Learning Based Model Predictive Control (LBMPC) For Optimum Control of Asynchronous Motor

Download as pdf or txt
Download as pdf or txt
You are on page 1of 91

Learning Based Model Predictive Control (LBMPC) for

Optimum Control of Asynchronous Motor

B.E (Electrical), Batch 2014-15


Final Year Project Report
Prepared By:

Muhammad Affan EE-14002


Syed Hamza Tariq EE-14005
Raphea Siddiqui EE-14026
Mehmood Ali EE-14038

INTERNAL ADVISOR EXTERNAL ADVISOR

Dr. Riazuddin Engr. Waleed Nasim

DEPARTMENT OF ELECTRICAL ENGINEERING

N.E.D UNIVERSITY OF ENGINEERING & TECHNOLOGY


KARACHI-75270
ii

ACKNOWLEDGEMENT

We are grateful to Almighty Allah for providing us the ability and aptness to work
as a dedicated team on this project and completing it on time with the assigned task. We
are indebted to our internal, Prof. Dr. Riazuddin for his continuous guidance and support
in technical and research aspects. We are thankful to the prestigious institution NED
University of Engineering and Technology, Electrical Engineering Department and all its
faculty members.

We are highly thankful to our external Engr. Waleed Naveed for his guidance
regarding the industrial applications of project and its future scope.

We would like to thank our parents for their encouragement and generous
collaboration throughout the project.
iii

ABSTRACT

The increasing importance of energy efficiency has made the industries think that
how to get the best from their motors for a particular application. Keeping this thing in
mind, the purpose of doing this project is to design an efficient and cost-effective solution
for industries to improve their motors’ performance (i.e. speed, torque etc.).

Motor industry is one of the major industries in the technological era; the modern
motors must fulfill the standards from both technical and economical perspective. About
90 percent of industry uses induction motors because of robustness (able to work in
extreme environmental conditions), economical, high efficiency and high power to weight
ratio. However, the control of induction motor is a challenging mathematical task because
of its nonlinear torque speed curve. The torque greatly depends on the slip, speed variation
depends on the load .With the development of power electronics, motor control techniques
like scalar method have evolved through the years but inability of handling load and noise
transients have forced the researchers to develop precise, mathematically complex
techniques like Field Oriented Control (FOC) and Direct Torque Control (DTC) has been
used in industries but some of the technical limitations have urged the researchers to
develop some advance algorithms like MPC. Model predictive Control is technique
inspired from 'human perception/anticipation' that uses model, constraints, performance
index and receding horizon to optimize the control problem using mathematical model of
that system.
iv

Table of Contents

List of Figures .................................................................................................................vii


List of Tables ................................................................................................................... ix
Nomenclature ................................................................................................................... x
CHAPTER 1 ..................................................................................................................... 1
INTRODUCTION............................................................................................................ 1
Brief Introduction............................................................................................................1
Problem Statement .........................................................................................................1
Objectives .......................................................................................................................2
Scope of the Work ..........................................................................................................2
Significance of Research .................................................................................................3
CHAPTER 2 ..................................................................................................................... 4
LITERATURE REVIEW ............................................................................................... 4
Induction Motor Working ...............................................................................................4
Reference Frame .............................................................................................................4
2.2.1 Direct Quadrature Zero (Dq0) Transformation .......................................................5
Advanced Vector Control Techniques ............................................................................8
2.3.1 Linear Quadratic Regulator (LQR) .........................................................................9
2.3.2 Field Oriented Control (FOC) .............................................................................. 10
2.3.3 Direct Torque Control (DTC) .............................................................................. 11
2.3.4 Model Predictive Control (MPC)......................................................................... 14
Kalman Filter (An Observer)......................................................................................... 17
Effect of Sampling Period on Control System .............................................................. 17
Effect of Weighting Matrices on System’s Performance ............................................. 19
2.6.1 Effect of Input Weighting Matrix (R) .................................................................. 19
2.6.2 Effect of Output Weighting Matrix (Q) ............................................................... 20
Artificial Neural Network ............................................................................................. 22
2.7.1 Introduction to Artificial Neural Networks .......................................................... 22
2.7.2 Equivalent Electrical Model of a Neuron ............................................................ 22
2.7.3 Types of Activation Functions ............................................................................. 23
Neuromorphic Algorithms ........................................................................................... 26
v

2.8.1 Brain Parts Concerned with Emotions ................................................................. 26


2.8.2 Control Mechanism.............................................................................................. 27
CHAPTER 3 ................................................................................................................... 28
METHODOLOGY/ OBSERVATION AND CALCULATION ................................ 28
Dynamic Model of an Induction Motor ....................................................................... 28
Control Derivation ....................................................................................................... 38
CHAPTER 4 ................................................................................................................... 42
RESULTS ....................................................................................................................... 42
Comparison of MPC with PID, Open Loop and DLQR .................................................. 42
Control of PMSM using MPC........................................................................................ 43
Induction Motor Parameters ....................................................................................... 43
Open Loop Simulation of Induction Motor .................................................................. 45
Simulation of Direct Torque Control (DTC) of Induction Motor .................................. 46
4.5.1 Fixed Speed and Fixed Torque ............................................................................ 46
4.5.2 Fixed Speed and Step changes in Torque ............................................................ 47
4.5.3 Step changes in Speed at Fixed Torque ............................................................... 48
4.5.4 Current THD at fixed Speed and Torque ............................................................. 49
4.6 Simulation of Field Oriented Control (FOC) of Induction Motor ................................. 50
4.6.1 Fixed Speed and Fixed Torque ............................................................................ 50
4.6.2 Fixed Speed and Step changes in Torque ............................................................ 51
4.6.3 Step changes in Speed at fixed Torque ................................................................ 52
4.6.4 Current THD at fixed Speed and Torque ............................................................. 53
4.7 Simulation of Rotating Frame Finite Control Set-Model Predictive Control (FCS-MPC)
of Induction Motor .................................................................................................................. 54
4.7.1 Fixed Speed and Fixed Torque ............................................................................ 54
4.7.2 Fixed Speed and Step changes in Torque ............................................................ 55
4.7.3 Step changes in Speed at fixed Torque ................................................................ 56
4.7.4 Current THD at fixed Speed and Torque ............................................................. 57
4.8 Simulation of Stationary Frame Finite Control Set-Model Predictive Control(FCS-MPC)
of Induction Motor .................................................................................................................. 58
4.8.1 Fixed Speed and Fixed Torque ............................................................................ 58
4.8.2 Fixed Speed and Step changes in Torque ............................................................ 59
4.8.3 Step changes in Speed at fixed Torque ................................................................ 60
4.8.4 Current THD at fixed Speed and Torque ............................................................. 61
vi

Learning Based Model Predictive Control of Induction Motor.................................... 61


CHAPTER 5 ................................................................................................................... 63
CONCLUSIONS, FUTURE RECOMMENDATIONS .............................................. 63
REFERENCES ............................................................................................................... 64
vii

List of Figures

Figure 2. 1 History of Transformation Techniques .........................................................................5


Figure 2. 2 Stator Current in abc, αβ and dq Reference Frame ......................................................6
Figure 2. 3 Relation of Stator Currect Space Vector with Phase Currents......................................7
Figure 2. 4 Current Space Vector in Stationary and Rotating Reference Frame ............................8
Figure 2. 5 Flow Diagram of FOC.................................................................................................. 10
Figure 2. 6 Working Diagram of DTC............................................................................................ 12
Figure 2. 7 A Voltage Source Inverter Feeding Three Coils.......................................................... 13
Figure 2. 8 Space Vector Diagram for 2-level Inverter................................................................. 13
Figure 2. 9 Generations of MPC ................................................................................................... 15
Figure 2. 10 Block Diagram of MPC ............................................................................................. 16
Figure 2. 11 Receding Horizon Control ........................................................................................ 16
Figure 2. 12 Kalman Filtering ....................................................................................................... 17
Figure 2. 13 Effect of Input Weighting Matrix on System............................................................ 20
Figure 2. 14 Effect of Output Weighting Matrix on System......................................................... 21
Figure 2. 15 Equivalent Computational Model of Neuron........................................................... 22
Figure 2. 16 Graphical Representation of Threshold Function .................................................... 24
Figure 2. 17 Graphical Representation of Signum Function ........................................................ 24
Figure 2. 18 Graphical Representation of Sigmoid Function ....................................................... 25
Figure 2. 19 Graphical Representation of Tangent Hyperbolic Function .................................... 26
Figure 2. 20 A Learning Organism ................................................................................................ 27

Figure 3. 1 Induction Motor Model ............................................................................................. 28

Figure 4. 1 Comparison between MPC, PID and DLQR for 2nd Order System on MATLAB .......... 42
Figure 4. 2 Simulation of Model Predictive Control of PMSM on MATLAB ................................. 43
Figure 4. 3 Simulation of an Open Loop IM on MATLAB ............................................................. 45
Figure 4. 4 (a) Simulink Block Diagram of DTC ............................................................................. 46
Figure 4. 5 (b) Simulink Block Diagram of DTC............................................................................. 46
Figure 4. 6 Fixed Reference Tracking by Induction Motor Controlled by DTC ............................ 47
Figure 4. 7 Impact of Torque Changes in Induction Machine Speed Controlled by DTC ............. 47
Figure 4. 8 Impact of Torque Changes in Induction Machine speed Controlled by DTC (Zoomed)
..................................................................................................................................................... 48
Figure 4. 9 Step Reference Speed tracking of DTC Controlled Induction Motor ......................... 48
Figure 4. 10 Absolute Error of above Reference Staircase Speed Tracking ................................. 49
Figure 4. 11 Current Frequency Spectrum of Induction Machine Controlled by DTC ................. 49
Figure 4. 12 Simulink Block Diagram of FOC ................................................................................ 50
Figure 4. 13 Unmasked-Simulink Block Diagram of FOC ............................................................. 50
Figure 4. 14 Fixed Reference Tracking by Induction Machine Controlled by FOC....................... 51
Figure 4. 15 Impact of Torque Changes on Induction Machine speed when Controlled by FOC 51
viii

Figure 4. 16 Impact of Torque Changes on Induction Machine Speed when Controlled by FOC
(Zoomed)...................................................................................................................................... 52
Figure 4. 17 Step Reference Speed Tracking of FOC Controlled Induction Motor ...................... 52
Figure 4. 18 Absolute Error of above Reference Staircase Speed Tracking ................................. 53
Figure 4. 19 Current Frequency Spectrum of Induction Machine Controlled by FOC ................. 53
Figure 4. 20 Simulink Block Diagram of MPC (αβ) ....................................................................... 54
Figure 4. 21 Fixed Reference Tracking by Induction Machine Controlled by MPC (αβ) .............. 54
Figure 4. 22 Impact of Torque Changes on Induction Machine Speed when Controlled by MPC
(αβ)............................................................................................................................................... 55
Figure 4. 23 Impact of Torque Changes on Induction Machine Speed when Controlled by MPC
(αβ) (Zoomed) .............................................................................................................................. 55
Figure 4. 24 Step Reference Speed Tracking of MPC (αβ) Controlled Induction Motor.............. 56
Figure 4. 25 Absolute Error of above Reference Staircase Speed Tracking ................................. 56
Figure 4. 26 Current Frequency Spectrum of Induction Machine Controlled by MPC (αβ) ........ 57
Figure 4. 27 Simulink Block Diagram of MPC (dq) ....................................................................... 58
Figure 4. 28 Fixed Reference Tracking by Induction Machine Controlled by MPC (dq) .............. 58
Figure 4. 29 Impact of Torque Changes on Induction Machine Speed when Controlled by MPC
(dq) ............................................................................................................................................... 59
Figure 4. 30 Impact of Torque Changes on Induction Machine Speed when Controlled by MPC
(dq) (Zoomed) .............................................................................................................................. 59
Figure 4. 31 Step Reference Speed tracking of MPC (dq) controlled Induction Motor ............... 60
Figure 4. 32 Absolute error of above reference staircase speed tracking ................................... 60
Figure 4. 33 Current Frequency Spectrum of Induction Machine Controlled by MPC (dq) ........ 61
Figure 4. 34 Learning Predictive Scheme ..................................................................................... 62
Figure 4. 35 Reference vs Actual Speed....................................................................................... 62
ix

List of Tables

Table 2. 1 Choice of Sampling Period for Digital Control System ................................................ 18

Table 4. 1 Parameters of Induction Motor .................................................................................. 43


x

Nomenclature

𝑣𝑎𝑠 = 𝑎 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑣𝑜𝑙𝑡𝑎𝑔𝑒 𝑖𝑛 𝑎𝑏𝑐 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒


𝑣𝑏𝑠 = 𝑏 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑣𝑜𝑙𝑡𝑎𝑔𝑒 𝑖𝑛 𝑎𝑏𝑐 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
𝑣𝑐𝑠 = 𝑐 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑣𝑜𝑙𝑡𝑎𝑔𝑒 𝑖𝑛 𝑎𝑏𝑐 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
γas = 𝑎 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑖𝑛 𝑎𝑏𝑐 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
γbs = 𝑏 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑖𝑛 𝑎𝑏𝑐 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
γcs = 𝑐 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑖𝑛 𝑎𝑏𝑐 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
𝑖𝑎𝑠 = 𝑎 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑖𝑛 𝑎𝑏𝑐 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
𝑖𝑏𝑠 = 𝑏 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑖𝑛 𝑎𝑏𝑐 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
𝑖𝑐𝑠 = 𝑐 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑖𝑛 𝑎𝑏𝑐 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
Ψas = 𝑎 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑓𝑙𝑢𝑥 𝑖𝑛 𝑎𝑏𝑐 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
Ψbs = 𝑏 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑓𝑙𝑢𝑥 𝑖𝑛 𝑎𝑏𝑐 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
Ψcs = 𝑐 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑓𝑙𝑢𝑥 𝑖𝑛 𝑎𝑏𝑐 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
𝑅𝑠 = 𝑠𝑡𝑎𝑡𝑜𝑟 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒
⃗⃗⃗
𝑉𝑠 = 𝑠𝑡𝑎𝑡𝑜𝑟 𝑣𝑜𝑙𝑡𝑎𝑔𝑒 𝑖𝑛 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟 𝑟𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛
⃗⃗⃗⃗⃗
Ψ𝑠 = 𝑠𝑡𝑎𝑡𝑜𝑟 𝑓𝑙𝑢𝑥 𝑖𝑛 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟 𝑟𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛
⃗⃗𝐼𝑠 = 𝑠𝑡𝑎𝑡𝑜𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑖𝑛 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟 𝑟𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛

⃗⃗⃗⃗⃗⃗
𝑉𝑟 ∗= 𝑟𝑜𝑡𝑜𝑟 𝑣𝑜𝑙𝑡𝑎𝑔𝑒 𝑖𝑛 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟 𝑟𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛
𝑅𝑟 = 𝑟𝑜𝑡𝑜𝑟 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒
⃗⃗⃗⃗⃗⃗
𝐼𝑟 ∗= 𝑟𝑜𝑡𝑜𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑖𝑛 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟 𝑟𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛
⃗⃗⃗⃗⃗⃗⃗⃗
Ψr ∗= 𝑟𝑜𝑡𝑜𝑟 𝑓𝑙𝑢𝑥 𝑖𝑛 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟 𝑟𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛
⃗⃗⃗⃗
𝑈𝑟 = 𝑟𝑜𝑡𝑜𝑟 𝑣𝑜𝑙𝑡𝑎𝑔𝑒 𝑟𝑒𝑓𝑒𝑟𝑟𝑒𝑑 𝑡𝑜 𝑠𝑡𝑎𝑡𝑜𝑟 𝑖𝑛 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟 𝑟𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛
⃗⃗𝐼𝑟 = 𝑟𝑜𝑡𝑜𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑟𝑒𝑓𝑒𝑟𝑟𝑒𝑑 𝑡𝑜 𝑠𝑡𝑎𝑡𝑜𝑟 𝑖𝑛 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟 𝑟𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛

⃗⃗⃗⃗
Ψ𝑟 = 𝑟𝑜𝑡𝑜𝑟 𝑓𝑙𝑢𝑥 𝑟𝑒𝑓𝑒𝑟𝑟𝑒𝑑 𝑡𝑜 𝑠𝑡𝑎𝑡𝑜𝑟 𝑖𝑛 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟 𝑟𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛
𝐿𝑠 = 𝑠𝑡𝑎𝑡𝑜𝑟 𝑖𝑛𝑑𝑢𝑐𝑡𝑎𝑛𝑐𝑒
xi

𝐿𝑟 = 𝑟𝑜𝑡𝑜𝑟 𝑖𝑛𝑑𝑢𝑐𝑡𝑎𝑛𝑐𝑒
𝐿ℎ = 𝑚𝑢𝑡𝑢𝑎𝑙 𝑖𝑛𝑑𝑢𝑐𝑡𝑎𝑛𝑐𝑒
𝜃𝑒 = 𝑒𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑎𝑙 𝑎𝑛𝑔𝑙𝑒
𝜔𝑒 = 𝑒𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑎𝑙 𝑎𝑛𝑔𝑢𝑙𝑎𝑟 𝑠𝑝𝑒𝑒𝑑
⃗⃗⃗⃗⃗
𝑢𝑠 ′ = 𝑠𝑡𝑎𝑡𝑜𝑟 𝑣𝑜𝑙𝑡𝑎𝑔𝑒 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟 𝑖𝑛 𝑑𝑞 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
⃗⃗⃗⃗
𝑖𝑠 ′ = 𝑠𝑡𝑎𝑡𝑜𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟 𝑖𝑛 𝑑𝑞 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
⃗⃗⃗⃗⃗⃗
𝛹𝑟 ′ = 𝑠𝑡𝑎𝑡𝑜𝑟 𝑓𝑙𝑢𝑥 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟 𝑖𝑛 𝑑𝑞 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑓𝑟𝑎𝑚𝑒
𝑢𝑠𝑑 = 𝑑 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑣𝑜𝑙𝑡𝑎𝑔𝑒 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟
𝑢𝑠𝑞 = 𝑞 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑣𝑜𝑙𝑡𝑎𝑔𝑒 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟

𝑖𝑠𝑑 = 𝑑 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟


𝑖𝑠𝑞 = 𝑞 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟

𝛹𝑟𝑑 = 𝑑 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑟𝑜𝑡𝑜𝑟 𝑓𝑙𝑢𝑥 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟


𝛹𝑟𝑞 = 𝑞 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑟𝑜𝑡𝑜𝑟 𝑓𝑙𝑢𝑥 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟

𝑈𝑠𝛼 = 𝛼 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑣𝑜𝑙𝑡𝑎𝑔𝑒 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟


𝑈𝑠𝛽 = 𝛽 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑣𝑜𝑙𝑡𝑎𝑔𝑒 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟

𝐼𝑠𝛼 = 𝛼 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟


𝐼𝑠𝛽 = 𝛽 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑠𝑡𝑎𝑡𝑜𝑟 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟

𝛹𝑟𝛼 = 𝛼 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑟𝑜𝑡𝑜𝑟 𝑓𝑙𝑢𝑥 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟


𝛹𝑟𝛽 = 𝛽 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑜𝑓 𝑟𝑜𝑡𝑜𝑟 𝑓𝑙𝑢𝑥 𝑠𝑝𝑎𝑐𝑒 𝑣𝑒𝑐𝑡𝑜𝑟
𝑇𝑠 = 𝑆𝑎𝑚𝑝𝑙𝑖𝑛𝑔 𝑇𝑖𝑚𝑒

MPC = Model Predictive Control


DTC = Direct Torque Control
FOC = Field Oriented Control
THD = Total Harmonic Distortions
Rpm= Revolutions per minute
DLQR=Discrete linear Quadratic Regulator
PMSM = Permanent Magnet Synchronous Motor
SVPWM =Space Vector Pulse Width Modulation
xii
CHAPTER 1

INTRODUCTION

Brief Introduction

As AC motors are dominating industrial applications, nowadays, because of many


reasons out of which the most notable one is their construction without slip rings or
commutators that make AC motors cheap and less prone to maintenance. These
advantages have determined induction motors as the execution element in development of
the electrical drives for all related aspects: starting, braking, speed reversal, speed change,
etc. Therefore, the theme of this project is to develop Learning Based Model Predictive
Control (LBMPC) strategy for vector control of an Induction Motor (IM) that provides
better performance and robustness.

AC induction motors are preferred because of their simple electrical construction


without brushes and commutators, hence there is lesser chance of sparks and fumes and
IM can work in toughest, closed and heat prone environment. These factors make the
control of induction motor difficult and therefore require some high-performance control
algorithms such as vector control. In vector control method, the motor model is valid for
transient conditions also. To achieve this design, it is important to know the concepts of
complex space vectors, Clarke and Park transformations as they form the basis of vector
control methods. Also, knowledge about working of various intelligent controllers will
play key role in successful execution of simulation of project.

Problem Statement

The major problem in the control of an induction motor is that during its control
the speed and torque cannot be controlled independently, vector control algorithm gives
the advantage of controlling the induction motor as the DC motor. The other techniques
of motor control are:

1. V\F Scalar Control


2. DTC (Direct Torque Control)
3. FOC (Field Oriented Control)
4. MPC
2

These techniques are less efficient because of the following reasons:

• Since scalar control (like Volt/Hertz) strategy for induction motors has some
limitations like it generates oscillations on the produced torque because motor
model is considered just for steady state and not for best performance during
transients. While in vector control method, the motor model is valid for transient
conditions also
• High noise at very low speed in DTC
• current and torque ripples in steady state in DTC
• Variable switching frequency behavior in DTC
• Poor dynamic performance of FOC
• High processing requirements in FOC
• High complexity due to additional hardware in FOC

Objectives

With the development of electronics technology and information technology, more


powerful digital signal processors (DSP) and microprocessors are available at reduced
cost. As a result, very sophisticated and advanced modern control theory can be applied
to AC motor drives, like Model Predictive Control (MPC).

Our objective is to design an efficient scheme of speed control: Model Predictive


Control (MPC). For high performance induction motor drives usually torque and stator
flux are selected as the control variables. We will be performing a simulation on
MATLAB Simulink, for controlling the speed of an induction motor using Model
Predictive Control and analyze the results obtained through this simulation. Both the
transient and steady state response will be examined. Since MPC relies on the accuracy of
system model and machine parameters, therefore proper values must be provided to the
model for accurate simulation results. The success of model based control strategies such
as MPC depends strongly on the availability of a reasonable accurate process model.
Consequently, model development is the most critical step in applying the MPC.

Scope of the Work

Induction motor is one of the widely used machine, about 90 percent of industrial
motors are induction. The above coated figure best explains the wide range of utilization
of induction motor. It issued almost in every industrial setup i.e. pumping, fans,
compressed air systems, conveyers, elevators, transportation and in all types of process
3

industries. The problem with the use of induction motor is the low starting torque as well
as high starting current which are undesirable, due to low starting torque it cannot be used
in lifting heavy loads and traction application. The performance of induction motor can
be enhanced by modern VFDs or VSDs.

Significance of Research

The use of conventional VFDs has been proved to be helpful for the years but
advancement in industrial arena demands the advancement in the motor industry which is
one of the biggest candidate of industrial performance. The use of induction motors over
the conventional DC motors has urged the researchers to develop the precise algorithms
and sophisticated techniques to improve the performance of induction motors. Although
the conventional VFDs are helpful from the perspective of speed control but they are in
efficient as far as dynamic performance and handling of transients are concerned, vector
control gives additional benefit of dynamic and transient response as well as the
independent control of speed and torque.
CHAPTER 2

LITERATURE REVIEW

Induction Motor Working

Induction motor works on the principle of electromagnetic induction. The supply


is given to the rotor which creates rotating magnetic field. This rotating magnetic field
when cuts the rotor bars voltage is induced and the current starts to flow in the closed
looped bars. The flow of current creates another magnetic field that is the rotor magnetic
field and the interaction of these two fields creates the torque and the motor starts rotating
in the same direction as that of RMF.IM do not need external supply of DC current to the
rotor hence the structure and construction of motor is greatly simplified the maintenance
cost of brushes and commutators is reduced [1].

Reference Frame

The dynamic model of induction motor, governing its performance, is based on


current and voltage equations that are differential in nature[4]. The presence of various
inductances (stator, rotor & mutual) in such time varying equations make it difficult to
examine the IM model because inductances are dependent on position of rotor and the
magnetic reluctance keeps varying due to the relative motion of electric circuit. Hence,
some remedy is required to cope up with this issue, therefore, transformation is used for
decoupling of variables and IM is modeled on an acceptable reference frame.
5

Several mathematical transformation techniques had been used in the analysis of


electrical machines. An overview of those can be seen in the following figure:

R. H. PARK TRANSFORMATION (1920s)


Rotor Reference Frame for Synchronous Machine

H. C. STANLEY TRANSFORMATION (1930s)


Stationary Reference Frame for Induction Machine

G. KRON TRANSFORMATION
Synchronous Reference Frame for Induction Machine

D. S. BRERETON TRANSFORMATION
Rotor Reference Frame for Induction Machine

KRAUSE AND THOMAS TRANSFORMATION


Arbitrary Reference Frame

E. CLARKE TRANSFORMATION

Figure 2. 1 History of Transformation Techniques

Advantages of Reference Frame Transformation:

• Magnetic coupling between phases is removed


• The representation of both rotor and stator in the same reference frame makes
mutual inductance independent of rotor position
• Comparatively less number of voltage equations need to be dealt with
• Change of voltage equations from time-variant to time-invariant
• Ease in implementation of control algorithms on digital signal processors

2.2.1 Direct Quadrature Zero (Dq0) Transformation

In electrical engineering, three phase circuits make use of DQ0 transform. It makes
their analysis and calculation more comprehensible because AC signals are treated as DC
signals by rotating the reference frame. Basically, DQ0 is the product of Clarke and Park
transforms[2]. First, the Clarke transform converts three phase quantities (say currents; 𝑖𝑎 ,
𝑖𝑏 and 𝑖𝑐 ) into two-phase orthogonal stator axis i.e. 𝑖𝛼 and 𝑖𝛽 . Second, in Park transform,
6

those two currents are translated into (d, q) reference frame as 𝑖𝑠𝑑 and 𝑖𝑠𝑞 current
components.

Figure 2. 2 Stator Current in abc, αβ and dq Reference Frame

The above figure shows relationship of stator current in 𝑎𝑏𝑐-reference frame, 𝛼𝛽-
reference frame (stationary) and 𝑑𝑞-reference frame (rotating).

2.2.1.1 Clarke Transformation (3-phase to 2-phase transformation)

The 3-phase quantities of IM’s stator are not independent and can be given by following
equations:

𝑖𝑠𝐴 (𝑡) + 𝑖𝑠𝐵 (𝑡) + 𝑖𝑠𝐶 (𝑡) = 0


𝑣𝑠𝐴 (𝑡) + 𝑣𝑠𝐵 (𝑡) + 𝑣𝑠𝐶 (𝑡) = 0
𝜑𝑠𝐴 (𝑡) + 𝜑𝑠𝐵 (𝑡) + 𝜑𝑠𝐶 (𝑡) = 0

Here;
𝑖𝑠 = stator phase current
𝑣𝑠 = stator phase voltage
𝜑𝑠 = stator phase flux linkages

The conversion of quantities from three to two-phase can be denoted in matrix as well as
compact form by following equations:
7

𝑖 (𝑡)
𝑖𝑠𝛼 (𝑡) 2 1 cos(𝛾) cos(2𝛾) 𝑠𝐴
[ ]= [ ] [𝑖 (𝑡)]
𝑖𝑠𝛽 (𝑡) 3 0 sin(𝛾) sin(2𝛾) 𝑠𝐵
𝑖𝑠𝐶 (𝑡)
𝐢s (𝑡) = 𝑖𝑠𝛼 (𝑡) + 𝑗𝑖𝑠𝛽 (𝑡)
2
𝐢s (𝑡) = [𝑖𝑠𝐴 (𝑡) + 𝒂 𝑖𝑠𝐵 (𝑡) + 𝒂𝟐 𝑖𝑠𝐶 (𝑡)]
3
Here;
2𝜋
𝛾= 3
𝐢s = stator current space vector
𝒂 = vector operator that produces vector rotation of 𝛾

Figure 2. 3 Relation of Stator Currect Space Vector with Phase Currents

In the above figure, showing relationship between stator current space vector and
phase currents, there is an angle 𝛾 between ABC axes. The real-axis (𝑠𝛼) comes together
in position with 𝑠𝐴-axis while the imaginary-axis (𝑗𝑠𝛽) is in quadrature with 𝑠𝛼-axis.
Also, 𝐢s has been illustrated at an arbitrary position.

2.2.1.2 Park Transformation (Vector Rotation)

Since, it is imperative to choose a common reference frame for both stator and
rotor that’s why, in vector rotation, the space vector quantities (current, voltage, flux
linkages) are rotated by a known angle 𝜃 and speed 𝜔 i.e. corresponding phase frequency.
Thus, the current space vector will now be given by:

𝐢𝑑𝑞 = 𝑖𝑠𝑑 + 𝑗𝑖𝑠𝑞 = 𝐢s 𝑒 −𝑗𝜃


8

In matrix form:

𝑖𝑠𝑑 cos(𝜃) sin(𝜃) 𝑖𝑠𝛼


[𝑖 ] = [ ][ ]
𝑠𝑞 −sin(𝜃) cos(𝜃) 𝑖𝑠𝛽

Following figure shows relation of current space vector in stationary and rotating
reference frame. In rotating reference frame (𝑑, 𝑞):

real component of 𝐢s = direct axis component


imaginary component of 𝐢s = quadrature axis component

Figure 2. 4 Current Space Vector in Stationary and Rotating Reference Frame

If 𝐢s is viewed w.r.t the 𝑑𝑞 reference frame then it would appear to be stationary


but if seen from 𝛼𝛽 reference frame, then it could be perceived as rotating along with the
𝑑 and 𝑞 axes.

Advanced Vector Control Techniques

The modern control techniques of motor control have gained popularity in control
arena because of precise control and lower power consumption. The two popular
techniques are:
9

• FOC
• DTC
• Tuned-PID
• FCS-MPC

The control of induction motor could be modeled as that of DC machine by


transformation of three phase time variant into two phase time invariant system by
advanced mathematical transformations like Clarke and Park. The precise control of
machine in both steady state and transient regions could improve the dynamic
performance of induction machine. The classical techniques of motor control possess
some inherent flaws like:

• The equivalent model on which three phase sinusoidal wave is applied is based on
steady state analysis, during the transient period this could induce the high spikes
of voltage and current in the system which could reduce the overall power
conversion efficiency, moreover a designer has to compensate these additional
transients by increasing the size of power components.
• The control of three phase circuit looks like three different single-phase circuits,
instead of single three phase controls.
• The three phase sinusoidal reference variables are difficult to control by PI
controllers because PI controllers will distort the sinusoidal reference

Vector control techniques are the solution to overcome the above shortcomings of
classical drives.

2.3.1 Linear Quadratic Regulator (LQR)

LQR is controller(optimal) which tracks the reference trajectory, the tracking is


quite accurate as compared to conventional controller like PID. An LQR is based on the
receding horizon concept such that future outputs are predicted at every time step to
minimize a global criterion/cost function. By estimating future outputs based on past
outputs, we can better regulate offset in tracking. Usually, the output horizon is infinite.

LQR is an accurate technique to design a controller for precise industrial


applications. The LQR minimizes the cost function by proposing an optimal control. The
cost function is configured by the two weight matrices Q and R, these are cost function
and system input respectively[5].
10

2.3.1.1 LQR for Induction Motor

The control of Induction Motor is difficult because of its multivariable control,


non-linearity and coupled structure. To achieve accurate control performance of position
control of IM, a linear quadratic regulator (LQR) is used.

The model of Induction motor is modelled as second order, in which the


mechanical speed and position are system states.

2.3.2 Field Oriented Control (FOC)

FOC controls the stator current represented by a vector, the three-phase time
variant variables are converted into time invariant quantities[6]. The FOC needs two
references the torque reference and flux reference, torque reference is aligned with q-
component and flux component is aligned with d-component. Since the transformation is
based on projections hence it can handle both transient and sinusoidal disturbances. Hence
torque can be independently controlled by manipulation the q component of stator current
keeping the rotor flux constant.

Figure 2. 5 Flow Diagram of FOC


11

Any two-phase currents of motor are passed throw Clarke transformation block
which transforms 𝑖𝑎, 𝑖𝑏 into 𝑖𝛼, 𝑖𝛽 . These two current components are input to the park
transformation block which converts it into d, q rotating reference frame. The 𝑖𝑠𝑞 and 𝑖𝑠𝑑
are compared with the reference torque and flux values PI controller the outputs of
regulator are 𝑣𝑠𝑞𝑟𝑒𝑓 , 𝑣𝑠𝑑𝑟𝑒𝑓 . 𝑡ℎ𝑒 𝑣𝑠𝑞𝑟𝑒𝑓 , 𝑣𝑠𝑑𝑟𝑒𝑓 are input to inverse park block which
converts it into 𝑣𝑠𝛼𝑟𝑒𝑓 , 𝑣𝑠𝛽𝑟𝑒𝑓 .These are the Clarke components of stator voltages. These
outputs are fed to SVPWM block which derives the three-phase inverter. The information
of rotor flux position is must in both park and inverse park transform. The basic
requirements to implement FOC are the information of two phase because the third phase
could be evaluated by simple mathematical equation in case of start connected, the
sampling of these currents by A/D converter and precise rotor flux position, the evaluation
of rotor flu position is difficult in case of induction motor because the rotor and rotor flux
are not perfectly aligned there is concept of slip speed[7].

2.3.3 Direct Torque Control (DTC)

Stator flux is controlled by appropriate selection of inverter states, do not require


coordinate transformation and pi regulators, PWM and position encoders as compared to
previous control techniques. As far as control of any induction motor is concerned the
major mechanical parameters are speed and torque which are to be controlled, to control
these two mechanical parameters the electrical quantities stator flux and torque are of
utmost importance. For the control purpose, the dependence of these two electrical
quantities on other parameters is necessary[7]. The independent control of speed and
torque of AC motors is also challenging task.

From basic electrical machines, we know that:

3×𝑝×𝐿𝑚
Te =4×𝐿 |𝜓𝑟 ||𝜓𝑠 | sin 𝜃𝑟𝑠
𝑟 ×𝐿𝑆

The equation shows the dependence of electromagnetic torque on the stator flux,
rotor flux and angular displacement between them. The angular displacement between
these two vectors greatly depend on the stator flux because slight change in stator flux
cause a significant change in the output torque so to control the output torque significant
application of voltage vectors to the inverters are necessary.
12

Figure 2. 6 Working Diagram of DTC

From basic machine design, it is evaluated that rate of change of flux results an
EMF, if the ohmic drop of stator windings are neglected then the below elation can be
evaluated:

𝑑𝜓𝑠
=𝑉𝑆
𝑑𝑡

Hence stator flux can be controlled by the proper selection of voltage vector, by
the switching of igbts the 8 voltage vectors can be evaluated, out of 8, two have zero sum
of voltages hence only six vectors are left and they can be applied in sequence to achieve
the control of IM, the 360-degree plane is divided into six segments each 60-degree wide.

360
Ie = 6
13

Figure 2. 7 A Voltage Source Inverter Feeding Three Coils

Figure 2. 8 Space Vector Diagram for 2-level Inverter

2.3.3.1 Characteristic Features

• Do not need modulation technique, PI regulators, PWM, co-ordinate transformer


• Controlling variables are magnetizing flux and torque
• Since it does not need any feedback mechanism neither tachometer nor encoder is
required
• Torque response is 10 times faster than AC and DC drives
• It need motor parameters to control the torque hence it is like that of DC drives
• Do not need any mechanical transducer in motor shaft
14

• Instead of pulse width modulation it uses hysteresis band modulation in which


there are two bands i.e. the upper band and the lower band. If the current exceeds
the upper band the upper switch is switched off as a result the current decays, and
if it reduces than lower band the upper switch gets on
• The only disadvantage of DTC is its poor performance in low speed and low
frequencies

2.3.4 Model Predictive Control (MPC)

Being among one of the categories of control schemes, a predictive control makes
use of system’s model to pre-calculate output values at future instants. These predicted
values are then used to optimize the behavior of respective plant by generating series of
control moves. The aim of these controlled calculations is to shift the predicted response
in direction of setpoint using optimal manner[7].

It would be more appropriate to define predictive control as a set of methodologies


and not a single technique because various mathematical translations can be utilized to
shape the problem which is under consideration and those formulations are basically the
reason that gives rise to numerous techniques under the umbrella of predictive control
strategy.

2.3.4.1 Brief History of MPC

The roots of advanced control techniques are found in 1960s when Linear
Quadratic Regulator (LQR) was mapped out by Kalman for minimization of an
unconstrained quadratic objective function based on states and inputs of system. But, it
couldn’t create a significant effect on the growth of control technology as all real-world
systems are subjected to constraints and non-linearities. Also, the control engineers of that
time considered this idea as unrealistic.

Articles on Model Predictive Heuristic Control (also called Model Algorithmic


Control (MAC)) by Richalet et al. and Dynamic Matrix Control (DMC) by Cutler and
Ramaker, in 1970s, opened ways for the implementation of MPC in industries. Both
strategies were established using dynamic model of system, advancement of
microprocessors was considered and stability was achieved through manipulation of
weights in cost function. Quadratic Dynamic Matrix Control (QDMC) by Garcia,
Morshedi in 1986 is regarded as the second generation of MPC. A constrained open loop
system was controlled using quadratic programming[11].
15

Side by side, adaptive control strategies for mono-variable processes also rise that
were relying on transfer function based models and Diophantine equation for
determination of future inputs. Minimum Variance Control (MVC) by Astron et al. in
1970 was the first proposal being made in this context. It was followed by Generalized
Minimum Variance Control (GMVC)[14], in which, penalized input was taken into
account for the cost function. Predictor Based Self Tuning Control by Peterka (in 1984)
encountered the limitations of horizon.

Generalized Predictive Control (GPC), put forward by Clarke et. al. in 1987, is the
one that is used by mostly nowadays. A summary of all generations of MPC is depicted
in the following figure:

Figure 2. 9 Generations of MPC

2.3.4.2 Components of MPC

The three basic elements, found in every algorithm of MPC, are:

• Predictive model
• Objective function
• Control law
16

2.3.4.3 Strategy of MPC

A block diagram of MPC is shown in figure. Mathematical model of system, as


accurate as possible, is used in MPTC for prediction of control variables. The residuals as
well as the differences between the actual and predicted outputs act as the feedback signal
to prediction block. Two kinds of calculations make use of these predicted values; set-
point and control[4].

Figure 2. 10 Block Diagram of MPC

The biggest advantage of MPC is receding horizon. At every sampling instant, a


set of input-values (manipulated variable), consisting of both current and future inputs, is
calculated. But, only the first move is applied and a new sequence of input-values need to
be obtained at the upcoming sampling instant[6]. In this manner, up-to-date data about ‘u’
w.r.t the latest measurement of output ‘y’ is utilized instantly. Hence, it could be said that
future instants are kept in mind for current instant and so the name Predicted Control has
been devised.

Figure 2. 11 Receding Horizon Control


17

2.3.4.4 Types of Model in MPC

There are two types of Model in MPC:

• Process Model
i. Impulse response (used by industry)
ii. Step response
iii. Transfer function (used by industries and academic institutions)
iv. State Space (used by researchers)

• Disturbance Model

Kalman Filter (An Observer)

Kalman filters are ideal for systems which are continuously changing. They have
the advantage that they are light on memory (they don’t need to keep any history other
than the previous state), and they are very fast, making them well suited for real time
problems and embedded systems.

Figure 2. 12 Kalman Filtering

Effect of Sampling Period on Control System

In order to choose a sampling frequency in control systems, following rule is used:


18

fS = (6 to 25) fbCL

here;
fs = sampling frequency
fbCL = closed loop system bandwidth

Table 2. 1 Choice of Sampling Period for Digital Control System

Type of Variable (of plant) Sampling Period (s)

Flow rate 1–3

Level 5 – 10

Pressure 1–5

Temperature 10 – 180

Distillation 10 – 180

Servo-Mechanism 0.001 – 0.05

Catalytic Reactors 10 – 45

Cement Plants 20 – 45

Dryers 20 – 45

Keeping the sampling period large has adverse effects on relative stability because
a greater sampling frequency/rate makes the maximum allowable gain larger.

An increased sampling period (decreasing sampling frequency) has a detrimental


effect on the steady state error.

Sampling rates are tried to be kept as low as possible due to economic reasons as
the control algorithm would get more time to execute and then it could be run on slow
processors as well. Moreover, system’s response is affected due to digitization of analog
control systems i.e. they become unstable at low frequencies. Using Nyquist criterion, the
sampling frequency should be twice as the bandwidth of the error signal which in turn is
19

bounded by the system’s bandwidth, therefore ωs ≥ 2 ωB. Factor of 10 to 20 is needed to


make sure about a satisfactory response.

If case of pathological sampling, controllability can also be lost.

Effect of Weighting Matrices on System’s Performance

Weighting matrices Q and R are one of the design parameters of MPC.

Q = output weighting matrix or weights of output variables


R = input weighting matrix or weights of manipulated variables

2.6.1 Effect of Input Weighting Matrix (R)

Increasing R is producing more deviations at output and the settling time is also
increasing because the input is being penalized heavily. In addition to this, MPC gain is
decreasing.
20

Figure 2. 13 Effect of Input Weighting Matrix on System

2.6.2 Effect of Output Weighting Matrix (Q)

Increasing Q is producing less deviations at output and the settling time is also decreasing.
In addition to this, MPC gain is increasing.
21

Figure 2. 14 Effect of Output Weighting Matrix on System


22

Artificial Neural Network

2.7.1 Introduction to Artificial Neural Networks

Neural Network is “highly interconnected system composed of individual units,


each having the tendency to store and use knowledge”.

NN mimics the human brain in two ways:

• Brian acquires the knowledge based on its observations


• Synaptic weights i.e. Inter-neuron connection’s gains are used.

2.7.2 Equivalent Electrical Model of a Neuron

Human brain is highly complex and it is nearly impossible to create it exact


electrical Model. However, the given equivalent neural network model can closely
resemble the actual and have lower processing requirements.

Assume x0, x1, x2…… xn are inputs provided to neurons.

Figure 2. 15 Equivalent Computational Model of Neuron


23

The inter-connection gains represent the strength of each connection. These are
called synapses in biological brain sciences.

W
KP, shows the strength of the connection from neuron p to neuron k.

For instance, WK0 denotes synaptic weight from neuron 0 to neuron k.

Here,

𝑛
𝑉𝑘 = ∑ = 1 𝑤𝑘𝑗 𝑥𝑗
𝑗

Therefore,

𝑌𝑘 = 𝜃𝑘 + 𝑉𝑘

The output Yk is determined by passing the weighted sum Vk through an activation


function.

2.7.3 Types of Activation Functions

2.7.3.1 Threshold Function

The Threshold function is provided with weighted sum Vk, its output Yk can be defined
as:
24

Figure 2. 16 Graphical Representation of Threshold Function

This shows that Vk ranges from -∞ to +∞ but Yk ranges from 0 to 1.

2.7.3.2 Signum Function

The values range from -1 to +1 called signum function.

Mathematically,

Figure 2. 17 Graphical Representation of Signum Function


25

2.7.3.3 Sigmoid Function

This one is nonlinear function promising better results. Mathematically,

Figure 2. 18 Graphical Representation of Sigmoid Function

For large positive numbers, sigmoid function is approximately 1


At 0, sigmoid function is 0.5
For large negative numbers, sigmoid function is approximately 0
This low and high output of neuron (0 and 1) allows smooth transition.

2.7.3.4 Tangent Hyperbolic Function

Ranges from 0 to 1 can be limited from -1 to +1 then the function is called Tangent
hyperbolic function. Mathematically,
26

Figure 2. 19 Graphical Representation of Tangent Hyperbolic


Function

Neuromorphic Algorithms

As Neural Network are inspired from the brain learning mechanism, Brain
Emotional Learning based Intelligence is one such evolutionary neuromorphic computing
technique that closely mimics the brain emotions generating, evaluating and reacting
part[21].

To develop a better understanding of this control technique, a brief overview of


emotional learning of human brain and role of each part is given below:

2.8.1 Brain Parts Concerned with Emotions

Emotions enable us to react with situations. The main part of brain concerned with
emotion (fear, anger, thirst, sexual drive, hunger etc.) is our limbic system. It deals with
emotion, memory and arousal (stimulation)[21]. Limbic system includes:

• Amygdala: play a role in fear and anger


27

• Hippocampus: helps preserve and retrieve memories

• Hypothalamus: control emotion response, sexual responses, hormones release


and regulate body temperature

• Thalamus: concerned with consciousness, sleep and alertness

Role of orbitofrontal cortex: area of prefrontal cortex, has extensive connection with
limbic system;

• Concerned with function ranging from olfaction and emotion to learning and
behavioral flexibility

• Involved in cognition (thinking) and decision making

Role of sensory cortex: located in parietal lobe of brain;

• Receives all sensory input from body like pain, visual or auditory stimuli and send
it to somatosensory cortex for process

2.8.2 Control Mechanism

The basic idea behind this controller is that it performed biomimicry of feedback,
evaluation and action just like human brain. The following diagram is the block level
implementation of this controller.

Figure 2. 20 A Learning Organism


CHAPTER 3

METHODOLOGY/ OBSERVATION AND CALCULATION

Dynamic Model of an Induction Motor

Figure 3. 1 Induction Motor Model [5]

𝑣𝑎𝑠 = γas 𝑖𝑎𝑠 + D Ψas


𝑣𝑏𝑠 = γbs 𝑖𝑏𝑠 + D Ψbs
𝑣𝑐𝑠 = γcs 𝑖𝑐𝑠 + D Ψcs

In terms of space vector, 3-phase voltages are defined as:

2𝜋 4𝜋
2
𝑣𝑠 = (𝑣𝑎 (𝑡) + 𝑣𝑏 (𝑡)𝑒 𝑗
⃗⃗⃗⃗ 3 + 𝑣𝑐 (𝑡)𝑒 𝑗 3 )
3

For a squirrel cage induction motor, the space vector voltage equation is:

⃗⃗⃗⃗
𝑉𝑠 = 𝑅𝑠 ⃗⃗𝐼𝑠 + D⃗⃗⃗⃗⃗⃗
Ψ𝑠 (3.1)
29

Where Rs is the stator resistance and 𝜓s is the space vector of stator flux. With respect to
its own winding system, the space vector voltage equation of the rotor is:

⃗⃗⃗⃗⃗⃗
𝑉𝑟 ∗ = 𝑅𝑟 ⃗⃗⃗⃗⃗⃗
𝐼𝑟 ∗ + D ⃗⃗⃗⃗⃗⃗⃗⃗
Ψr ∗ = 0 (3.2)

All the vectors are in rotor frame of reference.

For the synchronization of space vectors of both stator and rotor:

⃗⃗⃗⃗⃗
𝑈𝑟 = ⃗⃗⃗⃗⃗⃗⃗⃗
𝑈𝑟 ∗ 𝑒 +𝑗𝜃𝑒

⃗⃗⃗⃗⃗
𝐼𝑟 = ⃗⃗⃗⃗⃗⃗
𝐼𝑟 ∗ 𝑒 +𝑗𝜃𝑒

⃗⃗⃗⃗⃗⃗
Ψ𝑟 = ⃗⃗⃗⃗⃗⃗⃗⃗
Ψ𝑟 ∗ 𝑒 +𝑗𝜃𝑒

θe is the electrical angle. Since rotor winding is lagging the stator winding by θe, thus to
synchronize rotor winding and stator winding the rotor winding is advanced by θe.

⃗⃗⃗⃗⃗⃗⃗⃗
𝑈𝑟 ∗ 𝑒 +𝑗𝜃𝑒 𝑅𝑟 ⃗⃗⃗⃗⃗⃗
𝐼𝑟 ∗ 𝑒 +𝑗𝜃𝑒 + D ⃗⃗⃗⃗⃗⃗⃗⃗
Ψr ∗ 𝑒 +𝑗𝜃𝑒 = 0

𝑒 +𝑗𝜃𝑒
⃗⃗⃗⃗⃗
𝑈𝑟 = 𝑅𝑟 ⃗⃗𝐼𝑟 + D ⃗⃗⃗⃗
Ψr 𝑒 +𝑗𝜃𝑒

Ψ ⃗⃗⃗⃗⃗
⃗⃗⃗⃗⃗
𝑈𝑟 = 𝑅𝑟 ⃗⃗𝐼𝑟 + D [𝑒 +𝑗𝜃r 𝑒 ] ( 𝑒 +𝑗𝜃𝑒 )

+𝑗𝜃𝑒 D ⃗⃗⃗⃗⃗ Ψr D 𝑒 +𝑗𝜃𝑒


Ψr − ⃗⃗⃗⃗⃗
⃗⃗⃗⃗⃗
𝑈𝑟 = 𝑅𝑟 ⃗⃗𝐼𝑟 + [
𝑒
𝑒 +𝑗𝜃𝑒 ] (∵D𝜃=𝜔)
𝑒 2 𝑗𝜃𝑒

⃗⃗⃗⃗⃗
𝑈𝑟 = 𝑅𝑟 ⃗⃗𝐼𝑟 + D ⃗⃗⃗⃗
Ψr - ⃗⃗⃗⃗
Ψr j 𝜔𝑒 = 0 (3.3)
30

For Flux Representation:

⃗⃗⃗⃗⃗
Ψ𝑠 = 𝐿𝑠 ⃗⃗𝐼𝑠 +𝐿ℎ ⃗⃗𝐼𝑟 (3.4)

⃗⃗⃗⃗⃗⃗
Ψ𝑟 = 𝐿ℎ ⃗⃗𝐼𝑠 +𝐿𝑟 ⃗⃗𝐼𝑟 (3.5)

Equation (3.1) becomes:

⃗⃗⃗⃗
𝑉𝑠 = 𝑅𝑠 ⃗⃗𝐼𝑠 + DΨ
⃗⃗⃗⃗⃗𝑠

Substituting the value of ⃗⃗⃗⃗⃗


Ψ𝑠 from Equation (3.4):

⃗⃗⃗⃗
𝑉𝑠 = 𝑅𝑠 ⃗⃗𝐼𝑠 + D(𝐿𝑠 ⃗⃗𝐼𝑠 + 𝐿ℎ ⃗⃗𝐼𝑟 )

⃗⃗⃗⃗
𝑉𝑠 = 𝑅𝑠 ⃗⃗𝐼𝑠 + D𝐿𝑠 ⃗⃗𝐼𝑠 + D𝐿ℎ ⃗⃗𝐼𝑟

We should eliminate ⃗⃗𝐼𝑟 , substituting the value of ⃗⃗𝐼𝑟 from Equation (3.5):

Ψ ⃗⃗⃗⃗⃗⃗⃗ 𝐿ℎ
⃗⃗⃗⃗
𝑉𝑠 = 𝑅𝑠 ⃗⃗𝐼𝑠 + D𝐿𝑠 ⃗⃗𝐼𝑠 + D𝐿ℎ [ 𝐿 𝑟 − (𝐼⃗⃗𝑠 )]
𝑟 𝐿𝑟

𝐿 𝐿ℎ 2
⃗⃗⃗⃗
𝑉𝑠 = 𝑅𝑠 ⃗⃗𝐼𝑠 + D𝐿𝑠 ⃗⃗𝐼𝑠 +𝐿ℎ DΨ
⃗⃗⃗⃗⃗⃗𝑟 − 𝐷 (𝐼⃗⃗𝑠 )
𝑟 𝐿𝑟

𝐿ℎ 2 𝐿
⃗⃗⃗⃗
𝑉𝑠 = 𝑅𝑠 ⃗⃗𝐼𝑠 +D𝐼⃗⃗𝑠 [ 𝐿𝑠 − ⃗⃗⃗⃗⃗⃗𝑟
] + 𝐿ℎ DΨ
𝐿𝑟 𝑟

𝐿ℎ 2 𝐿 (3.6)
⃗⃗⃗⃗
𝑉𝑠 = 𝑅𝑠 ⃗⃗𝐼𝑠 +D𝐼⃗⃗𝑠 𝐿𝑠 [1 − ⃗⃗⃗⃗⃗⃗𝑟
] + 𝐿ℎ DΨ
𝐿𝑟 𝐿𝑠 𝑟

⃗⃗⃗⃗⃗
𝑈𝑟 = 𝑅𝑟 ⃗⃗𝐼𝑟 + D ⃗⃗⃗⃗
Ψr - ⃗⃗⃗⃗
Ψr j 𝜔𝑒 = 0
31

𝑅𝑟 ⃗⃗𝐼𝑟 + D ⃗⃗⃗⃗
Ψr - ⃗⃗⃗⃗
Ψr j 𝜔𝑒 = 0

⃗⃗⃗⃗⃗⃗⃗
Ψ 𝐿ℎ
𝑅𝑟 [ 𝐿 𝑟 − (𝐼⃗⃗𝑠 )] + D ⃗⃗⃗⃗
Ψr - ⃗⃗⃗⃗
Ψr j 𝜔𝑒 = 0
𝑟 𝐿𝑟

⃗⃗⃗⃗⃗⃗⃗𝑟 𝑅𝑟
Ψ 𝑅𝑟 𝐿ℎ
D ⃗⃗⃗⃗
Ψr = − + (𝐼⃗⃗𝑠 ) + ⃗⃗⃗⃗
Ψr j 𝜔𝑒
𝐿𝑟 𝐿𝑟

Substituting the value of D ⃗⃗⃗⃗


Ψr in Equation (3.6)

𝐿ℎ 2 𝐿 ⃗⃗⃗⃗⃗⃗⃗𝑟 𝑅𝑟
Ψ 𝑅𝑟 𝐿ℎ
⃗⃗⃗⃗
𝑉𝑠 = 𝑅𝑠 ⃗⃗𝐼𝑠 +D𝐼⃗⃗𝑠 𝐿𝑠 [1 − ] + 𝐿ℎ [− + (𝐼⃗⃗𝑠 ) + ⃗⃗⃗⃗⃗
Ψr j𝜔𝑒 ]
𝐿𝑟 𝐿𝑠 𝑟 𝐿𝑟 𝐿𝑟

𝐿ℎ 2 𝑅𝑟 𝐿ℎ 𝑅 𝐿 𝐿 2
⃗⃗⃗⃗
𝑉𝑠 = 𝑅𝑠 ⃗⃗𝐼𝑠 +D𝐼⃗⃗𝑠 𝐿𝑠 [1 − ]− ⃗⃗⃗⃗⃗⃗
Ψ𝑟 + 𝑟𝐿 2ℎ ⃗⃗𝐼𝑠 + 𝐿ℎ ⃗⃗⃗⃗⃗⃗
Ψr j𝜔𝑒
𝐿𝑟 𝐿𝑠 𝐿𝑟 2 𝑟 𝑟

𝐿ℎ 2 𝐿ℎ 2 𝑅𝑟 𝐿ℎ 𝐿ℎ (3.7)
⃗⃗⃗⃗
𝑉𝑠 = ⃗⃗𝐼𝑠 [𝑅𝑠 + 𝑅𝑟 ] + 𝐿𝑠 [1 − ] D𝐼⃗⃗𝑠 + ⃗⃗⃗⃗⃗⃗
Ψ𝑟 [− + + j𝜔𝑒 ]
𝐿𝑟 2 𝐿𝑟 𝐿𝑠 𝐿𝑟 2 𝐿𝑟

Let,

𝑅𝑟 𝐿ℎ 𝐿ℎ
T = ⃗⃗⃗⃗⃗⃗
Ψ𝑟 [− 𝐿𝑟 2
+ + 𝐿𝑟
j𝜔𝑒 ]

Equation (3.7) becomes:

𝐿ℎ 2 𝐿ℎ 2 (3.8)
⃗⃗⃗⃗
𝑉𝑠 = ⃗⃗𝐼𝑠 [𝑅𝑠 + 𝑅𝑟 2 ] + 𝐿𝑠 [1 − ] D𝐼⃗⃗𝑠 + 𝑇
𝐿𝑟 𝐿𝑟 𝐿𝑠

Using Equation (3.4) 𝑎𝑛𝑑 (3.5):

Ψ −𝐿 𝐼 ⃗⃗⃗⃗⃗⃗ ⃗⃗⃗
⃗⃗⃗⃗⃗⃗
Ψ𝑟 = 𝐿ℎ ⃗⃗𝐼𝑠 + 𝐿𝑟 [ 𝑠 𝐿 𝑠 𝑠 ]

𝐿 𝐿𝑟 𝐿𝑠
⃗⃗⃗⃗⃗⃗
Ψ𝑟 = 𝐿ℎ ⃗⃗𝐼𝑠 + 𝐿 𝑟 ⃗⃗⃗⃗⃗
Ψ𝑠 − ⃗⃗𝐼𝑠
ℎ 𝐿ℎ
32

𝐿𝑟 𝐿𝑠 𝐿
⃗⃗⃗⃗⃗⃗
Ψ𝑟 = [𝐿ℎ − ] ⃗⃗𝐼𝑠 + 𝐿 𝑟 ⃗⃗⃗⃗⃗
Ψ𝑠
𝐿ℎ ℎ

Substituting the value of ⃗⃗⃗⃗⃗⃗


Ψ𝑟 in T:

𝐿𝑟 𝐿𝑠 𝐿𝑟 𝑅𝑟 𝐿ℎ 𝐿ℎ
T = [[𝐿ℎ − ] ⃗⃗𝐼𝑠 + ⃗⃗⃗⃗⃗
Ψ𝑠 ] [− + + j𝜔𝑒 ]
𝐿ℎ 𝐿ℎ 𝐿𝑟 2 𝐿𝑟

(3.9)
Let,

𝐿
𝐿1 = 𝐿ℎ
𝑟

𝑅𝑟 𝐿1 𝐿 ⃗⃗⃗⃗⃗⃗
Ψ𝑠
T= [j𝜔𝑒 𝐿1 − ] [𝐿ℎ ⃗⃗𝐼𝑠 − 𝐿 𝑠 𝐼⃗⃗𝑠 + ]
𝐿𝑟 1 𝐿1

𝑅𝑟 𝐿1 𝑅𝑟 𝐿𝑠 𝑅𝑟
T = j 𝜔𝑒 𝐿1 𝐿ℎ ⃗⃗𝐼𝑠 − j 𝜔𝑒 𝐿𝑠 ⃗⃗𝐼𝑠 + j 𝜔𝑒 ⃗⃗⃗⃗⃗
Ψ𝑠 − 𝐿ℎ ⃗⃗𝐼𝑠 + ⃗⃗𝐼𝑠 − ⃗⃗⃗⃗⃗
Ψ𝑠
𝐿𝑟 𝐿𝑟 𝐿𝑟

Substituting the value of T in Equation (3.8):

𝐿ℎ 2 𝐿ℎ 2
⃗⃗⃗⃗
𝑉𝑠 = ⃗⃗𝐼𝑠 [𝑅𝑠 + 𝑅𝑟 2 ] + 𝐿𝑠 [1 − ] D𝐼⃗⃗𝑠 + [j 𝜔𝑒 𝐿1 𝐿ℎ ⃗⃗𝐼𝑠 − j 𝜔𝑒 𝐿𝑠 ⃗⃗𝐼𝑠 + j 𝜔𝑒 ⃗⃗⃗⃗⃗
Ψ𝑠 −
𝐿𝑟 𝐿𝑟 𝐿𝑠
𝑅𝑟 𝐿1 𝑅𝑟 𝐿𝑠 𝑅𝑟
𝐿ℎ ⃗⃗𝐼𝑠 + ⃗⃗𝐼𝑠 − ⃗⃗⃗⃗⃗
Ψ𝑠 ]
𝐿𝑟 𝐿𝑟 𝐿𝑟

𝐿ℎ 2 𝑅𝑟 𝐿1 𝑅𝑟 𝐿𝑠
⃗⃗⃗⃗
𝑉𝑠 = [𝑅𝑠 + 𝑅𝑟 + j 𝜔𝑒 𝐿1 𝐿ℎ − j 𝜔𝑒 𝐿𝑠 − 𝐿ℎ + ] ⃗⃗𝐼𝑠
𝐿𝑟 2 𝐿𝑟 𝐿𝑟

𝐿ℎ 2 𝑅𝑟
+ 𝐿𝑠 [1 − ] D𝐼⃗⃗𝑠 + [j 𝜔𝑒 − ] ⃗⃗⃗⃗⃗
Ψ𝑠
𝐿𝑟 𝐿𝑠 𝐿𝑟

Put,

𝐿
𝐿1 = 𝐿ℎ
𝑟

𝐿ℎ 2 𝐿ℎ 2 𝐿ℎ 2 𝑅𝑟 𝐿𝑠 𝐿ℎ 2
⃗⃗⃗⃗
𝑉𝑠 =[𝑅𝑠 + 𝑅𝑟 2 + j 𝜔𝑒 − j 𝜔𝑒 𝐿𝑠 − 𝑅𝑟 + ] ⃗⃗𝐼𝑠 + 𝐿𝑠 [1 − ] D𝐼⃗⃗𝑠 +
𝐿𝑟 𝐿𝑟 𝐿𝑟 2 𝐿𝑟 𝐿𝑟 𝐿𝑠
𝑅𝑟
[j 𝜔𝑒 − ] ⃗⃗⃗⃗⃗
Ψ𝑠
𝐿𝑟
33

𝐿ℎ 2 𝐿ℎ 2 𝑅𝑟 𝐿𝑠 𝐿ℎ 2 𝐿𝑟 𝐿𝑠 − 𝐿ℎ 2
⃗⃗⃗⃗
𝑉𝑠 = [𝑅𝑠 + 𝑅𝑟 2 − 2 𝑅𝑟 + ] ⃗⃗𝐼𝑠 + [+ j 𝜔𝑒 − j 𝜔𝑒 𝐿𝑠 ] ⃗⃗𝐼𝑠 + 𝐿𝑠 [ ]
𝐿𝑟 𝐿𝑟 𝐿𝑟 𝐿𝑟 𝐿𝑟 𝐿𝑠
j𝜔 𝐿 −𝑅
D𝐼⃗⃗𝑠 + [ 𝑒 𝐿𝑟 𝑟 ] ⃗⃗⃗⃗⃗
Ψ𝑠
𝑟

𝑅𝑟 𝐿𝑠 𝐿ℎ 2 𝐿𝑟 𝐿𝑠 – 𝐿ℎ 2 j 𝜔𝑒 𝐿𝑟 – 𝑅𝑟
⃗⃗⃗⃗
𝑉𝑠 = [𝑅𝑠 + ] ⃗⃗𝐼𝑠 + [+ j 𝜔𝑒 − j 𝜔𝑒 𝐿𝑠 ] ⃗⃗𝐼𝑠 + 𝐿𝑠 [ ] D𝐼⃗⃗𝑠 + [ ] ⃗⃗⃗⃗⃗
Ψ𝑠
𝐿𝑟 𝐿𝑟 𝐿𝑟 𝐿𝑠 𝐿𝑟

𝑅 𝐿 +𝑅 𝐿 + j 𝜔𝑒 𝐿ℎ 2 − j 𝜔𝑒 𝐿𝑠 𝐿𝑟 𝐿𝑟 𝐿𝑠 − 𝐿ℎ 2 j 𝜔𝑒 𝐿𝑟 − 𝑅𝑟
⃗⃗⃗⃗
𝑉𝑠 = [ 𝑠 𝑟 𝐿 𝑟 𝑠 ] ⃗⃗𝐼𝑠 + [ ] ⃗⃗𝐼𝑠 + 𝐿𝑠 [ ] D𝐼⃗⃗𝑠 +[ ] ⃗⃗⃗⃗⃗
Ψ𝑠
𝑟 𝐿𝑟 𝐿𝑟 𝐿𝑠 𝐿𝑟

𝑅 𝐿 +𝑅 𝐿 + j 𝜔𝑒 𝐿ℎ 2 − j 𝜔𝑒 𝐿𝑠 𝐿𝑟 𝐿𝑟 𝐿𝑠 − 𝐿ℎ 2 j 𝜔𝑒 𝐿𝑟 − 𝑅𝑟
⃗⃗⃗⃗
𝑉𝑠 = [ 𝑠 𝑟 𝐿 𝑟 𝑠 ] ⃗⃗𝐼𝑠 + [ ] ⃗⃗𝐼𝑠 + [ ] D𝐼⃗⃗𝑠 +[ ] ⃗⃗⃗⃗⃗
Ψ𝑠
𝑟 𝐿𝑟 𝐿𝑟 𝐿𝑟

𝑅 𝐿 +𝑅 𝐿 + j 𝜔 𝑒 𝐿ℎ 2 − j 𝜔 𝑒 𝐿𝑠 𝐿𝑟 2
⃗⃗⃗⃗
𝑉𝑠 = [ 𝑠 𝑟 𝑟 𝑠 ] ⃗⃗𝐼𝑠 + [𝐿𝑟 𝐿𝑠 − 𝐿ℎ ] D𝐼⃗⃗𝑠 + [j 𝜔𝑒 𝐿𝑟 − 𝑅𝑟 ] ⃗⃗⃗⃗⃗
Ψ𝑠
𝐿𝑟 𝐿 𝑟 𝐿 𝑟

𝑅 𝐿 +𝑅 𝐿 + j 𝜔𝑒 (𝐿ℎ 2 − 𝐿𝑠 𝐿𝑟 ) 𝐿𝑟 𝐿𝑠 − 𝐿ℎ 2 j 𝜔𝑒 𝐿𝑟 − 𝑅𝑟
⃗⃗⃗⃗
𝑉𝑠 = [ 𝑠 𝑟 𝑟 𝑠 ] ⃗⃗𝐼𝑠 + [ ] D𝐼⃗⃗𝑠 + [ ] ⃗⃗⃗⃗⃗
Ψ𝑠
𝐿𝑟 𝐿𝑟 𝐿𝑟

2− 𝐿
𝑅 𝐿 +𝑅 𝐿 𝐿 𝑠 𝐿𝑟 𝐿𝑟 𝐿𝑠 − 𝐿ℎ 2 j 𝜔𝑒 𝐿𝑟 − 𝑅𝑟
⃗⃗⃗⃗
𝑉𝑠 = [ 𝑠 𝑟𝐿 𝑟 𝑠 ] ⃗⃗𝐼𝑠 + [j 𝜔𝑒 ( ℎ )] ⃗⃗𝐼𝑠 + [ ] D𝐼⃗⃗𝑠 + [ ] ⃗⃗⃗⃗⃗
Ψ𝑠
𝑟 𝐿𝑟 𝐿𝑟 𝐿𝑟

1
𝛌=𝐿 2
𝑠 𝐿𝑟 − 𝐿ℎ

𝑅 𝐿 +𝑅 𝐿 j𝜔 1 j𝜔 𝐿 −𝑅
⃗⃗⃗⃗
𝑉𝑠 = [ 𝑠 𝑟𝐿 𝑟 𝑠 ] ⃗⃗𝐼𝑠 − [ 𝐿 𝑒λ ] ⃗⃗𝐼𝑠 + [𝐿 λ ] D𝐼⃗⃗𝑠 + [ 𝑒 𝐿𝑟 𝑟 ] ⃗⃗⃗⃗⃗
Ψ𝑠
𝑟 𝑟 𝑟 𝑟 (3.10)

Let,

U = ⃗⃗⃗⃗
𝑉𝑠
𝑋1 = ⃗⃗𝐼𝑠
𝑋́1 = D ⃗⃗𝐼𝑠
𝑋2 = ⃗⃗⃗⃗⃗
Ψ𝑠
𝑋2 = D ⃗⃗⃗⃗⃗
́ Ψ𝑠

D𝐼⃗⃗𝑠 = ⃗⃗⃗⃗
𝑉𝑠 𝐿𝑟 λ − [(𝑅𝑠 𝐿𝑟 + 𝑅𝑟 𝐿𝑠 )λ + j 𝜔𝑒 ] ⃗⃗𝐼𝑠 + 𝛌 (𝑅𝑟 − j 𝜔𝑒 𝐿𝑟 ) ⃗⃗⃗⃗⃗
Ψ𝑠

𝑋́1 = 𝑈 𝐿𝑟 λ − [(𝑅𝑠 𝐿𝑟 + 𝑅𝑟 𝐿𝑠 )λ + j 𝜔𝑒 ] 𝑋1 + 𝛌 (𝑅𝑟 − j 𝜔𝑒 𝐿𝑟 ) 𝑋2


34

Now for second term:

⃗⃗⃗⃗
𝑉𝑠 = ⃗⃗⃗⃗
𝑅𝑠 ⃗⃗𝐼𝑠 + D ⃗⃗⃗⃗⃗
Ψ𝑠

⃗⃗⃗⃗
𝑉𝑠 = 𝑅𝑠 𝑋1

𝑋́2 = 𝑈 − 𝑅𝑠 𝑋1

𝑋́ − [(𝑅𝑠 𝐿𝑟 + 𝑅𝑟 𝐿𝑠 )λ + j 𝜔𝑒 ] λ (𝑅𝑟 − j 𝜔𝑒 𝐿𝑟 ) 𝑋1 𝐿 λ
[ 1] = [ ] [ ]+[ 𝑟 ]U
́
𝑋2 − 𝑅𝑠 0 𝑋2 1

Representation in 𝛼𝛽 Reference Frame

Decomposing the space vector into real and imaginary axis:


𝑋́ − [(𝑅𝑠 𝐿𝑟 + 𝑅𝑟 𝐿𝑠 )λ + j 𝜔𝑒 ] λ (𝑅𝑟 − j 𝜔𝑒 𝐿𝑟 ) 𝑋1 𝐿 λ
[ 1] = [ ] [ ]+[ 𝑟 ]U
́
𝑋2 − 𝑅𝑠 ⃗⃗⃗⃗ 0 𝑋2 1
𝑈𝑠 = 𝑈𝑠𝛼 + 𝑗 𝑈𝑠𝛽 (3.11)
⃗⃗𝐼𝑠 = 𝐼𝑠𝛼 + 𝑗 𝐼𝑠𝛽
⃗⃗⃗⃗
𝛹𝑟 = 𝛹𝑟𝛼 + 𝑗 𝛹𝑟𝛽

𝑑 𝑅𝑟 𝑅𝑟 𝐿ℎ
⃗⃗⃗⃗
𝛹𝑟 = +𝑗𝜔𝑒 ⃗⃗⃗⃗
𝛹𝑟 − ⃗⃗⃗⃗
𝛹𝑟 + ⃗⃗𝐼𝑠
𝑑𝑡 𝐿𝑟 𝐿𝑟
𝑋́ − [(𝑅𝑠 𝐿𝑟 + 𝑅𝑟 𝐿𝑠 )λ + j 𝜔𝑒 ] λ (𝑅𝑟 − j 𝜔𝑒 𝐿𝑟 ) 𝑋1 𝐿 λ
[ 1] = [ ] [ ]+[ 𝑟 ]U
́
𝑋2 − 𝑅 0 𝑋 1
𝑠
𝑑 1 𝐿ℎ 2
⃗⃗⃗⃗
𝛹 = +𝑗𝜔𝑒 ⃗⃗⃗⃗
𝛹𝑟 − ⃗⃗⃗⃗
𝛹 + ⃗⃗𝐼
𝑑𝑡 𝑟 𝜏𝑟 𝑟 𝜏𝑟 𝑠

𝑑 ⃗⃗⃗⃗𝑟 − 𝛹
+𝑗𝜔𝑒 𝛹 ⃗⃗⃗⃗𝑟 + 𝐿ℎ ⃗⃗𝐼𝑠
⃗⃗⃗⃗
𝛹𝑟 =
𝑑𝑡 𝜏𝑟
𝑋́1 − [(𝑅𝑠 𝐿𝑟 + 𝑅𝑟 𝐿𝑠 )λ + j 𝜔𝑒 ] λ (𝑅𝑟 − j 𝜔𝑒 𝐿𝑟 ) 𝑋1 𝐿 λ
[ ]=[ ] [ ]+[ 𝑟 ]U
́
𝑋2 − 𝑅𝑠 0 𝑋 2 1
𝑑
⃗⃗⃗⃗
𝛹 𝜏 = +𝑗𝜔𝑒 ⃗⃗⃗⃗ 𝛹𝑟 − ⃗⃗⃗⃗𝛹𝑟 + 𝐿ℎ ⃗⃗𝐼𝑠
𝑑𝑡 𝑟 𝑟

𝑑 (3.9)
⃗⃗⃗⃗
𝛹𝑟 𝜏𝑟 + ⃗⃗⃗⃗
𝛹𝑟 = +𝑗𝜔𝑒 ⃗⃗⃗⃗
𝛹𝑟 + 𝐿ℎ ⃗⃗𝐼𝑠
𝑋́1 − [(𝑅 𝑑𝑡
𝑠 𝐿𝑟 + 𝑅𝑟 𝐿𝑠 )λ + j 𝜔𝑒 ] λ 𝑟 − j 𝜔𝑒 𝐿𝑟 ) 𝑋1
(𝑅 𝐿 λ
[ ]=[ ] [ ]+[ 𝑟 ]U
́
𝑋2 − 𝑅 𝑠 0 𝑋2 1
35

Now we have two equations in space vector form and we transform into 𝛼-𝛽 frame of
reference.

Let,

𝑋1 = 𝐼𝑠 𝑋2 = 𝛹𝑟 𝑈 = 𝑈𝑠

𝐾𝑟 1 𝑈
𝑋1 + 𝜏𝜎 ′ 𝑋1 ′ = ( − 𝑗𝜔𝑒 ) 𝑋2 +
𝑟𝛼 𝜏𝑟 𝑟𝜎

𝐾𝑟 1 𝑈
𝑋1 ′ (𝜏𝜎 ′ ) = −𝑋1 + ( − 𝑗𝜔𝑒 ) 𝑋2 +
𝑟𝛼 𝜏𝑟 𝑟𝜎

−1 𝐾𝑟 1 𝑈
𝑋1 ′ = ( ) 𝑋1 + ( − 𝑗𝜔𝑒 ) 𝑋2 + (3.12)
𝜏𝜎 𝑟𝛼 𝜏𝜎 ′ 𝜏𝑟 𝜏𝜎 ′ 𝑟𝜎

Using Equation (3.9):

𝑋2 + 𝜏𝑟 𝑋2 ′ = 𝑗 𝜔𝑒 𝜏𝑟 𝑋2 + 𝐿ℎ 𝑋1

𝑗 𝜔𝑒 𝜏𝑟 − 1 𝐿ℎ
𝑋2 ′ = ( ) 𝑋2 + 𝑋
𝜏𝑟 𝜏𝑟 1

−1 𝐾𝑟 1
′ ′ ′
( − 𝑗𝜔𝑒 ) 1
𝑋1 𝜏𝜎 𝑟𝛼 𝜏𝜎 𝜏𝑟 𝑋1
[ ′] = [ ] + [𝜏𝜎 ′ 𝑟𝜎 ]
𝑋2 𝐿ℎ 𝑗 𝜔𝑒 𝜏𝑟 − 1 𝑋2
( ) 0
[ 𝜏𝑟 𝜏𝑟 ]
36

Representation in dq Reference Frame

To change the reference frame to the d-q reference frame the reference axis is rotated and
having angular displacement 𝜃𝑠 :

⃗⃗⃗⃗⃗ 𝑢𝑠 𝑒 −𝑗𝜃𝑠 = 𝑢𝑠𝑑 + 𝑗𝑢𝑠𝑞


𝑢𝑠 ′ = ⃗⃗⃗⃗

⃗⃗⃗⃗
𝑖𝑠 ′ = ⃗⃗𝑖𝑠 𝑒 −𝑗𝜃𝑠 = 𝑖𝑠𝑑 + 𝑗𝑖𝑠𝑞

⃗⃗⃗⃗⃗⃗
𝛹𝑟 ′ = ⃗⃗⃗⃗⃗
𝛹𝑟 𝑒 −𝑗𝜃𝑠 = 𝛹𝑟𝑑 + 𝑗𝛹𝑟𝑞

where ⃗⃗⃗⃗⃗⃗ 𝑖𝑠 ′ and ⃗⃗⃗⃗⃗


𝑢𝑠 ′ , ⃗⃗⃗⃗ 𝛹𝑟 ′ are the vectors referred to d-q frame 𝜃𝑠 = 𝜔𝑠 𝑡 where 𝜔𝑠 is the
angular speed of rotating flux. The rotor flux is fixed to real axis. Therefore, the rotor flux
component ⃗⃗⃗⃗⃗
𝛹𝑟 ′ is zero.

Multiplying with the factor 𝑒 −𝑗𝜃𝑠 gives:

⃗⃗⃗⃗⃗
𝑑𝑖𝑠 𝑘𝑟 1 1 ′ (3.23)
⃗⃗⃗⃗
𝑖𝑠 ′ + 𝜏𝜎 ′ [ + 𝑗𝜔𝑠 ⃗⃗⃗⃗
𝑖𝑠 ′ ] = ( − 𝑗𝜔𝑒 ) ⃗⃗⃗⃗⃗⃗
𝛹𝑟 ′ + ⃗⃗⃗⃗⃗
𝑢
𝑑𝑡 𝑟𝜎 𝜏𝜎 𝑟𝜎 𝑠

Where the following equality is used:

𝑑𝑖⃗⃗𝑠 −𝑗𝜃 ⃗⃗⃗⃗⃗


𝑑𝑖𝑠
𝑒 𝑠= + 𝑗𝜔𝑠 ⃗⃗⃗⃗
𝑖𝑠 ′
𝑑𝑡 𝑑𝑡

By substituting in Equation (3.10), the dynamic model in the d-q frame of reference is
obtained:

𝑑𝑖𝑠𝑑 1 𝑘𝑟 1
= 𝑖𝑠𝑑 + 𝜔𝑠 𝑖𝑠𝑞 + 𝛹𝑟𝑑 + 𝑢
𝑑𝑡 𝜏𝜎 ′ ′
𝑟𝜎 𝜏𝜎 𝜏𝑟 𝑟𝜎 𝜏𝜎 ′ 𝑠𝑑
37

𝑑𝑖𝑠𝑞 1 𝑘𝑟 1
= −𝜔𝑠 𝑖𝑠𝑑 − ′ 𝑖𝑠𝑞 − 𝜔𝑒 𝛹𝑟𝑑 + 𝑢
𝑑𝑡 𝜏𝜎 𝑟𝜎 𝜏𝜎 ′ 𝑟𝜎 𝜏𝜎 ′ 𝑠𝑞

Similarly,

𝑑𝛹𝑟𝑑 𝐿ℎ 1 (3.13)
= 𝑖𝑠𝑑 − 𝛹
𝑑𝑡 𝜏𝑟 𝜏𝑟 𝑟𝑑

𝐿ℎ (3.44)
0= 𝑖 − (𝜔𝑠 − 𝜔𝑒 ) 𝛹𝑟𝑑
𝜏𝑟 𝑠𝑞

Where the q component of rotor flux 𝛹𝑟𝑞 = 0. The relationship is of slip estimation yield
from Equation (3.12) is:

𝐿ℎ 𝑖𝑠𝑞
𝜔𝑠 = 𝜔𝑒 +
𝜏𝑟 𝛹𝑟𝑑

𝜔𝑒 is the angular velocity (electrical) of rotor. Since, 𝛹𝑟𝑑 is not directly used hence current
𝑖𝑠𝑑 is used for the purpose:

1 𝑖𝑠𝑞
𝜔𝑠 = 𝜔𝑒 +
𝜏𝑟 𝑖𝑠𝑑

𝑑𝛹
Where 𝑑𝑡𝑟𝑑 = 0 is assumed in Equation (3.11) to obtain the steady-state solution of 𝑖𝑠𝑑
in relation to 𝛹𝑟𝑑 :

𝑢𝑠𝑑 cos 𝜃𝑠 sin 𝜃𝑠 𝑢𝑠𝛼


[𝑢 ] = [ ][ ]
𝑠𝑞 − sin 𝜃𝑠 cos 𝜃𝑠 𝑢𝑠𝛽

𝑖𝑠𝑑 cos 𝜃𝑠 sin 𝜃𝑠 𝑖𝑠𝛼


[𝑖 ] = [ ][ ]
𝑠𝑞 − sin 𝜃𝑠 cos 𝜃𝑠 𝑖𝑠𝛽
38

𝛹𝑟𝑑 cos 𝜃𝑠 sin 𝜃𝑠 𝛹𝑠𝛼


[𝛹 ] = [ ][ ]
𝑟𝑞 − sin 𝜃𝑠 cos 𝜃𝑠 𝛹𝑠𝛽

Control Derivation

𝑥𝑘+1 = 𝐴𝑥 𝑘 +𝐵𝑢 𝑘

𝑦𝑘 = 𝐷𝑥 𝑘

Using above state space model, prediction model for k=k+1 is formulated:

𝑦𝑘+1 = 𝐷𝑥 𝑘+1

Substituting values in above equation

𝑦𝑘+1 = D (𝐴𝑥 𝑘 +𝐵𝑢 𝑘 )

𝑦𝑘+1 =D 𝐴𝑥 𝑘 +𝐷𝐵𝑢 𝑘

For k=k+2:

𝑦𝑘+2 = D (𝐴𝑥 𝑘+1 +𝐵𝑢 𝑘+1 )

𝑦𝑘+2 = D𝐴2 𝑥 𝑘 +𝐵𝑢 𝑘 +𝐷𝐵𝑢 𝑘+1

For k=k+3:

𝑦𝑘+3 = D (𝐴𝑥 𝑘+2 +𝐵𝑢 𝑘+2 )

𝑦𝑘+3 = D𝐴3 𝑥 𝑘 +D𝐴2 𝐵𝑢 𝑘 +𝐷𝐴𝐵𝑢 𝑘+1 + 𝐷𝐵𝑢 𝑘+2


39

Writing above equation in matrix form as:

𝑦𝑘 𝐷 0 0 0 𝑢𝑘
𝑦𝑘+1 𝐷𝐴 𝐷𝐵 0 0 𝑢
[𝑦 ] = [ 2 ] 𝑥𝑘 + [ ] [ 𝑘+1 ]
𝑘+2 𝐷𝐴 𝐷𝐴𝐵 𝐵𝐵 0 𝑢
𝑘+2
𝑦𝑘+3 𝐷𝐴3 𝐷𝐴2 𝐵 𝐷𝐴𝐵 𝐷𝐵

Prediction Matrix formulation for prediction horizon of L=4:

𝑦𝑘
𝑦𝑘+1
[𝑦 ] = 𝑦𝑘│4
𝑘+2
𝑦𝑘+3

𝐷
𝐷𝐴
[ 2 ] = 𝑂4
𝐷𝐴
𝐷𝐴3

0 0 0
𝐷𝐵 0 0
[ ] = 𝐻α 4
𝐷𝐴𝐵 𝐵𝐵 0
𝐷𝐴2 𝐵 𝐷𝐴𝐵 𝐷𝐵

𝑢𝑘
𝑢
[ 𝑘+1 ] = 𝑢𝑘│3
𝑢𝑘+2

Where,

𝑦𝑘│4 = Data matrix of output variables


𝑢𝑘│3 = Data matrix of input variables
𝑂4 = Extended observability matrix for pair (𝐷, 𝐴)
𝐻 α 4 = Lower block triangular Toeplitz matrix for (𝐷, 𝐴, 𝐵) matrices
𝑦𝑘│4 = 𝑂4 𝑥𝑘 + 𝐻 α 4 𝑢𝑘│3

For prediction horizon of length L:

𝑦𝑘│𝐿 = 𝑂𝐿 𝑥𝑘 + 𝐻 α 𝐿 𝑢𝑘│𝐿−1
40

For k =k+1:

𝑦𝑘+1│𝐿 = 𝑂𝐿 𝑥𝑘+1 + 𝐻 α 𝐿 𝑢𝑘+1│𝐿−1

But,

𝑥𝑘+1 = 𝐴𝑥 𝑘 +𝐵𝑢 𝑘

𝑦𝑘+1│𝐿 = 𝑂𝐿 𝐴𝑥 𝑘 +𝑂𝐿 𝐵𝑢 𝑘 + 𝐻 α 𝐿 𝑢𝑘+1│𝐿−1

𝑢𝑘
𝑦𝑘+1│𝐿 = 𝑂𝐿 𝐴𝑥 𝑘 + [𝑂𝐿 𝐵 𝐻 α 𝐿 ] [𝑢 ]
𝑘+1│𝐿−1

Let,

𝑃𝐿 = 𝑂𝐿 𝐴𝑥 𝑘

𝐹𝐿 = [𝑂𝐿 𝐵 𝐻α 𝐿 ]

𝑢𝑘
𝑈𝑘│𝐿 = [𝑢 ]
𝑘+1│𝐿−1

Substituting these we get following prediction model

𝑦𝑘+1│𝐿 = 𝑃𝐿 + 𝐹𝐿 𝑈𝑘│𝐿

𝑇
𝐽𝑘 = (𝑦𝑘+1│𝐿 − 𝑟𝑘+1│𝐿 ) Q (𝑦𝑘+1│𝐿 − 𝑟𝑘+1│𝐿 ) + 𝑈 𝑇 𝑘│𝐿 P 𝑈𝑘│𝐿

Put the value of 𝑦𝑘+1│𝐿 :

𝑇
𝐽𝑘 = (𝑃𝐿 + 𝐹𝐿 𝑈𝑘│𝐿 − 𝑟𝑘+1│𝐿 ) Q (𝑃𝐿 + 𝐹𝐿 𝑈𝑘│𝐿 − 𝑟𝑘+1│𝐿 ) +𝑈 𝑇 𝑘│𝐿 P 𝑈𝑘│𝐿
41

𝑇
𝐽𝑘 = 𝑈 𝑇 𝑘│𝐿 (𝐹𝐿 𝑇 𝑄 𝐹𝐿 + 𝑃) 𝑈𝑘│𝐿 + 2𝐹𝐿 𝑇 𝑄 (𝑃𝐿 − 𝑟𝑘+1│𝐿 ) 𝑈𝑘│𝐿 + (𝑃𝐿 − 𝑟𝑘+1│𝐿 ) Q
(𝑃𝐿 − 𝑟𝑘+1│𝐿 )

𝐽𝑘 = 𝑈 𝑇 𝑘│𝐿 H 𝑈𝑘│𝐿 + 2 𝑓 𝑇 𝑈𝑘│𝐿 + 𝐽𝑂 (3.15)

Where,

H = 𝐹𝐿 𝑇 𝑄 𝐹𝐿 + 𝑃

f = 𝐹𝐿 𝑇 𝑄 (𝑃𝐿 − 𝑟𝑘+1│𝐿 )

𝑇
𝐽𝑂 = (𝑃𝐿 − 𝑟𝑘+1│𝐿 ) Q (𝑃𝐿 − 𝑟𝑘+1│𝐿 ) (3.16)

Differentiate Equation (3.13) w.r.t 𝑈𝑘│𝐿 :

𝑑 𝐽𝑘
= 𝑈 𝑇 𝑘│𝐿 H 𝑈𝑘│𝐿 + 2 𝑓 𝑇 𝑈𝑘│𝐿 + 𝐽𝑂
𝑑 𝑈𝑘│𝐿

𝑈 ∗ 𝑘│𝐿 = − 𝐻 −1 * f
CHAPTER 4

RESULTS

Comparison of MPC with PID, Open Loop and DLQR

A simple 2nd order system is controlled using MPC and its response was compared
with Open loop response of the system, PID control and DLQR (Discrete Linear Quadratic
Regulator). It was observed that the convergence time of the system with PID controller
is less when gain variables are effectively chosen, which is a tedious task. However, PID
controller is not efficient for multi model systems and is not capable of handling
constraints. DLQR can overcome the problems of PID effectively. Moreover, it also
provides better performance in the presence of noise by incorporating a few modifications
such as pre-compensator, filters etc. MPC is a more general approach towards the given
problem. It can handle multi model systems, provides noise rejection capabilities and can
handle multiple constraints.

The above discussed points were verified on MATLAB Simulink on a second


order unconstrained system. The control algorithm discussed in section 3.2 is used for
MPC control of the system.

Figure 4. 1 Comparison between MPC, PID and DLQR for 2nd Order System on MATLAB
43

Control of PMSM using MPC

The construction and working of PMSM (Permanent Magnet Synchronous Motor)


is very much similar to that of Induction Motor. However, the control of PMSM is
relatively simple. To better understand the control algorithm MPC, we simulated PMSM
control through MPC. As the basic control technique of MPC is same, the same technique
can be extended for IM control.

Simulation is done on MATLAB Simulink and it was observed that the MPC effectively
tracks the reference speed and transients.

Figure 4. 2 Simulation of Model Predictive Control of PMSM on MATLAB

Induction Motor Parameters

Table 4. 1 Parameters of Induction Motor

PI SPEED CONTROLLER PARAMETERS

TSC = 2x10-5
Ts _DTFC = 4x10-5
Ts = 2x10-5
44

freq_max = 20000
Sampling Time of PI Controller [s] TSW =7 x 2x10-5
Proportional Gain Kp = 3.016
Integrative Gain Ki = 0.141

MACHINE PARAMETERS

Moment of Inertia (kg.m2) J = 0.062


Pole Pairs P=1
Magnetizing Inductance [H] Lm = 170 x 10-3
Stator Inductance [Henry] Ls = 175 x 10-3
Rotor Inductance [Henry] Lr = 175 x 10-3
Stator Resistance [Ohm] Rs = 1.2
Rotor Resistance [Ohm] Rr = 1
Nominal Stator Flux [Wb] sf_nom = 0.71
Nominal Torque [N.m] T_nom = 20
DC Link Voltage [Volts] Vdc = 530

AUXILIARY CONSTANTS

𝐿𝑠
𝑡𝑠 =
𝑅𝑠

𝐿𝑟
𝑡𝑟 =
𝑅𝑟

𝐿2𝑚
𝑠𝑖𝑔𝑚𝑎 = 1 − ( )
𝐿𝑟 𝐿𝑠

𝐿𝑚
𝑘𝑟 =
𝐿𝑟

𝑟_𝑠𝑖𝑔𝑚𝑎 = 𝑅𝑠 + 𝑘𝑟2 𝑅𝑟

𝑠𝑖𝑔𝑚𝑎 ∗ 𝐿𝑠
𝑡_𝑠𝑖𝑔𝑚𝑎 = 𝑟_𝑠𝑖𝑔𝑚𝑎
45

T_nom
𝑙𝑎𝑚𝑏𝑑𝑎 =
sf_nom

Open Loop Simulation of Induction Motor

The MATLAB Simulink is used to simulate the open-loop induction motor. The
open loop simulation is performed to validate the state variable response. Induction motor
model is selected from the Simulink Toolbox along with IGBT inverter. Arbitrary control
voltages ‘Usd’ and ‘Usq’ are defined and applied to PWM generator after transformation.
The Pulse Width Modulation (PWM) with carrier frequency of 2 kHz, dc voltage of 520V
and no load are used to keep the open loop test simple. The position angle in DQ-frame is
obtained independently from frequency of repeating sequence generator of 25Hz
frequency.

Figure 4. 3 Simulation of an Open Loop IM on MATLAB


46

Simulation of Direct Torque Control (DTC) of Induction Motor

Figure 4. 4 (a) Simulink Block Diagram of DTC

Figure 4. 5 (b) Simulink Block Diagram of DTC

4.5.1 Fixed Speed and Fixed Torque

Induction motor running at 500rpm and 10 N-m


47

Figure 4. 6 Fixed Reference Tracking by Induction Motor Controlled by DTC

4.5.2 Fixed Speed and Step changes in Torque

Induction motor running at 500rpm and step torque changes at 2 sec (2 N-m), 0.5sec (10
N-m) and 2.5 sec (5 N-m)

Figure 4. 7 Impact of Torque Changes in Induction Machine Speed Controlled by DTC

Below is the zoomed view indicating the minor changes in speed due to torque change.
48

Figure 4. 8 Impact of Torque Changes in Induction Machine speed Controlled by DTC (Zoomed)

4.5.3 Step changes in Speed at Fixed Torque

Step changes in speed at 0th second to 150 rpm, 1500 rpm at 1.5 sec and 1000rpm at
3.5sec instant.

Figure 4. 9 Step Reference Speed tracking of DTC Controlled Induction Motor

While the below graph shows absolute error:


49

Figure 4. 10 Absolute Error of above Reference Staircase Speed Tracking

4.5.4 Current THD at fixed Speed and Torque

The given current THD is observed at 500rpm speed and 10N-m torque.

Figure 4. 11 Current Frequency Spectrum of Induction Machine Controlled by DTC


50

4.6 Simulation of Field Oriented Control (FOC) of Induction Motor

Figure 4. 12 Simulink Block Diagram of FOC

Figure 4. 13 Unmasked-Simulink Block Diagram of FOC

4.6.1 Fixed Speed and Fixed Torque

Induction motor running at 500rpm and 10 N-m.


51

Figure 4. 14 Fixed Reference Tracking by Induction Machine Controlled by FOC

4.6.2 Fixed Speed and Step changes in Torque

Induction motor running at 500rpm and step torque changes at 2sec (2N-m), 0.5sec (10
N-m) and 2.5 sec (5 N-m)

Figure 4. 15 Impact of Torque Changes on Induction Machine speed when Controlled by FOC

Below is the zoomed view indicating the minor changes in speed due to torque change.
52

Figure 4. 16 Impact of Torque Changes on Induction Machine Speed when Controlled by FOC (Zoomed)

4.6.3 Step changes in Speed at fixed Torque

Step changes in speed at 0th second to 150rpm, 1500rpm at 1.5 sec and 1000rpm at
3.5sec instant.

Figure 4. 17 Step Reference Speed Tracking of FOC Controlled Induction Motor

While the below graph shows absolute error.


53

Figure 4. 18 Absolute Error of above Reference Staircase Speed Tracking

4.6.4 Current THD at fixed Speed and Torque

The given current THD is observed at 500rpm speed and 10N-m torque.

Figure 4. 19 Current Frequency Spectrum of Induction Machine Controlled by FOC


54

4.7 Simulation of Rotating Frame Finite Control Set-Model Predictive


Control (FCS-MPC) of Induction Motor

Figure 4. 20 Simulink Block Diagram of MPC (αβ)

4.7.1 Fixed Speed and Fixed Torque

Induction motor running at 500rpm and 10 N-m

Figure 4. 21 Fixed Reference Tracking by Induction Machine Controlled by MPC (αβ)


55

4.7.2 Fixed Speed and Step changes in Torque

Induction motor running at 500rpm and step torque changes at 2sec (2N-m), 0.5sec (10
N-m) and 2.5 sec (5 N-m)

Figure 4. 22 Impact of Torque Changes on Induction Machine Speed when Controlled by MPC (αβ)

Below is the zoomed view indicating the minor changes in speed due to torque change.

Figure 4. 23 Impact of Torque Changes on Induction Machine Speed when Controlled by MPC (αβ) (Zoomed)
56

4.7.3 Step changes in Speed at fixed Torque

Step changes in speed at 0th second to 150rpm, 1500rpm at 1.5 sec and 1000rpm at
3.5sec instant.

Figure 4. 24 Step Reference Speed Tracking of MPC (αβ) Controlled Induction Motor

While the below graph shows absolute error.

Figure 4. 25 Absolute Error of above Reference Staircase Speed Tracking


57

4.7.4 Current THD at fixed Speed and Torque

The given current THD is observed at 500rpm speed and 10N-m torque

Figure 4. 26 Current Frequency Spectrum of Induction Machine Controlled by MPC (αβ)


58

4.8 Simulation of Stationary Frame Finite Control Set-Model Predictive


Control(FCS-MPC) of Induction Motor

Figure 4. 27 Simulink Block Diagram of MPC (dq)

4.8.1 Fixed Speed and Fixed Torque

Induction motor running at 500rpm and 10 N-m

Figure 4. 28 Fixed Reference Tracking by Induction Machine Controlled by MPC (dq)


59

4.8.2 Fixed Speed and Step changes in Torque

Induction motor running at 500rpm and step torque changes at 2sec (2N-m), 0.5sec (10
N-m) and 2.5 sec (5 N-m)

Figure 4. 29 Impact of Torque Changes on Induction Machine Speed when Controlled by MPC (dq)

Below is the zoomed view indicating the minor changes in speed due to torque change.

Figure 4. 30 Impact of Torque Changes on Induction Machine Speed when Controlled by MPC (dq) (Zoomed)
60

4.8.3 Step changes in Speed at fixed Torque

Step changes in speed at 0th second to 150rpm, 1500rpm at 1.5 sec and 1000rpm at
3.5sec instant.

Figure 4. 31 Step Reference Speed tracking of MPC (dq) controlled Induction Motor

While the below graph shows absolute error.

Figure 4. 32 Absolute error of above reference staircase speed tracking


61

4.8.4 Current THD at fixed Speed and Torque

The given current THD is observed at 500rpm speed and 10N-m torque.

Figure 4. 33 Current Frequency Spectrum of Induction Machine Controlled by MPC (dq)

Learning Based Model Predictive Control of Induction Motor

Here, the evolutionary learning technique i.e. brain emotional intelligence is


combined with FCS-MPC with the purpose to minimize the tuning or weighing gains.
This technique is utilized in outer-speed loop to control reference values effectively.
62

Figure 4. 34 Learning Predictive Scheme

Figure 4. 35 Reference vs Actual Speed


CHAPTER 5

CONCLUSIONS, FUTURE RECOMMENDATIONS

After the comparison of conventional model predictive control with other vector
control techniques, MPC was found to be the most optimal method. However,
computational requirements and excessive tuning limits its application to induction motor.
Therefore, it is combined with artificial intelligence to eliminate tuning requirements, but
as in-case of BELIC, it tends to be highly computationally expensive than MPC.
Therefore, other efficient artificial intelligence techniques should be explored to overcome
the weaknesses of the conventional or proposed learning based MPC. In proposed
technique, optimized BELIC-MPC will be used to control motor successfully control
induction motor.
REFERENCES

[1] M. Popescu, "Induction Motor Modeling for Vector Control Purposes," Helsinki University
of Technology, Laboratory of Electromechanics, Espoo 2000.

[2] A. Chourasia, V. Srivastava, A. Choudhary and S. Praliya, "Comparison Study of Vector


Control of Induction Motor Using Rotor Flux Estimation by Two Different Methods,"
International Journal of Electronic and Electrical Engineering, vol. 7, no. 3, pp. 201-206,
2014.

[3] 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.

[4] C. Bordons and E. F. Camacho, "A Generalized Predictive Controller for a wide class of
Industrial Processes," IEEE Transactions on Control Systems Technology, vol. 6, no. 3, pp.
372-387, 1998.

[5] S. e. al, "Model Predictive Control," in Process Dynamics and Controls, 2011.

[6] Y. Zhang, H. Yang and B. Xia, "Model Predictive Control of Induction Motor Drives: Torque
Control Versus Flux Control," IEEE Transactions on Industry Applications, vol. 52, no. 5, pp.
4050-4060 , 2016.

[7] T. Geyer, G. Papafotiou and M. Morari, "Model Predictive Direct Torque Control - Part I:
Concept, Algorithm and Analysis," IEEE Transactions on Industrial Electronics, vol. 56, no.
6, pp. 1894-1905 , 2009.
65

[8] T. Geyer and G. Papafotiou, "Direct Torque Control for Induction Motor Drives: A Model
Predictive Control Approach Based on Feasibility," 2005.

[9] R. Lee, P. Pillay, and R. Harley, “D,Q reference frames for the simulation of induction
motors,” Electric Power Systems Research, vol. 8, no. 1, pp. 15–26, 1984.

[10] J. Rodriguez and P. Estay, Predictive control of power converters and electrical drives.
Chichester, West Sussex, UK: Wiley, 2012.

[11] L. Wang, S. Chai, D. Yoo, L. Gan, and K. Ng, PID and predictive control of electrical drives
and power converters using MATLAB/Simulink. Solaris South Tower, Singapore: John Wiley
& Sons, Inc., 2005.

[12] Grü ne Lars and Pannek Jü rgen, Nonlinear Model Predictive Control: Theory and Algorithms.
Cham: Springer International Publishing, 2017.

[13] J. A. Rossiter, Model-Based Predictive Control A Practical Approach. Bosa Roca: CRC Press,
2013.

[14] L. Wang, Model predictive control system design and implementation using MATLAB.
London: Springer, 2010.

[15] E. F. Camacho and C. Bordons, Model predictive control. London: Springer, 2007.

[16] B. K. Bose, Modern power electronics and AC drives. New Delhi: PHI Learning, 2012.

[17] Z. Jun, H. Lili, C. Likuan, and L. Yongbiao, “Vector Control Based on SVPWM for ACIM,”
Journal of Networks, vol. 8, no. 5, 2013.
66

[18] H. Miranda, P. Cortes, J. Yuz, and J. Rodriguez, “Predictive Torque Control of Induction
Machines Based on State-Space Models,” IEEE Transactions on Industrial Electronics, vol.
56, no. 6, pp. 1916–1924, 2009.

[19] F. Salem and M. Mosaad, “A Comparison Between MPC and Optimal PID Controllers: Case
Studies,” Michael Faraday IET International Summit 2015, 2015

[20] A. Z. Diab and V. V. Pankratov, “Model predictive control of vector controlled induction
motor drive,” 2012 7th International Forum on Strategic Technology (IFOST), 2012.

[21] C. Remsing, “Linear Control,” in RHODES UNIVERSITY.

[22] Analog Devices Inc. (2000). Reference Frame Conversions with the ADMC331. [online].
Available at:
http://smd.hu/Data/Analog/DSP/Motorcontrol/application/CONTROL/refframe331.pdf/
[Accessed 1 Aug. 2018].

[23] Texas Instruments (2018). Clarke and Park Transforms on the TMS320C2xx. BPRA048.
[online]. Available at: http://www.ti.com/lit/an/bpra048/bpra048.pdf/ [Accessed 1 Aug.
2018].

[24] En.wikipedia.org. (2018). Direct-quadrature-zero transformation. [online]. Available at:


https://en.wikipedia.org/wiki/Direct-quadrature-
zero_transformation#Park's_transformation/ [Accessed 1 Aug. 2018].

[25] Quora.com. (2018). [online] Available at: https://www.quora.com/Which-parts-of-the-


brain-control-which-human-behaviours/ [Accessed 9 Sep. 2018].
67

[26] Courses. lumenlearning.com. (2018). Biology of Emotion| Boundless Psychology. [online]


Available at: https://courses.lumenlearning.com/boundless-psychology/chapter/biology-
of-emotion/ [Accessed 9 Sep. 2018].
Final Year Project Report (Latest)- EE
ORIGINALITY REPORT

16 %
SIMILARIT Y INDEX
8%
INT ERNET SOURCES
11%
PUBLICAT IONS
9%
ST UDENT PAPERS

PRIMARY SOURCES

1
Submitted to University of Newcastle upon
Tyne
1%
St udent Paper

2
Wang, Liuping, Shan Chai, Dae Yoo, Lu Gan,
and Ki Ng. "Modeling of AC Drives and Power
1%
Converter", PID and Predictive Control of
Electrical Drives and Power Converters using
Matlab®/Simulink®, 2015.
Publicat ion

3
Submitted to Higher Education Commission
Pakistan
1%
St udent Paper

4
www.springer.com
Int ernet Source 1%
5
Submitted to American University of Beirut
St udent Paper 1%
6
tii.ieee-ies.org
Int ernet Source <1%
7
tel.archives-ouvertes.fr
Int ernet Source <1%
8
Submitted to University of Sheffield
St udent Paper <1%
9
Submitted to The University of the South
Pacific
<1%
St udent Paper

10
Submitted to University of Liverpool
St udent Paper <1%
11
Rodriguez, J., M. Kazmierkowski, J. Espinoza,
P. Zanchetta, H. Abu-Rub, H. Young, and C.
<1%
Rojas. "State of the Art of Finite Control Set
Model Predictive Control in Power Electronics",
IEEE Transactions on Industrial Informatics,
2012.
Publicat ion

12
idus.us.es
Int ernet Source <1%
13
Ravindra Kumar Sharma, Vivek Sanadhya,
Laxmidhar Behera, S. Bhattacharya. "Vector
<1%
control of a permanent magnet synchronous
motor", 2008 Annual IEEE India Conference,
2008
Publicat ion

14
K. Kalanithi, H. Abirami, S. Malaisamy, M.
Mohammed Rafiq, K. Saravanan. "Model
<1%
identification and dynamic matrix control of
ethane-ethylene distillation column", 2016
International Conference on Electrical,
Electronics, and Optimization Techniques
(ICEEOT), 2016
Publicat ion

15
"Trends in Advanced Intelligent Control,
Optimization and Automation", Springer
<1%
Nature, 2017
Publicat ion

16
Submitted to SASTRA University
St udent Paper <1%
17
Submitted to Universiti Teknologi MARA
St udent Paper <1%
18
Adel Merabet, Jason Gu. "Robust nonlinear
predictive control with modeling uncertainties
<1%
and unknown disturbance for single-link
flexible joint robot", 2008 7th World Congress
on Intelligent Control and Automation, 2008
Publicat ion

19
Submitted to Leicester College
St udent Paper <1%
20
Submitted to University of Newcastle
St udent Paper <1%
21
Zhonggang Yin, Chao Du, Jing Liu, Xiangdong
Sun, Yanru Zhong. "Research on
<1%
Autodisturbance-Rejection Control of Induction
Motors Based on an Ant Colony Optimization
Algorithm", IEEE Transactions on Industrial
Electronics, 2018
Publicat ion

22
Submitted to Central Queensland University
St udent Paper <1%
23
Kassem, Ahmed M., and Almoataz Y.
Abdelaziz. "Functional Predictive Control for
<1%
Voltage Stability Improvements of Autonomous
Hybrid Wind–Diesel Power System", Electric
Power Components and Systems, 2014.
Publicat ion

24
Chen, J.B.. "A multisymplectic integrator for the
periodic nonlinear Schrodinger equation",
<1%
Applied Mathematics and Computation,
20051115
Publicat ion

25
Joseph Plasmans. "A Multi-Country Closed-
Economy MU Model", Dynamic Modeling and
<1%
Econometrics in Economics and Finance, 2006
Publicat ion

26
192.149.244.11
Int ernet Source <1%
27
Selvi, S., and S. Gopinath. "Vector control of
induction motor drive using optimized GA
<1%
technique", 2015 IEEE 9th International
Conference on Intelligent Systems and Control
(ISCO), 2015.
Publicat ion

28
Submitted to University of Malaya
St udent Paper <1%
29
V. Sabatini, L. Bigarelli, M. Di Benedetto, A.
Lidozzi, L. Solero, G. Brown. "FPGA-based
<1%
Model Predictive Control for High Frequency
Variable Speed Generating Units", 2018
International Symposium on Power Electronics,
Electrical Drives, Automation and Motion
(SPEEDAM), 2018
Publicat ion

30
Submitted to Heathfield High School
St udent Paper <1%
31
Submitted to Federal University of Technology
St udent Paper <1%
32
Submitted to University of Warwick
St udent Paper <1%
33
Submitted to Monash University
St udent Paper <1%
34
Submitted to VIT University
St udent Paper <1%
35
Submitted to Universiti Teknologi Petronas
St udent Paper <1%
36
e.bangor.ac.uk
Int ernet Source <1%
37
digital-library.theiet.org
Int ernet Source <1%
38
Di Ruscio, David. "A Method for Identification
of Combined Deterministic Stochastic
<1%
Systems", Lecture Notes in Statistics, 1997.
Publicat ion

39
eprints.whiterose.ac.uk
Int ernet Source <1%
40
A. Gosling, S. Machin, C. Meghir. "The
Changing Distribution of Male Wages in the
<1%
U.K.", The Review of Economic Studies, 2000
Publicat ion

41
eref.uni-bayreuth.de
Int ernet Source <1%
42
A. Balluchi, L. Benvenuti, S. Engell, T. Geyer et
al. "Hybrid Control of Networked Embedded
<1%
Systems", European Journal of Control, 2005
Publicat ion

43
Ehrenfeld, Stephan, Oliver Herbort, and Martin
V. Butz. "Modular neuron-based body
<1%
estimation: maintaining consistency over
different limbs, modalities, and frames of
reference", Frontiers in Computational
Neuroscience, 2013.
Publicat ion

digitool.library.colostate.edu
Int ernet Source

<1%
44

45
Jung Hun Park, Han Woong Yoo, Soohee Han,
Wook Hyun Kwon. "Receding horizon control
<1%
for input-delayed systems", 2007 46th IEEE
Conference on Decision and Control, 2007
Publicat ion

46
Adrian I. Nachman, Lassi Päivärinta, Ari Teirilä.
"On imaging obstacles inside inhomogeneous
<1%
media", Journal of Functional Analysis, 2007
Publicat ion

47
www.nide-machine.com
Int ernet Source <1%
48
J. A. Riveros, J. Prieto, F. Barrero, S. Toral, M.
Jones, E. Levi. "Predictive Torque Control for
<1%
five-phase induction motor drives", IECON
2010 - 36th Annual Conference on IEEE
Industrial Electronics Society, 2010
Publicat ion

49
Y. El-Ibiary. "An accurate low cost method for
determining electric motors' efficiency for the
<1%
purpose of plant energy management", Record
of Conference Papers. Industry Applications
Society. Forty-Ninth Annual Conference. 2002
Petroleum and Chemical Industry Technical
Conference, 2002
Publicat ion
50
G. Mino-Aguilar, A. Michelle Dominguez, R.
Maya, R. Alvarez et al. "A direct torque control
<1%
for a PMSM", 2010 20th International
Conference on Electronics Communications
and Computers (CONIELECOMP), 2010
Publicat ion

51
Ishizaka, . "Analytic hierarchy process", Multi-
Criteria Decision Analysis Methods and
<1%
Software, 2013.
Publicat ion

52
Submitted to University of Northumbria at
Newcastle
<1%
St udent Paper

53
Witold Pawlus, Surya Teja Kandukuri, Geir
Hovland, Martin Choux, Michael R. Hansen.
<1%
"EKF-based estimation and control of electric
drivetrain in offshore pipe racking machine",
2016 IEEE International Conference on
Industrial Technology (ICIT), 2016
Publicat ion

54
Yongchang Zhang, Suyu Gao, Wei Xu. "An
improved model predictive current control of
<1%
permanent magnet synchronous motor drives",
2016 IEEE Applied Power Electronics
Conference and Exposition (APEC), 2016
Publicat ion

Submitted to iGroup
St udent Paper

<1%
55

56
www.ijmer.com
Int ernet Source <1%
57
mail.sahrdaya.ac.in
Int ernet Source <1%
58
www.waset.org
Int ernet Source <1%
59
Sun, Bo, Chi-Seng Lam, Ning-Yi Dai, Man-
Chung Wong, Chi-Kong Wong, Sai-Weng Sin,
<1%
U Seng-Pan, R. P. Martins, and U-Fat Chio. "A
FPGA-based power electronics controller for
hybrid active power filters", 2011 Asia Pacific
Conference on Postgraduate Research in
Microelectronics & Electronics, 2011.
Publicat ion

60
Azuwien Aida Bohari, Wahyu Mulyo Utomo,
Zainal Alam Haron, Nooradzianie Muhd. Zin,
<1%
Sy Yi Sim, Roslina Mat Ariff. "Speed Tracking
of Indirect Field Oriented Control Induction
Motor Using Neural Network", Procedia
Technology, 2013
Publicat ion

61
Marques, André M.. "Investigando a hipótese
da paridade do poder de compra: um enfoque
<1%
não linear", Revista de Economia
Contemporânea, 2011.
Publicat ion

62
eprints.nottingham.ac.uk
Int ernet Source <1%
63
Submitted to Kolej Universiti Linton
St udent Paper <1%
64
Saadatpoor, Atieh, Reza Iravani, and
Ahmadreza Tabesh. "Space-vector slope-based
<1%
method for fast locating of switched capacitors
in power systems", IET Generation
Transmission & Distribution, 2013.
Publicat ion

65
eprints.kfupm.edu.sa
Int ernet Source <1%
66
Andrzej M. Trzynadlowski. "The Field
Orientation Principle in Control of Induction
<1%
Motors", Springer Nature America, Inc, 1994
Publicat ion

67
www.jpier.org
Int ernet Source <1%
68
riuma.uma.es
Int ernet Source <1%
69
Moorthy, Yamuna. K., Pournami S. Chandran,
and S. Rishidas. "Motor current signature
<1%
analysis by multi-resolution methods using
Support Vector Machine", 2011 IEEE Recent
Advances in Intelligent Computational
Systems, 2011.
Publicat ion

70
COMPEL: The International Journal for
Computation and Mathematics in Electrical and
<1%
Electronic Engineering, Volume 24, Issue 4
(2006-09-19)
Publicat ion

71
Yongchang Zhang, Jianguo Zhu, Wei Xu,
Youguang Guo. "A Simple Method to Reduce
<1%
Torque Ripple in Direct Torque-Controlled
Permanent-Magnet Synchronous Motor by
Using Vectors With Variable Amplitude and
Angle", IEEE Transactions on Industrial
Electronics, 2011
Publicat ion

72
www.wseas.us
Int ernet Source <1%
73
pdfs.semanticscholar.org
Int ernet Source <1%
74
digital.lib.washington.edu
Int ernet Source <1%
75
www.neduet.edu.pk
Int ernet Source <1%
76
K.L. Mok. "EMTP simulation of a STATCOM
using hysteresis current control", Proceedings
of the IEEE 1999 International Conference on <1%
Power Electronics and Drive Systems PEDS 99
(Cat No 99TH8475), 1999
Publicat ion

77
Kovacic, . "Fuzzy Controller Design",
Automation and Control Engineering, 2005.
<1%
Publicat ion

78
Ghezzi, L.. "Toroidalization of generating
sequences in dimension two function fields",
<1%
Journal of Algebra, 20060715
Publicat ion

79
Salem, F., and M.I. Mosaad. "A Comparison
Between MPC and Optimal PID Controllers:
<1%
Case Studies", Michael Faraday IET
International Summit 2015, 2015.
Publicat ion

80
Studies in Computational Intelligence, 2014.
Publicat ion <1%
81
Yuan. "Electrical Fundamentals",
Electrochemical Impedance Spectroscopy in
<1%
PEM Fuel Cells, 2010
Publicat ion

82
Modern Electrical Drives, 2000.
Publicat ion <1%

You might also like