝑣 + 𝜋 2𝜋 𝜋
𝜃𝑣 − 𝜃𝑣 − 𝜃𝑣 + 𝜃𝑣 +
3 3 3 3

65
CHAPTER 5 Deadbeat Predictive Torque Control with Space Vector Modulation

5.4 Simulation Results

To verify the effectiveness of the proposed method, a PMSM drive system is modeled
in MATLAB/SIMULINK environment as presented in Appendix B. The same
parameters of the dual airgap AFPM flywheel motor as given in Table 3.1 are used.
The simulation results of the proposed method are compared with those obtained by
classical MPTC. Noticeably, the controller sampling frequency is chosen to be 10kHz
for both control approaches and one-cycle implementation delay is added in all
simulations.

Firstly, reversed torque performance is tested at zero motor speed and the results of
both control approaches are given in Figure 5.5. By comparing Figure 5.5 (a) and
Figure 5.5 (b), it is seen that the torque ripples produced by the proposed method are
maintained to be as low as 16.7% of the ripples of classical MPTC. Moreover, the
response time of torque reversal is nearly the same by the two approaches. This means
the proposed method contributes to significantly attenuate the torque ripples without
degrading the dynamic response.

+/- 16.7%

(a) (b)

Figure 5.5 Reversed torque response of PMSM: (a) torque (up) and zoom in
(below) of the classical MPTC (b) torque (up) and zoomed in (below) of the
proposed method.

The performance of the PMSM speed step is shown in Figure 5.6, where the results
obtained by the classical MPTC and the proposed method are illustrated in Figure 5.6

66
CHAPTER 5 Deadbeat Predictive Torque Control with Space Vector Modulation

(a) and Figure 5.6 (b), respectively. From subplot 1 of Figure 5.6 (a) and (b), it can
be observed that the response time of speed reversal for the afore-mentioned two
methods is identical. Nonetheless, torque ripples of the proposed method are much
smaller than those of the classical MPTC as shown by subplot 2, indicating the
superiority of the proposed method.

(a) (b)

Figure 5.6 Reversed speed response of PMSM: (a) classical MPTC speed and
torque response (b) proposed MPTC speed and torque response

To further demonstrate the contribution of the proposed method on the reduction of

flux ripples and stator current total harmonic distortion (THD), simulations are
carried out and the results are given in Figure 5.7 and Figure 5.8. By comparing
Figure 5.7 (a) and Figure 5.8 (a), one can see that the flux ripples and current
harmonics produced by the proposed approach are much smaller than those of the
classical MPTC. This implies a much lower THD, i.e. higher efficiency of PMSM
drives.

The frequency spectrum and THD of stator current produced by the classical MPTC
and the proposed method are shown in Figure 5.7 (b) and Figure 5.8 (b), respectively.

67
CHAPTER 5 Deadbeat Predictive Torque Control with Space Vector Modulation

(a)

(b)

Figure 5.7 Steady-state stator flux and current response of the classical MPTC
under 0.3Nm load at 1200 rpm speed: (a) subplot 1: amplitude of stator flux
subplot 2: phase-a current; (b) frequency spectrum and THD of stator current

From these two figures, a comparison of the numerical values of stator current THD
substantiate the statement that the proposed method is superior in terms of THD.
Furthermore, from the frequency spectrums provided in these two figures, it is
noticeable that a constant switching frequency of 10kHz is obtained by the proposed
method compared to a variable switching frequency of the classical MPTC

68
CHAPTER 5 Deadbeat Predictive Torque Control with Space Vector Modulation

(a)

10kHz Switching
Frequency

(b)

Figure 5.8 Steady-state stator flux and current response of the proposed method
under 0.3Nm load at 1200 rpm speed: (a) subplot 1: amplitude of stator flux,
subplot 2: phase-a current; (b) frequency spectrum and THD of stator current.

5.5 Experiment Results

For further investigation and verification of the control system, the experimental
platform has been built, as shown in Figure 5.9(a). Hardware implementation of the
system is illustrated by functional blocks as shown in Figure 5.9(b). The dual airgap
AFPM flywheel motor is used for the experiment. The parameters of the AFPM
machine used in experiments are tabulated in Table 3.1. The upper and lower
windings are connected in series and the three phases are connected in Y-connection.
A 3-phase SiC MOSFET inverter from CREE Inc is employed for the power state.
The control algorithm is implemented in the dSPACE DS1103 controller. The
sampling and switching periods of the drive are set to be 100μs. The deadtime of the
MOSFET is set at 2us and the DC bus voltage is set at 100Vdc.

69
CHAPTER 5 Deadbeat Predictive Torque Control with Space Vector Modulation

(a)

(b)

Figure 5.9 Experiment platform and the flywheel motor. (a) photo of the overall
experiment setup; (b) system function block diagram

The system block diagram in Figure 5.1 is first simulated in Matlab/Simulink. Then
the model is generated into code by dSPACE coder and downloaded into the DS1103
main CPU for running. The control variable could be viewed and modified on-the-
fly via the dSPACE control desk software running on the PC. The PWM generation
is done by a TI DSP inside the dSPACE board. The ADC sampling is using the 16-
bit main ADC on the master board. Due to the synchronization issue, there’s noise on
the measured current, causing periodical switching noise on all the signals.

70
CHAPTER 5 Deadbeat Predictive Torque Control with Space Vector Modulation

For this chapter’s study, the optical encoder is not used. Instead, the stator flux
observer introduced in Chapter 3.3 is used for stator flux estimation, as well as torque
calculation and speed feedback.

Figure 5.10 Experiment Result of Classic MPTC: step from 600rpm to 1200rpm. (a)
reference and actual speed (b)torque response (c) stator flux

To demonstrate the better performance of the proposed method against the classic
MPTC method, the motor was first run at 600rpm and then step to 1200rpm at around
3 sec. Both are operating in a sensor-less way. Figure 5.10 shows the result of the
classic MPTC method and Figure 5.11 shows the result of the proposed method. By
comparing Figure 5.10 and Figure 5.11, one can see that the speed response of the
two methods is similar. However, the classic MPTC produces a very large speed
ripple, torque ripple, and flux ripple than that of the proposed method.

The zoom-in view of motor current at 1200rpm in the above two experiments as
shown in Figure 5.12. Comparing Figure 5.12 (a) and (b), It is very obvious that the

71
CHAPTER 5 Deadbeat Predictive Torque Control with Space Vector Modulation

proposed method produces much clean motor current than the classic MPTC. Thus,
the effectiveness of the proposed method is fully demonstrated.

Figure 5.11 Experiment Result of Proposed MPTC: step from 600rpm to 1200rpm.
(a) reference and actual speed (b)torque response (c) stator flux

72
CHAPTER 5 Deadbeat Predictive Torque Control with Space Vector Modulation

Figure 5.12 Experiment Result: Steady-state motor current at 1200rpm (a) classic
MPTC; (b) proposed MPTC.

5.6 Conclusion

In this chapter, a simplified deadbeat predictive torque control strategy is proposed

to improve the performance of PMSM drives with significantly reduced
computational burden. SVM technique is employed to mitigate torque and flux
ripples while obtaining constant switching frequency. Coordinate transformation is
not needed by the proposed method. Compared to the conventional model predictive
torque control, the proposed method reduces torque and ripples as well as current
THD to a large extent without degrading dynamic responses.

SVM modulation essentially employs 3 VVs (2 active VVs and 1 zero VV) to reduce
torque ripple for PTC. In the next chapter, 2 VVs (1 active and 1 zero VV) based
method will be proposed for FS-MPTC.

73
CHAPTER 6 Duty Cycle Based Model Predictive Direct Torque and Flux Control

CHAPTER 6 Duty Cycle Based Model Predictive Torque

Control for PMSM

6.1 Introduction

Finite Set-Model Predictive Toque Control (FS-MPTC, or simply MPTC) is

considered a significant improvement over the classic DTC as it can predict the future
steps and dynamic select the optimal voltage vector to satisfy certain cost function.
At the same time, it maintains the advantages of fast response and less rotor position
dependency.

As only one VV is used on the entire sampling cycle, the conventional FS-MPTC still
suffers relative high torque ripples despite the cost function includes the torque and
flux error constraints. Increasing switching frequency is able to alleviate this problem,
but the computation burden and switching loss also significantly increased. Some
recent efforts have been made to mitigate the torque ripple of FS-MPTC, including
offline torque ripple minimization [83], virtual voltage vector methods [52, 84], but
all at the cost of significantly increased complexity. The idea of applying a duty cycle
for the FS-MPTC was recently proposed by [85] for induction machines. It first
predicts the effect of zero VV on torque control and compensates for the possible
deficiency via a set of discretized new VVs. As the determination of the duty ratio is
not based on torque or flux error, this approach implies a degradation on its accuracy,
in addition to the increased computation burden.

To effectively address the torque ripple problem, this thesis proposes a method using
two VVs (including one active and one zero VV) in each sampling cycle. The
proposed method employs a PI controller to dynamically adjust the duty cycle of the
active VV. It contributes to overcoming the flux-deregulation problem mentioned in
[86]. Block diagram of the proposed method is illustrated in Figure 6.1. This method
applies to both IPMSM and PMSM.

74
CHAPTER 6 Duty Cycle Based Model Predictive Direct Torque and Flux Control

Figure 6.1 Block diagram of the proposed model predictive torque control.

6.2 Model and Conventional FS-MPTC

The prediction of torque and flux is based on the mathematical model of PMSM in
the stationary (𝛼 − 𝛽) reference frame as described in Chapter 2.1. Copied to here as:

𝑑𝝀𝒔
𝒗𝒔 = 𝑅𝑠 𝒊𝒔 + (6.1)
𝑑𝑡

3
𝑇𝑒 = 𝑃(𝝀𝒔 × 𝒊𝒔 ) (6.2)
2

The conventional FS-MPTC first does torque and flux prediction based on predicted
current, and then do iterative cost function minimization by comparing the numerical
value of cost function with all the possible voltage vectors. The detail of these two
steps as described in [86] are given below:

1) 𝑇𝑒 and 𝝀𝒔 prediction

To estimate 𝑇𝑒 (𝑘 + 1) and 𝝀𝒔 (𝑘 + 1), where (𝑘 + 1) represents the next sampling

cycle, the value of 𝝀𝒔 (𝑘) must be obtained first. In DTC, the stator flux can be
estimated by (6.3),

75
CHAPTER 6 Duty Cycle Based Model Predictive Direct Torque and Flux Control

𝝀𝒔 (𝑘) = 𝝀𝒔 (𝑘 − 1) + 𝑇𝑠 (𝒗𝒔 (𝑘 − 1) − 𝑅𝑠 𝒊𝒔 (𝑘 − 1))

(6.3)

where 𝑇𝑠 is the sampling period. Based on the measured stator current 𝒊𝒔 (𝑘) at the
current sampling cycle and 𝝀𝒔 (𝑘), 𝝀𝒔 (𝑘 + 1) is derived as

𝝀𝒔 (𝑘 + 1) = 𝝀𝒔 (𝑘) + 𝑇𝑠 (𝒗𝒔 (𝑘) − 𝑅𝑠 𝒊𝒔 (𝑘))

(6.4)

By referring to (6.2), it is seen that 𝒊𝒔 (𝑘 + 1) is also required for the prediction of

𝑇𝑒 (𝑘 + 1). Theoretically, stator current is predicted based on (6.5),

𝑑𝒊𝒔 𝑅𝑠 𝜔𝑟𝑒 1
= − 𝒊𝒔 + 𝒋(𝝀𝒔 − 𝐿𝑞 𝒊𝒔 ) + 𝒗𝒔
𝑑𝑡 𝐿𝑞 𝐿𝑞 𝐿𝑞 (6.5)

where 𝒋 represents +90𝑜 phase shift, 𝐿𝑞 denotes the stator inductance in 𝑞-axis and
𝜔𝑟𝑒 is rotor electrical speed. By using the Euler forward method, 𝒊𝒔 (𝑘 + 1) is derived
as:

𝑅𝑠
1− 𝑇 −𝜔𝑟𝑒 𝑇𝑠
𝑖𝑠𝛼 (𝑘 + 1) 𝐿𝑞 𝑠 𝑖𝑠𝛼 (𝑘)
[ ]= [ ]
𝑖𝑠𝛽 (𝑘 + 1) 𝑅𝑠 𝑖𝑠𝛽 (𝑘)
𝜔𝑟𝑒 𝑇𝑠 1− 𝑇
[ 𝐿𝑞 𝑠 ]
𝜔𝑟𝑒 (6.6)
0 𝑇
𝐿𝑞 𝑠 𝜆𝑠𝛼 (𝑘) 𝑇𝑠 𝑣𝑠𝛼 (𝑘)
+ 𝜔𝑟𝑒 [ ]+ [ ]
𝜆𝑠𝛽 (𝑘) 𝐿𝑞 𝑣𝑠𝛽 (𝑘)
− 𝑇𝑠 0
[ 𝐿𝑞 ]

Subsequently, 𝑇𝑒 (𝑘 + 1) is predicted as

3
𝑇𝑒 (𝑘 + 1) = 𝑃(𝝀𝒔 (𝑘 + 1) × 𝒊𝒔 (𝑘 + 1)) (6.7)
2

To compensate for one cycle digital delay, the torque and flux have to be further

76
CHAPTER 6 Duty Cycle Based Model Predictive Direct Torque and Flux Control

predicted for one more step ahead by shifting forward one sampling cycle:

3
𝑇𝑒 (𝑘 + 2) = 𝑃(𝝀𝒔 (𝑘 + 2) × 𝒊𝒔 (𝑘 + 2)) (6.8)
2

In (6.8) the predictions of 𝝀𝑘+2

𝒔 and 𝒊𝑘+2
𝒔 are obtained as:

𝝀𝒔 (𝑘 + 2) = 𝝀𝒔 (𝑘 + 1) + 𝑇𝑠 (𝒗𝒔 (𝑘 + 1) − 𝑅𝑠 𝒊𝒔 (𝑘 + 1))
(6.9)

𝑅𝑠
1− 𝑇 −𝜔𝑟𝑒 𝑇𝑠
𝑖𝑠𝛼 (𝑘 + 2) 𝐿𝑞 𝑠 𝑖𝑠𝛼 (𝑘 + 1)
[ ]= [ ]
𝑖𝑠𝛽 (𝑘 + 2) 𝑅𝑠 𝑖𝑠𝛽 (𝑘 + 1)
𝜔𝑟𝑒 𝑇𝑠 1− 𝑇
[ 𝐿𝑞 𝑠 ]
𝜔𝑟𝑒
0 𝑇
𝐿𝑞 𝑠 𝜆𝑠𝛼 (𝑘 + 1) (6.10)
+ 𝜔𝑟𝑒 [ ]
𝜆𝑠𝛽 (𝑘 + 1)
− 𝑇𝑠 0
[ 𝐿𝑞 ]
𝑇𝑠 𝑣𝑠𝛼 (𝑘 + 1)
+ [ ]
𝐿𝑞 𝑣𝑠𝛽 (𝑘 + 1)

2) Iterative cost function minimization

To obtain optimal torque and flux control, the value of cost function (6.11) has to be
evaluated with all the possible VVs 𝒗𝒔 (𝑘 + 1). In (6.11), notation 𝑘𝑓 represents the
weighting factor of stator flux error. For three-phase, two-level inverter, the number
of possible VVs is 7. It is noted that the parameter 𝑘𝑓 needs to be tuned to achieve
the best dynamic response. A good starting point to select 𝑘𝑓 is (6.12):

𝑟𝑒𝑓 𝑟𝑒𝑓
𝐽 = |𝑻𝒆 − 𝑻𝑘+2 𝑘+2
𝒆 | + 𝑘𝑓 ||𝝀𝒔 | − |𝝀𝒔 || (6.11)
𝑇𝑟𝑎𝑡𝑒𝑑
𝑘𝑓 = (6.12)
𝜆𝑟𝑎𝑡𝑒𝑑

where 𝑇𝑟𝑎𝑡𝑒𝑑 and 𝜆𝑟𝑎𝑡𝑒𝑑 are the rated torque and stator flux, respectively.

77
CHAPTER 6 Duty Cycle Based Model Predictive Direct Torque and Flux Control

6.3 Proposed Duty Ratio Controller and Improvement of Flux

Regulation

6.3.1 PI controller based duty ratio adjustment

As mentioned above, the application of only one VV in an entire sampling cycle

inevitably generates high torque ripples. To further diminish the torque ripples, two
voltage vectors, including one active VV and one zero VV, are used in the proposed
method. This means VV 𝒗𝒔 (𝑘 + 1) becomes the combination of two VVs
𝒗𝒔𝟏 (𝑘 + 1) and 𝒗𝒔𝟐 (𝑘 + 1). Define the time duration of active VV 𝒗𝒔𝟏 (𝑘 + 1) as 𝑡1 .
Then, the time duration of zero VV 𝒗𝒔𝟐 (𝑘 + 1) = 𝟎 is (𝑇𝑠 − 𝑡1 ) . In this way,
𝑇𝑒 (𝑘 + 2) can be re-derived from (6.8), (6.9), and (6.10) as

3
𝑇𝑒 (𝑘 + 2) = 𝑃 {(𝑀1 𝑁2 − 𝑀2 𝑁1 )
2
𝑀2
+ [𝑣𝑠𝛼 (𝑘 + 1) (𝑁2 − ) (6.13)
𝐿𝑞
𝑀1
+ 𝑣𝑠𝛽 (𝑘 + 1) (−𝑁1 + )] 𝑡1 }
𝐿𝑞

where

𝑅
𝐾1 = 1 − 𝐿 𝑠 𝑇𝑠 𝑀1 = 𝜆𝑠𝛼 (𝑘 + 1) − 𝑇𝑠 𝑅𝑠 𝑖𝑠𝛼 (𝑘 + 1)
{ 𝑞 ,{ ,
𝐾 =𝜔 𝑇 𝑀2 = 𝜆 𝑠𝛽 (𝑘 + 1) − 𝑇𝑠 𝑅𝑠 𝑖 𝑠𝛽 (𝑘 + 1)
2 𝑟𝑒 𝑠

𝐾2 𝜆𝑠𝛽 (𝑘+1)
𝑁1 = 𝐾1 𝑖𝑠𝛼 (𝑘 + 1) − 𝐾2 𝑖𝑠𝛽 (𝑘 + 1) + 𝐿𝑞
{ 𝐾2 𝜆𝑠𝛼 (𝑘+1)
𝑁2 = 𝐾2 𝑖𝑠𝛼 (𝑘 + 1) + 𝐾1 𝑖𝑠𝛽 (𝑘 + 1) − 𝐿𝑞

On the basis of (6.13), if the dead-beat based torque control principle 𝑇𝑒 (𝑘 + 2) =

𝑇𝑒∗ is applied [86], the resultant time duration of active VV can be derived as

2𝑇𝑒∗
𝑡1 = 3𝑃 − 𝑀1 𝑁2 + 𝑀2 𝑁1
𝑀 𝑀 (6.14)
𝑣𝑠𝛼 (𝑘 + 1) (𝑁2 − 𝐿 2 ) + 𝑣𝑠𝛽 (𝑘 + 1) (−𝑁1 + 𝐿 1 )
𝑞 𝑞

However, the direct application of (6.14) can lead to flux-deregulation as reported in

78
CHAPTER 6 Duty Cycle Based Model Predictive Direct Torque and Flux Control

[86]. This thesis proposes a PI controller based duty ratio determination mechanism
as shown in Figure 6.2. It is seen from this figure that the duty cycle d is dynamically
determined by two PI controllers. One PI controller regulates the torque error and the
other regulates the flux error. The larger output of the two PI controllers will be
selected as the final duty cycle output. In the worst case, when the duty cycle becomes
unity, it turns into a conventional MPTC with (k+2) step prediction. The time duration
of active VV is given by 𝑡1 = 𝑑 ∗ 𝑇𝑠 .

Figure 6.2 Proposed duty cycle control method.

6.3.2 Pseudo Algorithm of the Proposed Method

The proposed control algorithm can be summarized as follows:

Step 1. Measure the stator current 𝒊𝒔 (𝑘) and rotor speed 𝜔𝑟𝑒 (𝑘) , estimate the stator
flux 𝝀𝒔 (𝑘) using (6.3), and set 𝒗𝒔 (𝑘) as the optimal voltage vector applied in the
previous sampling cycle.

Step 2. Predict the stator flux 𝝀𝒔 (𝑘 + 1) and stator current 𝒊𝒔 (𝑘 + 1) using (6.4) and
(6.6), respectively.

Step 3. Use torque PI regulator and flux PI regulator to find a duty ratio d and 𝑡1 .

Step 4. Predict the torque 𝑇𝑒 (𝑘 + 2) and stator flux 𝝀𝒔(𝑘 + 2) using (6.8) to (6.9). Then,
evaluate the value of cost function 𝐽𝑖 using (6.11) for each VV 𝒗𝑖𝒔 (𝑘 + 1).

79
CHAPTER 6 Duty Cycle Based Model Predictive Direct Torque and Flux Control

Step 5. Find the minimum cost function among 𝐽𝑖 . Then, determine the corresponding
optimal VV 𝒗𝒔 (𝑘 + 1) with its time duration. Return to Step 1.

6.3.3 Potential flux de-regulation of dead-beat based control

Despite its simplicity, the deadbeat-based torque regulation in (6.14) produces the
potentially optimal time duration by considering only the minimization of torque error.
It ignores the flux regulation. Even though flux error is still taken into account in cost
function (6.11), the selected optimal VV, which originally satisfies both torque and
flux control as in the conventional MPTC, is already adjusted by deadbeat-based
torque control. By giving preference to torque control, flux de-regulation might be
caused since the limited control freedom of changing only the VV may not be
sufficient for flux regulation under some circumstances. In other words, the
application of the deadbeat principle equivalently increases the weighting of torque
control in the optimization. Moreover, the increased weighting is varying
dynamically with respect to the changing of operating conditions, such as load and
speed. Consequently, it is difficult to ensure proper flux regulation under all operating
conditions.

6.4 Simulation Results

To verify the proposed method, simulation studies are performed in Matlab/Simulink

(refer to Appendix B for the simulation model) and the results are presented in this
section. To compare the performance with that of [86] the same parameters of PMSM
are used and are given in Table 6.1. The sampling frequency of both conventional
MPTC and the proposed MPTC is set at 5kHz。

The first simulation evaluates the steady-state performance of the proposed method
against the conventional FCS-MPTC method. The torque, stator flux, and current
waveform harmonics are compared. The results of classic FS-MPTC are shown in
Figure 6.3(a) and Figure 6.4(a), and the results of proposed MPTC are shown in
Figure 6.3(b) and Figure 6.4(b), respectively.

80
CHAPTER 6 Duty Cycle Based Model Predictive Direct Torque and Flux Control

Table 6.1 PMSM parameter for simulation study

Rated torque 4 Nm
Number of pole pairs 2
Stator resistance 5.5 Ω
Permanent magnet flux linkage 0.297 Wb
d-axis inductance 0.0868 H
q-axis inductance 0.1697 H
DC link voltage 200 V

Comparing Figure 6.3 (a) and (b), it is observed that the proposed method
significantly reduces the torque ripples and stator current harmonics and at the same
time obtains a smoother flux response compared to the conventional MPTC. There’s
an approximately 40% torque ripple reduction by using the proposed method. What’s
more, the flux ripples are used by near 67% compared with the conventional method.
The effect of torque and flux ripple reduction could also be reflected in the stator
current. The total harmonic distortion (THD) analysis on the stator current produced
by two different methods are plotted in Figure 6.4. By comparing Figure 6.4 (a) and
(b), it is seen that the current THD of classic MPTC is estimated as 6.09% while the
proposed MPTC only at around 2.62%, leading to a 43% reduction on the THD.
Based on this, the THD of stator current is estimated as 6.09% and 2.62% for
conventional MPTC and the proposed method, respectively. Hence, the THD is
attenuated by roughly 43% using the proposed method. From the above discussions,
it can be concluded that the proposed duty cycle based MPTC method can
significantly improve the steady-state performance of PMSM drives.

The second simulation compares the transient response of the proposed method
against classical MPTC. The transient speed and torque response of classical MPTC
and proposed MPTC are presented in Figure 6.5 (a) and (b), respectively. By
comparing the speed response, both obtain a fast step speed response of
approximately 0.08s for a step speed change from 300rpm to 600rpm.

81
CHAPTER 6 Duty Cycle Based Model Predictive Direct Torque and Flux Control

Torque (Nm)
Torque (Nm)

Amplitude (Wb)
Amplitude (Wb)

Stator Flux
Stator Flux

Stator Current (A)

Stator Current (A)

(a) (b)

Figure 6.3 Steady-state response of conventional MPTC(a) and proposed MPTC (b)
at 300rpm with 3.2Nm load.

(a) (b)

Figure 6.4 Current harmonics of conventional MPTC(a) and proposed MPTC (b) at
300rpm with 3.2Nm load.

Thus, the transient speed response of the proposed method is equivalent to the classic
method. Nonetheless, if look at the torque ripple as shown in subplots 2 of Figure 6.5
(a) and (b), the proposed method is superior in torque ripple and flux ripple reduction
at low load condition and still 20% smaller torque and flux ripple when it reaches
torque limit region.

82
CHAPTER 6 Duty Cycle Based Model Predictive Direct Torque and Flux Control

Speed (rpm)

Speed (rpm)
0.08s 0.08s
Torque (Nm)

Torque (Nm)
1Nm 0.6Nm
Amplitude (Wb)

Amplitude (Wb)
Stator Flux

Stator Flux

(a) (b)

Figure 6.5 Performance of step speed test: (a) speed, torque, and flux responses of
PMSM controlled by classical MPTC, (b) speed, torque and flux responses of
PMSM controlled by the proposed method
Torque (Nm)

Torque (Nm)

0.013 s 0.013 s
Amplitude (Wb)

Amplitude (Wb)
Stator Flux

Stator Flux

(a) (b)

Figure 6.6 Performance of torque reversal test: (a) speed, torque, and flux responses
of PMSM controlled by classical MPTC, (b) speed, torque and flux responses of
PMSM controlled by the proposed method.

83
CHAPTER 6 Duty Cycle Based Model Predictive Direct Torque and Flux Control

In addition, the performance of the proposed method is also examined in the scenario
of dynamic torque reversal when the motor is held standstill. A full load of 4 Nm is
applied to test the full load performance at zero speed. The results are given in Figure
6.6. By comparing Figure 6.6 (a) and (b), it is visible that the dynamic torque response
is nearly the same under two different control approaches. However, the proposed
method generates smaller torque ripples and flux ripples at full load.
Speed (rpm)

Speed (rpm)
Torque (Nm)

Torque (Nm)
Amplitude (Wb)
Amplitude (Wb)

Stator Flux
Stator Flux

(a) (b)

Figure 6.7 Illustration of flux de-regulation problem of dead-beat based duty cycle
control in (a) and excellent flux regulation of PI-based duty cycle control in (b).
reference in red and actual in blue.

The third simulation evaluates the flux-deregulation problem seen in the dead-beat
based approach in [86]. In this simulation, a step speed increase from 0 to 300rpm is
implemented with 0.5Nm load; then at 0.5s, a step load of 3.2Nm is applied. At 0.6s,
the motor speed step from 300rpm to 600rpm, carrying the 3.2Nm load. The problem
of flux de-regulation is illustrated in Figure 6.7(a). It is observed from subplot 3 of
Figure 6.7 (a) that the stator flux goes out of control frequently under step load and
step speed changes. This is mainly caused by the neglecting of flux regulation in the
deadbeat-based duty ratio determination. In comparison, by using the proposed PI-

84
CHAPTER 6 Duty Cycle Based Model Predictive Direct Torque and Flux Control

based duty cycle determination method, the same simulation is carried out and a well-
regulated stator flux is obtained together with excellent torque and speed responses
as shown in Figure 6.7 (b). Therefore, it can be concluded that the proposed method
is capable of reducing torque ripples without sacrificing flux regulation performance.

6.5 Conclusion

This chapter proposes an improved model predictive torque control (MPTC)

approach by employing PI regulated duty ratio calculation. Compared to the
conventional MPTC, the proposed approach contributes to significantly reduce the
torque ripples and stator current harmonics for PMSM drives. Additionally, it
maintains the advantages of fast response and less rotor position dependency while
avoiding the undesirable flux-deregulation phenomenon

85
CHAPTER 7 Conclusion and Future Works

CHAPTER 7 Conclusion and Future Works

7.1 Conclusion

A. General Summary

This thesis focuses on the performance improvement control for direct torque
controlled PMSM drives, with flywheel as the potential application. The main
research area including torque ripple reduction, speed range extension, robustness
enhancement through sensor-less direct torque control and control methods
verification on the dual airgap AFPM flywheel motor.

In the first place, the torque ripple reduction of DTC is investigated. Emerging
techniques to address the torque ripple problem can be classified into three groups:
(1) Model Predictive Torque Control (MPTC) which select an optimal voltage
vector(VV) and apply it on the next full control cycle (one VV); (2) Duty cycle based
MPTC which finds an optimal active voltage vector and zero voltage vector (two
VVs) for the next control cycle and (3), Space Vector Modulation (SVM) MPTC
which essentially uses two active vectors and one zero vectors for the next control
cycle (three VVs). In Chapter 6, this thesis contributes an improved duty cycle based
MPTC which finds the optimal duty cycle by a PI controller. It obtains excellent
torque and flux regulation with fixed switching frequency. In addition, SVM based
DTFC methods and Deadbeat SVM based MPTC methods are also explored by this
thesis in Chapter 4 and Chapter 5 respectively. Especially, in Chapter 5, the predictive
deadbeat control is proposed in this thesis. Compared to the existing work, this
method is implemented in the stationary reference frame and has a simplified nature.
It produces very smooth torque and flux responses using SVM. Moreover, the usage
of prediction techniques compensates for the one cycle digital delay and thus, is
desirable for the digital control system.

In the second place, fast and stable speed control is generally preferred in a wide
speed range for high-speed PMSM drives, especially important for the flywheel
storage system. Field Weakening (FW) techniques are generally used to extend the

86
CHAPTER 7 Conclusion and Future Works

speed range to allow extra energy storage. Most of the existing FW methods for
DTFC based drives have constrained constant power speed range (CPSR) due to the
limited usage of DC-link voltage. Moreover, many of these FW approaches are
heavily dependent on machine parameters. This can lead to degraded performance of
drives or even control instability under some circumstances. To extend the CPSR and
simultaneously enhance the parameter robustness of FW operation, this thesis
proposes a new robust FW algorithm for DTFC based SPMSM drives in Chapter 4.
The proposed algorithm achieves extended CPSR and reduced parameter dependency.

It’s worth mentioning that most of the direct torque methods are all less dependent on
the accuracy of machine parameters and rotor position, which are inherently more
robust than the FOC based methods. However, the stator flux observer plays a vital
role in the accurate estimation of flux, torque, angle, and speed. Various sensor-less
observers have been proposed in the literature to save the cost of such sensors and to
improve the drives’ reliability. This thesis studied three stator flux observers and
evaluated its performance and compensator gain setting method in Chapter 3. Then
the flux observer-based control is experimentally tested out in Chapter 5. It could
potentially be used for all the DTC methods discussed in this thesis.

Lastly, the dual airgap AFPM flywheel machine is studied. The machine model is
being introduced in Chapter 2. The parameters are being used for simulation study
across Chapter 3 to Chapter 5, and it could also be applied to Chapter 6. In addition,
the flywheel prototype has been experimentally tested with classic MPTC and dead-
beat SVM based MPTC in Chapter 5.

B. Comparison of Methods from Chapter 4 to Chapter 6

Now that there are various different direct torque control methods presented from
Chapter 4 to Chapter 6, namely, DTFC-SVM in Chapter 4, DB-SVM PTC in Chapter
5, Duty Cycle PTC in Chapter 6, and the classic MPTC being used as benchmark
method in Chapter 5 and Chapter 6, it would be an interesting idea to compare these
methods in terms of torque ripple reduction, speed range extension, robustness
enhancement, computation simplicity, to show merits and demerits of each method.
Although Table 2.2 gives a simple comparison on the torque ripple and switching

87
CHAPTER 7 Conclusion and Future Works

frequency, it’s too general. In this section, since the simulation models for each of the
methods are available as in Appendix B, the AFPM flywheel motor model is being
used for simulation with each of the control methods, to give a more concrete
comparison. The comparisons are conducted as below.

(1). Torque ripple reduction

Since the torque ripple is affected by speed control parameters and the load condition,
the comparison is done by running the AFPM model in torque control mode, and
applying the 1.0 Nm torque for both torque reference and load torque. This shall make
the motor running in standstill condition, eliminating the impact of speed control. In
addition, the sampling, control and switching frequency is set at 10kHz and
simulation step length is 0.5us. The DC voltage is set at 150V. The same AFPM
parameters as shown in Table 3.1is used for all the simulations. Then the torque ripple
is quantified by using standard deviation. The results of the torque ripple plot are
shown in Figure 7.1:
Torque (Nm)
Torque (Nm)
Torque (Nm)
Torque (Nm)

Time (s)

Figure 7.1 Torque Regulation Ripple for Four Compared Methods under 1Nm Load
and Reference

The data from the above simulation plot are then analyzed to compare the steady-

88
CHAPTER 7 Conclusion and Future Works

state torque ripple. Because the four methods have slightly different dynamic
response time, data after 0.01s are used for comparison to make the comparison result
more representative. The mean and standard deviation for each method is calculated
in Table 7.1. Note that the benchmark method for this comparison is the classic MPTC.
The number “1” means the torque ripple obtained by using this method, while a
number lower than this means lower ripple, and vice versa. For example, 0.7 means
the torque ripple is 70% of the that obtained classic MPTC methods.

From this set of data, it is seen that, overall, all three methods achieved good torque
tracking performance, as the mean value is very close to the target torque reference 1
Nm. In terms of steady-state torque ripple, the DTFC method achieves the lowest
torque ripple while the classic MPTC is the highest. DB-SVM PTC is slightly lower
than DTFC and Duty-cycle MPTC ranks the third.

Table 7.1 Mean and Standard Deviation of Four Compared Methods

Control Method Classic DTFC; DB-SVM Duty

MPTC (Chapter 4) MPTC MPTC
(Chapter 5) (Chapter 6)
Mean (Nm) 0.997539 0.999939 1.004791 1.000474
Standard Deviation (Nm) 0.061431 0.002311 0.006953 0.015039
Compare with MPTC: 1 0.037624 0.113181 0.244812

(2). Speed range extension

The benchmark for speed range extension is based on no-load base speed which is
2170rpm for 150Vdc voltage. The conventional FW algorithm using (4.5) and (4.6)
as introduced in Chapter 4. From the data in Chapter 4, the max speed for using the
conventional field weakening method can achieve 4700rpm while the proposed field
weakening method can achieve 5500rpm. Given that the base speed is 2170rpm, the
proposed method extends the base speed by 2.53 times and the conventional method
extends the base speed by 2.16 times.

In fact, the speed extension range depends on max available bus voltage, the load,

89
CHAPTER 7 Conclusion and Future Works

maximum phase current, and machine parameters. Different parameters used for
simulation/experiment will result in a different extension range. Thus, a theoretic
analysis and calculation of the speed extension range is necessary and is given here.

The stator flux vector 𝝀𝒔 equation given in (7.2) is depicted in the dq frame as shown
in Figure 7.2. Note that the current vector 𝒊𝒔 is located at the origin and the 𝐿𝑠 𝒊𝒔 is
parallel to 𝒊𝒔 . The circle is given by 𝐿𝑠 𝒊𝒔_𝒎𝒂𝒙 . When the motor operates in MTPA
mode, the 𝐿𝑠 𝒊𝒔 term will rise along the DA line until the maximum current is reached
at point A. The vector length OA is the maximum stator flux amplitude |𝝀𝒔_𝒎𝒂𝒙 | given
by (7.2). The minimum stator flux length |𝝀𝒔_𝒎𝒎𝒊𝒏 | happens at point B given by (7.3).
During field weakening, the stator flux vector may be located at any point in the
sector ABD depending on the actual load condition and current angle.

𝝀𝒔 = 𝝀𝒓 + 𝐿𝑠 𝒊𝒔
(7.1)

Figure 7.2 Field Weakening Operating Region

|𝜆𝑠_𝑚𝑎𝑥 | = √𝜆𝑝𝑚 2 + (𝐿𝑠 𝐼𝑚𝑎𝑥 )2 (7.2)

|𝜆𝑠_𝑚𝑖𝑛 | = 𝜆𝑝𝑚 − 𝐿𝑠 𝐼𝑚𝑎𝑥

(7.3)

90
CHAPTER 7 Conclusion and Future Works

In steady state, the phase voltage is given by:

𝒗𝒔 (𝑗𝜔) = 𝑅𝑠 𝒊𝒔 (𝑗𝜔) + 𝑗𝜔𝝀𝒔 (𝑗𝜔) (7.4)

Then when the 𝝀𝒔 reaches the |𝝀𝒔_𝒎𝒂𝒙 |, the resulting maximum voltage is illustrated
as shown in Figure 7.3 (a). The max speed in this condition can be solved from (7.5):

𝑉𝑚𝑎𝑥 2 = (𝑅𝑠 𝐼𝑚𝑎𝑥 )2 + (𝜔𝜆𝑠_𝑚𝑎𝑥 )2 − 2𝑅𝑠 𝐼𝑚𝑎𝑥 𝜔𝜆𝑠_𝑚𝑎𝑥 cos (𝜋 − 𝛿) (7.5)

(a) maximum flux (b) minimum flux

Figure 7.3 Voltage at maximum and minimum stator flux point

In this thesis study, equation (7.6) is used which gives a smaller speed:

𝑉𝑚𝑎𝑥 − 𝑅𝑠 𝐼𝒎𝒂𝒙
𝜔𝑟𝑒 = (7.6)
|𝜆𝑠_𝑚𝑎𝑥 |

Similarly, when the stator flux is reduced to the minimum, the phase voltage is shown
in the triangle as depicted in Figure 7.3 (b), and the speed given by:

√𝑉𝑚𝑎𝑥 2 − (𝑅𝑠 𝐼𝒎𝒂𝒙 )2

(7.7)
𝜔𝑟𝑒 =
|𝜆𝑠_𝑚𝑖𝑛 |

91
CHAPTER 7 Conclusion and Future Works

In the real case, the maximum speed will be smaller than (7.7) as the available torque
becomes zero when the stator flux reaches point B.

In Chapter 4, without the field-weakening, the maximum phase voltage is limited by

the SVPWM in the linear modulation region. Thus, the maximum available phase
voltage is:

𝑉𝑚𝑎𝑥1 = 𝑉𝑑𝑐 /√3 (7.8)

The conventional FW also uses the 𝑉𝑑𝑐 /√3 as max voltage. The proposed FW is able
to push the maximum voltage to (7.9):

𝑉𝑚𝑎𝑥2 = 2𝑉𝑑𝑐 /𝜋 (7.9)

Without considering other factors, the available voltage using (7.9) will be 10.26%
higher than using (7.8):

𝑉𝑚𝑎𝑥2 /𝑉𝑚𝑎𝑥1 = 2√3/𝜋 (7.10)

The base speed without field-weakening can be derived by substituting (7.2) and (7.8)
into (7.6):

𝑉𝑑𝑐 /√3 − 𝑅𝑠 𝐼𝒎𝒂𝒙

𝜔𝑟𝑒_𝑏𝑎𝑠𝑒 =
(7.11)
√𝜆𝑝𝑚 2 + (𝐿𝑠 𝐼𝑚𝑎𝑥 )2

Due to the fact that the field weakening is meaningful only when the load is less than
the full load, and for the flywheel application, the steady-state load should be very
mall, the no-load base speed is preferred. This can be found by simply set Imax to
zero in (7.11):

𝑉𝑑𝑐 /√3
𝜔𝑟𝑒_𝑏𝑎𝑠𝑒_𝑛𝑜𝑙𝑜𝑎𝑑 = (7.12)
𝜆𝑝𝑚

92
CHAPTER 7 Conclusion and Future Works

The maximum speed for conventional field-weakening can be found by substituting

(7.2) and (7.8) into (7.7):

√(𝑉𝑑𝑐 /√3)2 − (𝑅𝑠 𝐼𝒎𝒂𝒙 )2

(7.13)
𝜔𝑟𝑒_max _𝑐𝑜𝑛𝑣𝐹𝑊 =
𝜆𝑝𝑚 − 𝐿𝑠 𝐼𝑚𝑎𝑥

The maximum speed for proposed field-weakening can be derived by substituting

(7.2) and (7.9) into (7.7):

√(2𝑉𝑑𝑐 /𝜋)2 − (𝑅𝑠 𝐼𝒎𝒂𝒙 )2

𝜔𝑟𝑒_max _𝑝𝑟𝑜𝑝𝐹𝑊 = (7.14)
𝜆𝑝𝑚 − 𝐿𝑠 𝐼𝑚𝑎𝑥

A numerical case study with different parameters is then summarized in Table 7.2
below. It is seen from the table that a single parameter change makes a different speed
extension ratio. The ratio will become more different from different load conditions.

Table 7.2 Speed Range Extension Numerical Case Study (No Load)

Parameters Case 1 Case 2 Case 3 Case 4

Vdc (V) 150 150 75 150
Imax (Apk) 5 2.5 5 5
Rs (Ohm) 4.65 4.65 4.65 4.65
Ls (mH) 26.8 26.8 26.8 13.4
𝜆𝑝𝑚 (V⸱s/rad) 0.1905 0.1905 0.1905 0.1905

𝜔𝑟𝑒_𝑏𝑎𝑠𝑒_𝑛𝑜𝑙𝑜𝑎𝑑 (rad/s) 454.5 454.5 227.3 454.5

𝜔𝑟𝑒_max _𝑐𝑜𝑛𝑣𝐹𝑊 (rad/s) 1120.8 607.0 354.7 512.9
𝜔𝑟𝑒_max _𝑝𝑟𝑜𝑝𝐹𝑊 (rad/s) 1278.1 679.0 433.4 584.8

𝜔𝑟𝑒_max _𝑐𝑜𝑛𝑣𝐹𝑊 /𝜔𝑟𝑒_𝑏𝑎𝑠𝑒_𝑛𝑜𝑙𝑜𝑎𝑑 2.47 1.34 1.56 1.13

𝜔𝑟𝑒_max _𝑝𝑟𝑜𝑝𝐹𝑊 /𝜔𝑟𝑒_𝑏𝑎𝑠𝑒_𝑛𝑜𝑙𝑜𝑎𝑑 2.81 1.49 1.91 1.29

𝜔𝑟𝑒_max _𝑝𝑟𝑜𝑝𝐹𝑊 /𝜔𝑟𝑒_max _𝑐𝑜𝑛𝑣𝐹𝑊 1.14 1.12 1.22 1.14

93
CHAPTER 7 Conclusion and Future Works

In summary, the proposed field-weakening method in Chapter 4 excels the

conventional field-weakening method for the following two reasons:

1). Proposed method pushes the voltage to the maximum of 2𝑉𝑑𝑐 /𝜋 which is 10.26%
higher than the maximum 𝑉𝑑𝑐 /√3 by conventional method. This is achieved through
the flux adjustment PI controller to achieve highest voltage.

2). Proposed method pushes the current to the maximum current limit to ensure the
flux can be reduced as low as possible, so that the maximum achievable speed can be
higher. This is achieved through the torque adjust PI. The conventional field-
weakening purely based on machine parameters and thus minimum stator flux cannot
guaranteed.

From the above summary, it is also seen that the principle applies to the control
methods in other chapters as well.

(3). Robustness enhancement

For robustness enhancement, all the four methods studied in this thesis uses a stator
flux observer and removed the encoder as position/speed feedback. The amount of
robustness enhancement is the same for all four methods in terms of position feedback
sensor dependency.

In terms of the field-weakening capability with respect to the machine parameters,

the proposed field-weakening is able to tolerate 20% stator resistance and 20%
inductance detune, as demonstrated in Chapter 4, Figure 4.9 to Figure 4.12.

(4). Computation simplicity

In terms of computation simplicity, the amount of calculations required for the

“TqFluxControlAndSwitching” block as in Appendix B.3 to B.6 is given in Table 7.3.
The data is extracting from the code generation report by generating the model into
generic C code using “ert.tlc” on Intel X86 CPU. The PWM generation part was
excluded as it is handled by CPU hardware.

94
CHAPTER 7 Conclusion and Future Works

Table 7.3 Computation resource report for the four compared methods

Control Method Classic DTFC; DB-SVM Duty

MPTC (Chapter 4) MPTC MPTC
(Chapter 5) (Chapter 6)
Lines of Code 228 181 236 271
Accumulated Stack Size
(bytes) 422 213 254 405
Complexity: 29 27 31 36

A screenshot of the generated step function report for the MPTC model is given in
Figure 7.4 as an example.

Figure 7.4 Code Generation Report Screenshot for MPTC

7.2 Future Works

Despite the progress made so far in the thesis, due to time constraints, there’re
potentially several other topics that are worth studying. Future works are planned in
the following areas:

A. Sensor-ed and Sensor-less Performance Difference Study

This thesis only explores one stator flux observer and the performance difference
from the control using encoder feedback is worth quantitively studied. In addition,
some other interesting stator flux observers, for example, the one introduced by [87],
could be comparatively evaluated. The same characteristic function analysis method

95
CHAPTER 7 Conclusion and Future Works

could be applied to this observer.

B. Axial direction displacement control

So far this thesis focus on the rotating direction torque and flux control for the dual
air-gap AFPM flywheel motor. The control of the axial direction force was only being
addressed by [34] with classical PI control. Special sensor is required to measure the
z-axis displacement (for example, Micro-Epsilon Co. DT110-T). All other control
methods have not been addressed yet. It would be interesting to combine the z-axis
control with the torque and flux control, or at least, to quantify to how much extent
the two have coupled.

C. Flywheel in the energy storage system

This thesis has been focus on the control of the flywheel motor itself. The control of
the flywheel as part of a larger system needs to be studied. For example, when the
flywheel charges and discharges rapidly, the dc bus voltage need to be regulated. As
pointed out by [10], the operating point will change under rapid discharge control,
this leads to the degraded performance with the PI controller.

96
APPENDIX

Appendix A Derivation of (5.7)

The deadbeat based DTFC target to let k+2 step prediction meets the reference value:

𝑇𝑒∗ = 𝑇𝑒 (𝑘 + 2) (A.1)

|𝜆∗𝑠 | = √𝜆2𝑠𝛼 (𝑘 + 2) + 𝜆2𝑠𝛽 (𝑘 + 2) (A.2)

From A.2, we have

𝜆2𝑠𝛼 (𝑘 + 2) + 𝜆2𝑠𝛽 (𝑘 + 2) − |𝜆∗𝑠 |2 = 0 (A.3)

To predict 𝜆𝑠𝛼𝛽 (𝑘 + 2) and 𝑇𝑒 (𝑘 + 2) , 𝑖𝑠𝛼𝛽 (𝑘 + 2) , 𝜆𝑠𝛼𝛽 (𝑘 + 1) and 𝑖𝑠𝛼𝛽 (𝑘 + 1)

are required:

𝜆𝑠𝛼 (𝑘 + 1) 𝜆𝑠𝛼 (𝑘) 𝑣𝑠𝛼 (𝑘) 𝑖𝑠𝛼 (𝑘)

[ ]=[ ] + 𝑇𝑠 ([ ] − 𝑅𝑠 [ ]) (A.4)
𝜆𝑠𝛽 (𝑘 + 1) 𝜆𝑠𝛽 (𝑘) 𝑣𝑠𝛽 (𝑘) 𝑖𝑠𝛽 (𝑘)

𝑖𝑠𝛼 (𝑘 + 1) 𝑇𝑠 𝑅𝑠 𝑖𝑠𝛼 (𝑘)

[ ] = (1 − )[ ]
𝑖𝑠𝛽 (𝑘 + 1) 𝐿𝑞 𝑖𝑠𝛽 (𝑘)
(A.5)
𝑇𝑠 𝑤𝑟𝑒 −(𝜆𝑠𝛽 (𝑘) − 𝐿𝑞 𝑖𝑠𝛽 (𝑘)) 𝑇𝑠 𝑣𝑠𝛼 (𝑘)
+ [ ]+ [ ]
𝐿𝑞 (𝜆𝑠𝛼 (𝑘) − 𝐿𝑞 𝑖𝑠𝛼 (𝑘)) 𝐿𝑞 𝑣𝑠𝛽 (𝑘)

(A.4) and (A.5) can be calculated from the k-step known variable. To find
𝑖𝑠𝛼𝛽 (𝑘 + 2) , A prediction from (A.4) and (A.5) and the unknown 𝑣𝑠𝛼𝜌 (𝑘 + 1) is
conducted as shown in (A.6):

97
APPENDIX

𝑖𝑠𝛼 (𝑘 + 2) 𝑇𝑠 𝑅𝑠 𝑖𝑠𝛼 (𝑘 + 1)
[ ] = (1 − )[ ]
𝑖𝑠𝛽 (𝑘 + 2) 𝐿𝑞 𝑖𝑠𝛽 (𝑘 + 1)

𝑇𝑠 𝑤𝑟𝑒 −(𝜆𝑠𝛽 (𝑘 + 1) − 𝐿𝑞 𝑖𝑠𝛽 (𝑘 + 1))

+ [ ] (A.6)
𝐿𝑞 (𝜆𝑠𝛼 (𝑘 + 1) − 𝐿𝑞 𝑖𝑠𝛼 (𝑘 + 1))

𝑇𝑠 𝑣𝑠𝛼 (𝑘 + 1)
+ [ ]
𝐿𝑞 𝑣𝑠𝛽 (𝑘 + 1)

As the 𝜆𝑠𝛼𝛽 (𝑘 + 2) prediction is given by (A.7)

𝜆𝑠𝛼 (𝑘 + 2) 𝜆𝑠𝛼 (𝑘 + 1)
[ ]=[ ]
𝜆𝑠𝛽 (𝑘 + 2) 𝜆𝑠𝛽 (𝑘 + 1)
(A.7)
𝑣𝑠𝛼 (𝑘 + 1) 𝑖𝑠𝛼 (𝑘 + 1)
+ 𝑇𝑠 ([ ] − 𝑅𝑠 [ ])
𝑣𝑠𝛽 (𝑘 + 1) 𝑖𝑠𝛽 (𝑘 + 1)

Now, bring (A.7) back to (A.3):

[𝜆𝑠𝛼 (𝑘 + 1)+𝑇𝑠 𝑣𝑠𝛼 (𝑘 + 1) − 𝑇𝑠 𝑅𝑠 𝑖𝑠𝛼 (𝑘 + 1)]2

+ [𝜆𝑠𝛽 (𝑘 + 1)+𝑇𝑠 𝑣𝑠𝛽 (𝑘 + 1) (A.8)

− 𝑇𝑠 𝑅𝑠 𝑖𝑠𝛽 (𝑘 + 1)]2 − |𝜆∗𝑠 |2 = 0

Define the following two auxiliary variables:

𝑀1 = 𝜆𝑠𝛼 (𝑘 + 1) − 𝑇𝑠 𝑅𝑠 𝑖𝑠𝛼 (𝑘 + 1)

𝑀2 = 𝜆𝑠𝛽 (𝑘 + 1) − 𝑇𝑠 𝑅𝑠 𝑖𝑠𝛽 (𝑘 + 1)

Then (A.8) can be simplified as:

98
APPENDIX

[𝑀1+𝑇𝑠 𝑣𝑠𝛼 (𝑘 + 1)]2 + [𝑀2+𝑇𝑠 𝑣𝑠𝛽 (𝑘 + 1)]2 − |𝜆∗𝑠 |2

(A.9)
=0

On the other hand, the k+2 torque prediction is given by:

3
𝑇𝑒∗ = 𝑇𝑒 (𝑘 + 2) = 𝑃 (𝜆𝑠𝛼 (𝑘 + 2)𝑖𝑠𝛽 (𝑘 + 2)
2
(A.10)
− 𝜆𝑠𝛽 (𝑘 + 2)𝑖𝑠𝛼 (𝑘 + 2))

Substitute (A.10) with (A.6) and (A.7), and re-arrange, we have:

[𝑀1+𝑇𝑠 𝑣𝑠𝛼 (𝑘 + 1)]

𝑇𝑠 𝑅𝑠
∗ [(1 − ) 𝑖𝑠𝛽 (𝑘 + 1)
𝐿𝑞

𝑇𝑠 𝑤𝑟𝑒
− (𝜆𝑠𝛼 (𝑘 + 1) − 𝐿𝑞 𝑖𝑠𝛼 (𝑘 + 1))
𝐿𝑞

𝑇𝑠
+ 𝑣 (𝑘 + 1)] − [𝑀2+𝑇𝑠 𝑣𝑠𝛽 (𝑘 + 1)]
𝐿𝑞 𝑠𝛽 (A.11)

𝑇𝑠 𝑅𝑠
∗ [(1 − ) 𝑖𝑠𝛼 (𝑘 + 1)
𝐿𝑞

𝑇𝑠 𝑤𝑟𝑒
+ (𝜆𝑠𝛽 (𝑘 + 1) − 𝐿𝑞 𝑖𝑠𝛽 (𝑘 + 1))
𝐿𝑞

𝑇𝑠 2𝑇 ∗
+ 𝑣𝑠𝛼 (𝑘 + 1)] =
𝐿𝑞 3𝑃

We then define:

𝑇𝑠 𝑅𝑠 (𝜔𝑟𝑒 𝑇𝑠 ) 𝜔𝑟𝑒 𝑇𝑠
𝑁1 = (1 − ) 𝑖𝑠𝛼 (𝑘 + 1) + 𝑖𝑠𝛽 (𝑘 + 1) − 𝜆 (𝑘 + 1)
𝐿𝑞 1 𝐿𝑞 𝑠𝛽

99
APPENDIX

𝑇𝑠 𝑅𝑠 𝑇𝑠 𝑤𝑟𝑒
𝑁2 = (1 − ) 𝑖𝑠𝛽 (𝑘 + 1) + (𝜆𝑠𝛼 (𝑘 + 1) − 𝐿𝑞 𝑖𝑠𝛼 (𝑘 + 1))
𝐿𝑞 𝐿𝑞

The torque equation in (A.11) can be simplified as:

𝑇𝑠
[𝑀1+𝑇𝑠 𝑣𝑠𝛼 (𝑘 + 1)] [𝑁2 + 𝑣 (𝑘 + 1)]
𝐿𝑞 𝑠𝛽

− [𝑀2+𝑇𝑠 𝑣𝑠𝛽 (𝑘 + 1)] [𝑁1 (A.12)

𝑇𝑠 2𝑇 ∗
+ 𝑣𝑠𝛼 (𝑘 + 1)] =
𝐿𝑞 3𝑃

Let V𝛼 = 𝑇𝑠 𝑣𝑠𝛼 (𝑘 + 1) and V𝛽 = 𝑇𝑠 𝑣𝑠𝛽 (𝑘 + 1), (A.12) turns into the form of (5.7):

2𝑇 ∗ 𝑀2
−𝑁2
+𝑀2𝑁1−𝑀1𝑁2 𝐿𝑞
3𝑃
V𝛽 = 𝑀1
+ 𝑀1
V𝛼 (A.13)
−𝑁1 −𝑁1
𝐿𝑞 𝐿𝑞

100
APPENDIX

Appendix B Simulation Models

B.1 Top Level Overall Model shared across Chapter 3 to Chapter 6

101
APPENDIX

B.2 Speed Control and Field Weakening Model

This part of the model sets the speed reference, torque reference, and flux reference. The conventional field weakening part in which servers

adjust the torque and flux reference is also included here. This model is also shared from Chapter 3 to Chapter 6.

102
APPENDIX

B.3 DTFC Model for Chapter 3 and Chapter 4

This is the “TqFluxControlAndSwitching” block for DTFC in Chapter 3 and 4.

103
APPENDIX

B.4 MPTC Model for Chapter 5 and Chapter 6

This is the “TqFluxControlAndSwitching” block for MPTC which is used as a benchmark control method in Chapter 5 and 6.

104
APPENDIX

B.5 DB-SVM-PTC Model for Chapter 5

This is the “TqFluxControlAndSwitching” block for the Deadbeat-SVM PTC method in Chapter 5.

105
APPENDIX

B.6 DutyCycle-PTC Model for Chapter 6

This is the “TqFluxControlAndSwitching” block for the Duty Cycle PTC method in Chapter 6.

106
APPENDIX

B.7 Stator Flux Torque Speed and Angle Observers for Chapter 3 to Chapter 6

This is the “StatorFlux_Tq_Angle_Speed_Observer” block for Chapter 3 to Chapter 6. The Observer 3 is selected in the variant subsystem

107
APPENDIX

Observer 1 Model

108
APPENDIX

Observer 2 Model

109
APPENDIX

Observer 3 Model and the code:

110
APPENDIX

function [Iap_est,Ibt_est,Flux_Ap,Flux_Bt,Flux_Act_Ap,Flux_Act_Bt,Theta] =
Observer_dq(Valpha,Vbeta,Ialpha,Ibeta,ModelParam)

global Rs,
global flux_r,
global Ld,
global Lq,
global Ts,
%value declaired as persistent cannot be output. so _k1 means previous
%value.
persistent Flux_Ap_k1,
persistent Flux_Bt_k1,
persistent IntQ_Tempk1,
persistent IntD_Tempk1,
persistent IntQ_Satk1,
persistent IntQ_UnSatk1,
persistent IntD_Satk1,
persistent IntD_UnSatk1,

persistent Theta_est,

if isempty(Flux_Ap_k1)
Flux_Ap_k1=(0);
Flux_Bt_k1=(-flux_r);

111
APPENDIX

IntQ_Tempk1=(0);
IntD_Tempk1=(0);

IntQ_Satk1=(0);
IntQ_UnSatk1=(0);

IntD_Satk1=(0);
IntD_UnSatk1=(0);

Theta_est=(0);
end

Kp_Apest=ModelParam(1);%kp_obs
Ki_Apest=ModelParam(2); %ki_obs
Kp_Btest=ModelParam(1);%kp_obs
Ki_Btest=ModelParam(2);%kp_obs

Int_Max=3*Rs; %%2*Rs
Int_Min=-3*Rs; %%-2*Rs
CosEst=cos(Theta_est);
SinEst=sin(Theta_est);
%%Group 1
Flux_d_est=Flux_Ap_k1*CosEst + Flux_Bt_k1*SinEst; %%Correct Flux_d_est
Flux_q_est=-Flux_Ap_k1*SinEst + Flux_Bt_k1*CosEst; %%Correct Flux_q_est

112
APPENDIX

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Id_est=(Flux_d_est - flux_r)/Ld;
Iq_est=Flux_q_est/Lq;
%%Group 1
Iap_est=Id_est*CosEst - Iq_est*SinEst; %%Correct Iap_est
Ibt_est=Id_est*SinEst + Iq_est*CosEst; %%Correct Ibt_est
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Iap_est_err = (Ialpha - Iap_est);

Ibt_est_err = (Ibeta - Ibt_est);
PropQ_Temp=Kp_Apest*Iap_est_err;
PropD_Temp=Kp_Btest*Ibt_est_err;
IntQ_Temp=IntQ_Tempk1 + (Ki_Apest*Iap_est_err + (IntQ_Satk1 - IntQ_UnSatk1))*Ts;
%%Vap_comp_k1=IntQ_Satk1;
IntD_Temp=IntD_Tempk1 + (Ki_Btest*Ibt_est_err + (IntD_Satk1 - IntD_UnSatk1))*Ts;
%%Vbt_comp_k1=IntD_Satk1;
if(IntQ_Temp>Int_Max)
IntQ_Temp=Int_Max;
end
if(IntQ_Temp<Int_Min)
IntQ_Temp=Int_Min;
end
if(IntD_Temp>Int_Max)
IntD_Temp=Int_Max;
end

113
APPENDIX

if(IntD_Temp<Int_Min)
IntD_Temp=Int_Min;
end
IntQ_UnSatk=PropQ_Temp + IntQ_Temp;
IntD_UnSatk=PropD_Temp + IntD_Temp;
if(IntQ_UnSatk>Int_Max)
IntQ_Satk=Int_Max; %%Vap_comp=IntQ_Satk;
else
IntQ_Satk=IntQ_UnSatk;
end
if(IntQ_UnSatk<Int_Min)
IntQ_Satk=Int_Min; %%Vap_comp=IntQ_Satk;
else
IntQ_Satk=IntQ_UnSatk;
end
if(IntD_UnSatk>Int_Max)
IntD_Satk=Int_Max; %%Vbt_comp=IntD_Satk;
else
IntD_Satk=IntD_UnSatk;
end
if(IntD_UnSatk<Int_Min)
IntD_Satk=Int_Min; %%Vbt_comp=IntD_Satk;
else
IntD_Satk=IntD_UnSatk;
end
CompAp_k=Valpha - Ialpha*Rs + IntQ_Satk*1;
CompBt_k=Vbeta - Ibeta*Rs + IntD_Satk*1;

114
APPENDIX

Flux_Ap= Flux_Ap_k1 + (CompAp_k)*Ts;

Flux_Bt= Flux_Bt_k1 + (CompBt_k)*Ts;
%%Active Flux Estimation
Flux_Act_Ap=Flux_Ap - Lq*Ialpha; %Debug
Flux_Act_Bt=Flux_Bt - Lq*Ibeta; %Debug
Theta=atan2(Flux_Act_Bt,Flux_Act_Ap);

%update k1 value

Flux_Ap_k1=Flux_Ap;
Flux_Bt_k1=Flux_Bt;
IntQ_Tempk1=IntQ_Temp;
IntD_Tempk1=IntD_Temp;

IntQ_Satk1=IntQ_Satk;
IntQ_UnSatk1=IntQ_UnSatk;

IntD_Satk1=IntD_Satk;
IntD_UnSatk1=IntD_UnSatk;

Theta_est=Theta;
end

115
APPENDIX

B.8 Advanced FieldWeakening

This is the “AdvancedFW” block mainly studied in Chapter 4. The output ports are feedback into the “SpeedRegulator” block for enhanced

field weakening

116
APPENDIX

Appendix C Clarke and Park Transforms

The transform of motor vectors (here we use current to illustrate) form a-b-c frame
to  −  reference frame is called Clarke Transform given by:

i  1 − 12 − 12  ia 
i  = 2  0 3 
− 23  ib 
 3 2

 i0  1 1 1 
 
2 2 2   ic 

where i0 is always 0 in a balanced three-phase system. The coefficient 2/3 is to make

energy conservation. Conversely, the inverse Clarke transformation is defined as

ia   1 0 1 i 
i  =  − 1 3 
1 i 
 b  2 2

 ic   − 1 − 3
1  i0 
 2 2

If expressed in non-matrix form, and drop the zero sequence component, the Clarke
transform can be expressed as

i = ia

i = ( 2ib + ia ) / 3

And the inverse transform is:

ia = i

ib = − 2 i +
1 3
i
2 

 ic = − 2 i −
1 3
i
2 

Park Transform is used to convert a vector from the static  −  reference frame to

the rotating d − q reference frame, given by

117
APPENDIX

id   cos e sin e  i 

i  =   
 q   − sin e cos e  i 

Note that the angle  e is defined as the angle between the d-axis and  axis. The
inverse Park transform is simply a change of sign at the sine components:

i  cos e − sin e  id 

i  =   
    sin e cos e  iq 

118
AUTHOR’S PUBLICATIONS

Author’s Publications

Journal Papers Submitted for Possible Publication

1. H. Yang, X. Zhang, Y. Wang, and G. Foo, “A Duty Cycle based Finite-set Model
Predictive Torque Control for Permanent Magnet Synchronous Motor Drives”

2. H. Yang, X. Zhang, Y. Wang “A Simplified Deadbeat Predictive Torque Control

Strategy for PMSM Drives using SVM”.

Conference Papers

3. H. Yang, X. Zhang, Y. Wang, and G. Foo, “A Duty Cycle based Finite-set Model
Predictive Torque Control for Permanent Magnet Synchronous Motor Drives”,
(Accepted by and Presented on Southern Power Electronics Conference (SPEC)
2018, Singapore)

Publications During Ph.D. Period but not Covered by Thesis

4. H. Yang, Y. Wang, and F. Lim, "Design of an energy sharing dual-escalator system

based on the dual active bridge converter," in Electric Machines & Drives
Conference (IEMDC), 2013 IEEE International, 2013, pp. 1424-1429: IEEE.

5. H. Yang, P. Wu, and Y. Wang, "Design Issues for the active clamp forward
converter in micro-inverter applications," in Power Electronics for Distributed
Generation Systems (PEDG), 2015 IEEE 6th International Symposium on, 2015,
pp. 1-4: IEEE.

6. L. Teng, H. Yang, and Y. Wang, “Model reference tracking control of linear motor
with dead-zone via switched systems subjected to time-varying delay,” In 42nd
Annual Conference of the IEEE Industrial Electronics Society, Florence, Italy,
2016.

119
BIBLIOGRAPHY

Bibliography

[1] S.-K. Sul, Control of electric machine drive systems. John Wiley & Sons, 2011.

[2] F.-J. Lin and S.-L. Chiu, "Adaptive fuzzy sliding-mode control for PM
synchronous servo motor drives," IEE Proceedings-Control Theory and
Applications, vol. 145, no. 1, pp. 63-72, 1998.

[3] C. Y. Lin and Y. C. Liu, "Precision Tracking Control and Constraint Handling
of Mechatronic Servo Systems Using Model Predictive Control,"
IEEE/ASME Transactions on Mechatronics, vol. 17, no. 4, pp. 593-605, 2012.

[4] K. Baoquan, L. Chunyan, and C. Shukang, "Flux-weakening-characteristic

analysis of a new permanent-magnet synchronous motor used for electric
vehicles," IEEE Transactions on Plasma Science, vol. 39, no. 1, pp. 511-515,
2011.

[5] D.-x. Shang, Y.-c. Liu, F.-j. Sun, and H.-y. Zhang, "Study on DTC-SVM of
PMSM based on propeller load characteristic," in Intelligent Control and
Automation, 2008. WCICA 2008. 7th World Congress on, 2008, pp. 6445-
6449: IEEE.

[6] S.-A. Kim, S.-I. Byun, S.-G. Lee, and Y.-H. Cho, "Boost and flux-weakening
control of engine direct connection type IPMSG over wide-speed range," in
Industrial Electronics and Applications (ICIEA), 2015 IEEE 10th Conference
on, 2015, pp. 1307-1312: IEEE.

[7] Y. Zhao, C. Wei, Z. Zhang, and W. Qiao, "A review on position/speed

sensorless control for permanent-magnet synchronous machine-based wind
energy conversion systems," IEEE Journal of Emerging and Selected Topics
in Power Electronics, vol. 1, no. 4, pp. 203-216, 2013.

[8] T. D. Nguyen, G. Foo, K. J. Tseng, and D. M. Vilathgamuwa, "Modeling and

Sensorless Direct Torque and Flux Control of a Dual-Airgap Axial Flux
Permanent-Magnet Machine With Field-Weakening Operation," IEEE/ASME
Transactions on Mechatronics, vol. 19, no. 2, pp. 412-422, 2014.

[9] A. A. K. Arani, H. Karami, G. B. Gharehpetian, and M. S. A. Hejazi, "Review

of Flywheel Energy Storagee Systems structures and applications in power
systems and microgrids," Renewable and Sustainable Energy Reviews, vol.
69, pp. 9-18, 2017/03/01/ 2017.

[10] X. Zhang and J. Yang, "A Robust Flywheel Energy Storage System Discharge
Strategy for Wide Speed Range Operation," IEEE Transactions on Industrial
Electronics, vol. 64, no. 10, pp. 7862-7873, 2017.

[11] T. D. Nguyen and G. Foo, "Sensorless control of a dual-airgap axial flux

permanent magnet machine for flywheel energy storage system," IET Electric

120
BIBLIOGRAPHY

Power Applications, vol. 7, no. 2, pp. 140-149, 2013.

[12] K. B. Bimal, Modern power electronics and AC drives. Prentice-Hall, 2002.

[13] J. I. Itoh, N. Nomura, and H. Ohsawa, "A comparison between V/f control
and position-sensorless vector control for the permanent magnet synchronous
motor," in Proceedings of the Power Conversion Conference-Osaka 2002
(Cat. No.02TH8579), 2002, vol. 3, pp. 1310-1315 vol.3.

[14] Z. Tang, X. Li, S. Dusmez, and B. Akin, "A New V/f-Based Sensorless MTPA
Control for IPMSM Drives," IEEE Transactions on Power Electronics, vol.
31, no. 6, pp. 4400-4415, 2016.

[15] P. C. Perera, F. Blaabjerg, J. K. Pedersen, and P. Thogersen, "A sensorless,

stable V/f control method for permanent-magnet synchronous motor drives,"
IEEE Transactions on Industry Applications, vol. 39, no. 3, pp. 783-791, 2003.

[16] A. Moldovan, F. Blaabjerg, and I. Boldea, "Active-flux-based, V/f-with-

stabilizing-loops versus sensorless vector control of IPMSM Drives," in 2011
IEEE International Symposium on Industrial Electronics, 2011, pp. 514-519.

[17] T.-H. Liu and J.-J. Huang, "Improved efficiency of a fan drive system without
using an encoder or current sensors," The Journal of Engineering, vol. 2018,
no. 4, pp. 222-229, 2018.

[18] S. Shinn-Ming and H. Tsai-Wang, "Minimum copper loss control for

sensorless v/f controlled IPMSM drives," in 2012 IEEE International
Symposium on Industrial Electronics, 2012, pp. 708-712.

[19] S. C. Agarlita, C. E. Coman, G. D. Andreescu, and I. Boldea, "Stable V/f

control system with controlled power factor angle for permanent magnet
synchronous motor drives," IET Electric Power Applications, vol. 7, no. 4, pp.
278-286, 2013.

[20] J.-J. Moon, W.-S. Im, and J.-M. Kim, "Novel phase advance method of BLDC
motors for wide range speed operations," in Applied Power Electronics
Conference and Exposition (APEC), 2013 Twenty-Eighth Annual IEEE, 2013,
pp. 2343-2348: IEEE.

[21] S.-M. Sue, K.-L. Wu, J.-S. Syu, and K.-C. Lee, "A phase advanced
commutation scheme for IPM-BLDC motor drives," in 2009 4th IEEE
Conference on Industrial Electronics and Applications, 2009, pp. 2010-2013:
IEEE.

[22] F. Blaschke, "The principle of field orientation as applied to the new

transvector closed-loop system for rotating-field machines," Siemens review,
vol. 34, no. 3, pp. 217-220, 1972.

[23] M. P. Kazmierkowski, L. García Franquelo, J. Rodriguez, M. Perez, and J. I.

León Galván, "High-performance motor drives," IEEE Industrial Electronic

121
BIBLIOGRAPHY

Magazine, 5 (3), 6-26, 2011.

[24] I. Takahashi and T. Noguchi, "A new quick-response and high-efficiency

control strategy of an induction motor," IEEE Transactions on Industry
Applications, no. 5, pp. 820-827, 1986.

[25] M. Depenbrock, "Direct self-control (DSC) of inverter fed induktion

machine," in Power Electronics Specialists Conference, 1987 IEEE, 1987, pp.
632-641: IEEE.

[26] C. French and P. Acarnley, "Direct torque control of permanent magnet

drives," IEEE Transactions on Industry Applications, vol. 32, no. 5, pp. 1080-
1088, 1996.

[27] L. Zhong, M. F. Rahman, W. Y. Hu, and K. Lim, "Analysis of direct torque

control in permanent magnet synchronous motor drives," IEEE Transactions
on Power Electronics, vol. 12, no. 3, pp. 528-536, 1997.

[28] G. Foo, S. Sayeef, and M. Rahman, "Low-speed and standstill operation of a

sensorless direct torque and flux controlled IPM synchronous motor drive,"
IEEE Transactions on Energy Conversion, vol. 25, no. 1, pp. 25-33, 2010.

[29] T. G. Habetler, F. Profumo, M. Pastorelli, and L. M. Tolbert, "Direct torque

control of induction machines using space vector modulation," IEEE
Transactions on Industry Applications, vol. 28, no. 5, pp. 1045-1053, 1992.

[30] G. S. Buja and M. P. Kazmierkowski, "Direct torque control of PWM inverter-

fed AC motors-a survey," IEEE Transactions on Industrial Electronics, vol.
51, no. 4, pp. 744-757, 2004.

[31] P. Correa, M. Pacas, and J. Rodriguez, "Predictive Torque Control for

Inverter-Fed Induction Machines," IEEE Transactions on Industrial
Electronics, vol. 54, no. 2, pp. 1073-1079, 2007.

[32] H. Miranda, P. Cortés, J. I. Yuz, and J. Rodríguez, "Predictive torque control

of induction machines based on state-space models," IEEE Transactions on
Industrial Electronics, vol. 56, no. 6, pp. 1916-1924, 2009.

[33] W. Fengxiang, L. Shihua, M. Xuezhu, X. Wei, J. Rodriguez, and R. M. Kennel,

"Model-Based Predictive Direct Control Strategies for Electrical Drives: An
Experimental Evaluation of PTC and PCC Methods," IEEE Transactions on
Industrial Informatics, vol. 11, no. 3, pp. 671-681, 2015.

[34] T. D. Nguyen, "Dual air-gap axial flux permanent magnet machines for
flywheel energy storage systems," Ph.D, Nanyang Technological University,
2012.

[35] M. Koc, T. Sun, and J. Wang, "Performance improvement of direct torque

controlled interior mounted permanent magnet drives by employing a linear
combination of current and voltage based flux observers," IET Power

122
BIBLIOGRAPHY

Electronics, vol. 9, no. 10, pp. 2052-2059, 2016.

[36] L. Wang, S. Chai, D. Yoo, L. Gan, and K. Ng, PID and predictive control of
electrical drives and power converters using MATLAB/Simulink. John Wiley
& Sons, 2015.

[37] T. D. Nguyen, G. F. H. Beng, K.-J. Tseng, D. M. Vilathgamuwa, and X. Zhang,

"Modeling and position-sensorless control of a dual-airgap axial flux
permanent magnet machine for flywheel energy storage systems," Journal of
Power Electronics, vol. 12, no. 5, p. 10, 2012.

[38] T. Abe, T. G. Habetler, F. Profumo, and G. Griva, "Evaluation of a high

performance induction motor drive using direct torque control," in Power
Conversion Conference, 1993. Yokohama 1993., Conference Record of the,
1993, pp. 444-449: IEEE.

[39] N. R. N. Idris and A. H. M. Yatim, "Direct torque control of induction

machines with constant switching frequency and reduced torque ripple,"
IEEE Transactions on Industrial Electronics, vol. 51, no. 4, pp. 758-767, 2004.

[40] Y. Zhang and J. Zhu, "Direct torque control of permanent magnet synchronous
motor with reduced torque ripple and commutation frequency," IEEE
Transactions on Power Electronics, vol. 26, no. 1, pp. 235-248, 2011.

[41] L. Tang, L. Zhong, M. F. Rahman, and Y. Hu, "A novel direct torque
controlled interior permanent magnet synchronous machine drive with low
ripple in flux and torque and fixed switching frequency," IEEE Transactions
on Power Electronics, vol. 19, no. 2, pp. 346-354, 2004.

[42] Y. Zhang and H. Yang, "Model predictive torque control of induction motor
drives with optimal duty cycle control," IEEE Transactions on Power
Electronics, vol. 29, no. 12, pp. 6593-6603, 2014.

[43] R. Morales-Caporal and M. Pacas, "Encoderless predictive direct torque

control for synchronous reluctance machines at very low and zero speed,"
IEEE Transactions on Industrial Electronics, vol. 55, no. 12, pp. 4408-4416,
2008.

[44] X. Zhang and G. H. B. Foo, "A constant switching frequency-based direct

torque control method for interior permanent-magnet synchronous motor
drives," IEEE/ASME Transactions on Mechatronics, vol. 21, no. 3, pp. 1445-
1456, 2016.

[45] I. Boldea, M. C. Paicu, G.-D. Andreescu, and F. Blaabjerg, "“Active Flux”

DTFC-SVM sensorless control of IPMSM," IEEE Transactions on Energy
Conversion, vol. 24, no. 2, pp. 314-322, 2009.

[46] J. S. Lee, C.-H. Choi, J.-K. Seok, and R. D. Lorenz, "Deadbeat-direct torque
and flux control of interior permanent magnet synchronous machines with
discrete time stator current and stator flux linkage observer," IEEE

123
BIBLIOGRAPHY

Transactions on Industry Applications, vol. 47, no. 4, pp. 1749-1758, 2011.

[47] D. Xiao and M. F. Rahman, "Sensorless direct torque and flux controlled IPM
synchronous machine fed by matrix converter over a wide speed range," IEEE
Transactions on Industrial Informatics, vol. 9, no. 4, pp. 1855-1867, 2013.

[48] Y. Zhang, J. Zhu, and W. Xu, "Analysis of one step delay in direct torque
control of permanent magnet synchronous motor and its remedies," in
Electrical Machines and Systems (ICEMS), 2010 International Conference on,
2010, pp. 792-797: IEEE.

[49] W. Xie et al., "Finite-control-set model predictive torque control with a

deadbeat solution for PMSM drives," IEEE Transactions on Industrial
Electronics, vol. 62, no. 9, pp. 5402-5410, 2015.

[50] M. Habibullah, D. D.-C. Lu, D. Xiao, J. E. Fletcher, and M. F. Rahman, "Low

complexity predictive torque control strategies for a three-level inverter
driven induction motor," IET Electric Power Applications, vol. 11, no. 5, pp.
776-783, 2017.

[51] M. Habibullah, D. D.-C. Lu, D. Xiao, and M. F. Rahman, "A simplified finite-
state predictive direct torque control for induction motor drive," IEEE
Transactions on Industrial Electronics, vol. 63, no. 6, pp. 3964-3975, 2016.

[52] Y. Wang et al., "Deadbeat model-predictive torque control with discrete

space-vector modulation for PMSM drives," IEEE Transactions on Industrial
Electronics, vol. 64, no. 5, pp. 3537-3547, 2017.

[53] G. Buja, M. Candela, and R. Menis, "A novel direct control scheme for SVM
inverter-fed induction motor drives," in Industrial Electronics, 1999. ISIE'99.
Proceedings of the IEEE International Symposium on, 1999, vol. 3, pp. 1267-
1272: IEEE.

[54] R. Nalepa and T. Orlowska-Kowalska, "Optimum trajectory control of the

current vector of a nonsalient-pole PMSM in the field-weakening region,"
IEEE Transactions on Industrial Electronics, vol. 59, no. 7, pp. 2867-2876,
2012.

[55] D. Lu and N. C. Kar, "A review of flux-weakening control in permanent

magnet synchronous machines," in Vehicle Power and Propulsion Conference
(VPPC), 2010 IEEE, 2010, pp. 1-6: IEEE.

[56] J. Simanek, J. Novak, O. Cerny, and R. Dolecek, "FOC and flux weakening
for traction drive with permanent magnet synchronous motor," in Industrial
Electronics, 2008. ISIE 2008. IEEE International Symposium on, 2008, pp.
753-758: IEEE.

[57] K. Jang-Mok and S. Seung-Ki, "Speed control of interior permanent magnet

synchronous motor drive for the flux weakening operation," IEEE
Transactions on Industry Applications, vol. 33, no. 1, pp. 43-48, 1997.

124
BIBLIOGRAPHY

[58] M. F. Rahman, L. Zhong, and K. W. Lim, "A direct torque-controlled interior

permanent magnet synchronous motor drive incorporating field weakening,"
IEEE Transactions on Industry Applications, vol. 34, no. 6, pp. 1246-1253,
1998.

[59] Y. Inoue, S. Morimoto, and M. Sanada, "Comparative study of PMSM drive

systems based on current control and direct torque control in flux-weakening
control region," in Electric Machines & Drives Conference (IEMDC), 2011
IEEE International, 2011, pp. 1094-1099: IEEE.

[60] P. L. Jansen and R. D. Lorenz, "A physically insightful approach to the design
and accuracy assessment of flux observers for field oriented induction
machine drives," in Conference Record of the 1992 IEEE Industry
Applications Society Annual Meeting, 1992, pp. 570-577 vol.1.

[61] G.-D. Andreescu, C. I. Pitic, F. Blaabjerg, and I. Boldea, "Combined flux

observer with signal injection enhancement for wide speed range sensorless
direct torque control of IPMSM drives," IEEE Transactions on Energy
Conversion, vol. 23, no. 2, pp. 393-402, 2008.

[62] I. Boldea, M. C. Paicu, and G. D. Andreescu, "Active Flux Concept for

Motion-Sensorless Unified AC Drives," IEEE Transactions on Power
Electronics, vol. 23, no. 5, pp. 2612-2618, 2008.

[63] M. Paicu, I. Boldea, G.-D. Andreescu, and F. Blaabjerg, "Very low speed
performance of active flux based sensorless control: interior permanent
magnet synchronous motor vector control versus direct torque and flux
control," IET Electric Power Applications, vol. 3, no. 6, pp. 551-561, 2009.

[64] G. Foo and M. Rahman, "Sensorless direct torque and flux-controlled IPM
synchronous motor drive at very low speed without signal injection," IEEE
Transactions on Industrial Electronics, vol. 57, no. 1, pp. 395-403, 2010.

[65] W. Xu and R. D. Lorenz, "Reduced Parameter Sensitivity Stator Flux Linkage

Observer in Deadbeat-Direct Torque and Flux Control for IPMSMs," IEEE
Transactions on Industry Applications, vol. 50, no. 4, pp. 2626-2636, 2014.

[66] M. Koç, J. Wang, and T. Sun, "An Inverter Nonlinearity-Independent Flux

Observer for Direct Torque-Controlled High-Performance Interior Permanent
Magnet Brushless AC Drives," IEEE Transactions on Power Electronics, vol.
32, no. 1, pp. 490-502, 2017.

[67] I. Boldea et al., "DTFC-SVM motion-sensorless control of a PM-assisted

reluctance synchronous machine as starter-alternator for hybrid electric
vehicles," IEEE Transactions on Power Electronics, vol. 21, no. 3, pp. 711-
719, 2006.

[68] K. Jang-Hwan, C. Jong-Woo, and S. Seung-Ki, "Novel rotor-flux observer

using observer characteristic function in complex vector space for field-

125
BIBLIOGRAPHY

oriented induction motor drives," IEEE Transactions on Industry Applications,

vol. 38, no. 5, pp. 1334-1343, 2002.

[69] B. Singh, S. Jain, and S. Dwivedi, "Torque ripple reduction technique with
improved flux response for a direct torque control induction motor drive," IET
Power Electronics, vol. 6, no. 2, pp. 326-342, 2013.

[70] X. Zhang and J. Yang, "A DC-link voltage fast control strategy for high-speed
PMSM/G in flywheel energy storage system," IEEE Transactions on Industry
Applications, vol. 54, no. 2, pp. 1671-1679, 2017.

[71] D. Zhang and J. G. Jiang, "Sensorless control of PMSM for DC micro-grid

flywheel energy storage based on EKF," The Journal of Engineering, vol.
2019, no. 16, pp. 1227-1231, 2019.

[72] J. S. Lee and R. D. Lorenz, "Robustness Analysis of Deadbeat-Direct Torque

and Flux Control for IPMSM Drives," IEEE Transactions on Industrial
Electronics, vol. 63, no. 5, pp. 2775-2784, 2016.

[73] A. Tripathi, A. M. Khambadkone, and S. K. Panda, "Dynamic control of

torque in overmodulation and in the field weakening region," IEEE
Transactions on Power Electronics, vol. 21, no. 4, pp. 1091-1098, 2006.

[74] T. Sun and J. Wang, "Extension of virtual-signal-injection-based MTPA

control for interior permanent-magnet synchronous machine drives into the
field-weakening region," IEEE Transactions on Industrial Electronics, vol.
62, no. 11, pp. 6809-6817, 2015.

[75] K. Chen, Y. Sun, and B. Liu, "Interior permanent magnet synchronous motor
linear field-weakening control," IEEE Transactions on Energy Conversion,
vol. 31, no. 1, pp. 159-164, 2016.

[76] T. Sun, J. Wang, and M. Koc, "Self-Learning Direct Flux Vector Control of
Interior Permanent-Magnet Machine Drives," IEEE Transactions on Power
Electronics, vol. 32, no. 6, pp. 4652-4662, 2017.

[77] E. Levi and M. Wang, "A speed estimator for high performance sensorless
control of induction motors in the field weakening region," IEEE
Transactions on Power Electronics, vol. 17, no. 3, pp. 365-378, 2002.

[78] S.-Y. Jung, C. C. Mi, and K. Nam, "Torque control of IPMSM in the field-
weakening region with improved DC-link voltage utilization," IEEE
Transactions on Industrial Electronics, vol. 62, no. 6, pp. 3380-3387, 2015.

[79] P. Matic and S. N. Vukosavic, "Voltage angle direct torque control of

induction machine in field-weakening regime," IET Electric Power
Applications, vol. 5, no. 5, pp. 404-414, 2011.

[80] P. Cortes, J. Rodriguez, C. Silva, and A. Flores, "Delay compensation in

model predictive current control of a three-phase inverter," IEEE

126
BIBLIOGRAPHY

Transactions on Industrial Electronics, vol. 59, no. 2, pp. 1323-1325, 2012.

[81] N. R. N. Idris and A. H. M. Yatim, "An improved stator flux estimation in

steady-state operation for direct torque control of induction machines," IEEE
Transactions on Industry Applications, vol. 38, no. 1, pp. 110-116, 2002.

[82] J. Hu and B. Wu, "New integration algorithms for estimating motor flux over
a wide speed range," IEEE Transactions on Power Electronics, vol. 13, no. 5,
pp. 969-977, 1998.

[83] M. H. Vafaie, B. M. Dehkordi, P. Moallem, and A. Kiyoumarsi, "A new

predictive direct torque control method for improving both steady-state and
transient-state operations of the PMSM," IEEE Transactions on Power
Electronics, vol. 31, no. 5, pp. 3738-3753, 2016.

[84] Z. Zhou, C. Xia, Y. Yan, Z. Wang, and T. Shi, "Torque ripple minimization of
predictive torque control for PMSM with extended control set," IEEE
Transactions on Industrial Electronics, vol. 64, no. 9, pp. 6930-6939, 2017.

[85] M. R. Nikzad, B. Asaei, and S. O. Ahmadi, "Discrete Duty-Cycle-Control

Method for Direct Torque Control of Induction Motor Drives With Model
Predictive Solution," IEEE Transactions on Power Electronics, vol. 33, no. 3,
pp. 2317-2329, 2018.

[86] X. Zhang, G. Foo, and T. Ngo, "An Improved Finite-set Model Predictive
Torque Control for Interior Permanent Magnet Synchronous Motor Drives,"
in 2018 International Power Electronics Conference (IPEC-Niigata 2018-
ECCE Asia), 2018, pp. 1724-1729: IEEE.

[87] G. H. B. Foo and M. Rahman, "Direct torque control of an IPM-synchronous

motor drive at very low speed using a sliding-mode stator flux observer,"
IEEE Transactions on Power Electronics, vol. 25, no. 4, pp. 933-942, 2010.

127