an improved battery/ultracapacitor … · storage system management strategy for electric vehicles...

229
A AN IMPROVED BATTERY/ULTRACAPACITOR HYBRID ENERGY STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY DEPARTMENT OF MECHANICAL ENGINEERING NATIONAL UNIVERSITY OF SINGAPORE 2017 Supervisors: Associate Professor Hong Geok Soon, Main Supervisor Associate Professor Lu Wen Feng, Co-Supervisor Examiners: Associate Professor Lee Kim Seng Associate Professor Chen Chao Yu, Peter Associate Professor Chen Xiaoqi, University of Canterbury

Upload: vuhuong

Post on 17-Sep-2018

229 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

A

AN IMPROVED BATTERY/ULTRACAPACITOR HYBRID ENERGY

STORAGE SYSTEM MANAGEMENT STRATEGY FOR

ELECTRIC VEHICLES

SO KAI MAN

(B.Eng. (Hons.), NUS)

A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

DEPARTMENT OF MECHANICAL ENGINEERING

NATIONAL UNIVERSITY OF SINGAPORE

2017

Supervisors:

Associate Professor Hong Geok Soon, Main Supervisor

Associate Professor Lu Wen Feng, Co-Supervisor

Examiners:

Associate Professor Lee Kim Seng

Associate Professor Chen Chao Yu, Peter

Associate Professor Chen Xiaoqi, University of Canterbury

Page 2: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

ii

DECLARATION PAGE

I hereby declare that this thesis is my original work and it has been written by me in its

entirety. I have duly acknowledged all the sources of information which have been used in

the thesis.

This thesis has also not been submitted for any degree in any university previously.

_______________________________________

So Kai Man

22 December 2017

Page 3: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

iii

ACKNOWLEDGMENTS

Firstly, the author would like to express his sincere gratitude to his supervisors – Assoc. Prof

Hong Geok Soon, Assoc. Prof Lu Wen Feng and Prof Wong Yoke San – for their

supervision, guidance and advice throughout the project.

Secondly, the author would like to extend his appreciation to his laboratory mates for their

help, suggestions and recommendations. They are Mr Kawsar Ali and Ms Sindhu Shetty from

Electrical Machines and Drives Lab, Department of Electrical & Computer Engineering, Mr

Lihil Uthpala Subasinghe and Mr Manikandan Balasundaram from Thermal Process Lab,

Department of Mechanical Engineering, Dr Zhang Ming, Ms See Hian Hian, Mr Ch’ng Chin

Boon, and Mr Yedige Tlegenov from Control Lab, Department of Mechanical Engineering.

Thirdly, the author would like to thank the staff at Control Labs 1 & 2, and Thermal Process

Lab 1, Department of Mechanical Engineering for their kind assistance rendered.

Lastly, special mention must be given to the author’s friends for their moral support.

Page 4: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

iv

TABLE OF CONTENTS

DECLARATION PAGE ......................................................................................................... ii

ACKNOWLEDGMENTS ..................................................................................................... iii

TABLE OF CONTENTS ....................................................................................................... iv

SUMMARY .............................................................................................................................. x

LIST OF TABLES ................................................................................................................. xii

LIST OF FIGURES .............................................................................................................. xiv

LIST OF SYMBOLS ............................................................................................................. xx

LIST OF ACRONYMS ........................................................................................................ xxi

1 INTRODUCTION............................................................................................................ 1

1.1 Problems of Electric Vehicles ..................................................................................... 1

1.2 Battery/UC HESS ........................................................................................................ 2

1.3 Energy Management Strategy & Power Management Strategy .................................. 3

1.4 Criticism of Battery/UC HESS ................................................................................... 4

1.5 Objective ..................................................................................................................... 4

1.6 Scope ........................................................................................................................... 5

1.7 Structure of Thesis ...................................................................................................... 6

1.8 Contributions ............................................................................................................... 8

1.9 Publications ................................................................................................................. 9

2 LITERATURE REVIEW ............................................................................................. 10

2.1 Overview ................................................................................................................... 10

2.2 Hardware: HESS Topologies .................................................................................... 11

2.3 Software: HESS Management Strategies .................................................................. 14

2.3.1 Energy Management Strategies ......................................................................... 14

2.3.2 Power Management Strategies ........................................................................... 16

Page 5: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

v

2.4 Case Studies .............................................................................................................. 18

2.4.1 Dixon, Ortuzar & Moreno.................................................................................. 18

2.4.2 Avelino, Garcia, Ferreira & Pomilio.................................................................. 20

2.4.3 Choi, Lee & Seo ................................................................................................. 20

2.5 Limitations of Existing Works and Proposed Approach ........................................... 21

3 VEHICLE & HESS MODELLING ............................................................................. 23

3.1 Selected HESS Topology .......................................................................................... 23

3.2 Drive Cycle ............................................................................................................... 24

3.3 Simulation Approach................................................................................................. 26

3.4 Motor & Inverter Model ............................................................................................ 27

3.5 Auxiliary Loads ......................................................................................................... 30

3.6 Vehicle Model ........................................................................................................... 30

3.7 Battery Model ............................................................................................................ 37

3.8 Ultracapacitor Model................................................................................................. 40

3.9 DC/DC Converter Model .......................................................................................... 42

3.9.1 Boost Mode Duty Cycle..................................................................................... 43

3.9.2 Boost Mode Efficiency ...................................................................................... 46

3.9.3 Buck Mode Duty Cycle ..................................................................................... 47

3.9.4 Buck Mode Efficiency ....................................................................................... 50

3.9.5 Combined Duty Cycle and Efficiency ............................................................... 51

3.10 Battery Cycle Life Model .......................................................................................... 52

3.11 Parameters for Modelling .......................................................................................... 54

3.11.1 General ............................................................................................................... 54

3.11.2 Using UR18650W Batteries .............................................................................. 57

Page 6: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

vi

4 HESS: IMPROVED ENERGY & POWER MANAGEMENT STRATEGIES ...... 59

4.1 Power Management Strategy Pt. 1: Battery Limits ................................................... 59

4.1.1 Battery Power at Constant Speeds ..................................................................... 60

4.1.2 Final Battery Limit Curve .................................................................................. 61

4.2 Energy Management Strategy ................................................................................... 63

4.2.1 Sufficient Space in UC for Regenerative Braking ............................................. 64

4.2.2 Sufficient Energy in UC for Acceleration ......................................................... 70

4.2.3 Selected Braking and Acceleration Torque Values ........................................... 74

4.2.4 Target UC Energy Band..................................................................................... 79

4.2.5 Summary ............................................................................................................ 85

4.3 Power Management Strategy Pt. 2: Implementation ................................................. 85

4.4 Summary ................................................................................................................... 88

5 HESS SIMULATIONS .................................................................................................. 91

5.1 Implementation.......................................................................................................... 91

5.2 EMS Comparison: Target UC Energy Band ............................................................. 93

5.3 PMS Comparison: Battery Power Limits .................................................................. 96

5.4 Drive Cycles .............................................................................................................. 97

5.4.1 LA92 Drive Cycle .............................................................................................. 97

5.4.2 EUDC ............................................................................................................... 103

5.4.3 FTP-75 City Drive Cycle ................................................................................. 104

5.5 Battery Cycle Life ................................................................................................... 107

5.5.1 Description ....................................................................................................... 107

5.5.2 Drive Cycle Comparison.................................................................................. 107

5.5.3 Simulation ........................................................................................................ 108

5.5.4 Drive Cycle Selection for Experiment 2 .......................................................... 112

Page 7: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

vii

5.6 Summary ................................................................................................................. 112

6 HESS EXPERIMENTS ............................................................................................... 114

6.1 Setup ........................................................................................................................ 114

6.1.1 Energy Storage Components............................................................................ 117

6.1.2 DC/DC Converter ............................................................................................ 117

6.1.3 Sensors ............................................................................................................. 120

6.1.4 Safety Components .......................................................................................... 124

6.2 Scaling ..................................................................................................................... 126

6.3 Scale Factor k .......................................................................................................... 127

6.3.1 Selecting k Based on Drive Cycle ................................................................... 127

6.3.2 Selecting k Based on Simulation/Experiment Energy Ratio ........................... 128

6.3.3 Selecting k Based on Battery Limitations ........................................................ 129

6.3.4 Consequences of k = 160 ................................................................................. 130

6.4 Experiment 0 ........................................................................................................... 132

6.4.1 Objective .......................................................................................................... 132

6.4.2 Procedure ......................................................................................................... 132

6.4.3 Results .............................................................................................................. 134

6.4.4 Summary .......................................................................................................... 140

6.5 Software Implementation ........................................................................................ 140

6.5.1 Algorithm Curve Fitting .................................................................................. 140

6.5.2 Syncing Speed with Power .............................................................................. 144

6.5.3 Problems with Maccor ..................................................................................... 144

6.5.4 Battery Power vs. Current Thresholds ............................................................. 145

6.5.5 Integral Controller for UC Voltage Control..................................................... 146

6.6 Experiment 1 ........................................................................................................... 149

Page 8: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

viii

6.6.1 Objective .......................................................................................................... 149

6.6.2 EUDC Drive Cycle .......................................................................................... 149

6.6.3 FTP-75 City Drive Cycle ................................................................................. 153

6.6.4 Average Currents ............................................................................................. 158

6.6.5 Summary .......................................................................................................... 159

6.7 Experiment 2 ........................................................................................................... 160

6.7.1 Objective .......................................................................................................... 160

6.7.2 Description ....................................................................................................... 160

6.7.3 Procedure ......................................................................................................... 161

6.7.4 Battery Discharge Capacity Test ..................................................................... 162

6.7.5 Battery-only Setup ........................................................................................... 165

6.7.6 Initial Results ................................................................................................... 166

6.7.7 Battery-only Undervoltage During Demanding Sections ................................ 167

6.7.8 Battery-only Contact Resistance ...................................................................... 168

6.7.9 Battery-only Setup, Revised ............................................................................ 172

6.7.10 Final Results..................................................................................................... 174

6.7.11 Summary .......................................................................................................... 176

7 HESS REDUCED-SCALED SIMULATIONS ......................................................... 177

7.1 Differences between Full and Reduced-scale Simulation ....................................... 177

7.2 EUDC ...................................................................................................................... 178

7.3 FTP-75 City Drive Cycle ........................................................................................ 179

7.4 Total Energy Use ..................................................................................................... 181

7.5 Summary ................................................................................................................. 183

8 CONCLUSION & FUTURE WORKS ...................................................................... 184

8.1 Conclusion ............................................................................................................... 184

Page 9: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

ix

8.2 Future Works ........................................................................................................... 186

8.2.1 Optimization to Extend Battery Cycle Life ..................................................... 186

8.2.2 Full-scale Implementation ............................................................................... 187

8.2.3 SuPower Battery Cycle Life Curve Fitting ...................................................... 188

8.2.4 Cost-Benefit Analysis of UCs .......................................................................... 189

8.2.5 Improvement to Experiments ........................................................................... 189

BIBLIOGRAPHY ................................................................................................................ 190

APPENDIX ........................................................................................................................... 201

A Charging Procedure .................................................................................................... 201

B Scaling......................................................................................................................... 202

C Sensor Circuits ............................................................................................................ 206

Page 10: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

x

SUMMARY

Electric Vehicles (EVs) still have not been adopted by the masses yet. Common grievances

include the high cost of EVs as well as the limited driving range. The EV battery is an

expensive component, so it is desirable to extend the batteries’ cycle life. One reason which

shortens battery lifespan is high charge and discharge rates. A solution is to adopt a Hybrid

Energy Storage System (HESS), with an Ultracapacitor (UC) assisting the battery. The high

power density of the UC can relieve the battery of high charge and discharge rates, extending

the battery’s cycle life.

There are two parts to the HESS – energy management strategy (EMS) and power

management strategy (PMS). Most existing works focus on PMS. The existing EMS are quite

empirical, such as a fixed target UC energy level regardless of loading conditions. This can

be further improved.

In this work, a novel HESS management strategy is proposed. The EMS involves a more

comprehensive method of setting the target UC energy level using a speed-dependent band,

which considers worst case scenarios and real-life drive cycles. This is the first contribution

of this work. With the proposed EMS, the UC can achieve two goals – to contain sufficient

energy required for future accelerations, and to have sufficient space to store energy captured

from future regenerative braking.

The PMS has two goals – to ensure the EMS (target UC energy level) is followed, and to

ensure the battery charge/discharge rates do not exceed the limits. A novel method of setting

the battery power limit based on speed is proposed, which is the second contribution of this

work. This also has two goals – better utilisation of the UC and to ensure that when the EV

travels at constant speed, the power is supplied mainly by the battery.

Page 11: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

xi

Simulations were performed using the proposed battery/UC HESS management strategy on a

mid-sized EV sedan. The results show that the proposed strategy achieves the four goals of

the EMS and PMS mentioned above, as well as the two speed-dependent battery limit goals.

Further simulations show that existing published works cannot achieve all the goals

simultaneously unless their UCs are sized twice as large, increasing weight and costs.

Subsequently, battery cycle life simulations were performed to observe the battery capacity

fade for the proposed battery/UC HESS, and for a battery-only system. Almost 30%

reduction in capacity loss due to cycling was seen for the proposed battery/UC HESS as

compared to the battery-only system when running three FTP-75 city drive cycles daily over

10 years.

Afterwards, a reduced-scale experiment was constructed. This experiment verified the

proposed strategy could work physically as intended. Then the experiment was compared to a

reduced-scale simulation. They behave similarly time-wise and in terms of energy consumed.

Lastly, another set of experiment was performed to compare the battery cycle life of the

battery/UC HESS to a battery-only system. Each setup was cycled continuously with the

FTP-75 city drive cycle. As only 190 cycles have been completed, the results are too close to

call.

Page 12: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

xii

LIST OF TABLES

Table 3-1 Simulation parameters for the vehicle model. ......................................................... 55

Table 3-2 Simulation parameters for the EV motor................................................................. 56

Table 3-3 Parameters of new components to be installed in the EV. ...................................... 57

Table 3-4 Parameters of Nissan Leaf and modified batteries. ................................................. 58

Table 3-5 Parameters for battery voltage and battery cycle life curve fitting. ........................ 58

Table 5-1 Comparison of no. of drive cycles to hit 50km. .................................................... 108

Table 5-2 Simulated cycle life capacity losses over 10 years. ............................................... 110

Table 5-3 Simulated cycle life for 20% capacity losses. ....................................................... 111

Table 5-4 Simulated cycle life for 80% capacity losses, including calendar loss. ................ 112

Table 6-1 Energy storage components. .................................................................................. 117

Table 6-2 DC/DC converter components. ............................................................................. 119

Table 6-3 Sensors and supporting components. .................................................................... 121

Table 6-4 Safety & miscellaneous components. .................................................................... 125

Table 6-5 Symbols and their meanings for scaling derivation. ............................................. 127

Table 6-6 Battery and UC energy specifications. .................................................................. 129

Table 6-7 Value of k which allows 3x FTP-75 city drive cycle to complete. ....................... 130

Table 6-8 UC scaling specifications. ..................................................................................... 131

Table 6-9 Battery scaling specifications. ............................................................................... 132

Table 6-10 DC/DC converter efficiency test for one data point. ........................................... 133

Table 6-11 Reduced-scale simulation parameters. ................................................................ 137

Table 6-12 Average of absolute battery currents for experiments ......................................... 159

Table 6-13 Discharge capacity test procedure. ...................................................................... 162

Table 6-14 FTP-75 city drive cycles before battery undervoltage. ....................................... 169

Table 7-1 Experiment battery energy use. ............................................................................. 182

Page 13: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

xiii

Table 7-2 Comparison of battery energy use between experiments and reduced-scale

simulations. ............................................................................................................................ 182

Table A-1 Symbols and their meanings for scaling derivation. ............................................. 202

Page 14: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

xiv

LIST OF FIGURES

Figure 1-1 Ragone plot illustrating power and energy densities [9]. ......................................... 2

Figure 2-1 Passive configuration [32]...................................................................................... 11

Figure 2-2 Nissan Leaf Li-ion battery pack discharge curve [34]. .......................................... 12

Figure 2-3 Partially-decoupled configuration [32]. ................................................................. 12

Figure 2-4 Partially-decoupled configuration 2 [32]. .............................................................. 13

Figure 2-5 Fully-decoupled cascaded configuration [32]. ....................................................... 13

Figure 2-6 Fully-decoupled multiple converter configuration [32]. ........................................ 14

Figure 2-7 Power management strategies [11]. ....................................................................... 16

Figure 2-8 Target UC SOC values [47]. .................................................................................. 19

Figure 3-1 Single DC/DC converter between battery and UC. ............................................... 23

Figure 3-2 FTP-75 city drive cycle (Transient) [55]. .............................................................. 24

Figure 3-3 FTP-75 HWFET drive cycle (Transient) [55]. ....................................................... 25

Figure 3-4 LA92 drive cycle (Transient) [55]. ........................................................................ 25

Figure 3-5 ECE-15 drive cycle (Modal) [55]. ......................................................................... 26

Figure 3-6 EUDC (Modal) [55]. .............................................................................................. 26

Figure 3-7 2011 Nissan Leaf combined motor/inverter efficiency [59]. ................................. 28

Figure 3-8 Combined motor/inverter efficiencies.................................................................... 29

Figure 3-9 Block diagram for vehicle model updating. ........................................................... 31

Figure 3-10 EUDC input. ......................................................................................................... 35

Figure 3-11 Forces over EUDC: intermediate output from model. ......................................... 36

Figure 3-12 Torque over EUDC: intermediate output from model. ........................................ 36

Figure 3-13 Power over EUDC: model output. ....................................................................... 36

Figure 3-14 Battery model. ...................................................................................................... 37

Figure 3-15 Block diagram for battery model updating. ......................................................... 37

Page 15: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

xv

Figure 3-16 UC model. ............................................................................................................ 40

Figure 3-17 Block diagram for UC model updating. ............................................................... 40

Figure 3-18 Electrical model of DC/DC converter. ................................................................. 42

Figure 3-19 Boost mode, Q2 on. .............................................................................................. 43

Figure 3-20 Boost mode, Q2 off. ............................................................................................. 44

Figure 3-21 Buck mode, Q1 on................................................................................................ 47

Figure 3-22 Buck mode, Q1 off. .............................................................................................. 48

Figure 3-23 Combined buck-boost duty cycle. ........................................................................ 51

Figure 3-24 Combined buck-boost efficiency. ........................................................................ 51

Figure 4-1 Battery power required at constant vehicle speed. ................................................. 61

Figure 4-2 Speed-dependent PMS battery limit curve. ............................................................ 62

Figure 4-3 Block diagram of overall EMS design ................................................................... 64

Figure 4-4 EMS block diagram to anticipate UC space required or energy generated. .......... 65

Figure 4-5 (a) Kinematics (b) torques (c) powers (d) SOCs during regenerative braking. ..... 69

Figure 4-6 (a) Kinematics (b) torques (c) powers (d) SOCs during acceleration. ................... 73

Figure 4-7 Target UC SOC for varying brake torques and start velocities. ............................ 76

Figure 4-8 Brake torques corresponding to max. recovered UC energy for each velocity. .... 76

Figure 4-9 Target UC SOC for varying brake torques and start velocities. ............................ 78

Figure 4-10 Target UC SOC band vs. speed, 6 UC modules. ................................................. 80

Figure 4-11 Target UC SOC band vs. speed, 5 UC modules. ................................................. 81

Figure 4-12 Target UC SOC for varying brake torques and start velocities, 5 UC modules. . 82

Figure 4-13 Target UC SOC band vs. speed, no battery power loosening .............................. 83

Figure 4-14 Block diagram of PMS algorithm. ....................................................................... 86

Figure 5-1 Complete simulation block diagram. ..................................................................... 92

Figure 5-2 Target UC SOC band vs. speed, 6 UC modules. ................................................... 93

Page 16: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

xvi

Figure 5-3 Target UC SOC band vs. speed, 13 UC modules. ................................................. 95

Figure 5-4 Speed-dependent PMS battery limit curve comparison. ........................................ 96

Figure 5-5 LA92 drive cycle power required. ......................................................................... 98

Figure 5-6 LA92 torque profile for the mid-sized EV. ............................................................ 98

Figure 5-7 LA92 power distribution of battery and UC. ......................................................... 99

Figure 5-8 LA92 UC SOC (target and actual) and battery SOC. ............................................ 99

Figure 5-9 LA92 torque profile for the mid-sized EV, zoomed 820-920s. ........................... 100

Figure 5-10 LA92 power distribution of battery and UC, zoomed 820-920s. ....................... 100

Figure 5-11 LA92 UC and battery SOC, zoomed 820-920s. ................................................. 101

Figure 5-12 EUDC power distribution of battery and UC. .................................................... 103

Figure 5-13 EUDC UC and battery SOC. .............................................................................. 103

Figure 5-14 FTP-75 city power distribution of battery and UC. ........................................... 105

Figure 5-15 FTP-75 city UC and battery SOC. ..................................................................... 105

Figure 5-16 FTP-75 city power distribution of battery and UC, zoomed 150-250s. ............. 106

Figure 5-17 FTP-75 city UC and battery SOC, zoomed 150-250s. ....................................... 106

Figure 5-18 Battery capacity loss curve for battery/UC system over FTP-75 city. ............... 109

Figure 6-1 Electrical diagram of experiment setup................................................................ 115

Figure 6-2 Photo of experiment setup (front). ....................................................................... 115

Figure 6-3 Photo of experiment setup (top). .......................................................................... 116

Figure 6-4 Photo of experiment setup (side, auxiliary equipment). ...................................... 116

Figure 6-5 Photo of DC/DC converter. .................................................................................. 118

Figure 6-6 Voltage sensor calibration. ................................................................................... 123

Figure 6-7 Current sensor calibration. ................................................................................... 124

Figure 6-8 Experiment 0 setup. .............................................................................................. 133

Figure 6-9 DC/DC converter boost efficiency. ...................................................................... 134

Page 17: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

xvii

Figure 6-10 DC/DC converter buck efficiency. ..................................................................... 135

Figure 6-11 Interpolated DC/DC converter efficiency from experiment. ............................. 136

Figure 6-12 DC/DC converter efficiency from simulation. ................................................... 139

Figure 6-13 DC/DC converter efficiency, experimental minus simulation output. ............... 139

Figure 6-14 Reduced-scale combined motor/inverter efficiency. .......................................... 141

Figure 6-15 Battery power to maintain EV at constant speed for k=160. ............................. 141

Figure 6-16 Target UC SOC band for k=160. ....................................................................... 142

Figure 6-17 EUDC speed profile after scaling with k=160. .................................................. 149

Figure 6-18 Battery & UC currents from experiment, EUDC. .............................................. 150

Figure 6-19 Battery & UC voltages from experiment, EUDC. ............................................. 151

Figure 6-20 Target UC voltage addition term vuc,tar,add, EUDC. ............................................ 151

Figure 6-21 Power and time sync check, EUDC. .................................................................. 153

Figure 6-22 FTP-75 city speed profile after scaling with k=160. .......................................... 153

Figure 6-23 Battery & UC currents from experiment, FTP-75 city. ..................................... 154

Figure 6-24 Battery & UC voltages from experiment, FTP-75 city. ..................................... 154

Figure 6-25 Target UC voltage addition term vuc,tar,add, FTP-75 city. .................................... 155

Figure 6-26 Battery & UC currents from experiment, FTP-75 city, zoomed 180-280s. ....... 155

Figure 6-27 Battery & UC voltages from experiment, FTP-75 city, zoomed 180-280s. ...... 156

Figure 6-28 Target UC voltage addition term vuc,tar,add, FTP-75 city, zoomed 180-280s. ..... 156

Figure 6-29 Power and time sync check, FTP-75 city. .......................................................... 157

Figure 6-30 Battery-only setup. Currents from experiment, FTP-75 city run 1. ................... 158

Figure 6-31 (a) Discharging current from ACS3 (b) coulomb counting. .............................. 164

Figure 6-32 Discharging voltage. .......................................................................................... 164

Figure 6-33 Battery discharging capacity. ............................................................................. 164

Figure 6-34 Battery discharge capacity tests over 60 cycles. ................................................ 166

Page 18: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

xviii

Figure 6-35 Battery-only voltage for FTP-75 city, tripped, k=160. ...................................... 167

Figure 6-36 Battery-only current for FTP-75 city, tripped, k=160. ....................................... 167

Figure 6-37 Contact resistance experiment connections diagram. ........................................ 170

Figure 6-38 Current profile to find voltage drop. .................................................................. 171

Figure 6-39(a) Measured voltages (b) voltage drop. ............................................................. 171

Figure 6-40 Equivalent resistance (contact resistance and other losses). .............................. 171

Figure 6-41 Photo of battery-only setup. ............................................................................... 172

Figure 6-42(a) Measured voltages (b) voltage drop. ............................................................. 173

Figure 6-43 Equivalent resistance (contact resistance and others). ....................................... 173

Figure 6-44 Battery capacity of battery-only and battery/UC setup. ..................................... 174

Figure 6-45 Relative capacity. Cycle capacity divided by initial capacity. ........................... 174

Figure 6-46 Capacity “gained” due to high contact resistance .............................................. 175

Figure 7-1 Battery & UC currents from simulation, EUDC (compare with experiment in

Figure 6-18). .......................................................................................................................... 178

Figure 7-2 Battery & UC voltages from simulation, EUDC (compare with experiment in

Figure 6-19). .......................................................................................................................... 178

Figure 7-3 Battery & UC currents from simulation, FTP-75 city (compare with experiment in

Figure 6-23). .......................................................................................................................... 179

Figure 7-4 Battery & UC voltages from simulation, FTP-75 city (compare with experiment in

Figure 6-24). .......................................................................................................................... 180

Figure 7-5 Battery & UC currents from simulation, FTP-75 city, zoomed 180-280s (compare

with experiment in Figure 6-26). ........................................................................................... 180

Figure 7-6 Battery & UC voltages from simulation, FTP-75 city, zoomed 180-280s (compare

with experiment in Figure 6-27). ........................................................................................... 181

Figure A-1 Charging current (b) coulomb counting. ............................................................. 201

Page 19: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

xix

Figure A-2 Charging voltage. ................................................................................................ 201

Figure A-3 Voltage sensors and filters. ................................................................................. 206

Figure A-4 Current sensor filters, precision voltage reference and thermistor. ..................... 207

Figure A-5 Relays. ................................................................................................................. 208

Page 20: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

xx

LIST OF SYMBOLS

ρ Air density g Gravitational acceleration

ωx Angular velocity h Height of CG

A, B, K,

Vbatt,0

Battery constants Rx Internal resistance, where x =

batt / uc / etc.

a, b, c, d, e Battery cycle life constants m Mass

C Capacitance of UC Px Power, where x = batt / uc / etc.

Qx Capacity, where x = batt /

uc / etc.

μs Static friction coefficient

Ahthroughput Charge used SOCx State of Charge, where x = batt

/ uc / etc.

ix Current, where x = batt /

uc / etc.

T Temperature

Lb Distance from rear wheel

to CG

t, j, k Time / Loop iteration

Cd Drag coefficient τx Torque

ηx Efficiency vwh Velocity of wheel

Ex Energy, where x = batt / uc

/ etc.

vx Voltage, where x = batt / uc /

etc.

Af Frontal area rwh Wheel radius

Gr Gear ratio L Wheelbase length

θ Gradient angle of road

Page 21: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

xxi

LIST OF ACRONYMS

BMS Battery Management System

CC Constant Current

CV Constant Voltage

DoD Depth of Discharge

ECE Economic Commission for Europe

EMS Energy Management Strategy

EUDC Extra-Urban Driving Cycle

EV Electric Vehicle

FTP Federal Test Procedures

HESS Hybrid Energy Storage System

IGBT Insulated-gate bipolar transistor

Li-ion Lithium-ion

MOSFET Metal-oxide-semiconductor field-effect transistor

NEDC New European Driving Cycle

OCV Open Circuit Voltage

PMS Power Management Strategy

PMSM Permanent Magnet Synchronous Motor

SOC State of Charge

UC Ultracapacitor

Page 22: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

1

1 INTRODUCTION

1.1 Problems of Electric Vehicles

Electric Vehicles (EVs) have already been commercially available for a number of years.

Well-known models include the Nissan Leaf, Mitsubishi i-MiEV, and the high-performance

Tesla Model S. However, they have not been adopted by the masses yet – most of the EVs

today are only used for trial or research purposes. Reasons for this situation include the high

costs of EVs, as well as their limited driving range.

In Singapore, a brand new Nissan Leaf for private use costs S$200k with Certificate of

Entitlement (COE) at May 2014 prices (EVs for research purposes are exempted from COEs)

[1]. As a comparison, the petrol powered equivalent Nissan Sylphy 1.6L costs S$110k [1],

making the Nissan Leaf almost twice as expensive. In the United States, the Nissan Leaf

costs US$21.5k [2] while a Nissan Sentra 1.8L (a rebadged Sylphy) costs US$16k [3] as of

December 2014, making the Nissan Leaf still 35% more expensive.

The Lithium-ion (Li-ion) battery is an expensive component of the EV. As commercial EVs

are still relatively new, there is not much data on the reliability of EV batteries yet. But in the

reasonably mature hybrid car scene, consumers often complain about the costs of replacing

the battery, which usually last only about 10 years, sometimes even shorter [4] [5]. The

Nissan Leaf battery pack costs US$5.5k (and has an eight year warranty) [6], a quarter of a

brand new US$21.5k Nissan Leaf, which is a substantial amount.

Therefore, it is desirable to extend the lifespan of the battery in order to reduce replacement

costs. One reason which causes a reduction of battery lifespan is high charge and discharge

rates [7] [8].

Page 23: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

2

1.2 Battery/UC HESS

A solution is to adopt a Hybrid Energy Storage System (HESS), with an additional energy

storage device to assist the battery. A good choice for the additional device is the

Ultracapacitor (UC), also known as the Supercapacitor, which has a higher capacitance than

traditional electrolytic capacitors, but lower voltage limits.

The UC has a high power density, but a low energy density. The Li-ion battery has

contrasting specifications – low power density but high energy density. This is illustrated in

Figure 1-1 with a Ragone plot.

Figure 1-1 Ragone plot illustrating power and energy densities [9].

Therefore, the UC and battery complement each other perfectly in an HESS. Furthermore, the

UC has a longer cycle life. Generally, Li-ion batteries achieve 500 to 1000 full

charge/discharge cycles, while UCs achieve 1 million cycles [10]. Therefore, the UC can

handle the high charging and discharging power peaks due to its high power density and

longer cycle life, resulting in lower battery charge and discharge rates, and therefore longer

battery cycle life.

Page 24: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

3

There are no commercial vehicles with a battery/UC HESS at the moment as much research

still needs to be done to address the energy and power management issues in the HESS.

As a side note, there are two main purposes of a battery/UC HESS system. The first purpose

is to extend battery cycle life (which this thesis focuses on). In this case, the battery is sized

such that the battery can solely meet the required energy and power of the EV design. Then

the UC is added to the system to reduce the peak battery power so that the battery cycle life

can be extended.

The second purpose is to provide additional power that the battery is unable to provide on its

own. For example, the battery of the EV could be halved, and the UC provides the other half

of the power. The benefits of this purpose as compared to the first purpose is that weight and

costs have been reduced, as only half the original battery is required. However, the drawback

is that battery cycle life is not extended, and since the battery is halved, the EV range is also

halved. As range is a big concern of EVs today, it is not suggested to reduce the range

further. Therefore, this work focuses on the first purpose, which is to extend battery cycle

life.

1.3 Energy Management Strategy & Power Management Strategy

Although energy management strategy (EMS) and power management strategy (PMS) are

sometimes used interchangeably in the literature, they are actually two distinct concepts [11].

EMS involves managing the energy levels in the HESS, while PMS means managing the

power flow within the HESS. Both management strategies must be implemented together for

an HESS to work.

There are many such battery/UC HESS strategies discussed in the literature. Most works

focus on PMS. The EMS are generally empirical or quite rudimentary, such as attempting to

maintain a constant energy level in the UC, regardless of loading. Therefore, the UC has to be

Page 25: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

4

sized larger in order to contain sufficient energy for acceleration, and have sufficient

remaining space to store energy recovered from regenerative braking (which are two goals

the UC should have).

UCs of today are relatively expensive, with the UC suggested in this proposed work – six

Maxwell general purpose UC modules [12] – costing approximately US$7.2k (as of

December 2017) [13]. Therefore, it is desirable to minimize the UC size.

1.4 Criticism of Battery/UC HESS

Some critics of the battery/UC HESS argue the extra money could be used to buy more

battery modules instead of a UC to relieve battery stress. However, this would greatly

increase the EV’s weight for a similar power density. The battery pack in this proposed work

(similar to a Nissan Leaf battery) is 294kg and rated for 90kW, while the six Maxwell UC

bank is 62kg and rated for 331kW. The supporting components (e.g. inductor, DC/DC

converter, etc.) are 22kg, resulting in a theoretical combined rating of 421kW at 378kg. If a

duplicate battery pack were added instead of the UC, the twin battery packs would have a

combined rating of 180kW at 588kg (still far from the 421kW).

Despite the extra range the second battery pack offers, it is far heavier than the battery/UC

HESS option. The extra weight is equivalent to almost three passengers (assuming each

passenger is 75kg). This results in sluggish vehicle performance and larger energy

consumption from hauling the extra weight. Therefore, the battery/UC HESS combination is

better for relieving battery stress and increasing the battery cycle life, while not significantly

affecting vehicle performance.

1.5 Objective

The objective of this work is to develop a new HESS management strategy, consisting of an

EMS and a PMS, which performs better than existing works in terms of achieving four goals.

Page 26: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

5

The first two goals are that the UC should have sufficient energy in case of future

accelerations, and should have sufficient space in case of future regenerative braking. These

two goals would be realized by the EMS. The next two goals are that the battery should

supply the steady state constant speed power, and that the UC should be utilized even during

low power demands to extend battery cycle life. These two goals are realized by the PMS.

Existing works do not ensure these goals are met.

The reasons why these four goals are necessary are described in Chapter 2 Literature Review

and Chapter 4, which explains the EMS and PMS design. Ultimately, the goal of the HESS is

to extend battery cycle life.

1.6 Scope

First, a literature review is performed to examine the current state of the art. Second, the new

HESS management strategy for EVs is developed. In order to demonstrate the new HESS

management strategy, simulations and experiments are carried out. To run the simulations,

models have to be built. So the third step consists of examining and selecting existing models

for use in the simulations. Modifications are carried out where necessary to make it suitable

for this work.

The fourth step consists of running the simulations. One simulation each compares the EMS

and PMS to existing works and shows why the proposed strategy is better. Another

simulation demonstrates running the proposed HESS management strategy over drive cycles

and that the four HESS goals are met. The last simulation compares the battery cycle life

degradation of the battery/UC HESS system to a battery-only system. The purpose of this last

simulation is to show that the ultimate goal of the HESS is achieved, which is to extend

battery cycle life.

Page 27: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

6

The fifth step consists of running reduced-scale experiments. One experiment verifies the

algorithm is able to work as intended physically. Another experiment compares the battery

cycle life of actual batteries for the proposed battery/UC HESS to a battery-only system.

As it is difficult to compare the reduced-scale experiments with the full-scale EV simulations

directly, the last step involves creating a reduced-scale simulation. This allows direct

comparison of the reduced-scale experiment with the reduced-scale simulation to determine if

they perform similarly.

1.7 Structure of Thesis

The structure of this thesis is as follows.

Chapter 2 comprises a review of existing works. Firstly, an overall view of HESS is

presented, covering various other applications besides EVs. Subsequently, the scope is

narrowed to EVs, where the hardware is discussed, specifically, various battery/UC HESS

topologies. Afterwards, software is discussed, specifically, various battery/UC HESS EMS

and PMS. Drawbacks in existing works are also highlighted.

Chapter 3 discusses the models used for simulating the proposed strategy in Matlab. This

includes the dynamics of the car, as well as models of the electrical components used. The

battery cycle life model is also illustrated. The models are mostly from existing works.

Chapter 4 describes the EMS and PMS in detail with full mathematical formulation. First, the

proposed EMS is described. The method of calculating the speed-dependent target UC energy

band is explained in detail. It is based on averaged worst case (most) energy recovered during

braking, worst case energy requirement during acceleration, as well as with real-life drive

cycles. This type of justification is not seen in existing works. Since it already considers the

worst case scenarios, the proposed EMS does not need future knowledge of the route or drive

profile.

Page 28: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

7

In the chapter, the PMS is also explained in detail. The PMS ensures the EMS works

correctly by controlling the power flow between the battery and UC to regulate the UC

energy level, and also enforces speed-dependent battery charge/discharge limits, with priority

placed in the latter. A speed-dependent battery limit is not seen in existing works.

Chapter 5 comprises the simulation results. Two main simulations were performed in Matlab.

A specific design case study was considered in the simulations, where the models were based

on equipping a mid-sized EV sedan with an HESS. This EV sedan is based on a Nissan Leaf.

A mid-sized EV sedan was chosen for this work as it is designed for regular everyday use by

the masses (as opposed to high performance racing) and will therefore have a bigger

audience.

The first two simulations compare the EMS and PMS with existing works, showing that they

cannot achieve the goals simultaneously. A third simulation shows that the proposed strategy

works over a drive cycle and can achieve the EMS and PMS goals.

A fourth simulation on battery cycle life was performed to observe the fall in battery capacity

due to cycling the proposed battery/UC HESS. This was subsequently compared to a battery-

only system.

Chapter 6 explains the experiment setup and the scaling. The experiments are reduced-scale

bench setups and only considers the electrical components, specifically, the battery, UC and a

custom-built DC/DC converter. This setup is connected directly to a programmable load to

represent the motor loading.

Two main experiments were performed. The first experiment was to verify the algorithm

works as designed on a physical setup. The second experiment compares the battery cycle life

of the battery/UC HESS to a battery-only system.

Page 29: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

8

In Chapter 7, a reduced-scale simulation was created and compared to the experiment. They

are compared in terms of energy consumption and how they behave time-wise.

Lastly, Chapter 8 contains the conclusion and future works.

1.8 Contributions

The main contribution of this work to the state of the art is a new management strategy for a

battery/UC HESS.

As mentioned earlier, the EMS has two goals for the UC – ensuring sufficient space in the

UC for capturing energy during future regenerative braking, and ensuring the UC has

sufficient remaining energy for future accelerations. The first contribution in this work is a

new method involving a speed-dependent target band to ensure the two goals are met, where

multiple factors are considered, such as worst case (most) energy recovered during braking,

worst case energy requirement during acceleration, and real-life drive cycles. In existing

works, no such justification and rigorous calculations have been performed. Most are based

empirically or on experience, resulting in situations where the two goals may not be met.

Also, the EMS design and procedures are explained in detail, providing a framework for

designing such a system.

As mentioned earlier, the PMS has two goals – to ensure the EMS is followed, and that

battery limits are not exceeded. However, this is not the contribution as existing PMS works

already perform this. The PMS contribution lies in the speed-dependent battery limit, which

is not seen in existing literature. It is a simple but useful method to ensure better utilisation of

the UC, and this is compared with other rule-based deterministic battery limits to show the

benefits of the proposed algorithm.

Page 30: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

9

1.9 Publications

An early version of the proposed EMS was simulated and presented in a conference [14]. The

full proposed HESS management strategy (EMS and PMS) including both simulation and

experimental results, but excluding cycle life results, have been submitted to a journal [15]

and is under review at the time of submission of this thesis.

Page 31: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

10

2 LITERATURE REVIEW

In this chapter, existing works in HESS development are highlighted. First, a broad overview

of HESS is provided. Second, common topologies on battery/UC HESS (hardware) specific

to EVs are introduced. Third, existing EMS and PMS works (software) for EVs from current

researchers are discussed. Subsequently, some specific cases are considered. Lastly, the

limitations of existing works are examined, leading to the motivation of the work proposed in

this thesis.

2.1 Overview

Although this work focuses on battery/UC HESS for EVs, there have been other research

applications of a battery/UC HESS. One common application is for a power grid with

renewable energy sources [16] [17] [18]. For example, wind energy is an intermittent energy

source, with varying wind speeds encountered throughout the day. Therefore, the UC in the

HESS handles transients from the energy generated from wind turbines, while the batteries

handle the main load. There has also been research for battery/UC HESS in other types of

transportation, for example, ships [19] [20] [21] or tramways [22] [23].

Specific to the EV case, the battery/UC HESS is sometimes supplemented by other energy

generating devices, such as fuel cells [22] [23] [24] [25] [26] or solar cells [27]. The HESS

algorithm attempts to control the power and energy flow in all the components.

HESS can also be found in traditional internal combustion engine (ICE) vehicles with a non-

electric drive train. In these cases, ICEs are integrated with other energy storage devices, such

as batteries for hybrid vehicles, or a flywheel energy storage for Formula One racing [28]

[29] [30]. A hydraulic axial piston unit has also been paired with a diesel engine in one case

[31], known as a hybrid hydraulic vehicle.

Page 32: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

11

2.2 Hardware: HESS Topologies

Next, some of the common HESS topologies encountered in the literature specific to

battery/UC HESS EVs [32] [33] are discussed.

Figure 2-1 Passive configuration [32].

Figure 2-1 shows a passive connection, where the battery and UC are connected directly in

parallel, without any DC/DC converters. The voltages of the battery and UC must always be

equal, and the UC essentially acts as a low pass filter. Advantages for this method are its

simplicity and avoidance of expensive DC/DC converters. The disadvantage is that the UC

cannot vary its voltage sufficiently to deliver its stored energy. This is because the battery and

UC voltages have been clamped together, despite having vastly different charge/discharge

voltage profiles.

The Li-ion battery voltage profile is relatively flat throughout its operating region (see Figure

2-2), while the UC voltage profile varies proportionally to the amount of charge remaining

(Quc = C Vuc).

Subsequent topologies are active configurations, involving at least one DC/DC converter,

allowing controllability in power flow between the battery and UC. In simple terms, a

DC/DC converter can be thought of as a transformer for DC circuits (in these cases, an

multiple-tap transformer controlled by a microprocessor).

Page 33: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

12

Figure 2-2 Nissan Leaf Li-ion battery pack discharge curve [34].

Figure 2-3 is the most common single DC/DC converter configuration used in the literature,

with the battery directly connected to the DC link, and a DC/DC converter between the

battery and UC. This allows the UC to vary its voltage sufficiently within its operating range.

A drawback of this method is difficulty in balancing the battery cells. Since the battery is

connected to the high voltage dc-link, the balancing of the cells will have to be done at high

voltages. Another drawback is that the battery is directly exposed to the current and power

fluctuations from the inverter if the DC/DC converter is not controlled well.

Figure 2-3 Partially-decoupled configuration [32].

Swapping the battery and UC positions produces Figure 2-4. The benefit in this configuration

is that the battery can be connected at a lower voltage, allowing easier cell balancing. In

addition, the UC is directly exposed to the current fluctuation instead of the battery, which is

Page 34: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

13

desired. However, there is a big variation in the DC link voltage, so the inverter must be able

to accept such large variations with reasonable efficiency levels.

Figure 2-4 Partially-decoupled configuration 2 [32].

Figure 2-5 shows a cascaded configuration with two DC/DC converters. Here, the battery and

UC are both fully decoupled from the DC link. In this situation, the battery can be balanced at

a lower voltage, and the UC can have large variations in voltage without affecting the DC

link. Therefore, the design and specifications of the battery and UC can have more flexibility.

The drawback is an additional DC/DC converter, which is expensive, and requires another

layer of control. There will also be further power losses due to the additional DC/DC

converter. The battery and UC positions in Figure 2-5 can be swapped to create another

configuration.

Figure 2-5 Fully-decoupled cascaded configuration [32].

The most popular twin DC/DC converter configuration in the literature is shown in Figure

2-6. The advantages and disadvantages are similar to the cascaded configuration in Figure

Page 35: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

14

2-5, but the topology in Figure 2-6 has an additional advantage – independent control of the

battery and UC, providing more flexibility and more accurate control, as compared to the

cascaded configuration.

Figure 2-6 Fully-decoupled multiple converter configuration [32].

In recent times, there has been research on integrating the various components (instead of

using discrete components) to get a higher efficiency. In [35], the battery, inverter, and

DC/DC converter are merged. While [36] has a topology which bypasses the DC/DC

converter when regenerating energy to the UC.

2.3 Software: HESS Management Strategies

In this section, existing EMS in the literature are discussed, followed by existing PMS.

2.3.1 Energy Management Strategies

In general, EMS can be classified into three strategies, where the target UC energy is – a

constant level, a constant band, or a variable level.

2.3.1.1 Constant Target UC Energy Level

Existing EMS are generally quite rudimentary, such as maintaining a constant energy level in

the UC, regardless of loading [22] [23] [25] [26] [37] [38] [39] [40] [41] [42] [43] [44]. After

handling high power peaks, the UC energy level is restored to the pre-determined constant

target. The pre-determined target level varies between different works, for example, a UC

Page 36: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

15

state of charge (SOC) of 75% in Torreglosa, et al.’s work [22] or 87.5% in Avelino, et al.’s

work [37] or 25% in Yu, et al.’s work [42].

Usually, a medium SOC value is chosen empirically for two goals – such that the UC can

capture energy if regenerative braking is encountered in future, as well as supply energy if

acceleration is required in future. However, in the simulations section, this medium SOC is

shown to be unable to meet the two goals for certain situations, unless their UC is sized twice

as large as that in this proposed work.

2.3.1.2 Constant Target UC Energy Band

Another work, Cao & Emadi propose a target energy band [32], which is known as the

thermostat on-off strategy, where the UC is kept within a fixed band of 25.5V to 28.5V

(approx. 78% to 88% SOC). When the UC energy reaches the bottom of the band, it is

charged until it reaches the top of the band. Again, it is unable to meet the two goals in

certain situations, unless their UC is sized larger.

2.3.1.3 Variable Target UC Energy Level

Other works propose a variable target UC energy level [45] [46] [47] [48] [49] dependent on

speed. In general, at higher vehicle speeds, capturing energy recovered from regenerative

braking is more probable in the future. Therefore, more room in the UC is required, and the

target UC energy level is set lower. In contrast, at lower vehicle speeds, supplying energy for

acceleration is more probable. Therefore, the target UC energy level is set higher. The

relationship between target UC energy level and speed is usually set empirically or by

intuition, such as the target UC energy level being inversely related to speed due to the

kinetic energy equation E=0.5mv2 [46] [47] [48]. However, not all kinetic energy can be

recovered in braking. Also, during acceleration, more than just kinetic energy is needed to

overcome friction losses.

Page 37: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

16

Carter, et al. [48] use an empirical formula,

𝑉𝑡𝑎𝑟 = 𝑉𝑢𝑐,𝑚𝑎𝑥√1 − 0.01875 𝑣𝑚𝑎𝑥 (2-1)

where Vtar is the target UC voltage level, Vuc,max is the maximum UC voltage level and vmax is

the maximum speed of the car, which is 40mph. When the car is stopped, Vtar = Vuc,max, while

if the car is at maximum speed, then Vtar = 0.5Vuc,max. The 0.5 value was selected to prevent

the UC voltage from falling too low, leading to large DC/DC converter inefficiencies.

A more advanced variable target UC energy level strategy is seen in the works of Choi, et al.

[45], which estimates the energy recovered from future regenerative braking, ensuring the UC

always has sufficient space to store that energy. However, in their algorithm, all energy is

anticipated to be charged to the UC – none to the battery – increasing UC size. As mentioned

earlier, UCs of today are relatively expensive, so it is desirable to minimize UC size.

Moreover, they do not consider the worst case scenarios.

2.3.2 Power Management Strategies

PMS can be divided into rule-based and optimization-based strategies [11] [33] [50] [51] as

shown in Figure 2-7. These strategies apply not only to EVs with an HESS, but also to hybrid

vehicles.

Figure 2-7 Power management strategies [11].

Page 38: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

17

2.3.2.1 Rule-based

Rule-based algorithms are effective for real-time implementation in a vehicle as they are

simple and do not require much online computation power. The rules can be determined

based on heuristics, intuition, human expertise or mathematical models and simulations [33].

Usually the driving profile is not known in advance. The output of the rule-based approach

determines the mode of operation for the battery and UC. Rule-based approaches can be

deterministic or fuzzy.

In deterministic methods, the rules are usually derived from the present power demanded, the

velocity of the vehicle and the SOC of the battery and UC. For instance, a rule may be – if

driving power demanded is more than maximum battery power threshold, and the SOC of UC

is high, use the UC to assist the battery. They are generally implemented via look-up tables.

Works include [25] [32] [36] [47]. This technique is also implemented in many commercial

hybrid cars of today [50]. Some incorporate this idea into a model-predictive control based

approach, for example [22] [23] [26] [41] [44], while others use a low-pass/high-pass power

split approach, such as [35] [37] [38] [40].

Present works have a fixed maximum battery power threshold. For example, Cao & Emadi

[32] use a fixed battery limit of 12kW. If the power required exceeds 12kW, the battery

supplies 12kW while the UC supplies the remainder. The simulations in Chapter 5 later show

that this does not have good utilisation of the UC.

In fuzzy-based techniques, two or more operating modes are used to control the power

management between the battery and UC in a fuzzy logic manner. Therefore, the transition

between one mode and another does not happen at a specific moment of time, but in a

continuous time manner. It can be considered an extension of deterministic rule-based

methods. The advantages of fuzzy logic are that they are robust and tolerant to imprecise

measurements. They can also be easily tuned. Works include [23] [24] [39].

Page 39: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

18

The problem with rule-based strategies is that the drivetrain may not be optimized for

efficiency, as there may be substantial power loss in the power electronic converters.

2.3.2.2 Optimization-based

On the other hand, optimization-based algorithms can account for power losses.

Optimization-based algorithms can be divided into real-time and global optimization. Here,

well-known optimization techniques are used to optimize the system. Common objectives are

to minimize losses in the system, as well as minimize battery power variation and magnitude.

Real-time optimization is based on the system variables at that current point of time, which is

an instantaneous cost function. This does not require knowledge of the future driving profile.

Works include [43] [45]. However, a large amount of online computation is required, so there

has not been any experimental work directly implementing this strategy.

On the other hand, global optimization requires knowledge of the future drive profile in order

to find the global optimum. So this technique can only be used if the vehicle is driven the

same route repeatedly with similar driving patterns, such as garbage collection vehicles or

public buses. Otherwise, it can only be used as a basis for designing rules for online

implementation, or for comparison with other strategies [51], such as in [52]. Some works

implement this strategy as a neural network, where the results of an offline global

optimization are used to train an online network [46] [53].

2.4 Case Studies

Next, some interesting examples are examined in detail.

2.4.1 Dixon, Ortuzar & Moreno

The grandfathers of the battery/UC HESS are probably Dixon & Ortuzar [47]. Although they

were not the first to suggest the battery/UC HESS idea, they were one of the first to

Page 40: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

19

successfully implement it on a full-size Chevrolet LUV truck in 2002. Their EMS is

dependent on speed – the higher the speed of the vehicle and the battery SOC, the lower the

target voltage level of the UC. It ranges from SOC 100% to 15% (300V to 45V) as shown in

Figure 2-8. However, no detail was given in their work on how the curves were derived.

Figure 2-8 Target UC SOC values [47].

Their PMS is a simple rule-based approach. When the vehicle accelerates, the battery delivers

the required amount of current. If this current exceeds a threshold, the UC provides the

difference. Regenerative braking is similar, where the current is delivered into the battery. If

that current exceeds a threshold, then the UC captures the rest of the current. They used a

partially-decoupled configuration (Figure 2-3).

The strength of their technique is that it is simple to implement. The weakness is that the UC

is underutilized. The UC is only used when the battery exceeds the thresholds, leading to the

UC being idle during low power phases. In this work’s proposed algorithm, the UC is utilised

even during lower power phases to relieve the battery stress as much as possible.

Subsequently in 2006 and 2007, they published further works with an additional author,

Moreno, where they implemented global optimization using neural networks [46] [54]. The

objectives were to reduce battery current, reduce losses, and ensure the UC SOC at the end of

Page 41: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

20

the drive cycle is equal to the SOC at the start (i.e. unchanged). The last goal is related to

EMS, while the first two goals are related to PMS.

2.4.2 Avelino, Garcia, Ferreira & Pomilio

Avelino, et al. (2013) have built a working go-kart using a battery/UC HESS [37] with a

fully-decoupled configuration (Figure 2-6). Their EMS is a fixed target UC voltage level at

42V (SOC 87.5%). Their PMS is a rule-based strategy, where the UC takes the fast transients

as long as the UC SOC is within a target band, and the battery takes the rest. This was

implemented via two controllers with differing bandwidths. The UC controller reacts faster to

the desired power changes. The battery controller reacts slower, reducing the battery current

fluctuation, and slowly restoring the UC to the target voltage level. The advantage is simple

real-time implementation. However, they used a fixed UC target energy level. This will be

shown to be insufficient in subsequent simulations in this thesis.

A similar work has been done independently by Hredzak, et al (2014) [40], where a low-pass

filter (rule-based) was used to divide the power flow between the battery and UC in a

reduced-scale experiment. The EMS again consists of a fixed target UC voltage level at 5V

(SOC 41.6%). They used a fully-decoupled configuration (Figure 2-6).

2.4.3 Choi, Lee & Seo

In the work of Choi, et al (2014) [45], a real-time optimization strategy was simulated using a

partially-decoupled configuration (Figure 2-3). Their EMS is not a fixed UC level, but

variable, dependent on the speed of the vehicle. Their algorithm estimated the amount of

regenerative braking energy that would be generated if the vehicle braked at the maximum

deceleration at the next moment. Then it ensures sufficient space in the UC such that the UC

is able to fully capture this energy.

Page 42: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

21

Their PMS involved a real-time optimization strategy, where the objectives were to minimize

battery power magnitude and variation, reduce power losses, and to ensure the UC follows

the target energy level. Their work is one of the latest and most comprehensive strategies in

the literature today.

However, there are still some problems. In the EMS for example, the vehicle is always

assumed to brake with maximum deceleration, which is not the worst case scenario for

energy regeneration as energy is lost to friction brakes. Since real-time optimization was

used, it is expected to be computationally expensive. Only simulations were performed

without experiments. There will be more discussion on the shortcomings of Choi’s strategy in

later sections of this thesis, where it will be compared with this work’s proposed strategy.

2.5 Limitations of Existing Works and Proposed Approach

As mentioned earlier, most of the existing EMS are quite rudimentary, such as attempting to

maintain a fixed target voltage level in the UC. This target SOC was usually determined by

intuition, with no explanation on why that specific SOC was selected. These strategies will be

shown to be insufficient in simulations later.

In this work, a more comprehensive and rigorous method of setting the target UC energy

level is used. A new variable target UC energy band is proposed which varies with the speed

of the car. In the simulations later, when using the proposed strategy, the UC size can be

halved in terms of energy stored as compared to the fixed UC energy level strategies, saving

costs, yet still achieving the two goals. The proposed EMS can be considered an extension

and improvement of Choi’s EMS [45].

In addition, a new PMS with a speed-dependent battery limit is proposed, which is not seen in

existing works. It is a simple but useful method to ensure better utilisation of the UC. A rule-

based deterministic PMS is developed as it allows simple real-time implementation for a

Page 43: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

22

physical setup. In the simulations later, the proposed PMS is compared with other rule-based

deterministic PMS to show the benefits of the proposed work.

Page 44: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

23

3 VEHICLE & HESS MODELLING

In order to demonstrate the proposed novel HESS management strategy, foremost, a model of

the car and the powertrain components needs to be created. In this chapter, the modelling,

which was implemented in Matlab, is discussed.

First, the selected hardware topology is explained, then the concept of drive cycles is

introduced. Subsequently, the modelling of each component is explained, such as the vehicle

model, the battery model, etc. In addition, simulation demonstrations of the more complicated

models are provided to facilitate easier understanding. Lastly, the battery cycle life model is

discussed, which is used for characterizing the battery capacity loss over time.

These models are derived or modified from existing works in the literature.

3.1 Selected HESS Topology

This work proposes adding the UC to a mid-sized sedan EV as a specific design case study,

using the partially-decoupled single DC/DC converter topology. The mid-sized sedan EV is

modelled on a 2013 Nissan Leaf. To avoid major changes to the powertrain design from

conventional EV norms, the original battery is left connected to the DC link. Then a DC/DC

converter is connected to the DC link, and the UC is connected to the DC/DC converter as

shown in Figure 3-1.

Figure 3-1 Single DC/DC converter between battery and UC.

Bidirectional

DC/DC

Converter

Bidirectional

DC/AC

Converter

M

Battery UC

Braking

Chopper

DC Link

New Parts Existing Parts

Page 45: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

24

Using a single DC/DC converter saves costs, and allows the conventional existing EV

hardware to be relatively intact (as opposed to using the twin converter topologies). In

addition, as the Li-ion battery voltage is relatively constant throughout most of the operating

range, connecting it directly to the DC link is suitable. The UC voltage will fluctuate

significantly during usage, so it is better to place it after the DC/DC converter to reduce

voltage fluctuations in the DC link.

3.2 Drive Cycle

Drive cycles are created by various countries and organizations to test the performance of

vehicles, for example, fuel consumption or emissions. A drive cycle is defined by a series of

data points which represent the velocity of the vehicle over a period of time. There are two

types of drive cycles – transient and modal. Transient cycles involve much fluctuation in

speed and represent real-world driving scenarios. Examples include the American Federal

Test Procedures FTP-75 city, FTP-75 HWFET highway and LA92 drive cycles [55].

Figure 3-2 FTP-75 city drive cycle (Transient) [55].

Page 46: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

25

Figure 3-3 FTP-75 HWFET drive cycle (Transient) [55].

Figure 3-4 LA92 drive cycle (Transient) [55].

On the other hand, modal drive cycles are highly stylized drive cycles, involving constant

accelerations and constant speeds, which are hard to achieve in the real world. Examples

include the Economic Commission for Europe ECE-15 and Extra-Urban Driving Cycle

(EUDC) [55]. The New European Driving Cycle (NEDC) consists of four repeated ECE-15

followed by one EUDC.

Page 47: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

26

Figure 3-5 ECE-15 drive cycle (Modal) [55].

Figure 3-6 EUDC (Modal) [55].

3.3 Simulation Approach

There are two types of simulation approaches – backward approach and forward approach.

In the backward facing approach, the performance of a vehicle is evaluated over a drive

cycle, assuming the vehicle follows the drive cycle perfectly [56]. There is no driver model.

Instead, the force required to accelerate the vehicle based on the drive cycle is calculated

first. The calculations move backwards through the drivetrain, computing the torques at each

upstream component such as at the motor shaft. Then finally, the total motor power, the total

battery and UC power required to follow the drive cycle is determined.

Page 48: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

27

This is a convenient, simple and fast way to calculate power required. Most, if not all,

literature on battery/UC HESS use this approach. However, the drawback of this approach is

that the drive cycle is assumed to be followed perfectly and there are no transients in the

system. If the car is underpowered such that it cannot follow the drive cycle, then the

simulation becomes invalid.

In contrast, a forward-facing approach includes a driver model [56]. The driver gives

appropriate throttle and brake commands to follow the drive cycle (usually implemented with

a Proportional-Integral (PI) controller). The calculations go forward through the drivetrain,

computing the resulting acceleration and velocity due to the driver’s input, as well as the

power consumption. This is a more realistic scenario as transients are included, but is

extremely time-consuming in simulations. Also, the drive cycle may not be perfectly

followed due to the transients, making it difficult for comparison across different works.

Therefore, a backward approach is implemented in this work. Later in the EMS description in

section 4.2, there will be some elements of the forward approach for offline calculations.

3.4 Motor & Inverter Model

The motor used on a Nissan Leaf is an AC Permanent Magnet Synchronous Motor (PMSM).

In a backward approach, the dynamics of the motor and inverter and how they operate is not

crucial. Instead, these two units are viewed together as a whole. Only the physical limits of

the motor and the combined efficiencies of the motor and inverter are required. Specifically,

the maximum torque τm,max or minimum torque τm,min of the motor is computed from the

angular motor velocity ωm. In addition, the motor efficiency ηm is found from ωm and the

motor torque τm.

Page 49: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

28

From official Nissan data [57], the 2013 Nissan Leaf motor is rated at 80kW, with 254Nm

maximum torque and a base speed of 3008rpm. Thus, the maximum motor torque/speed

curve under variable speed operation for first quadrant motoring can be derived easily.

However, the speed at which the motor transitions from constant power to constant torque

during braking in the second quadrant is not known. This speed depends on the speed

regulation of the motor. The better the speed regulation, the closer it is to base speed. From

data in a technical book [58], this speed is estimated to be 10rpm higher than the base speed

of 3008rpm.

A technical magazine has the combined motor/inverter efficiency of an older 2011 Nissan

Leaf motor [59] as shown in Figure 3-7. However, there is no curve-fitted data, so the

efficiency map is recreated with some approximations.

Figure 3-7 2011 Nissan Leaf combined motor/inverter efficiency [59].

From Figure 3-7, the efficiency peaks near the centre of the figure, so it is approximated with

a hump equation given by (3-1) to (3-3). The efficiency for both the forward motoring

(upper) and forward braking (lower) quadrants are assumed symmetrical about the speed

axis.

Page 50: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

29

𝜂 = (𝜂𝑚𝑎𝑥)(ℎ) [𝑟

2 − (𝜔 − 𝜔𝑐𝑒𝑛)2 + (

𝜔𝑚𝑎𝑥𝜏𝑚𝑎𝑥

𝜏𝑠𝑐𝑎𝑙𝑒)2

(𝜏 − 𝜏𝑐𝑒𝑛)2]

(3-1)

𝜏𝑐𝑒𝑛 = 𝜏𝑚𝑎𝑥2

+ 𝜏𝑎𝑑𝑗 𝑖𝑓 𝜏 ≥ 0

−𝜏𝑚𝑎𝑥2

− 𝜏𝑎𝑑𝑗 𝑖𝑓 𝜏 < 0

(3-2)

𝜔𝑐𝑒𝑛 = 𝜔𝑚𝑎𝑥2

+ 𝜔𝑎𝑑𝑗 (3-3)

Equation (3-4) defines the maximum torque (edge of curve). Below base speed, maximum

torque is constant, while above base speed, maximum torque decreases inversely with angular

velocity.

𝜏𝑚𝑎𝑥 =

𝑃

𝜔 𝑖𝑓 𝜔 ≥ 𝜔𝑏𝑎𝑠𝑒

𝑃𝑚𝑎𝑥𝜔𝑏𝑎𝑠𝑒,𝑝𝑜𝑠

𝑖𝑓 𝜏 ≥ 0

𝑃𝑚𝑎𝑥𝜔𝑏𝑎𝑠𝑒,𝑛𝑒𝑔

𝑖𝑓 𝜏 < 0

𝑖𝑓 𝜔 < 𝜔𝑏𝑎𝑠𝑒

(3-4)

Figure 3-8 Combined motor/inverter efficiencies.

Page 51: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

30

The parameters of the hump equation have been tuned to match the 2011 Nissan Leaf

motor/inverter efficiency in Figure 3-7 as closely as possible. As the parameters are

decoupled, each parameter was tuned iteratively in a process similar to the bisection method.

The resultant parameters used are given in Table 3-2 in section 3.11, and the resultant

efficiency chart is shown in Figure 3-8, which gives a reasonable match. For example, the

maximum and minimum for both charts are at 95% and 85% (Note: 1rad/s = 9.55rpm ~

10rpm).

3.5 Auxiliary Loads

Auxiliary loads in a car include power steering, headlights, the radio, etc. It also includes the

heater and air-conditioning, which consumes the most power. Fleetcarma [60] performed

some experiments on a Nissan Leaf to evaluate the auxiliary power loads. The load varies

from 0.2kW to 5kW, depending on driver habits as well as the surrounding temperature.

However, for the drive cycle specifications, auxiliary loads are ignored. Therefore, in the

model, although there is a provision for the auxiliary load, it is set to zero.

3.6 Vehicle Model

In this section, the vehicle dynamic model is discussed, which is based on the works of [61].

As this work focuses mainly on the HESS and powertrain, the vehicle model is restricted to

one-dimensional movement – driving forwards or backwards. Other movement, such as the

lateral forces involved when cornering are not considered. This vehicle dynamic model uses

all the models discussed previously, from sections 3.2 to 3.5.

The drive cycles mentioned earlier contains the vehicle velocity v(j) (where j is the drive

cycle time in seconds) and is fed into this model at 1Hz [55]. v(j) is differentiated to get

vehicle acceleration a(j). In the drive cycles discussed earlier, the gradient is zero, so θ(j) = 0.

Page 52: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

31

Therefore, the inputs to the vehicle model are v(j) and θ(j). The output of the vehicle model is

the power required to follow the drive cycle Pdr+aux(j) at that instant. Figure 3-9 shows the

vehicle model block diagram. The numbers inside the blocks refer to the equation numbers,

which will be discussed subsequently.

Figure 3-9 Block diagram for vehicle model updating.

The aerodynamic drag Fw(j), rolling resistance Fr(j), and grading resistance Fg(j) can be

calculated from v(j), a(j) and θ(j), using the following formulas,

𝐹𝑤(𝑗) = 0.5 𝜌 𝐶𝑑 𝐴𝑓 (𝑣(𝑗))2 (3-5)

𝐹𝑟(𝑗) = 𝑚 𝑔 cos(𝜃(𝑗)) 𝐶𝑟(𝑗) 𝑜𝑛(𝑗) (3-6)

𝐹𝑔(𝑗) = 𝑚 𝑔 sin(𝜃(𝑗)) 𝑜𝑛(𝑗) (3-7)

Motor torque/speed,

efficiency

(Figure 3-8)

ηm(τm,ωm)

Compute max.

tractive effort

(3-14)

Compute

tractive effort

(3-5) to (3-13)

Ensure no

slipping

(3-15)

Compute motor

torque

(3-16), (3-17)

Clip braking

motor torque

(3-19)

Compute motor

ang. velocity

(3-20), (3-21)

Compute drive

cycle power

(3-22)

Compute total power

(3-23)

v(j)

θ(j) Ft(j) Ft(j)

τm(j)

ωm(j)

τm(j)

Pdr(j) Pdr+aux(j)

Ft,f,slip(j)

m, mk, ρ,

Cd, A, g

rwh, Gr

rwh, Gr, ηp

m, μ, Lb, h, rwh, g

Cr(j)

If regen. If acc.

Output

Ensure valid

motor torque

(3-18)

Page 53: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

32

where on(j) ensures the grading resistance and rolling resistance are involved only if the

vehicle is moving or starts to move (the brakes should hold the vehicle stationary on a

gradient instead of the motor),

𝑜𝑛(𝑗) =

0 𝑖𝑓 (𝑣(𝑗) = 0 𝑎𝑛𝑑 𝑎(𝑗) = 0)1 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

(3-8)

and rolling resistance coefficient Cr(j) is dependent on velocity,

𝐶𝑟(𝑗) = 𝐶𝑟,𝑎 (1 +

3.6

100 𝑣(𝑗))

(3-9)

Constant m is the gross vehicle weight, so this means the mid-sized EV is fully loaded in the

simulation. Other parameters such as the air density ρ, etc. can be found in Table 3-1 in

section 3.11.

The total resistance force Fv(j) is calculated by summing the three forces together,

𝐹𝑣(𝑗) = 𝐹𝑤(𝑗) + 𝐹𝑟(𝑗) + 𝐹𝑔(𝑗) (3-10)

Next the acceleration force Fa(j) is calculated by,

𝐹𝑎(𝑗) = (𝑚 +𝑚𝑟)(𝑎(𝑗)) (3-11)

where the effect of the rotating component inertias in the power train is translated to the car

body as mr. A rough approximation of mr is

𝑚𝑟 = 0.04 𝑚𝑘 (3-12)

where mk is the kerb weight (unloaded weight) of the car.

Total tractive effort Ft(j) is calculated by,

𝐹𝑡(𝑗) = 𝐹𝑎(𝑗) + 𝐹𝑣(𝑗) (3-13)

Page 54: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

33

Next, the car should not slip or skid. Otherwise, the backward approach would be inaccurate

as the drive cycle cannot be followed and the dynamic equations above would be invalid. The

maximum tractive effort can be derived from a free body diagram.

For a front wheel drive car, the maximum tractive effort is defined as

𝐹𝑡,𝑓,𝑠𝑙𝑖𝑝(𝑗) =

𝜇 𝑚 𝑔 cos 𝜃(𝑗) [𝐿𝑏 + 𝐶𝑟(𝑗) (ℎ − 𝑟𝑤ℎ)]

𝐿 + 𝜇 ℎ

(3-14)

The following equation (3-15) must hold true, otherwise slipping occurs and the simulation is

invalid.

|𝐹𝑡(𝑗)| ≤ 𝐹𝑡,𝑓,𝑠𝑙𝑖𝑝(𝑗) (3-15)

After accounting for the wheel radius rwh, the torque at the wheel is calculated by,

𝜏𝑤ℎ(𝑗) = 𝑟𝑤ℎ 𝐹𝑡(𝑗) (3-16)

After accounting for the gear ratio Gr and drivetrain mechanical losses ηp due to friction, etc.,

the torque which the motor must deliver is calculated by,

𝜏𝑚(𝑗) =

𝜏𝑤ℎ

(𝑗)

𝐺𝑟 𝜂𝑝 𝑖𝑓 𝜏𝑤ℎ(𝑗) ≥ 0

𝜏𝑤ℎ(𝑗) 𝜂𝑝

𝐺𝑟 𝑖𝑓 𝜏𝑤ℎ(𝑗) < 0

(3-17)

During acceleration, the torque required by the drive cycle should be less than the maximum

motor torque. The maximum torque/speed curve of the motor was shown in Figure 3-8.

Therefore, given a motor speed ωm(j), the maximum torque τm,max(ωm(j)) can be found. If

(3-18) is violated, then the car is underpowered and the backward approach simulation is

invalid.

𝜏𝑚(𝑗)+ ≤ 𝜏𝑚,𝑚𝑎𝑥(𝜔𝑚(𝑗)) (3-18)

Page 55: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

34

Deceleration is more complex. There are three different brake control strategies for

regenerative braking – series braking with optimal feel, series braking with optimal energy

recovery, and parallel braking [61].

The Nissan Leaf has an Electric Driven Intelligent Brake (EDIB) system [62], which is a

brake-by-wire system. According to Nissan’s description, it uses a series braking strategy. A

computer splits the braking torque between the motor and mechanical brakes electrically.

This is more complicated than parallel braking, where a physical link exists between the

brake pedal and the mechanical brakes.

Series braking with optimal energy recovery is the best case scenario for energy recovery

among the three brake control strategies, so this strategy was selected. The motor provides as

much braking torque as possible. If the braking torque exceeds the maximum torque that the

motor can handle, then the motor produces the maximum braking torque, and the remaining

braking torque is met by the mechanical braking system.

Again, the maximum braking torque (or alternatively, minimum motor torque) τm,min(ωm(j))

(which is a negative number) can be found from Figure 3-8. Therefore, the motor provides

the following braking torque,

𝜏𝑚(𝑗)− = max (𝜏𝑚(𝑗)−, 𝜏𝑚,𝑚𝑖𝑛(𝜔𝑚(𝑗))) (3-19)

At this point, the torque experienced by the motor τm(j) has been computed. Next, the motor

speed ωm(j) is calculated. Subsequently, these two terms are multiplied to determine the

power required.

First, the angular velocity of the wheel is calculated,

𝜔𝑤ℎ(𝑗) =

𝑣(𝑗)

𝑟𝑤ℎ

(3-20)

Next, the angular velocity of the motor is calculated,

Page 56: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

35

𝜔𝑚(𝑗) = 𝜔𝑤ℎ(𝑗) 𝐺𝑟 (3-21)

Finally, by considering the motor/inverter efficiency ηm(j) at that instant, the power required

by the drive cycle Pdr(j) can be calculated. ηm(j) can be read from Figure 3-8.

𝑃𝑑𝑟(𝑗) =

𝜏𝑚(𝑗) 𝜔𝑚(𝑗)

𝜂𝑚(𝑗) 𝑓𝑜𝑟 𝜏𝑚(𝑗)+

𝜏𝑚(𝑗) 𝜔𝑚(𝑗) 𝜂𝑚(𝑗) 𝑓𝑜𝑟 𝜏𝑚(𝑗)−

(3-22)

The total power which the HESS must supply Pdr+aux(j) includes both the power required by

the drive cycle and the auxiliary load,

𝑃𝑑𝑟+𝑎𝑢𝑥(𝑗) = 𝑃𝑑𝑟(𝑗) + 𝑃𝑎𝑢𝑥 (3-23)

Note that a positive Pdr+aux(j) value means the HESS must supply power to the car, while a

negative Pdr+aux(j) value means the HESS captures power from the car during regenerative

braking.

The four following figures illustrate a working example of the vehicle model. Figure 3-10

shows the drive cycle inputs v(j) and a(j) for the EUDC. Figure 3-11 and Figure 3-12 show

the forces and torques computed by the model, while Figure 3-13 shows the model output,

Pdr+aux.

Figure 3-10 EUDC input.

Page 57: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

36

Figure 3-11 Forces over EUDC: intermediate output from model.

Figure 3-12 Torque over EUDC: intermediate output from model.

Figure 3-13 Power over EUDC: model output.

Page 58: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

37

3.7 Battery Model

The battery model consists of a battery with open circuit voltage (OCV) vbatt,ocv in series with

the internal resistance as shown in Figure 3-14.

Figure 3-14 Battery model.

The inputs to the battery model are battery power demanded Pbatt(j) and battery open circuit

voltage (OCV) vbatt,ocv(j) at that moment. It also requires the starting charge in the battery

Qbatt,start, as well as the maximum possible charge Qbatt,max.

The outputs are the terminal voltage vbatt(j) and current flow ibatt(j) of the battery at that

moment, as well as the remaining charge in the next instant Qbatt(j+1) after discharging the

present ibatt(j), and battery OCV in the next instant vbatt,ocv(j+1).

Figure 3-15 shows a block diagram of the battery model.

Figure 3-15 Block diagram for battery model updating.

vbatt,ocv

Rbatt

vbatt, Pbatt

ibatt

vbatt(j)

Qbatt(j+1) Compute

SOC at

(j+1)

(3-31) Pbatt(j)

vbatt,ocv(j)

Rbatt

ibatt(j)

vbatt,ocv(j+1) Qdis(j+1) SOCbatt(j+1)

Qdis(j) Qbatt,max,

Qbatt,start,

Vbatt,0, K,

A, B

Qbatt,max

Compute terminal

volt. & current at (j)

(3-26), (3-27)

Compute OCV &

charge at (j+1)

(3-28) to (3-30)

Output

Page 59: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

38

Now, the model is explained in detail. Based on Figure 3-14, the battery current demanded

can be calculated from the power demanded,

𝑖𝑏𝑎𝑡𝑡(𝑗) =

𝑃𝑏𝑎𝑡𝑡(𝑗)

𝑣𝑏𝑎𝑡𝑡(𝑗)=

𝑃𝑏𝑎𝑡𝑡(𝑗)

𝑣𝑏𝑎𝑡𝑡,𝑜𝑐𝑣(𝑗) − 𝑅𝑏𝑎𝑡𝑡 𝑖𝑏𝑎𝑡𝑡(𝑗)

(3-24)

Rearranging, a quadratic equation is obtained,

𝑅𝑏𝑎𝑡𝑡 𝑖𝑏𝑎𝑡𝑡(𝑗)2 − 𝑣𝑏𝑎𝑡𝑡,𝑜𝑐𝑣(𝑗) 𝑖𝑏𝑎𝑡𝑡(𝑗) + 𝑃𝑏𝑎𝑡𝑡(𝑗) = 0 (3-25)

𝑖𝑏𝑎𝑡𝑡(𝑗) =

𝑣𝑏𝑎𝑡𝑡,𝑜𝑐𝑣(𝑗) − √𝑣𝑏𝑎𝑡𝑡,𝑜𝑐𝑣(𝑗)2 − 4 𝑅𝑏𝑎𝑡𝑡 𝑃𝑏𝑎𝑡𝑡(𝑗)

2 𝑅𝑏𝑎𝑡𝑡

(3-26)

Note that the more negative root of the quadratic equation is always used. During charging,

Pbatt(j) is negative. Therefore, ibatt(j) consists of one negative root and one positive root. Since

the battery voltage vbatt(j) is always positive, the negative ibatt(j) root is always selected such

that Pbatt(j) is negative, based on (3-24).

On the other hand, during discharging, Pbatt(j) is positive. There will be two positive ibatt(j)

roots. Again, the more negative (smaller) root is used to ensure the power losses due to ibatt(j)2

Rbatt are minimized.

If imaginary roots are present in the solution, it means the required battery power cannot be

delivered. For example, the internal resistance Rbatt might be so large that all the power is lost

in the internal resistance.

Subsequently, the battery voltage at the present instant is calculated by,

𝑣𝑏𝑎𝑡𝑡(𝑗) = 𝑣𝑏𝑎𝑡𝑡,𝑜𝑐𝑣(𝑗) − 𝑅𝑏𝑎𝑡𝑡 𝑖𝑏𝑎𝑡𝑡(𝑗) (3-27)

Next, the OCV vbatt,ocv(j), which varies with charging and discharging, is calculated. This is

based on a model by Tremblay, et al. [63]. In their work, an easy-to-use battery model was

described. They also presented a method of extracting parameters from the battery

Page 60: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

39

manufacturer’s discharge curve, such that the discharge curve can be replicated in a

simulation.

The US Department of Energy, Vehicle Technologies Program has performed experiments

with a Nissan Leaf battery to obtain the battery discharge curve as well as internal resistance

(see earlier Figure 2-2) [34]. Therefore, Tremblay’s methods have been applied to extract the

battery parameter constants A, B, K, and Vbatt,0 from the discharge curve.

Now, the battery model updating based on Tremblay’s work is discussed. First, the total

battery discharge since the start of the simulation Qdis(j+1) in the next instant after

discharging the present ibatt(j) is calculated. Then, the remaining charge in the battery

Qbatt(j+1) is calculated,

𝑄𝑑𝑖𝑠(𝑗 + 1) = ∫ 𝑖𝑏𝑎𝑡𝑡(𝑗) 𝑑𝑗

𝑗

0

= 𝑄𝑑𝑖𝑠(𝑗) + 𝑖𝑏𝑎𝑡𝑡(𝑗) ∆𝑗 (3-28)

𝑄𝑏𝑎𝑡𝑡(𝑗 + 1) = 𝑄𝑏𝑎𝑡𝑡,𝑠𝑡𝑎𝑟𝑡 − 𝑄𝑑𝑖𝑠(𝑗 + 1) (3-29)

where Qdis (1) = 0. Next, the battery OCV is calculated by,

𝑣𝑏𝑎𝑡𝑡,𝑜𝑐𝑣(𝑗 + 1) = 𝑉𝑏𝑎𝑡𝑡,0 − 𝐾

𝑄𝑏𝑎𝑡𝑡,𝑚𝑎𝑥

𝑄𝑏𝑎𝑡𝑡(𝑗 + 1)+ 𝐴 𝑒−𝐵(𝑄𝑏𝑎𝑡𝑡,𝑚𝑎𝑥−𝑄𝑏𝑎𝑡𝑡(𝑗+1))

(3-30)

Now, the battery model has been updated over a time instant. vbatt(j), ibatt(j), Qbatt(j+1) and

vbatt,ocv(j+1) have been calculated. The SOC of the battery can also be easily calculated by,

𝑆𝑂𝐶𝑏𝑎𝑡𝑡(𝑗) =

𝑄𝑏𝑎𝑡𝑡(𝑗)

𝑄𝑏𝑎𝑡𝑡,𝑚𝑎𝑥

(3-31)

Page 61: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

40

3.8 Ultracapacitor Model

The UC model consists of a capacitor with OCV vuc,ocv in series with the equivalent series

resistance (internal resistance) [64], similar to the previous battery model.

Figure 3-16 UC model.

The inputs to the UC model are UC power demanded Puc,ocv(j) and UC OCV vuc,ocv(j) at that

instant. It also requires the starting UC energy Euc,start. The outputs are the UC current at that

moment iuc(j), as well as the remaining energy Euc(j+1) and UC OCV vuc,ocv(j+1) in the next

moment.

Figure 3-17 shows a block diagram of the UC model.

Figure 3-17 Block diagram for UC model updating.

From the power demanded, the required UC current can be calculated by,

𝑖𝑢𝑐(𝑗) =

𝑃𝑢𝑐,𝑜𝑐𝑣(𝑗)

𝑣𝑢𝑐,𝑜𝑐𝑣(𝑗)

(3-32)

Compute

SOC at

(j+1)

(3-36)

Compute OCV &

energy at (j+1)

(3-33) to (3-35) Puc,ocv(j)

vuc,ocv(j)

Compute current

at (j)

(3-32)

iuc(j) vuc,ocv(j+1) Euc(j+1) SOCuc(j+1)

Euc(j),

Euc,start

C, Ruc Vuc,ocv,max

Output

Ruc

vuc,ocv, Puc,ocv vuc, Puc

iuc

Page 62: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

41

In the earlier battery model, battery internal resistance was included. But in this model, the

UC internal resistance Ruc is ignored, as it will be included in the next section, 3.9 DC/DC

Converter Model.

Next, the OCV vuc,ocv(j+1), which varies with charging and discharging, is calculated. The

energy of the UC Euc(j) at the present instant is given by,

𝐸𝑢𝑐(𝑗) = 0.5 𝐶 𝑣𝑢𝑐,𝑜𝑐𝑣(𝑗)2 (3-33)

After that, the energy in the UC at the next instant is calculated,

𝐸𝑢𝑐(𝑗 + 1) = −∫ 𝑃𝑢𝑐,𝑜𝑐𝑣(𝑗) 𝑑𝑗

𝑗

0

= 𝐸𝑢𝑐(𝑗) − 𝑃𝑢𝑐,𝑜𝑐𝑣(𝑗) ∆𝑗 (3-34)

where Euc (1) = Euc,start. Finally, vuc,ocv(j+1) is calculated,

𝑣𝑢𝑐,𝑜𝑐𝑣(𝑗 + 1) = √2 𝐸𝑢𝑐(𝑗 + 1)

𝐶

(3-35)

Now, the UC model has been updated over a time instant. iuc(j), Euc(j+1), and vuc,ocv(j+1) has

been calculated. The SOC of the UC can also be easily calculated by,

𝑆𝑂𝐶𝑢𝑐(𝑗) =

𝑄𝑢𝑐(𝑗)

𝑄𝑢𝑐,𝑚𝑎𝑥=𝐶 𝑣𝑢𝑐,𝑜𝑐𝑣(𝑗)

𝐶 𝑉𝑢𝑐,𝑚𝑎𝑥=𝑣𝑢𝑐,𝑜𝑐𝑣(𝑗)

𝑉𝑢𝑐,𝑚𝑎𝑥

(3-36)

As the DC/DC converter becomes inefficient at low voltages (see next section), the minimum

voltage of the UC is limited to be approximately half that of the maximum, i.e.

𝑉𝑢𝑐,𝑚𝑖𝑛 ≈ 0.5 𝑉𝑢𝑐,𝑚𝑎𝑥 (3-37)

This allows 75% of the energy in the UC to be utilised.

Page 63: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

42

3.9 DC/DC Converter Model

The bidirectional two-quadrant buck-boost DC/DC converter is shown in Figure 3-18. The

converter operates in buck mode (step down) when transferring power from the DC link to

the UC, and in boost mode (step up) when transferring power from the UC to the DC link.

Figure 3-18 Electrical model of DC/DC converter.

The power input on the high-side (left side) of the DC/DC converter Puc,H is the difference

between the required power and battery power,

𝑃𝑢𝑐,𝐻(𝑗) = 𝑃𝑑𝑟+𝑎𝑢𝑥(𝑗) − 𝑃𝑏𝑎𝑡𝑡(𝑗) (3-38)

Just like the motor/inverter model previously, knowledge of the DC/DC converter’s internal

operation is not critical. Only the efficiency of the DC/DC converter is required, such that the

drivetrain power flow can be modelled accurately.

After accounting for DC/DC converter losses (ɳDC/DC,bo for boost or ɳDC/DC,bu for buck) , the

power that the UC encounters Puc,ocv(j) (right side of the DC/DC converter in Figure 3-18) is

given by,

Puc,ocv

Puc,H DC Link

Pdr+aux

Q1 D1

Q2 D2

Pbatt

Page 64: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

43

𝑃𝑢𝑐,𝑜𝑐𝑣(𝑗) =

𝑃𝑢𝑐,𝐻(𝑗)

𝜂𝐷𝐶/𝐷𝐶,𝑏𝑜(𝑗) 𝑖𝑓 𝑃𝑢𝑐,𝐻(𝑗) ≥ 0

𝜂𝐷𝐶/𝐷𝐶,𝑏𝑢(𝑗) 𝑃𝑢𝑐,𝐻(𝑗) 𝑖𝑓 𝑃𝑢𝑐,𝐻(𝑗) < 0

(3-39)

The models are based on the works of [65]. The required duty cycle of the converter at each

instant j needs to be computed first, before the efficiency can be determined.

3.9.1 Boost Mode Duty Cycle

First, boost mode is considered. In boost mode, there are two subintervals of operation –

when insulated-gate bipolar transistor (IGBT) switch Q2 is on and when it is off (Q1 operates

complementary to Q2).

Figure 3-19 Boost mode, Q2 on.

In the subinterval when Q2 is on, the equivalent circuit is seen in Figure 3-19. From

Kirchhoff’s Voltage Law (KVL), the following can be formed,

𝑣𝐿(𝑡) = 𝑉𝑢𝑐,𝑜𝑐𝑣 − 𝑉𝑖𝑔𝑏𝑡 − 𝐼𝐿(𝑅𝐿 + 𝑅𝑢𝑐 + 𝑅𝑖𝑔𝑏𝑡) (3-40)

Page 65: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

44

Figure 3-20 Boost mode, Q2 off.

In the subinterval when Q2 is off, the equivalent circuit is seen in Figure 3-20. From KVL,

the following can be formed,

𝑣𝐿(𝑡) = 𝑉𝑢𝑐,𝑜𝑐𝑣 − 𝑉𝑑 − 𝑉𝑏𝑎𝑡𝑡 − 𝐼𝐿(𝑅𝐿 + 𝑅𝑢𝑐 + 𝑅𝑑) (3-41)

Then the principle of inductor volt-second balance is invoked. By averaging the two

subintervals with the duty cycle D2, the average inductor current IL can be determined,

𝑉𝐿 =1

𝑇𝑠∫ 𝑣𝐿(𝑡) 𝑑𝑡

𝑇𝑠

0

= 0

(3-42)

𝐷2 [𝑉𝑢𝑐,𝑜𝑐𝑣 − 𝑉𝑖𝑔𝑏𝑡 − 𝐼𝐿(𝑅𝐿 + 𝑅𝑢𝑐 + 𝑅𝑖𝑔𝑏𝑡)]

+ 𝐷2′ [𝑉𝑢𝑐,𝑜𝑐𝑣 − 𝑉𝑑 − 𝑉𝑏𝑎𝑡𝑡 − 𝐼𝐿(𝑅𝐿 + 𝑅𝑢𝑐 + 𝑅𝑑)] = 0

(3-43)

𝐼𝐿 =

𝑉𝑢𝑐,𝑜𝑐𝑣 − 𝐷2𝑉𝑖𝑔𝑏𝑡 − 𝐷2′(𝑉𝑑 + 𝑉𝑏𝑎𝑡𝑡)

𝑅𝐿 + 𝑅𝑢𝑐 + 𝐷2𝑅𝑖𝑔𝑏𝑡 + 𝐷2′𝑅𝑑

(3-44)

Page 66: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

45

As IGBT switches Q1 and Q2 are complementary (i.e. Q1 is off when Q2 is on, and vice-

versa), the following equation can be attained, where D1 and D2 are the duty cycles of

switches Q1 and Q2 respectively.

𝐷2′ = 𝐷1

𝐷2 = 𝐷1′

(3-45)

Next, switching losses are considered. Again, two subintervals are present, when Q2 is on

and when it is off.

When Q2 is on, switch Q1 needs to block the following voltage Vblock,bo,1,

𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑜,1 = 𝑉𝑏𝑎𝑡𝑡 − 𝑉𝑖𝑔𝑏𝑡 − 𝑅𝑖𝑔𝑏𝑡𝐼𝐿 (3-46)

When Q2 is off, switch Q2 needs to block the following voltage Vblock,bo,2,

𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑜,2 = 𝑉𝑏𝑎𝑡𝑡 + 𝑉𝑑 + 𝑅𝑑𝐼𝐿 (3-47)

By averaging the blocking voltage with the duty cycle, the average blocked voltage by the

switches is Vblock,bo,

𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑜 = 𝐷2𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑜,1 + 𝐷2′𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑜,2

= 𝑉𝑏𝑎𝑡𝑡 − 𝐷1′𝑉𝑖𝑔𝑏𝑡 + 𝐷1𝑉𝑑 + 𝐼𝐿[𝐷1𝑅𝑑 − 𝐷1

′𝑅𝑖𝑔𝑏𝑡]

(3-48)

The total switching power loss is given by Psw,bo, which depends on the switching frequency

fs and IGBT turn on and off times,

𝑃𝑠𝑤,𝑏𝑜 = 𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑜𝛼𝐼𝐿 (3-49)

𝛼 = 𝑓𝑠(𝑡𝑠𝑤,𝑜𝑛 + 𝑡𝑠𝑤,𝑜𝑓𝑓) (3-50)

Equation (3-51) is achieved by including the switching loss into the original inductor current

IL in (3-44). Subsequently, the required duty cycle D1 can be found by (3-52) if the voltages

Page 67: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

46

at the two ends of the DC/DC converter (Vuc,ocv, Vbatt), as well as the current IL flowing

through it are known.

𝐼𝐿 =

𝑉𝑢𝑐,𝑜𝑐𝑣 − 𝐷1𝑉𝑏𝑎𝑡𝑡 − (𝐷1′𝑉𝑖𝑔𝑏𝑡 + 𝐷1𝑉𝑑) − 𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑜𝛼

(𝑅𝐿 + 𝑅𝑢𝑐) + 𝐷1𝑅𝑑 + 𝐷1′𝑅𝑖𝑔𝑏𝑡

(3-51)

𝐷1𝐼𝐿[𝑅𝑑(1 + 𝛼) − 𝑅𝑖𝑔𝑏𝑡(1 − 𝛼)] + 𝑉𝑏𝑎𝑡𝑡 + 𝑉𝑑(1 + 𝛼) − 𝑉𝑖𝑔𝑏𝑡(1 − 𝛼)

+ 𝐼𝐿[(𝑅𝐿 + 𝑅𝑢𝑐) + 𝑅𝑖𝑔𝑏𝑡(1 − 𝛼)] + 𝑉𝑖𝑔𝑏𝑡(1 − 𝛼) − 𝑉𝑢𝑐,𝑜𝑐𝑣

+ 𝑉𝑏𝑎𝑡𝑡𝛼 = 0

(3-52)

3.9.2 Boost Mode Efficiency

To obtain the efficiency, the inductor current term IL should be replaced, because this is not

known. As the power on the high-side of the DC/DC converter Puc,H is known, (3-53) is

substituted into the boost duty cycle equation (3-52), resulting in (3-54) and (3-55).

𝐼𝐿 =

𝑃𝑢𝑐,𝐻𝑉𝑢𝑐,𝑜𝑐𝑣 𝜂𝐷𝐶/𝐷𝐶,𝑏𝑜

=𝑃𝑢𝑐,𝐻𝑉𝑏𝑎𝑡𝑡𝐷1

(3-53)

𝐷1

=−

𝑃𝑢𝑐,𝐻𝑉𝑏𝑎𝑡𝑡𝐷1

[(𝑅𝐿 + 𝑅𝑢𝑐) + 𝑅𝑖𝑔𝑏𝑡(1 − 𝛼)] − 𝑉𝑖𝑔𝑏𝑡(1 − 𝛼) + 𝑉𝑢𝑐,𝑜𝑐𝑣 − 𝑉𝑏𝑎𝑡𝑡𝛼

𝑃𝑢𝑐,𝐻𝑉𝑏𝑎𝑡𝑡𝐷1

[𝑅𝑑(1 + 𝛼) − 𝑅𝑖𝑔𝑏𝑡(1 − 𝛼)] + 𝑉𝑏𝑎𝑡𝑡 + 𝑉𝑑(1 + 𝛼) − 𝑉𝑖𝑔𝑏𝑡(1 − 𝛼)

(3-54)

𝐷12𝑉𝑏𝑎𝑡𝑡 + 𝑉𝑑(1 + 𝛼) − 𝑉𝑖𝑔𝑏𝑡 (1 − 𝛼)

+ 𝐷1 𝑉𝑖𝑔𝑏𝑡 (1 − 𝛼) − 𝑉𝑢𝑐,𝑜𝑐𝑣 + 𝑉𝑏𝑎𝑡𝑡𝛼

+𝑃𝑢𝑐,𝐻𝑉𝑏𝑎𝑡𝑡

[𝑅𝑑(1 + 𝛼) − 𝑅𝑖𝑔𝑏𝑡(1 − 𝛼)]

+ 𝑃𝑢𝑐,𝐻𝑉𝑏𝑎𝑡𝑡

[(𝑅𝐿 + 𝑅𝑢𝑐) + 𝑅𝑖𝑔𝑏𝑡(1 − 𝛼)] = 0

(3-55)

Page 68: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

47

The boost efficiency can be found by taking the positive root of D1 in (3-55), and then

substituting D1 into (3-56),

𝜂𝐷𝐶/𝐷𝐶,𝑏𝑜 =

𝑃𝑜𝑢𝑡𝑃𝑖𝑛

=𝑃𝑢𝑐,𝐻𝑉𝑢𝑐,𝑜𝑐𝑣𝐼𝐿

=𝑉𝑜𝑢𝑡𝐼𝑜𝑢𝑡𝑉𝑢𝑐,𝑜𝑐𝑣𝐼𝐿

=𝑉𝑏𝑎𝑡𝑡𝐷1𝐼𝐿𝑉𝑢𝑐,𝑜𝑐𝑣𝐼𝐿

=𝑉𝑏𝑎𝑡𝑡𝐷1𝑉𝑢𝑐,𝑜𝑐𝑣

(3-56)

3.9.3 Buck Mode Duty Cycle

Next, buck mode is considered, which is analysed in a similar way. Again, there are two

subintervals of operation – when IGBT switch Q1 is on and when it is off.

Figure 3-21 Buck mode, Q1 on.

In the subinterval when Q1 is on, the equivalent circuit is seen in Figure 3-21. From KVL, the

following can be formed,

𝑣𝐿(𝑡) = 𝑉𝑢𝑐,𝑜𝑐𝑣 + 𝑉𝑖𝑔𝑏𝑡 − 𝑉𝑏𝑎𝑡𝑡 − 𝐼𝐿(𝑅𝐿 + 𝑅𝑢𝑐 + 𝑅𝑖𝑔𝑏𝑡) (3-57)

Page 69: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

48

Figure 3-22 Buck mode, Q1 off.

In the subinterval when Q1 is off, the equivalent circuit is seen in Figure 3-22. From KVL,

the following can be formed,

𝑣𝐿(𝑡) = 𝑉𝑢𝑐,𝑜𝑐𝑣 + 𝑉𝑑 − 𝐼𝐿(𝑅𝐿 + 𝑅𝑢𝑐 + 𝑅𝑑) (3-58)

Then the principle of inductor volt-second balance is invoked. By averaging the two

subintervals with the duty cycle, the average inductor current IL is determined,

𝑉𝐿 =1

𝑇𝑠∫ 𝑣𝐿(𝑡) 𝑑𝑡

𝑇𝑠

0

= 0

(3-59)

𝐷1 [𝑉𝑢𝑐,𝑜𝑐𝑣 + 𝑉𝑖𝑔𝑏𝑡 − 𝑉𝑏𝑎𝑡𝑡 − 𝐼𝐿(𝑅𝐿 + 𝑅𝑢𝑐 + 𝑅𝑖𝑔𝑏𝑡)]

+ 𝐷1′ [𝑉𝑢𝑐,𝑜𝑐𝑣 + 𝑉𝑑 − 𝐼𝐿(𝑅𝐿 + 𝑅𝑢𝑐 + 𝑅𝑑)] = 0

(3-60)

Page 70: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

49

𝐼𝐿 =

𝑉𝑢𝑐,𝑜𝑐𝑣 + 𝐷1(𝑉𝑖𝑔𝑏𝑡 − 𝑉𝑏𝑎𝑡𝑡) + 𝐷1′𝑉𝑑

𝑅𝐿 + 𝑅𝑢𝑐 + 𝐷1𝑅𝑖𝑔𝑏𝑡 + 𝐷1′𝑅𝑑

(3-61)

Next, switching losses are considered. Again, two subintervals are present, when Q1 is on

and when it is off.

When Q1 is on, switch Q2 needs to block the following voltage Vblock,bu,1,

𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑢,1 = 𝑉𝑏𝑎𝑡𝑡 − 𝑉𝑖𝑔𝑏𝑡 + 𝑅𝑖𝑔𝑏𝑡𝐼𝐿 (3-62)

When Q1 is off, switch Q1 needs to block the following voltage Vblock,bu,2,

𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑢,2 = 𝑉𝑏𝑎𝑡𝑡 + 𝑉𝑑 − 𝑅𝑑𝐼𝐿 (3-63)

By averaging the blocking voltage with the duty cycle, the average blocked voltage by the

switches is Vblock,bu,

𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑢 = 𝐷1𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑢,1 + 𝐷1′𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑢,2

= 𝑉𝑏𝑎𝑡𝑡 + 𝐷1′𝑉𝑑 − 𝐷1𝑉𝑖𝑔𝑏𝑡 + 𝐼𝐿[𝐷1𝑅𝑖𝑔𝑏𝑡 − 𝐷1

′𝑅𝑑]

(3-64)

The total switching power loss is given by Psw,bu, where α was calculated earlier in (3-50).

𝑃𝑠𝑤,𝑏𝑢 = 𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑢𝛼𝐼𝐿 (3-65)

Equation (3-66) is achieved by including the switching loss into the original inductor current

IL in (3-61). Subsequently, the required duty cycle D1 can be found by (3-67) if the voltages

at the two ends of the DC/DC converter (Vuc,ocv, Vbatt), as well as the current IL flowing

through it are known.

𝐼𝐿 =

𝑉𝑢𝑐,𝑜𝑐𝑣 + 𝐷1𝑉𝑖𝑔𝑏𝑡 + 𝐷1′𝑉𝑑 − 𝐷1𝑉𝑏𝑎𝑡𝑡 + 𝑉𝑏𝑙𝑜𝑐𝑘,𝑏𝑢𝛼

(𝑅𝐿 + 𝑅𝑢𝑐) + 𝐷1𝑅𝑖𝑔𝑏𝑡 + 𝐷1′𝑅𝑑

(3-66)

Page 71: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

50

𝐷1𝐼𝐿[𝑅𝑖𝑔𝑏𝑡(1 − 𝛼) − 𝑅𝑑(1 + 𝛼) − 𝑅𝑡ℎ𝛼] + 𝑉𝑏𝑎𝑡𝑡 − 𝑉𝑖𝑔𝑏𝑡(1 − 𝛼)

+ 𝑉𝑑(1 + 𝛼)

+ 𝐼𝐿[(𝑅𝐿 + 𝑅𝑢𝑐) + 𝑅𝑑(1 + 𝛼)] − 𝑉𝑑(1 + 𝛼) − 𝑉𝑢𝑐,𝑜𝑐𝑣

− 𝑉𝑏𝑎𝑡𝑡𝛼 = 0

(3-67)

3.9.4 Buck Mode Efficiency

To get the buck efficiency, similar to the earlier boost case, the inductor current IL should be

replaced. (3-68) is substituted into the buck duty cycle (3-67), resulting in (3-69) and (3-70).

𝐼𝐿 =

𝑃𝑢𝑐,𝐻 𝜂𝐷𝐶/𝐷𝐶,𝑏𝑢

𝑉𝑢𝑐,𝑜𝑐𝑣=

𝑃𝑢𝑐,𝐻𝑉𝑏𝑎𝑡𝑡𝐷1

(3-68)

𝐷1

=−

𝑃𝑢𝑐,𝐻𝑉𝑏𝑎𝑡𝑡𝐷1

[(𝑅𝐿 + 𝑅𝑢𝑐) + 𝑅𝑑(1 + 𝛼)] + 𝑉𝑑(1 + 𝛼) + 𝑉𝑢𝑐,𝑜𝑐𝑣 + 𝑉𝑏𝑎𝑡𝑡𝛼

𝑃𝑢𝑐,𝐻𝑉𝑏𝑎𝑡𝑡𝐷1

[𝑅𝑖𝑔𝑏𝑡(1 − 𝛼) − 𝑅𝑑(1 + 𝛼)] + 𝑉𝑏𝑎𝑡𝑡 − 𝑉𝑖𝑔𝑏𝑡(1 − 𝛼) + 𝑉𝑑(1 + 𝛼)

(3-69)

𝐷12𝑉𝑏𝑎𝑡𝑡 − 𝑉𝑖𝑔𝑏𝑡 (1 − 𝛼) + 𝑉𝑑(1 + 𝛼)

+ 𝐷1 −𝑉𝑑 (1 + 𝛼) − 𝑉𝑢𝑐,𝑜𝑐𝑣 − 𝑉𝑏𝑎𝑡𝑡𝛼

+𝑃𝑢𝑐,𝐻𝑉𝑏𝑎𝑡𝑡

[𝑅𝑖𝑔𝑏𝑡(1 − 𝛼) − 𝑅𝑑(1 + 𝛼)]

+ 𝑃𝑢𝑐,𝐻𝑉𝑏𝑎𝑡𝑡

[(𝑅𝐿 + 𝑅𝑢𝑐) + 𝑅𝑑(1 + 𝛼)] = 0

(3-70)

The buck efficiency can be found by making D1 the subject in (3-70), and then substituting

D1 into (3-71),

𝜂𝐷𝐶/𝐷𝐶,𝑏𝑢 =

𝑃𝑜𝑢𝑡𝑃𝑖𝑛

=𝑉𝑢𝑐,𝑜𝑐𝑣𝐼𝐿𝑉𝑏𝑎𝑡𝑡𝐷1𝐼𝐿

=𝑉𝑢𝑐,𝑜𝑐𝑣𝑉𝑏𝑎𝑡𝑡𝐷1

(3-71)

Page 72: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

51

3.9.5 Combined Duty Cycle and Efficiency

Figure 3-23 shows the duty cycle of Q1 for both buck and boost modes combined into a

single chart for the case of Vbatt = 370V. This figure was achieved by varying the high-side

UC power Puc,H and UC voltage Vuc,ocv and then computing the resulting efficiency for the

buck and boost case separately using the equations earlier. Then the two charts were

combined. Similarly, Figure 3-24 shows the combined buck-boost efficiency in a single

diagram.

Figure 3-23 Combined buck-boost duty cycle.

Figure 3-24 Combined buck-boost efficiency.

Page 73: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

52

A positive Puc,H value means the UC is being discharged (boost), while a negative value

means the UC is being charged (buck). In general, the DC/DC converter is more efficient

around the 0kW mark (low demand) and at high UC voltages.

The top left corners in Figure 3-23 and Figure 3-24 have missing data. In that region,

equation (3-55) has no real solution. This means that all power discharged by the UC is lost

in the DC/DC converter, therefore the UC is unable to supply power to the HESS.

As a side note, the battery voltage (i.e. high-side DC/DC converter voltage) has minimal

effect on the efficiency, e.g. a large change in battery voltage Vbatt leads only to a marginal

change in efficiency. For a simple demonstration, an ideal synchronous converter has the

following properties, Rd – Rigbt = 0 and Vd – Vigbt = 0. Substituting this to solve the efficiency

equations in (3-55), (3-56), (3-70) and (3-71) and ignoring switching losses (α = 0), it is

found that the Vbatt term disappears from the equations altogether, so the efficiency does not

depend on Vbatt. However, in real-life, Rd – Rigbt and Vd – Vigbt are not equal to zero, but is

very small. Therefore, Vbatt only has a small effect on DC/DC converter efficiency.

In summary, the various models of the car and the powertrain have been discussed. These are

building blocks for the proposed HESS management strategy.

3.10 Battery Cycle Life Model

Next the battery cycle life model is considered, which is used to simulate the battery capacity

loss. The battery cycle life model is a modified version of the works of Wang, et al (2014)

[7]. In their works, they have cycled commercially available 1.5Ah Sanyo UR18650W (2007

technology) batteries to monitor their capacity loss. From fitting the resultant data, they have

created a semi-empirical model. Although the term battery cycle life is used as a generic term

to mean the battery lifetime, there are actually two components in their battery capacity loss

model – calendar life and cycle life. These are the equations,

Page 74: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

53

𝑄𝑙𝑜𝑠𝑠,%,𝑐𝑎𝑙𝑒𝑛𝑑𝑎𝑟 = 𝑓 √𝑡 𝑒

[−𝐸𝑎𝑅 𝑇

]

(3-72)

𝑄𝑙𝑜𝑠𝑠,%,𝑐𝑦𝑐𝑙𝑒 = (𝑎 𝑇2 + 𝑏 𝑇 + 𝑐) 𝑒[(𝑑 𝑇+𝑒) 𝐶𝑟𝑎𝑡𝑒)] 𝐴ℎ𝑡ℎ𝑟𝑜𝑢𝑔ℎ𝑝𝑢𝑡 (3-73)

𝑄𝑙𝑜𝑠𝑠,% = 𝑄𝑙𝑜𝑠𝑠,%,𝑐𝑎𝑙𝑒𝑛𝑑𝑎𝑟 + 𝑄𝑙𝑜𝑠𝑠,%,𝑐𝑦𝑐𝑙𝑒 (3-74)

Equation (3-72) is the percentage capacity loss due to calendar life, while (3-73) is the

percentage capacity loss due to cycle life. Equation (3-74) is the total percentage capacity

loss. Parameters a to f are empirical fit parameters, while t represents time in days, T for

temperature in Kelvin, and Ahthroughput for total charge used. Ea and R are the activation

energy and gas constant respectively.

For the calendar life equation (3-72), they have adopted a square root of time relation to

account for the diffusion limited capacity loss, and an Arrhenius correlation to capture the

influence of temperature. For the cycle life equation (3-73), the C-rate is exponential, while

that of time (or charge throughput) is linear.

In their tests, the C-rates were constant throughout, i.e. a constant 0.5C discharge rate. In this

work, the EV is simulated over drive cycles, which are far from constant C-rates. Therefore,

the cycle life equation is modified to have an incremental capacity loss for every incremental

change in time. This is similar to the works of [66], which modifies an earlier 2011 cycle life

model by the same authors, Wang, et al. [67] to work incrementally. These are the resultant

equations,

𝑄𝑙𝑜𝑠𝑠,%,𝑐𝑦𝑐𝑙𝑒(𝑡) =

𝑑𝑄𝑙𝑜𝑠𝑠,%,𝑐𝑦𝑐𝑙𝑒(𝑡)

𝑑𝑡∆𝑡 + 𝑄𝑙𝑜𝑠𝑠,%,𝑐𝑦𝑐𝑙𝑒(𝑡 − 1)

(3-75)

𝑑𝑄𝑙𝑜𝑠𝑠,%,𝑐𝑦𝑐𝑙𝑒(𝑡)

𝑑𝑡= (𝑎 𝑇2 + 𝑏 𝑇 + 𝑐) 𝑒[(𝑑 𝑇+𝑒) 𝐶𝑟𝑎𝑡𝑒(𝑡))]

𝑑𝐴ℎ𝑡ℎ𝑟𝑜𝑢𝑔ℎ𝑝𝑢𝑡(𝑡)

𝑑𝑡

(3-76)

𝑑𝐴ℎ𝑡ℎ𝑟𝑜𝑢𝑔ℎ𝑝𝑢𝑡(𝑡)

𝑑𝑡=𝐼(𝑡) ∆𝑡

3600

(3-77)

Page 75: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

54

𝐶𝑟𝑎𝑡𝑒(𝑡) =

𝐼(𝑡)

𝑄𝑏𝑎𝑡𝑡

(3-78)

From equations (3-76) to (3-78), I(t) and ∆t are required to compute the incremental cycle-

life loss dQloss,%,cycle(t)/dt over ∆t. Then, the incremental losses are summed to compute the

total loss using (3-75).

The model does not distinguish between charge and discharge. So the absolute value is taken

for charging currents, which is negative in the convention in this work, resulting in both

discharge and charge currents being positive for running the cycle life model.

3.11 Parameters for Modelling

3.11.1 General

The next two tables comprise the parameters used for modelling the EV and its various

components as discussed in the earlier sections.

Page 76: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

55

Table 3-1 Simulation parameters for the vehicle model.

Parameter Value Source

Car Properties

Gross vehicle weight m 1962 kg Kerb weight + new components

Kerb weight mk 1503 kg [57]

Weight of driver & 4

passengers

75 * 5 kg [55]

Gear ratio Gr 8.1938 [57]

Wheel radius rwh 0.31595 m Tyre size: 205/55R16 [57]

Frontal area Af 2.31 m2 [68]

Wheelbase L 2.7 m [57]

Height of CG h 0.54 m Estimated from another sedan,

Hyundai Sonata [69]

Distance from rear wheel to

CG Lb

1.6 m Estimated from another sedan,

Hyundai Sonata [69]

Top speed 144 km/h [57]

Drag coefficient Cd 0.29 [57]

Rolling resistance coefficient

parameter Cr,a

0.006 Tuned from [61]

Drivetrain efficiency ηp 0.98 Estimated from [61]

Auxiliary power Paux 0 -

Environment Properties

Gravitational acceleration g 9.81 m/s2 -

Air density ρ 1.2041 kg/m3 At 20oC

Static friction coefficient μs 0.85 Asphalt & Concrete Road [61]

Page 77: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

56

Table 3-2 Simulation parameters for the EV motor.

Parameter Value Source

Type AC PMSM [34]

Base speed (positive torque) ωbase,pos 3008 rpm [57]

Base speed (negative torque) ωbase,neg 3018 rpm Estimated from [58]

Max speed ωmax 10500 rpm [57]

Max torque τmax 254 Nm [57]

Max power 80 kW [57]

Radius of hump r (21000)(2π/60) Tuning parameter

Height of hump h 1/r2 Tuning parameter

Maximum efficiency ɳmax 1 Tuning parameter

Centre of hump, angular velocity-wise ωcen

(rad/s)

- Tuning parameter

Centre of hump, torque-wise τcen (Nm) - Tuning parameter

Shifts centre of hump up/down τadj 25 Nm Tuning parameter

Shifts centre of hump left/right ωadj 250 rad/s Tuning parameter

Compresses/elongates hump up/down τscale 1.1 Tuning parameter

The following table on the next page contains the parameters of the new components to be

installed in the mid-sized EV. The selected UC are six Maxwell 48V general purpose

modules, connected in series, totalling 288V, 13.8F [12].

Page 78: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

57

Table 3-3 Parameters of new components to be installed in the EV.

Parameter Value Source

UC Properties

Rated voltage Vuc,max 288 V [12]

Rated capacitance C 13.83 F [12]

Internal resistance Ruc 0.06 Ω [12]

Weight 62 kg [12]

DC/DC Converter Properties

IGBT forward voltage drop

Vce = Vigbt + IL Rigbt

Vigbt + IL Rigbt =

0.857 + IL(0.00285) V

[70]

IGBT reverse voltage drop

Vec = Vd + IL Rd

Vd + IL Rd =

0.88 + IL(0.00184) V

[70]

Inductor internal resistance RL 0.037 Ω [54]

Weight 22 kg [47]

3.11.2 Using UR18650W Batteries

Table 3-4 shows the parameters for the original Nissan Leaf battery. However, the parameters

cannot be used directly if battery cycle life simulations are to be included in this work. Earlier

in section 3.10, the battery cycle life model from Wang, et al [7] was discussed. They

performed cycle life experiments using Sanyo UR18650W batteries and empirically fitted a

model with their data. Therefore, the empirical parameters only suit the UR18650W batteries.

However, there is no data on the Nissan Leaf battery cycle life. Therefore, for the mid-sized

EV in this work, the 1.5Ah Sanyo UR18650W batteries are assembled into a 98S44P

configuration to have a similar capacity and nominal voltage to the original Nissan Leaf

battery. The modified configuration values are also shown in Table 3-4.

Table 3-5 lists the battery voltage curve and battery cycle life model parameters.

Page 79: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

58

Table 3-4 Parameters of Nissan Leaf and modified batteries.

Parameter Value Source Modified Value for

UR18650W

Type Li-ion

Cathode: LiMn2O4 with

LiNiO2

Anode: Graphite

[34] Li-ion

Cathode: LiMn1/3Ni1/3Co1/3

+ LiMn2O4 Anode:

Graphite

Cell configuration 96S 2P [34] 98S 44P

Maximum voltage 403.2 V [34] 411.6 V

Nominal voltage 364.8 V [34] 362.6 V

Minimum voltage 240 V [34] 245 V

Rated capacity Qmax 66.2 Ah (24 kWh) [34] 66 Ah (23.93 kWh)

Table 3-5 Parameters for battery voltage and battery cycle life curve fitting.

Parameter Value Source

Battery Voltage Curve

Constant A 16.89 V Calculated with [63]

Constant B 6.0423e-05 (Ah)-1 Calculated with [63]

Constant K 8.919 V Calculated with [63]

Constant Vbatt,0 395.229 V Calculated with [63]

Internal resistance Rbatt 0.14 Ω [34]

Battery Cycle Life Curve

Constant a 8.6124e-06; Ah-1 K-2 [7]

Constant b -5.1252e-03 Ah-1 K-1 [7]

Constant c 7.6292e-01 Ah-1 [7]

Constant d -6.7150e-03 K-1 C-rate-1 [7]

Constant e 2.3467 C-rate-1 [7]

Constant f 14876 day-0.5 [7]

Activation energy Ea 24.5 kJ mol-1 [7]

Gas constant R 8.31446; J mol-1 K-1 [7]

Temperature T 323.15 K -

Page 80: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

59

4 HESS: IMPROVED ENERGY & POWER MANAGEMENT

STRATEGIES

As mentioned earlier, there are two parts to an HESS management strategy – energy

management strategy (EMS) and power management strategy (PMS). In this chapter, the

algorithms of the proposed EMS and PMS are discussed. In order to discuss these algorithms,

the models derived in Chapter 3 earlier are used.

As a side note, the EMS and PMS were designed in an iterative fashion, i.e. EMS was

designed first, then PMS designed second. Subsequently, the EMS was revised to suit the

PMS, and PMS was also further revised to suit the new EMS. And the cycle continued.

Therefore, it is challenging to explain the EMS and PMS in a linear fashion in this thesis. Part

of the PMS is discussed first to introduce the concept of battery limits, before the EMS is

described in full. Finally, the remainder of the PMS is discussed, which focuses on its

implementation.

4.1 Power Management Strategy Pt. 1: Battery Limits

The PMS decides how the power flow should be split between the battery and UC. The PMS

has two goals,

To ensure the actual UC energy level follows the target UC energy level

To ensure the battery power limits are not exceeded

By ensuring the first PMS goal is met, the two EMS goals will also be met. The two EMS

goals and the method of calculating this target UC energy level is explained later in EMS

section 4.2.

The focus in this section is on the second and more urgent goal, which is to ensure that

battery power is restricted to self-imposed limits [Pbatt,min, Pbatt,max]. Based on the works of

Page 81: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

60

Wang, et al. [7], a lower battery charge/discharge rate leads to a longer battery cycle life. So

by reducing battery power with the self-imposed limits, the battery cycle life can be extended.

The following sections explain how the limits were calculated.

Note that as mentioned earlier in section 1.8, the contribution of this work to the state of the

art is the speed dependent battery limits, not the two PMS goals, as these two goals are

common in existing works.

4.1.1 Battery Power at Constant Speeds

In general, the battery is used to supply the steady state power, while the UC handles the

transients. For example, if the mid-sized EV is maintaining a constant speed of 100km/h, the

battery needs to supply all the power required at steady state. This is because the UC does not

have enough energy to supply power for an extended time, so it only assists during the

transients (acceleration or braking).

From this logic, a speed-dependent battery power limit was developed, and it is based on the

power required to maintain the mid-sized EV at a constant speed. A speed-dependent battery

power limit achieves two functions – to utilise the UC even during low power demands (such

as low speeds) to further reduce battery use, and to allow the battery to supply power during

steady state.

First, the power required to maintain the EV at a constant speed is studied. This is calculated

using the vehicle model simulation discussed in section 3.6. By setting the acceleration to

zero, and varying the velocity input v(j), the power required to maintain the EV at the

specified velocity inputs can be computed. The result is shown in Figure 4-1.

An upward sloping curve with increasing gradient is observed. This is expected as the

aerodynamic drag Fw(j) in equation (3-5) earlier is dependent on velocity squared. The

maximum power required occurs at the EV’s top speed, 40.3kW at 144km/h, which

Page 82: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

61

corresponds to 1.69C. This is far under the 90kW rated Nissan Leaf battery, as the Nissan

Leaf battery needs to handle transients as well. Limiting the battery power to approximately

half its rated power would benefit battery cycle life.

Figure 4-1 Battery power required at constant vehicle speed.

In short, Figure 4-1 shows what the maximum battery discharge power limit should be for the

proposed algorithm design, where the battery supplies enough power to maintain the mid-

sized EV at constant speed, and the UC supplies the extra transient acceleration power.

4.1.2 Final Battery Limit Curve

However, upon further simulations, the UC needs to be much larger in order to supply the

required transient power. This is further explained when discussing the EMS in section

4.2.4.3. Therefore, the battery discharge power limit restriction was loosened by a factor of

Pbatt,max,scale = 3. How this factor of 3 was calculated is also discussed in section 4.2.4.3.

By reflecting Figure 4-1 across its x-axis (making it upside down), the minimum allowed

battery power (or maximum charge power) is attained. On a Nissan Leaf, the maximum

allowed regenerative braking is a constant 30kW (1.25C), as seen on its digital Energy

Information dashboard screen [71]. Therefore, this -30kW is set as the absolute minimum

value.

Page 83: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

62

Figure 4-2 shows the final battery limit curve of the proposed algorithm design, which is

solely dependent on speed. The maximum battery limit has been loosened by a factor of

Pbatt,max,scale = 3. However, it has been clipped to a maximum of 40.6kW or 1.7C (slightly

more than the original 1.69C to give a buffer). Similarly, the minimum battery limit has been

clipped to the -30kW or 1.25C.

Figure 4-2 Speed-dependent PMS battery limit curve.

At lower speeds, the limits have been opened up to 0.4C (9.6kW). This allows the battery to

charge or discharge the UC to the required target UC energy level at those speeds. For

example, if the battery limits were zero at zero speed (as shown by the original curve in

Figure 4-1), then the UC would be unable to charge when the vehicle is not moving. 0.4C

was chosen as it is a gentle discharge on the battery, yet able to top-up the UC from 90%

SOC to full in about 10 seconds (the difference in SOC is due to deviation from the EMS

algorithm, which uses average values).

With this proposed speed-dependent band, the higher the vehicle speed, the more battery

power can be used.

The final speed-dependent battery limit curve can be expressed mathematically by,

Page 84: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

63

𝑃𝑏𝑎𝑡𝑡,𝑚𝑎𝑥(𝑣(𝑡)) = max(9.56,𝑚𝑖𝑛(𝑃𝑏𝑎𝑡𝑡,𝑚𝑎𝑥,𝑠𝑐𝑎𝑙𝑒 𝑃𝑏𝑎𝑡𝑡,𝑐𝑜𝑛𝑠𝑡(𝑣(𝑡)) , 40.6)) (4-1)

𝑃𝑏𝑎𝑡𝑡,𝑚𝑖𝑛(𝑣(𝑡)) = min(−9.56,𝑚𝑎𝑥(−𝑃𝑏𝑎𝑡𝑡,𝑐𝑜𝑛𝑠𝑡(𝑣(𝑡)),−30)) (4-2)

To summarise, Figure 4-2 shows the speed-dependent battery limit curve of the proposed

algorithm design, which is a simple rule-based deterministic design. The higher the vehicle

speed, the higher the limits, and more battery power can be used. This speed-dependent

battery power limit permits two goals to be achieved – to utilise the UC even during low

power demands to reduce battery use, and to allow the battery to supply power during steady

state. This will be further illustrated in the simulations in section 5.3. The simulations also

show that existing constant battery limits from literature are unable to meet these two goals

simultaneously.

4.2 Energy Management Strategy

The self-imposed battery power limits in the PMS was introduced in the previous section. In

this section, the EMS is discussed. There are two goals in the proposed EMS,

To ensure sufficient space in the UC to absorb energy during future regenerative

braking.

To ensure sufficient energy remaining in the UC for future accelerations.

These two goals are achieved by setting a target UC energy band. The upper limit of the band

defines the maximum allowed UC energy, which is calculated from the space required to

capture regenerative braking energy. The lower limit defines the minimum allowed UC

energy, which is calculated from the energy required for future accelerations. As long as the

actual UC energy level is within the band, then the two goals are achieved.

The contribution of this proposed EMS to the state of the art is the rigorous methods for

determining this band, which considers multiple factors, such as worst case scenarios and

Page 85: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

64

real-life drive cycles. In other works, the band, or mostly, just a target UC energy level is set

based on experience.

In the following sections, the upper and lower limits of the UC energy band are calculated

and explained. Figure 4-3 shows the overall flow of the proposed EMS design. From the

vehicle speed, the target UC energy band can be calculated.

Figure 4-3 Block diagram of overall EMS design

4.2.1 Sufficient Space in UC for Regenerative Braking

In this section, the ‘Anticipate UC space required for regen.’ block in Figure 4-3 is discussed.

Figure 4-4 presents a detailed diagram of this block. The ‘Anticipate UC energy required for

acc.’ block is shared in Figure 4-4 as they are similar. It will be discussed later in section

4.2.2.

The first EMS goal is to ensure sufficient space (i.e. sufficiently low energy or SOC) in the

UC for regenerative braking, by estimating the energy absorbed by the UC during

regenerative braking and therefore generating the initial maximum allowed starting UC SOC.

Upper

band limit

Lower band

limit

Anticipate UC space

required for regen.

(section 4.2.1, Figure

4-4)

Consider worst case

scenario

(section 4.2.3.2) UC sizing

(section

4.2.4.2) Worst case brake torque

Target UC

band

(section

4.2.4.1) Anticipate UC energy

required for acc.

(section 4.2.2, Figure

4-4)

Consider worst case

scenario

(section 4.2.3.3)

Worst case acc. torque

Battery

limits from

PMS

(section 4.1,

Figure 4-2)

Vehicle

Speed

Page 86: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

65

The general idea of the algorithm is a forward approach simulation – a constant braking

torque is applied, and then kinematic equations are used to predict the future velocity profile

of the car and the motor. With the motor velocity and torque, the power and energy generated

during regenerative braking can be computed. Now, the algorithm is explained in detail.

Figure 4-4 EMS block diagram to anticipate UC space required or energy generated.

The inputs to this algorithm are present motor speed ωm(j), target braking torque at the motor

τbr,tar, battery power at previous instant Pbatt(j-1). The output of this algorithm is the estimated

Motor torque/speed,

efficiency

(Figure 3-8)

ηm(τm,ωm)

τm,tar,

ωm,tar

Ptar+aux

Compute

motor ang.

velocity

(4-8)

Compute

motor

torque

(4-9)

Compute total

power

(4-10), (4-11)

Compute batt./UC

power split

(4-12) to (4-14)

Compute energy

generated

(4-15)

Compute energy

required

(4-18)

If regen. If acc.

If regen. If acc.

Puc,H,tar

Euc,regen,tar(j)

Euc,acc,tar(j)

Pbatt(j-1)

rwh, Gr

ωm,tar

ωm,tar

SOCuc,regen,tar(j)

SOCuc,acc,tar(j)

Compute max.

UC SOC

(Figure 3-17)

Compute min.

UC SOC

(Figure 3-17)

ηDC/DC(k)

ωm(j)

τbr,tar

τacc ratio

vwh,tar

τwh,tar(k)

Compute wheel

torque

(4-3)

Compute wheel

& motor torque

(4-16), (4-17)

m, mr, rwh,

Gr, ηp

If regen. If acc.

τwh,tar(k)

vwh,tar(k)

For k, loop until target speed

reached

τm,tar

Compute wheel

velocity

(4-4) to (4-7)

DC/DC converter

efficiency

(Figure 3-24)

Output

Page 87: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

66

energy sent to the UC on the low-side of the DC/DC converter Euc,regen,tar(j) and the maximum

initial UC SOC SOCuc,regen,tar(j).

In the first instant at k=1, the target brake torque τbr,tar at the motor is applied (The reason and

value for this target brake torque is explained later in section 4.2.3). The brake torque at the

wheel τwh,tar(k) is computed as follows,

𝜏𝑤ℎ,𝑡𝑎𝑟(𝑘) =

𝐺𝑟 𝜏𝑏𝑟,𝑡𝑎𝑟𝜂𝑝

(4-3)

The tractive effort Ft,tar(k=1) and resistance force Fv,tar(k=1) at that point of time is also

calculated. Then the acceleration awh,tar(k=1) can be determined as follows,

𝐹𝑡,𝑡𝑎𝑟(𝑘) =

𝜏𝑤ℎ,𝑡𝑎𝑟(𝑘)

𝑟𝑤ℎ

(4-4)

𝐹𝑣,𝑡𝑎𝑟(𝑗) = 𝐹𝑤,𝑡𝑎𝑟(𝑗) + 𝐹𝑟,𝑡𝑎𝑟(𝑗) + 𝐹𝑔,𝑡𝑎𝑟(𝑗) (4-5)

𝑎𝑤ℎ,𝑡𝑎𝑟(𝑘) =

𝐹𝑡,𝑡𝑎𝑟(𝑘) − 𝐹𝑣,𝑡𝑎𝑟(𝑘)

𝑚 +𝑚𝑟

(4-6)

The future velocity vwh,tar(k) can be computed with the acceleration from the previous instant

as follows,

𝑣𝑤ℎ,𝑡𝑎𝑟(𝑘) = 𝑣𝑤ℎ,𝑡𝑎𝑟(𝑘 − 1) + 𝑎𝑤ℎ,𝑡𝑎𝑟(𝑘)∆𝑡 (4-7)

Now, computations for the first iteration k=1 have been completed. Subsequently, equations

(4-3) to (4-7) are looped for iterations of k until the vehicle velocity vwh,tar drops to zero,

where the car has stopped moving and no further regenerative braking can occur.

As a side remark, k corresponds to the inner loop forward approach kinematic equation

iterations until the car has reached the desired velocity (zero in this case), while j corresponds

to the outer loop drive cycle in the backwards approach simulation (see Figure 4-4).

At the end of the loop, the motor speed ωm,tar at every iteration k is computed by,

Page 88: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

67

𝜔𝑚,𝑡𝑎𝑟 =

𝐺𝑟 𝑣𝑤ℎ,𝑡𝑎𝑟𝑟𝑤ℎ

(4-8)

The target braking torque as mentioned previously is τbr,tar. But before the power generated by

regenerative braking can be computed, the motor must be checked to see if it can handle the

high braking torques encountered.

As discussed in section 3.6 Vehicle Model, the maximum braking torque (alternatively,

minimum motor torque) τm,min(ωm(k)) (which is a negative number) as well as the

motor/inverter efficiency ηm,tar(k) can be found from a specified motor speed ωm,tar(k) using

Figure 3-8. Therefore, the motor provides the following braking torque,

𝜏𝑚,𝑡𝑎𝑟(𝑘)− = max (𝜏𝑏𝑟,𝑡𝑎𝑟(𝑘), 𝜏𝑚,𝑚𝑖𝑛(𝜔𝑚(𝑘))) (4-9)

The rest of the braking torque comes from the mechanical brakes. Now, the total power

generated during regenerative braking operation can be computed after considering

motor/inverter efficiencies,

𝑃𝑡𝑎𝑟(𝑘) =

𝜏𝑚,𝑡𝑎𝑟(𝑘) 𝜔𝑚,𝑡𝑎𝑟(𝑘)

𝜂𝑚,𝑡𝑎𝑟(𝑘) 𝑓𝑜𝑟 𝜏𝑚,𝑡𝑎𝑟(𝑘)+

𝜏𝑚,𝑡𝑎𝑟(𝑘) 𝜔𝑚,𝑡𝑎𝑟(𝑘) 𝜂𝑚,𝑡𝑎𝑟(𝑘) 𝑓𝑜𝑟 𝜏𝑚,𝑡𝑎𝑟(𝑘)−

(4-10)

Note that in regenerative braking, Ptar(k) is a negative value. After including the auxiliary

loads Paux, the total regenerative power becomes,

𝑃𝑡𝑎𝑟+𝑎𝑢𝑥(𝑘) = 𝑃𝑡𝑎𝑟(𝑘) + 𝑃𝑎𝑢𝑥 (4-11)

Some of this power is charged to the UC and some to the battery. This power split is decided

by the self-imposed battery power limits [Pbatt,min, Pbatt,max] discussed in section 4.1 earlier.

The regenerative power is charged to the battery as long as the anticipated battery power

Pbatt,tar is within limits, i.e. Pbatt,max(k) > Pbatt,tar(k) > Pbatt,min(k). The remaining power which

Page 89: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

68

cannot be captured by the battery will be charged to the UC (high-side UC power Puc,H,tar).

This can be expressed mathematically as follows,

𝑃𝑏𝑎𝑡𝑡,𝑡𝑎𝑟(𝑘) =

min(𝑃𝑏𝑎𝑡𝑡,𝑡𝑎𝑟(𝑘), 𝑃𝑏𝑎𝑡𝑡,𝑚𝑎𝑥(𝑘)) 𝑖𝑓 𝑃𝑏𝑎𝑡𝑡,𝑡𝑎𝑟 ≥ 0

max(𝑃𝑏𝑎𝑡𝑡,𝑡𝑎𝑟(𝑘), 𝑃𝑏𝑎𝑡𝑡,𝑚𝑖𝑛(𝑘)) 𝑖𝑓 𝑃𝑏𝑎𝑡𝑡,𝑡𝑎𝑟 < 0

(4-12)

𝑃𝑢𝑐,𝐻,𝑡𝑎𝑟(𝑘) = 𝑃𝑡𝑎𝑟+𝑎𝑢𝑥(𝑘) − 𝑃𝑏𝑎𝑡𝑡,𝑡𝑎𝑟(𝑘) (4-13)

where the starting value of Pbatt,tar(1) is,

𝑃𝑏𝑎𝑡𝑡,𝑡𝑎𝑟(1) = 𝑃𝑏𝑎𝑡𝑡(𝑗 − 1) (4-14)

Note that for the regenerative braking case, only Pbatt,tar < 0 is considered. Pbatt,tar > 0 is also

shown above for completion. It will be used in the acceleration case later.

Now, the anticipated (target) battery power Pbatt,tar and high-side UC power (before DC/DC

converter losses) Puc,H,tar has been computed for one instant of j, which comprises multiple

instances of k. Figure 4-5 shows a demonstration of this algorithm, where the mid-sized EV

was braked from 120km/h (33m/s) to zero. Figure 4-5(a) shows the kinematics, Figure 4-5(b)

shows the torques, Figure 4-5(c) shows the powers, and Figure 4-5(d) shows the

corresponding UC and battery SOC values.

The target braking torque τbr,tar in this demonstration is 244Nm (more on how this value was

selected later). The motor cannot handle this torque above the base speed, so the mechanical

brakes assist in the braking as seen in the first 6s of Figure 4-5(b). Note that in reality, the

mechanical brake torque acts on the wheel, but the torque in Figure 4-5(b) shows the

mechanical brake torque as if it occurred at the motor shaft for an easier comparison with the

motor torque.

In Figure 4-5(c), the power split is shown. As the total regenerative power Ptar+aux exceeds the

Pbatt,min battery limit, the battery charges at the limit while the UC handles the rest of the

power, given by Puc,tar (low-side power) and Puc,H,tar (high-side power).

Page 90: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

69

Figure 4-5 (a) Kinematics (b) torques (c) powers (d) SOCs during regenerative braking.

Assuming the UC is fully charged at the end of braking, the maximum corresponding UC

energy level or SOC in order to absorb this energy can be calculated by working backwards.

Page 91: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

70

Note, fully charged is defined as 98.8% UC SOC. In the experiment setup later, the UC was

not charged to 100% for safety reasons, but to 98.8%. So the simulation here was revised to

standardise the conditions.

By integrating the high-side UC power Puc,H,tar(k) over time k, as well as considering the

DC/DC converter losses, the energy Euc,regen,tar(j) sent to the UC can be computed for the

above demonstration,

𝐸𝑢𝑐,𝑟𝑒𝑔𝑒𝑛,𝑡𝑎𝑟(𝑗) =

𝑃𝑢𝑐,𝐻,𝑡𝑎𝑟(𝑘)

𝜂𝐷𝐶/𝐷𝐶,𝑏𝑜(𝑘)𝑑𝑘

𝑘

0

𝑖𝑓 𝑃𝑢𝑐,𝐻,𝑡𝑎𝑟(𝑘) ≥ 0

∫ 𝜂𝐷𝐶/𝐷𝐶,𝑏𝑢(𝑘) 𝑃𝑢𝑐,𝐻,𝑡𝑎𝑟(𝑘)𝑘

0

𝑑𝑘 𝑖𝑓 𝑃𝑢𝑐,𝐻,𝑡𝑎𝑟(𝑘) < 0

(4-15)

Therefore, the UC needs to have this amount of space available to absorb the regenerative

braking if the car were to brake to a standstill from 120km/h with a braking torque of 244Nm.

Inserting these values into the UC model discussed in section 3.8, the corresponding

maximum UC SOC values can be computed.

Figure 4-5(d) shows the SOCs of the battery and UC. The maximum allowed starting

SOCuc,regen,tar value is 48.9% for this demonstration (as shown by the orange arrow) to ensure

sufficient space in the UC.

4.2.2 Sufficient Energy in UC for Acceleration

The second EMS goal is to ensure sufficient energy in the UC for acceleration, by estimating

the energy required from the UC during acceleration and therefore generating the minimum

allowed initial starting UC SOC. The general idea of this algorithm is similar to the previous

regenerative braking algorithm, where kinematic equations are used to predict the future

velocity profile of the car and the motor. Then the required power can be calculated.

The difference between this algorithm and the regenerative braking algorithm is that the

regenerative braking algorithm allows a brake torque which is both constant and exceeds the

Page 92: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

71

motor limits since the mechanical brakes can assist. But in this acceleration algorithm, the

acceleration torque is solely provided by the motor, so it will not exceed the motor limits. In

addition, the torque may not be constant when the motor is operating in the constant power

region. Therefore, instead of specifying a constant acceleration torque, a constant acceleration

torque ratio τacc ratio = τm,tar(k)/τm,max(ωm(k)) is specified, where τm,max(ωm(k)) is obtained from

Figure 3-8. What this means is that by holding the accelerator pedal steady (constant τacc ratio),

the torque output τm,tar(k) will be directly proportional to the maximum torque of the motor

τm,max(ωm(k)) at that specific speed.

Now, the algorithm is explained in detail. The inputs to the algorithm are present motor speed

ωm(j), target acceleration torque ratio τacc ratio, battery power at previous instant Pbatt(j-1). The

output is the estimated energy required by the UC on the low-side of the DC/DC converter

Euc,acc,tar(j) and the minimum initial UC SOC SOCuc,acc,tar(j) (see Figure 4-4).

In the first instant k=1, the target acceleration torque τm,tar(k) at the motor is applied. Then the

acceleration torque at the wheel τwh,tar(k) is computed by,

𝜏𝑚,𝑡𝑎𝑟(𝑘) = 𝜏𝑎𝑐𝑐 𝑟𝑎𝑡𝑖𝑜 𝜏𝑚,𝑚𝑎𝑥(𝜔𝑚(𝑘)) (4-16)

𝜏𝑤ℎ,𝑡𝑎𝑟(𝑘) = 𝐺𝑟 𝜏𝑚,𝑡𝑎𝑟(𝑘) 𝜂𝑝 (4-17)

Then, the tractive effort Ft,tar(k) and resistance force Fv,tar(k) at that point of time are

calculated using earlier equations (4-4) and (4-5). Subsequently, the acceleration awh,tar(k) is

calculated using (4-6). Finally, the future velocity vwh,tar(k) is computed using equation (4-7).

Equations (4-16), (4-17), (4-4), (4-5), (4-6), (4-7), in this specific order, are looped until the

vehicle velocity vwh,tar increases to 120km/h (33m/s). A highway speed of 120km/h has been

selected instead of the top speed of the mid-sized EV, 144km/h, as under normal driving

conditions, the top speed would not be attained. Therefore, the algorithm ensures the UC has

sufficient energy just to get the car to highway speeds (more on limitations later).

Page 93: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

72

At the end of the loop, the motor speed ωm,tar at every iteration k can be computed with (4-8).

Next, the total power required for acceleration after considering motor/inverter efficiencies is

calculated using (4-10) and (4-11).

Some of this power will be supplied by the UC and some by the battery, where the battery

power is restricted to the speed-dependent limit discussed in section 4.1. The remaining

power which cannot be supplied by the battery due to the restriction will be supplied by the

UC, and this is expressed mathematically using equations (4-12) to (4-14).

Figure 4-6 shows a demonstration of this algorithm, where the car was originally travelling at

30km/h (8m/s) and suddenly required an acceleration torque ratio of 0.5.

Figure 4-6(c) shows the power split. As the total required acceleration power Ptar+aux exceeds

the Pbatt,max battery limit, the battery discharges at the limit while the UC handles the rest of

the power.

Assuming the UC SOC is at the minimum level (50%) at the end of the acceleration, the

minimum corresponding initial UC energy level or SOC to provide this acceleration can be

calculated by working backwards.

By integrating the high-side UC power Puc,H,tar(k) over time k, as well as considering the

DC/DC converter losses, the energy Euc,acc,tar(j) required from the UC for the demonstration

can be computed by,

𝐸𝑢𝑐,𝑎𝑐𝑐,𝑡𝑎𝑟(𝑗) =

𝑃𝑢𝑐,𝐻,𝑡𝑎𝑟(𝑘)

𝜂𝐷𝐶/𝐷𝐶,𝑏𝑜(𝑘)𝑑𝑘

𝑘

0

𝑖𝑓 𝑃𝑢𝑐,𝐷𝐶/𝐷𝐶,𝑡𝑎𝑟(𝑘) > 0

∫ 𝜂𝐷𝐶/𝐷𝐶,𝑏𝑢(𝑘) 𝑃𝑢𝑐,𝐻,𝑡𝑎𝑟(𝑘)𝑑𝑘𝑘

0

𝑖𝑓 𝑃𝑢𝑐,𝐷𝐶/𝐷𝐶,𝑡𝑎𝑟(𝑘) < 0

(4-18)

Therefore, the UC needs to have this amount of energy if the car were to accelerate from

30km/h to 120km/h with an acceleration torque ratio of 0.5. Inserting these values into the

UC model discussed in section 3.8, the corresponding SOCuc,acc,tar values can be computed.

Page 94: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

73

Figure 4-6 (a) Kinematics (b) torques (c) powers (d) SOCs during acceleration.

Page 95: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

74

Figure 4-6(d) shows the SOCs of the battery and UC. The minimum allowed starting

SOCuc,acc,tar value is 94.8% for this demonstration (as shown by the orange arrow) to ensure

sufficient energy in the UC for acceleration.

4.2.3 Selected Braking and Acceleration Torque Values

In this section, the braking torque τbr,tar and acceleration torque ratio τacc ratio values mentioned

in the previous sections are discussed.

The general idea here is to select the worst case regenerative braking scenario (recovers most

UC energy, therefore requires most space), and the worst case acceleration scenario

(consumes most UC energy). This is only done once and is not part of the simulation loop.

4.2.3.1 Worst Case Braking Torque

To recover the most energy during braking, the car should be stopped as quickly as possible,

with a braking torque just before the wheel skids. This is because less energy would be lost to

the resistance forces Fv if the distance d that the car travels is shorter,

𝐸𝑣,𝑙𝑜𝑠𝑠 = 𝐹𝑣 𝑑 (4-19)

However, this means the mechanical brakes need to be used, and energy will be dissipated, so

the braking forces need to be restrained to within the motor limits. Therefore, the most energy

can be recovered during regenerative braking when the braking torque is controlled to follow

the maximum torque/speed curve in Figure 3-8 (i.e. follow the edge).

However, there are a few problems with following the curve. Firstly, it is difficult for a driver

to operate the brake pedals such that it follows the curve exactly. Secondly, this goes against

conventional driving techniques. In driving school, learners are usually taught to brake harder

and firmer at the start of the deceleration to bring the speed of the car down quickly, and then

ease up on the brakes as the car slows. In contrast, to recover the most energy, the driver

Page 96: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

75

needs to apply a small braking torque at the start of the deceleration, and then progressively

increase the brake torque to the motor maximum as the car slows. This may be dangerous as

the driver may underestimate the braking effort required and overshoot.

In addition, the motor/inverter efficiency has not been considered yet. So in fact, following

the edge of the motor torque/speed curve does not give the worst case braking torque. The

worst case braking torque consists of a path within the torque/speed curve which attains the

highest overall efficiency. This further adds to the difficulty, and it is highly unlikely that a

driver would apply the brakes in such a pattern.

4.2.3.2 Worst Case Constant Brake Torque

To simplify the problem, the strategy in this proposed work is restricted to a constant brake

torque value, which is also more realistic in terms of driving patterns. Next, the constant

brake torque value that corresponds to the maximum energy absorbed by the UC is found.

The forward approach simulation discussed in section 4.2.1 is run repeatedly, varying the

braking torque value τbr,tar, as well as the starting vehicle velocity vwh. The battery power at

the previous instant Pbatt(j-1) is calculated to be the power required for maintaining the initial

starting vehicle velocity vwh (shown earlier in Figure 4-1). Then a target UC SOC value

SOCuc,regen,tar for every τbr,tar and vwh is obtained, and this is represented by a contour plot in

Figure 4-7. In general, the higher the starting velocity of the car, the more UC regenerative

braking energy it can recover.

At each velocity vwh, the brake torque τbr,tar that recovers the most energy (i.e. lowest starting

target UC SOC) is found, and this is highlighted by the red line in Figure 4-7. Figure 4-8

shows this line isolated.

Page 97: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

76

Figure 4-7 Target UC SOC for varying brake torques and start velocities.

Figure 4-8 Brake torques corresponding to max. recovered UC energy for each velocity.

In Figure 4-8, at high starting velocities, a smaller (less negative) constant braking torque

generates more energy. In contrast, at low starting velocities, a larger (more negative)

constant braking torque generates more energy. This seems counter intuitive, but can be

explained easily.

At high velocities, especially in the constant power region of the motor, the amount of

braking torque that the motor can absorb is low as shown in Figure 3-8. Therefore, a smaller

constant braking torque is used to reduce energy dissipated by the mechanical brakes.

Page 98: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

77

Conversely, when the starting velocity is low, the motor would be in the constant torque

region, providing more braking torque. Therefore, a braking torque value close to the motor

limits can be used.

Since regenerative braking is most often used in city start-stop driving, where the top speed is

at most 60km/h, the selected constant brake torque for this proposed strategy is 244Nm,

which is an average of the 0 – 60km/h section of Figure 4-8. Therefore, below 60km/h, there

will be sufficient space in the UC for regenerative braking. But at speeds above 60km/h, if

the driver coincidentally applies the corresponding worst case braking torque to stop the car,

the UC will not be able to absorb the energy completely. The excess will be directed to the

battery instead. If the driver is travelling above 60km/h, he is likely to be travelling on a

highway and the use of brakes is not as frequent as city driving. Furthermore, in normal

situations, he is unlikely to brake to a standstill in the middle of the highway. Therefore, this

is a good compromise to ensure minimal sizing of the UC (more on this in section 4.2.4.2 UC

Sizing). Although there are some compromises, it will be shown in the simulations later that

this is not a problem as the algorithm can tolerate harsh driving for short durations.

This was how the target braking torque τbr,tar = 244Nm was selected in section 4.2.1.

4.2.3.3 Worst Case Constant Acceleration Torque

In this section, the constant acceleration torque ratio that corresponds to the maximum energy

which needs to be supplied by the UC is determined.

Again, a constant acceleration torque ratio is used for simplicity. Similar to the previous

section, the forward simulation discussed in section 4.2.2 is run repeatedly, varying the

acceleration torque ratio τacc ratio, as well as the starting vehicle velocity vwh. Then a target UC

SOC value SOCuc,acc,tar for every τacc ratio and vwh is obtained, shown in Figure 4-9.

Page 99: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

78

Figure 4-9 Target UC SOC for varying brake torques and start velocities.

In general, the higher the desired acceleration and the lower the initial car speed, the more

UC energy is required to bring the car to a highway speed of 120km/h.

The white line in Figure 4-9 demarcates the UC 100% SOC crossing. To the bottom right of

the line with high acceleration ratios and low initial vehicle speeds, the UC needs to be

charged beyond 100% to enable that acceleration (i.e. insufficient energy in the UC). This

will not be a problem as explained later (Although the algorithm is not designed for

aggressive acceleration, it can tolerate it).

Again, at each velocity vwh in Figure 4-9, the acceleration torque ratio τacc ratio that consumes

the most UC energy is found. This corresponds to the right most edge, τacc ratio = 1 for all

speeds, i.e. to floor the accelerator.

However, in normal driving conditions, it is highly uncommon to floor the accelerator pedal

from 0 to 120km/h in a single shot. Therefore, the region in the bottom right corner is

unlikely to occur. Other methods of determining the worst case acceleration energy under

normal driving conditions are considered instead.

Page 100: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

79

Comparing the NEDC, FTP-75 city and FTP-75 HWFET regular driving cycles, the

maximum (worst case) torque for acceleration occurs in the FTP-75 city drive cycle with

123Nm at 37.2 km/h, corresponding to a motor speed of 2557rpm on the Nissan Leaf (below

motor base speed). Using these figures, the acceleration torque ratio was calculated as

follows,

𝜏𝑎𝑐𝑐 𝑟𝑎𝑡𝑖𝑜(𝑘) =

𝜏𝑚,𝑡𝑎𝑟𝜏max(2557 𝑟𝑝𝑚)

= 123 𝑁𝑚

254 𝑁𝑚= 0.48

(4-20)

Therefore, τacc ratio was selected at 0.5 (to allow some buffer). The implication is that the UC

only has sufficient energy to get to 120km/h for acceleration torque ratios below 0.5. This is a

good compromise as it ensures minimal sizing of the UC, while allowing sufficient

acceleration under normal driving conditions, such as those seen in the standard drive cycles.

It will be shown later that the algorithm can tolerate aggressive driving cycles such as the

LA92, as there is no flooring the accelerator from 0 to 120km/h.

This was how the target acceleration torque ratio τacc ratio = 0.5 was selected in section 4.2.2.

4.2.4 Target UC Energy Band

In this section, the creation of the target UC energy band is discussed. In earlier sections 4.2.1

and 4.2.2, the maximum and minimum target UC SOC values – SOCuc,regen,tar(j) and

SOCuc,acc,tar(j) respectively were calculated.

4.2.4.1 Target UC SOC Band

These maximum and minimum target SOC values define a UC SOC band. If the present UC

SOC SOCuc(j) is within the band, i.e.

𝑆𝑂𝐶𝑢𝑐,𝑎𝑐𝑐,𝑡𝑎𝑟(𝑗) ≤ 𝑆𝑂𝐶𝑢𝑐(𝑗) ≤ 𝑆𝑂𝐶𝑢𝑐,𝑟𝑒𝑔𝑒𝑛,𝑡𝑎𝑟(𝑗) (4-21)

then there would always be sufficient space and energy in the UC, meeting the goals of the

proposed EMS. As the braking torque τbr,tar and acceleration torque ratios τacc ratio have already

Page 101: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

80

been determined and fixed in the previous section, the band limits only depend on vehicle

velocity vwh.

Figure 4-10 shows the final result of the target UC SOC band versus vehicle velocity curve,

with the upper limit (calculated from regenerative braking) in blue and the lower limit

(calculated from acceleration) in red.

Figure 4-10 Target UC SOC band vs. speed, 6 UC modules.

It is observed that both curves slope downwards. For the blue regenerative braking curve, the

higher the vehicle speed, the more energy will be captured during regenerative braking,

resulting in the downward sloping curve. Similarly, for the red acceleration curve, the higher

the vehicle speed, the less energy required to get it to the highway speed of 120km/h.

A substantial change in the gradient is observed at about 90km/h for the red acceleration

curve. This corresponds to the flat section for the blue Pbatt,max curve in Figure 4-2. Speeds

above 45km/h correspond to the constant power region of the motor. At speeds above

90km/h, as the battery power limit is at the maximum (flat section), less UC energy is

required for a constant power. Therefore, the gradient of the red acceleration curve flattens

out.

Page 102: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

81

The following sections explain how Figure 4-10 was finalised.

4.2.4.2 UC Sizing

Six Maxwell 48V general purpose UC modules in series, totalling 288V and 13.8F were

used. The UC was sized based on being able to capture regenerative braking energy. The

lowest point of the blue regenerative braking curve in Figure 4-10 is 48.9% as shown by the

orange arrow. Earlier in section 3.8 Ultracapacitor Model, it was mentioned that the UC

should not be discharged below 50% SOC as the DC/DC converter becomes inefficient at

low UC voltages. Although 48.9% is slightly below 50%, it is still acceptable as it is not a

hard rule.

Another perspective was shown in earlier Figure 4-7, which was simulated with six UC

modules. In that figure, the white arrow points to the maximum starting 48.9% SOC with a

selected brake torque of 244Nm.

If only five UC modules were used, Figure 4-11 shows the results (ignore the red curve for

now). From the orange arrow, the lowest point of the blue regenerative braking curve is

37.0%, which is too low. Another perspective is shown in Figure 4-12.

Figure 4-11 Target UC SOC band vs. speed, 5 UC modules.

Page 103: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

82

Figure 4-12 Target UC SOC for varying brake torques and start velocities, 5 UC modules.

The white arrow in Figure 4-12 points to the maximum starting 37.0% SOC with a selected

brake torque of 244Nm. The white semi-circle at the top centre is a part which is

unachievable (imaginary roots to the DC/DC converter quadratic equations), where the

starting UC SOC drops too low, and has insufficient power to compensate for the DC/DC

converter losses.

Therefore, six UC modules is the minimum number of UC modules to fully capture the

regenerative braking energy.

In Figure 4-10, the two limits cross at about 115km/h. That means beyond 115km/h, one of

the two UC goals (sufficient energy for acceleration or space for braking) cannot be achieved.

An additional UC module would widen the band such that they no longer cross. But as it is a

very short section, six UC modules is a good compromise to avoid another UC module,

which costs approximately US$1.2k [13]. An alternative is to define the minimum of the red

acceleration curve as 48.9% SOC in earlier section 4.2.2, i.e. to shift the red acceleration

curve downwards slightly, such that the two curves no longer cross.

Page 104: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

83

4.2.4.3 Tuning Battery Limits

Earlier in section 4.1.2 on battery limits, loosening the battery restriction by a factor of

Pbatt,max,scale = 3 was mentioned. The reason for that is discussed here. If the battery power at

constant speed curve from Figure 4-1 was used directly as the battery limits, Figure 4-13

shows the result.

Figure 4-13 Target UC SOC band vs. speed, no battery power loosening

In Figure 4-13, the red acceleration curve shoots off the top of the figure at low speeds,

meaning more than 100% UC SOC is required to perform the acceleration. This is because

more energy is required for acceleration to a certain speed than energy generated from

braking from that speed, due to electrical losses (DC/DC converter, motor/inverter, etc.) and

mechanical losses (friction, etc.). There are two ways to solve this. The first is to increase the

UC size. However, this is an expensive solution. The second way is to loosen the self-

imposed battery restriction (i.e. sacrificing some cycle life), which allows the battery to be

used more.

In order to fit the red acceleration curve under the blue regenerative curve, the self-imposed

battery restriction was loosened by increasing Pbatt,max,scale iteratively in an approach similar to

Page 105: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

84

the bisection method. The value which fits the acceleration curve just under the regenerative

curve is Pbatt,max,scale = 3, resulting in the band shown in Figure 4-10.

In short, this loosening compensates for the differences in energy required for acceleration

and energy generated from braking. It allows the UC to be sized appropriately for both

regenerative braking and acceleration.

4.2.4.4 Target UC SOC Level

In Figure 4-10, the target UC SOC band was shown. As long as the actual UC SOC is within

the band, the two UC goals of sufficient energy for acceleration and space for braking are

satisfied. In this section, a specific point within the band is selected to reduce the band to a

target UC SOC level for implementation.

The maximum possible SOC of the two limits is selected, i.e.

𝑆𝑂𝐶𝑢𝑐,𝑡𝑎𝑟(𝑣𝑤ℎ(𝑡)) = max (𝑆𝑂𝐶𝑢𝑐,𝑎𝑐𝑐,𝑡𝑎𝑟(𝑣𝑤ℎ(𝑡)), 𝑆𝑂𝐶𝑢𝑐,𝑟𝑒𝑔𝑒𝑛,𝑡𝑎𝑟(𝑣𝑤ℎ(𝑡))) (4-22)

This results in SOCuc,tar following the blue SOCuc,regen,tar curve in Figure 4-10 below 115km/h

and the red SOCuc,acc,tar curve above 115km/h. This implies that more emphasis is placed on

acceleration as the EMS acceleration goal is always met. The UC has both sufficient space

for braking and energy for acceleration at speeds below 115km/h. Above 115km/h, the UC

has the minimum energy required for acceleration but slightly insufficient space for braking.

When travelling above 115km/h, the EV is likely to be on a highway and braking would not

be frequent, so this is not a major concern.

The reason for placing emphasis on acceleration is because acceleration power comes solely

from the battery and UC. While for braking, in the worst case scenario, the excess energy can

be dumped to the braking chopper or the mechanical brakes.

Page 106: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

85

4.2.5 Summary

In summary, comprehensive calculations which consider the worst case scenarios have been

provided to justify the proposed EMS. The upper limit of the target UC SOC band is

determined by the amount of energy which would be generated in the averaged worst case

regenerative braking. The lower limit of the band is determined by the amount of energy

required for the averaged worst case future acceleration and real-life drive cycles. Since the

worst case scenarios have already been considered, knowledge of future drive profile is not

required. This is an advantage as the proposed HESS management strategy can work on any

route, even new ones (In contrast, PMS based on global optimization as discussed in the

literature review require the drive profile beforehand).

As long as the actual UC SOC is within the proposed target UC SOC band, the UC will

always have sufficient space for regenerative braking and sufficient energy for acceleration,

meeting the two EMS goals (except for a very short section above 115km/h). An added

benefit is that during the algorithm design process, the UC is also sized appropriately to

reduce costs.

Although the algorithm has been designed using averaged worst case drive cycles and some

compromises have been made to reduce UC size, the algorithm can still tolerate harsh driving

cycles as shown in the simulation results later. The proposed target UC SOC band will also

be compared with other EMS works in the simulations.

4.3 Power Management Strategy Pt. 2: Implementation

The concept of the target UC energy (or SOC) level has been explained in the previous

section. Here, the implementation of the PMS is discussed. To recap, the PMS decides how

the power flow should be split between the battery and UC. It has two goals,

Page 107: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

86

To ensure the actual UC energy level follows the target UC energy level

To ensure the battery power limits are not exceeded

The PMS prioritises the second goal. That means it will ensure the UC follows the target UC

energy level normally, but if the self-imposed battery limits are exceeded, it will override

following the target UC energy level and get the battery to be within limits. As the target UC

energy level has been designed for averaged worst case scenarios in section 4.2, there will be

instances such as during hard acceleration or hard braking where the target UC voltage is

insufficient to keep the battery within its limits. Therefore, the second goal is used to always

keep the battery within the limits.

With the knowledge from the previous sections, the PMS is expressed mathematically here.

The inputs to the algorithm are target UC energy Euc,tar(j-1), actual UC energy level Euc(j-1) at

previous instant, power required for drive cycle Pdr+aux(j-1), and actual battery power Pbatt(j-

1). The outputs are battery power Pbatt(j), high-side UC power Puc,H(j).

Figure 4-14 Block diagram of PMS algorithm.

The PMS implementation algorithm is summarised in Figure 4-14. In earlier equation (4-22),

the target UC energy level was stated in terms of SOC. The corresponding energy levels is

calculated by,

𝐸𝑢𝑐,𝑡𝑎𝑟 =

1

2𝐶(𝑆𝑂𝐶𝑢𝑐,𝑡𝑎𝑟 𝑉𝑢𝑐,𝑚𝑎𝑥)

2

(4-23)

Puc,ocv,1(j) UC power

required

(4-24)

Battery power

restriction

(4-25) to (4-33)

System Euc,tar(j) Euc(j) Puc,H(j) +

_

Euc(j-1)

Page 108: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

87

In order to bring Euc(j) to the desired level Euc,tar(j), power should be channelled into the UC.

This required low-side UC power is calculated by,

𝑃𝑢𝑐,𝑜𝑐𝑣,1(𝑗) = [𝐸𝑢𝑐,𝑡𝑎𝑟(𝑗 − 1) − 𝐸𝑢𝑐(𝑗 − 1)]/𝑑𝑡 (4-24)

Then the output faces a restriction block which enforces the battery power limits. First, the

DC/DC converter losses are accounted for to get the high-side UC power,

𝑃𝑢𝑐,𝐻,1(𝑗) =

𝜂𝐷𝐶/𝐷𝐶,𝑏𝑜(𝑗) 𝑃𝑢𝑐,𝑜𝑐𝑣,1(𝑗) 𝑖𝑓 𝑃𝑢𝑐(𝑗) ≥ 0

𝑃𝑢𝑐,𝑜𝑐𝑣,1(𝑗)

𝜂𝐷𝐶/𝐷𝐶,𝑏𝑢(𝑗) 𝑖𝑓 𝑃𝑢𝑐(𝑗) < 0

(4-25)

Second, the battery power Pbatt,1(j) that ensures the target UC energy level is perfectly met is

determined by,

𝑃𝑏𝑎𝑡𝑡,1(𝑗) = 𝑃𝑑𝑟+𝑎𝑢𝑥(𝑗) − 𝑃𝑢𝑐,𝐻,1(𝑗) (4-26)

Third, the battery power is brought within the limits. Similar to equations (4-12) and (4-13)

earlier, it is enforced by,

𝑃𝑏𝑎𝑡𝑡(𝑗) =

min(𝑃𝑏𝑎𝑡𝑡,1(𝑗), 𝑃𝑏𝑎𝑡𝑡,𝑚𝑎𝑥(𝑗)) 𝑖𝑓 𝑃𝑏𝑎𝑡𝑡,1 ≥ 0

max(𝑃𝑏𝑎𝑡𝑡,1(𝑗), 𝑃𝑏𝑎𝑡𝑡,𝑚𝑖𝑛(𝑗)) 𝑖𝑓 𝑃𝑏𝑎𝑡𝑡,1 < 0

(4-27)

Fourth, the UC power Puc,H(j) is recalculated after the battery restriction is enforced,

𝑃𝑢𝑐,𝐻(𝑗) = 𝑃𝑑𝑟+𝑎𝑢𝑥(𝑗) − 𝑃𝑏𝑎𝑡𝑡(𝑗) (4-28)

Now, the desired UC and battery powers have been determined.

Fifth, two exceptions to the rule are considered. If the UC is fully charged, yet regenerative

braking is required (e.g. driving downhill or under a harsh driving cycle), then the

regenerative energy will be charged to the battery only, which may exceed the battery limits.

Therefore, priority is placed in not wasting energy (short term goals) over battery cycle life

(long term goals). This is expressed mathematically by,

Page 109: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

88

𝑃𝑢𝑐,𝐻(𝑗) =

0 𝑖𝑓 (𝑃𝑢𝑐,𝐻(𝑗) ≤ 0 𝑎𝑛𝑑 𝑆𝑂𝐶𝑢𝑐(𝑗) ≥ 1)

𝑃𝑢𝑐,𝐻(𝑗) 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

(4-29)

𝑃𝑏𝑎𝑡𝑡(𝑗) = 𝑃𝑑𝑟+𝑎𝑢𝑥(𝑗) − 𝑃𝑢𝑐,𝐻(𝑗) (4-30)

The second exception is if the battery is also fully charged (driving a long distance downhill),

then rheostatic braking will be used. The power generated from the motor will be dumped

into a braking chopper, so that the braking power of the motor is still utilized and not

switched off completely. This can be expressed mathematically as

𝑃𝑏𝑎𝑡𝑡,𝑜𝑛(𝑗) =

0 𝑖𝑓 (𝑃𝑏𝑎𝑡𝑡(𝑗) ≤ 0 𝑎𝑛𝑑 𝑆𝑂𝐶𝑏𝑎𝑡𝑡(𝑗) ≥ 1)1 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

(4-31)

𝑃𝑏𝑎𝑡𝑡(𝑗) = 𝑃𝑏𝑎𝑡𝑡,𝑜𝑛 𝑃𝑏𝑎𝑡𝑡(𝑗) (4-32)

𝑃𝑏𝑐(𝑗) =

𝑃𝑏𝑎𝑡𝑡(𝑗) 𝑖𝑓 𝑃𝑏𝑎𝑡𝑡,𝑜𝑛 = 0

0 𝑖𝑓 𝑃𝑏𝑎𝑡𝑡,𝑜𝑛 = 1

(4-33)

Now, the power split has been computed, namely, battery power Pbatt(j), high-side UC power

Puc,H(j), and braking chopper power Pbc(j), which are the outputs of the proposed power

management algorithm. Examples will be shown in the simulations section later.

4.4 Summary

To summarise this chapter, the proposed HESS management strategy which comprises the

EMS and PMS has been explained.

The EMS computes the target UC energy band and has two goals – to ensure sufficient UC

space to absorb energy during future regenerative braking, and to ensure sufficient UC energy

for future accelerations. Comprehensive calculations and justifications based on averaged

worst case scenarios have been provided to explain how this target UC energy band was

developed, which is not seen in existing works, and is one of the contributions of this thesis.

As the calculations have already considered the worst case scenarios, knowledge of the future

Page 110: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

89

drive profile is not required, which allows the proposed strategy to work on any route,

including new ones.

The PMS decides the power split between the battery and UC and also has two goals, which

are to ensure the actual UC energy level follows the target UC energy level, and to ensure the

battery power limits are not exceeded, with priority placed in the latter.

The speed-dependent self-imposed battery power limit is the second contribution of this

thesis, which was calculated based on the power required by the EV to maintain a constant

speed. It achieves two functions – to utilise the UC even during low power demands to

reduce battery use, and to allow the battery to supply power during steady state. Later, in the

simulations, it will be clear how the two goals are achieved.

In short, a framework for designing the HESS has been provided. This framework includes

the appropriate sizing of the UC to reduce costs. In this case, six UC modules have been used.

As mentioned earlier, the EMS and PMS were not designed linearly. The EMS was designed

first, followed by the PMS, and the EMS was revised again, and so on. For example, only

five UC modules were used in the original PMS design. This was found to be insufficient for

the EMS. So the number of UC modules was revised to six, and the simulations were re-run

to include the additional energy, voltage, weight, etc. Also, further revisions were performed

after the experiments in Chapter 6 were run, for example the maximum UC SOC was limited

to 98.8% in the experiments for safety reasons. So the maximum UC SOC in the simulation

was revised to 98.8% to standardise the conditions.

Unless stated otherwise, all the plots and figures from this chapter were from the final

simulation configuration.

Before the simulations are presented, the author has some final remarks about the proposed

HESS management strategy. Although a mid-sized EV sedan based on the Nissan Leaf was

Page 111: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

90

used to develop the HESS management strategy, the strategy is not limited to a sedan. The

parameters for modelling in section 3.11 can be updated to reflect non-sedans, such as

electric buses or electric goods vehicles.

As this work provides a framework for HESS designing, the procedures mentioned in this

chapter for creating the EMS and PMS can be repeated for the new vehicle. For example, the

minimum braking torque for an electric bus may not be τbr,tar = 244Nm, so the target UC

energy band will look different but will suit the electric bus specifications.

Page 112: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

91

5 HESS SIMULATIONS

In this chapter, simulation results of the HESS are presented. There are four simulations. The

first simulation compares the proposed EMS to that of other works, and shows why the

proposed EMS is better. The second simulation compares the proposed PMS to that of other

works. Similarly, it shows how the proposed PMS is better than other rule-based

deterministic strategies.

The third simulation involves running the mid-sized EV over drive cycles using the proposed

HESS management strategy to show that it can achieve all the desired goals. The fourth

simulation is a battery cycle life simulation to show the improvement of battery cycle life of

the proposed battery/UC HESS setup as compared to a battery-only setup.

5.1 Implementation

Before discussing the simulation results, first an overview of the simulation setup is shown.

The models and algorithms discussed earlier have been integrated to form Figure 5-1, which

shows the complete backward approach simulation, with separate blocks for EMS and PMS

algorithms, models, and drive cycle.

One data point of the drive cycle (v(j) and θ(j)) is fed into the vehicle model to calculate the

total power required for that data point. The EMS generates the target UC energy level based

on the vehicle speed to meet the two goals of sufficient energy for acceleration and sufficient

space for regenerative braking. Subsequently, these data are fed into the PMS, which decides

on the power split between the battery and UC. It performs the two goals of following the

target UC energy level and limiting battery power, with priority in the latter. Once the power

split has been determined, the battery and UC models are updated to reflect the loss or gain in

energy for that data point j. Then this cycle is repeated for the subsequent drive cycle j data

points.

Page 113: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

92

Figure 5-1 Complete simulation block diagram.

In the experiments later in Chapter 6, a similar approach is used, except that the electrical

models have been replaced with the actual hardware, as can be seen in the ‘output to’ and

‘input from’ hardware boxes in Figure 5-1.

Input from

Simulation

Model/

Hardware

SOCbatt(j-1)

SOCuc(j-1)

Pbatt(j-1)

DC/DC Converter

Model

(Figure 3-24)

Puc,ocv(j)

SOCbatt(j)

SOCuc(j) UC Model

(Figure 3-17)

Battery Model

(Figure 3-15)

Electrical Models

Euc,regen,tar(j) Acc./Regen.

UC Energy

(Figure 4-4)

Loop for k

Euc,acc,tar(j)

Target UC

Energy

(4-22),

(4-23)

Euc,tar(j)

EMS

Loop for j

PMS

(Figure 4-14)

Output to

Hardware

Pbatt(j)

Puc,H(j)

v(j)

θ(j) Vehicle

Model

(Figure 3-9)

Loop for j

ωm

Pdr+aux

Drive

Cycle

Pbc(j)

Page 114: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

93

5.2 EMS Comparison: Target UC Energy Band

In this section, the proposed target UC energy band is compared to other works and shown to

be better.

Figure 5-2 Target UC SOC band vs. speed, 6 UC modules.

Earlier in Figure 4-10, the proposed target UC SOC band against speed curve was presented.

Figure 5-2 shows the target UC SOC band again, where the blue line represents the upper

limit and the red line the lower limit. Comparisons are made with other works, where the

EMS algorithm of others are applied onto this work’s mid-sized EV. As mentioned earlier in

the literature review, there are a few types of EMS algorithms, such as a constant target UC

energy level or a speed-dependent target UC energy level. The exact value of the target UC

energy level differs between different works.

For example, Avelino, et al. use a constant 87.5% target SOC level [37] coloured green in

Figure 5-2. It exceeds the upper limit at speeds above 55km/h, meaning the UC has

insufficient space if the worst case τbr,tar = 244Nm of regenerative braking is applied at those

Above band

Below band

Page 115: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

94

speeds. A similar case can be seen for Torreglosa’s constant 75% target SOC [22] in light

blue, which falls under the lower limit below 70km/h, meaning the UC has insufficient

energy for an acceleration of τacc ratio = 0.5 at those speeds.

Choi, et al. use a speed-dependent target UC energy curve [45]. It is coloured orange and

rises above the upper limit at speeds above 45km/h. Below 45km/h, Choi’s curve stays within

or close to the band. However, their method of computation is different from the proposed

work, with the proposed method being more realistic. Choi uses the maximum deceleration

rate to anticipate energy recovered from future regenerative braking. However, the maximum

deceleration rate does not correspond to the worst case energy recovered. This is because

mechanical friction brakes activate at high deceleration rates, dissipating energy as explained

earlier in section 4.2.3, where the worst case scenarios was discussed. In addition, they

charge all regenerative power to the UC. In the proposed algorithm, a constant deceleration

brake torque τbr,tar which recovers the most energy is used, and that energy is charged to both

the battery and UC, reducing required UC size.

It turns out that these two design decisions by Choi reduce the effect of each other. As the

worst case energy was not used, less UC space is required as compared to this work. But

since they charge all regenerative power to the UC, more UC space is required as compared

to this work. This results in Choi’s curve being quite close to the upper limit of the proposed

target UC energy band.

Carter, et al. also use a speed-dependent target UC energy curve [48], coloured purple (Their

design was based on the works of [47]). It falls below and rises above the band at different

points. As discussed earlier in section 2.3.1.3 their curve was computed based on the

assumption that all kinetic energy is recovered to the UC. Similar to the Choi case, this is not

true as not all kinetic energy can be recovered, as some will be dissipated via the mechanical

brakes and some are lost as electrical losses or mechanical losses. In contrast, the proposed

Page 116: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

95

EMS algorithm in this work uses a far more rigorous method of computation, which

considers the worst case (averaged) scenarios and realistic drive cycles.

Although not shown in Figure 5-2, the constant target UC band of Cao & Emadi [32] suffers

from the same problem as the constant target UC level designs, where the two goals cannot

be satisfied at all times as it is speed-independent. The UC in their work is charged when the

bottom of the band is reached, and stops charging at the top of the band, which is not

dependent on speed.

As shown in Figure 5-3, if the UC energy storage size was increased by 2.2 times to 13

modules, the height of the band would increase and a constant target UC SOC of 76.5% can

be fitted within the band at all times. Only then would the constant UC SOC strategy meet the

two goals. But this would be an expensive and inconvenient solution due to the UC cost and

size.

Figure 5-3 Target UC SOC band vs. speed, 13 UC modules.

Page 117: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

96

5.3 PMS Comparison: Battery Power Limits

In this section, the proposed speed-dependent battery limit is compared to existing constant

battery limit works and is shown to be better.

To recall, the proposed speed-dependent battery power limit allows two goals, to utilise the

UC even during low power demands to reduce battery use, and to allow the battery to supply

power during steady state. Other deterministic rule-based designs discussed earlier in

literature review section 2.3.2.1, for example [22] [25] [32] [47], have a fixed (non-speed-

dependent) battery limit, and the battery limits differ between different works. In those cases,

the UC only activates when the battery exceeds a certain threshold. For example, in Cao &

Emadi’s work [32] where a simulation was performed, the battery was limited to 12kW

(0.39C), and the UC handles the excessive power beyond 12kW. While in Thounthong, et

al’s work [25] where both simulation and experiment were performed, the discharge battery

current was limited to 20A (0.29C), and the UC handles the excessive power beyond that.

Figure 5-4 Speed-dependent PMS battery limit curve comparison.

Earlier in Figure 4-2 in PMS section 4.1.2, the final proposed speed-dependent battery limit

was presented. Here in Figure 5-4, the limits are presented again, but only considering the

discharge power (positive, upper half only) as it is sufficient to show this comparison. It also

shows a constant 0.39C threshold from Cao & Emadi [32], and a 1.7C threshold at the other

Page 118: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

97

extreme for comparison purposes. Other works have a threshold which lies between these two

extremes.

If the 0.39C (9.3kW for this work’s mid-sized EV) threshold was used, the battery would not

be able to supply enough power for steady state at speeds above 45km/h. For example, from

earlier Figure 4-1, 20kW is required to maintain a constant 100km/h, and 9.3kW is far from

enough. Since the UC would need to constantly supply 20 – 9.3 = 10.7kW of power at steady

state, it would be drained of energy quickly.

On the other hand, if the 1.7C (40.6kW) threshold was used, the UC would be underutilized

during low power demands. As long as the battery usage does not exceed 40.6kW, the UC

will not be used. However, in the proposed algorithm, the UC is used even at low power

demands to reduce the battery usage to extend the battery cycle life. For example, at 20kmh,

the proposed algorithm uses the UC once the power demanded exceeds 9.6kW, as compared

to 40.6kW for the 1.7C threshold.

5.4 Drive Cycles

In this section, the mid-sized EV was simulated driving over three drive cycles – LA92,

EUDC and FTP-75 city. The two EMS goals of sufficient space and energy in the UC, and

the two PMS goals of enforcing the EMS and enforcing the battery limits are shown to be

met. Also, the two goals of the proposed speed-dependent battery limit are shown to be

achieved.

5.4.1 LA92 Drive Cycle

Earlier in EMS section 4.2.3, where the algorithm brake torque and acceleration torques were

selected, there were some compromises. For example, if the driver coincidentally applies the

corresponding worst case braking torque at speeds above 60 km/h to stop the car, there is

insufficient space in the UC and it will not be able to absorb the energy completely.

Page 119: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

98

Alternatively, if the driver accelerates harsher than an acceleration torque ratio of 0.5, he may

not have sufficient energy to get to highway speeds.

Therefore, an aggressive LA92 drive cycle (Figure 5-5) has been used to demonstrate that

although there are some compromises in the algorithm design, it can still tolerate an

aggressive drive cycle as the aggressive parts are usually short and not continuous.

Figure 5-5 LA92 drive cycle power required.

Figure 5-6 LA92 torque profile for the mid-sized EV.

Figure 5-6 shows the torque profile for the mid-sized EV running the LA92 drive cycle.

Some parts of the required torque (blue) are more than 0.5 of the motor limit (yellow). This

Page 120: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

99

means that the car is accelerating with more than the designed acceleration torque ratio τacc

ratio = 0.5, although just for short periods. Figure 5-7 shows the corresponding power profile

and the power split between the battery and UC using the proposed HESS management

strategy, and Figure 5-8 shows the SOC values of the UC and the UC band.

Figure 5-7 LA92 power distribution of battery and UC.

Figure 5-8 LA92 UC SOC (target and actual) and battery SOC.

Page 121: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

100

As the LA92 is a transient drive cycle, it is difficult to read the plots, so a zoomed-in section

from 820s to 920s with high torque and power demand is presented in the following figures.

Figure 5-9 LA92 torque profile for the mid-sized EV, zoomed 820-920s.

Figure 5-9 shows the zoomed-in torque profile, where the required torque exceeds the

selected acceleration torque ratio τacc ratio = 0.5 at approximately 855s.

Figure 5-10 LA92 power distribution of battery and UC, zoomed 820-920s.

Page 122: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

101

Figure 5-11 LA92 UC and battery SOC, zoomed 820-920s.

Figure 5-10 shows the corresponding power profile Pdr+aux and the power split between the

battery and UC using the proposed HESS management strategy. As the power demanded is

above the self-imposed battery limits (Pbatt,max), the battery power Pbatt is clipped to its

maximum Pbatt,max, with the excess power being handled by the UC (Puc,ocv for DC/DC

converter low-side power and Puc,H for DC/DC converter high-side power). This is one of the

goals of the PMS – to restrict battery current to its limits.

Figure 5-11 shows the target UC SOC band (defined by SOCuc,regen,tar and SOCuc,acc,tar), the

target UC SOC level SOCuc,tar and the actual UC SOC SOCuc, which stays within the band

most of the time. This is another goal of the PMS – to enforce the EMS by following the

target UC energy level. It also shows the actual battery SOC SOCbatt.

At approximately 855 seconds, the red SOCuc curve deviates from the target SOC SOCuc,tar in

Figure 5-11. This is because the battery limit Pbatt,max is exceeded in Figure 5-10. Since the

algorithm prioritises preventing the battery limits from being exceeded, the UC supplies the

extra power initially, resulting in its SOC dipping below the band. Once power demand is not

Page 123: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

102

as heavy, the UC SOC is recovered to the target SOC level as seen at approximately 875s

where the UC is charged (power is negative).

From these figures, the two PMS goals were shown to be met, which are to follow the target

UC energy level and to enforce battery limits, with priority placed in the latter. By

successfully following the target UC energy level, the UC would always have sufficient space

for future braking and sufficient energy for future acceleration, meeting the two EMS goals.

In addition, the aggressive LA92 drive cycle has shown the proposed HESS management

strategy can tolerate heavy demands (τacc ratio > 0.5) despite not being designed specifically for

the absolute worst case scenarios as the heavy demands are mostly short. Although not shown

here, heavy regenerative braking behaves similar to the heavy acceleration, just that the UC

SOC would rise above the target UC band to capture more energy instead of below the target

UC band in this heavy acceleration scenario.

It must be noted that in the event the driver attempts to do the absolute worst case

acceleration, that is to floor the vehicle from zero to full speed in a single sustained run, it

does not mean that the EV is unable to meet the power demand. When the UC is emptied

midway through the acceleration, the algorithm takes the power from the battery instead,

worsening its cycle life, but still satisfying the required power demand. This is a reasonable

compromise to reduce the UC size. In a conventional petrol-powered vehicle, it is understood

that flooring the car constantly leads to high fuel consumption and high wear and tear, so it

should be avoided. Similarly, for a battery/UC HESS EV, it would be accepted that

constantly flooring the car would lead to a shorter battery cycle life.

Lastly, another point to note is that future drive profile knowledge was not required. The

proposed HESS management strategy had no information on the next instance of the drive

cycle. It responded based on present data only.

Page 124: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

103

5.4.2 EUDC

In this section, the simulation is run with the Extra-Urban Driving Cycle (EUDC). This drive

cycle is used to demonstrate the proposed speed-dependent battery limit.

Figure 5-12 EUDC power distribution of battery and UC.

Figure 5-13 EUDC UC and battery SOC.

Page 125: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

104

Figure 5-12 shows the corresponding EUDC power profile Pdr+aux and the power split

between the battery and UC using the proposed HESS management strategy. Figure 5-13

shows the various SOC values.

Earlier in section 4.1, it was mentioned that the speed-dependent battery power limit achieves

two functions – to utilise the UC even during low power demands to reduce battery use, and

to allow the battery to supply power during steady state.

As the EUDC is a modal cycle with constant speed or acceleration sections, these two goals

can be seen clearly. From 250s to 290s in Figure 5-12, the battery limits are not exceeded as

Pdr+aux < Pbatt,max. However, the UC is still utilised (Puc,H > 0), which satisfies the first goal of

using the UC even during low demands to reduce battery usage.

From 290s to 320s in Figure 5-12, a constant speed of 100km/h is maintained (see Figure 3-6

for speed profile), and it is observed that the battery is supplying all the power at steady state,

satisfying the second goal of the speed-dependent battery limit.

5.4.3 FTP-75 City Drive Cycle

The four goals of the proposed HESS management strategy have already been demonstrated

in the LA92 drive cycle above. The LA92 drive cycle has also shown that the proposed

strategy is able to tolerate aggressive drive cycles despite not being designed for it.

Also, the EUDC has shown the two goals of the proposed speed-dependent battery limits.

Here, the FTP-75 city drive cycle is run for reference only, as the FTP-75 city drive cycle is

used for the cycle life tests in the experiments. Figure 5-14 shows the corresponding power

profile and the power split between the battery and UC using the proposed HESS

management strategy. Figure 5-15 shows the various SOC values.

Page 126: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

105

Figure 5-14 FTP-75 city power distribution of battery and UC.

Figure 5-15 FTP-75 city UC and battery SOC.

As the FTP-75 city is a transient drive cycle, a zoomed in section between 150s and 250s is

shown in the following two figures, where there is high power demand.

Page 127: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

106

Figure 5-16 FTP-75 city power distribution of battery and UC, zoomed 150-250s.

Figure 5-17 FTP-75 city UC and battery SOC, zoomed 150-250s.

Similar to the earlier LA92 drive cycle simulation, Figure 5-16 shows the battery power Pbatt

clipped to its maximum Pbatt,max around the 200s mark, with the excess power being handled

by the UC.

Page 128: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

107

In Figure 5-17, the red SOCuc curve deviates from the target SOC SOCuc,tar near the 200s

mark as the battery limit Pbatt,max is exceeded. Since the algorithm prioritises preventing the

battery limits from being exceeded, the UC supplies the extra power initially, resulting in its

SOC dipping below the target, but remains within the band. A smaller deviation is seen here

as compared to the earlier aggressive LA92 case, as the algorithm was designed with the

worst case scenario of the FTP-75 city drive cycle. Therefore, the deviation remains within

the band.

5.5 Battery Cycle Life

5.5.1 Description

Ultimately, the goal of a battery/UC HESS is to extend battery cycle life. So in this section,

the battery cycle life of the battery/UC HESS system is compared to that of a battery-only

system.

Earlier in section 3.10, the battery cycle life model was presented. In section 3.11.2,

configuring Sanyo UR18650W batteries to a 98S44P formation to approximate a Nissan Leaf

battery in terms of energy and nominal voltage was discussed. This allows the battery cycle

life model and the curve-fitted parameters to be used directly from Wang, et al [7].

From the drive cycle simulations in section 5.4 earlier, the battery currents were shown. The

battery currents are the required input to the battery cycle life model.

5.5.2 Drive Cycle Comparison

First, the best approach to represent a daily driving scenario is decided. In Singapore, a drive

from the suburbs (e.g. Jurong) to the city centre (e.g. Orchard) and back is approximately

50km. This represents a daily driving scenario of a person travelling between home and his

workplace. So 50km is set as the target distance to be driven. Table 5-1 shows the distances

Page 129: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

108

of the common drive cycles and the number of such repeated cycles to get as close as

possible to the 50km target.

Table 5-1 Comparison of no. of drive cycles to hit 50km.

FTP-75 City FTP-75 HWFET LA92 NEDC

Distance (km) 17.77 16.51 15.80 10.93

Time (min) 31.23 12.75 24.08 19.67

No. of cycles to hit

50km per day

3 3 3 5

Total distance (km) 53.31 49.52 47.40 54.65

5.5.3 Simulation

Next, the drive cycle simulations for each of the drive cycles mentioned in Table 5-1 is run,

and the battery current data is logged. At the end of each day, the battery is charged back to

its starting SOC of 80% at 0.5C and this data is appended to the original battery current data.

For example, a set of battery current data for the FTP-75 city drive cycle consists of three

consecutive FTP-75 city cycles, followed by charging the battery back to 80% SOC in

preparation for the next day’s driving.

This battery current data was divided by 44 as each parallel branch of the 98S44P battery

pack configuration is only subject to 1/44th the total current. Then it was fed into the battery

life model for 10 years of driving. The ambient temperature was set at 50oC to represent a

relatively hot battery compartment with frequent usage under the sun.

Figure 5-18 shows an example of the capacity loss curve for the case of three FTP-75 city

drive cycles per day over 10 years for the proposed battery/UC system.

Page 130: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

109

Figure 5-18 Battery capacity loss curve for battery/UC system over FTP-75 city.

From the chart, the capacity loss due to cycling is 17.9% and the capacity loss due to calendar

life is 98.5%, resulting in a total capacity loss of 116.5% over 10 years, i.e. the battery is

more than dead. A quick verification was performed to see if the calendar life model was

implemented correctly.

In Wang, et al’s work, the Sanyo UR18650W battery cycled at 46oC (the closest condition to

this work) over 1 year lost 30% of capacity due to calendar losses, and was extrapolated to

lose 40% over 2 years. Re-examining this work’s simulation in Figure 5-18, losing 98.5%

capacity in terms of calendar life over 10 years is a reasonable output of the model.

The reason for such large losses may be because the tested Sanyo UR18650W batteries are

from 2007, which is older technology. Another reason is the selected temperature of 50oC. If

the temperature was reduced to 20oC, the calendar losses would be halved.

Since the battery/UC system and battery-only system are subject to the same 10 years of

aging, the capacity loss due to calendar life would be the same for both cases (neglecting

minor temperature differences). In addition, calendar life loss is not within the algorithm’s

control – it is only dependent on time. Because of these reasons, the calendar loss is excluded

from the results. Only the capacity loss due to cycling is considered, which is controllable by

the proposed HESS algorithm.

Page 131: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

110

Table 5-2 Simulated cycle life capacity losses over 10 years.

FTP-75

City x3

FTP-75

HWFET x3

LA92

x3

NEDC

x5

Battery/UC Q-loss (%) 17.9 19.7 21.7 21.4

Avg. of absolute C-rate 0.25 0.46 0.36 0.27

Ah Throughput (Ah) 2735 2964 3226 3197

Battery-only Q-loss (%) 25.5 21.4 29.5 25.5

Avg. of absolute C-rate 0.35 0.50 0.45 0.32

Ah Throughput (Ah) 3770 3178 4232 3770

% Reduction in Q-loss 29.5 7.7 26.4 16.3

Table 5-2 shows the results, where the battery cycle life is quantified by the capacity loss

over 10 years. The best reduction of capacity loss occurs for the FTP-75 city drive cycle,

where the battery in the battery/UC system loses only 17.9% of its capacity due to cycling,

while the battery-only system loses 25.5% of its capacity. This is a 29.5% improvement when

using the proposed HESS algorithm.

It is observed the proposed HESS algorithm works best for city drive cycles, where there are

many transients and start-stop driving, such as the FTP-75 city and LA92 drive cycles. The

FTP-75 HWFET is a highway driving cycle. If the car is maintained at high speeds at steady

state, then there is not much utilization of the UC as the battery supplies the steady state

power. Similarly, the NEDC is a modal drive cycle, which is very ideal with smooth

accelerations and much constant speed driving, so there is less utilization of the UC as well.

Therefore, the proposed HESS system works best for city driving.

Page 132: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

111

Table 5-3 Simulated cycle life for 20% capacity losses.

FTP-75

City x3

FTP-75

HWFET x3

LA92

x3

NEDC

x5

Battery/UC Days 4070 3705 3364 3420

Ah Throughput (Ah) 3048 3008 2971 2995

Battery-only Days 2868 3419 2474 2860

Ah Throughput (Ah) 2961 2976 2867 2953

% Battery Cycle Life Extension 41.9% 8.4% 36.0% 19.6%

Table 5-3 offers another perspective, where the battery cycle life is quantified by the number

of days until 20% capacity loss due to cycling. Again, the longest battery cycle life extension

occurs for the FTP-75 city drive cycle, where the battery in the battery/UC system lasts 4070

days (11.1 years) until 20% capacity loss is experienced, while the battery-only system lasts

2868 days (7.9 years). This is a 41.9% extension of battery cycle life when considering only

cycling losses.

Note that if calendar life losses are also included, the improvements would be reduced

because the battery/UC system lasts longer, therefore experiencing more calendar losses also.

Table 5-4 shows such a scenario, which counts the number of days until 80% total loss (cycle

+ calendar) is experienced. It has been extended to 80% to allow for a longer cycling time for

a better comparison so that calendar life losses are less dominating.

As the calendar life of the Sanyo UR18650W is poor, the best battery cycle life improvement

(FTP-75 city case) is now 8.5% (over 5 years). In future works section 8.2.3, it is suggested

to extract model parameters from more modern batteries, which have better calendar life.

Page 133: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

112

Table 5-4 Simulated cycle life for 80% capacity losses, including calendar loss.

FTP-75

City x3

FTP-75

HWFET x3

LA92

x3

NEDC

x5

Battery/UC Days 1883 1846 1805 1812

Ah Throughput (Ah) 1410 1498 1594 1586

Battery-only Days 1735 1812 1666 1734

Ah Throughput (Ah) 1791 1577 1931 1790

% Battery Cycle Life Extension 8.5% 1.9% 8.3% 4.5%

5.5.4 Drive Cycle Selection for Experiment 2

To summarise the simulation results, when considering cycling losses only, the FTP-75 city

drive cycle shows the greatest improvement to battery life for the battery/UC system as

compared to the battery-only system. In terms of capacity loss due to cycling over 10 years,

an almost 30% reduction was seen for the battery/UC system. In terms of time to reach 20%

capacity loss due to cycling, a 40% improvement was seen for the battery/UC system.

Therefore, the FTP-75 city drive cycle was selected for the battery cycle life experiments in

Experiment 2 later.

5.6 Summary

To summarise, from the first simulation on EMS comparison, it was demonstrated that if six

UC modules are used, only the proposed EMS achieves the two goals of sufficient energy for

acceleration and sufficient energy for braking. Other existing works cannot always achieve

the two UC goals simultaneously unless their UCs are sized up to twice as large, increasing

weight and costs.

Page 134: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

113

In the second simulation, the proposed PMS speed-dependent battery limit allows better UC

utilization and allows the battery to be the main energy provider during constant speed

driving. Again, it was shown that other rule-based deterministic PMS are unable to achieve

these two goals simultaneously.

In the third simulation, the mid-sized EV equipped with the proposed HESS management

strategy was driven over various drive cycles.

The four goals of the proposed HESS management strategy were demonstrated in the LA92

drive cycle. The two EMS goals are to ensure sufficient space in the UC for future

regenerative braking and to ensure sufficient energy in the UC for future accelerations. The

two PMS goals are to ensure the target UC energy level (EMS) is followed, and to enforce

battery power limits, with priority on the latter. The LA92 drive cycle has also shown that the

proposed strategy is able to tolerate aggressive drive cycles despite not being designed for it.

Also, knowledge of the future drive profile was not required.

In addition, the EUDC demonstrated the two goals of the proposed speed-dependent battery

limit, which are UC utilization even during low power demands and that the battery should be

the main energy provider during constant speed driving.

Lastly, battery cycle life simulations were performed to observe the fall in battery capacity

for the proposed battery/UC HESS, and for the battery-only system. Almost 30% reduction in

capacity loss due to cycling was seen for the battery/UC HESS as compared to the battery-

only system when running the FTP-75 city drive cycle over 10 years.

Page 135: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

114

6 HESS EXPERIMENTS

As there was no budget to purchase an actual EV to implement the HESS, a reduced-scale

experiment was built instead. There are three distinct experiments, each with different

objectives. Experiment 0, a preliminary experiment, was used to investigate the efficiency of

the custom-built DC/DC converter. Experiment 1 was used to prove the proposed algorithm

worked as intended by running the algorithm over a drive cycle. Experiment 2 was used to

compare the battery cycle life for the battery/UC HESS configuration to a battery-only

configuration.

In this chapter, first the setup is described, then the scaling is discussed. Subsequently, each

of the three experiments are explained and analysed individually.

6.1 Setup

Figure 6-1 shows a detailed electrical diagram of the setup, while Figure 6-2 to Figure 6-4 are

photos of the setup. To simplify the setup, a programmable load replaced the EV motor and

inverter. This programmable load is a Maccor Series 4000 battery tester. Drive cycles in

terms of power were programmed into the Maccor tester, and it would sink or source power

accordingly. This machine has a maximum output of 1.6kW, versus 80kW of the mid-sized

EV. Therefore, a reduced-scale experiment has to be implemented. Scaling is discussed in a

subsequent section.

In general, the experiment setup was designed for a maximum of 20V and 30A. The Maccor

tester is limited to 20V, while current sensors above 30A are not as common and a substantial

increase in price.

Page 136: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

115

Figure 6-1 Electrical diagram of experiment setup.

Figure 6-2 Photo of experiment setup (front).

Fan for DC/DC

converter Switches

for relays

Power supply for aux.

equipment, e.g. sensors

Battery

Maccor

Thermistor

vuc

vbatt

to μC

to μC

to μC

to μC

to μC

Page 137: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

116

Figure 6-3 Photo of experiment setup (top).

Figure 6-4 Photo of experiment setup (side, auxiliary equipment).

UC

Battery

Ferrite

cores

DC/DC

converter Current

sensors

Ferrite

cores

Arduino

Voltage

sensor

filters

Current

sensor

filters

Power supply for aux.

equipment, e.g. sensors

Precision

voltage

reference

Ceramic

capacitors

Page 138: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

117

6.1.1 Energy Storage Components

The specifications for the reduced-scale battery and UC are listed in Table 6-1. Three

identical batteries were purchased, labelled 1 to 3. Battery 1 was used for preliminary testing

of the setup, and for running Experiment 1. Battery 2 was used for the battery-only

configuration and battery 3 for the battery/UC configuration in the Experiment 2 cycle life

tests.

Table 6-1 Energy storage components.

Component Model Specifications Quantity

Battery SuPower Li-Ion

18650 18V

7.8Ah, 5S3P, Li(NiCoMn)O2

14V (min) - 18.5V (nom) - 21V (max)

With built-in PCB protection

3

Ultracapacitor Tecate PowerBurst

PBD-58/16.2K

16.2V, 58F 1

6.1.2 DC/DC Converter

Initially, it was desired to purchase a controllable bidirectional DC/DC converter off the

shelf, as DC/DC converter development is not the highlight of this work. However, there are

no such DC/DC converters available in the market with the desired specifications at the

moment. Therefore, the bidirectional two-quadrant buck-boost DC/DC converter (Figure 6-5)

was self-made, powered with two MOSFETs and controlled by an Arduino microcontroller.

The microcontroller produces an 8-bit Pulse Width Modulation (PWM) signal, which

controls the two MOSFETs in a complementary manner. A dead time of 852ns was

programmed, where both MOSFETs are in off state to prevent a short between switching.

The 852ns value was calculated from the turn-on and turn-off times from the MOSFET

datasheet. The specifications for the DC/DC converter components are listed in Table 6-2.

Page 139: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

118

MOSFETs were used here instead of IGBTs in the full-scale simulations as MOSFETs are

better suited to low power applications such as this reduced-scale experiment, and IGBTs for

high power applications like an actual EV.

Figure 6-5 Photo of DC/DC converter.

Inductor

(part)

Electrolytic

Capacitor

(part)

MOSFET

driver

MOSFET &

heat sink

Page 140: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

119

Table 6-2 DC/DC converter components.

Component Model Quantity Specifications

Microcontroller Arduino MEGA 2560 1 Based on Atmel

ATmega2560

MOSFET Vishay IRFP048PBF N-channel,

enhancement mode

2 60V, 70A

MOSFET

driver

Cree CRD-001 2 -

Inductor Coilcraft AGP4233-333ME 2 in series 33µH, 24A

Capacitors

(Electrolytic)

TDK B41560A9158M000 2 100V, 1500µF

Panasonic 5306C1 1 500V, 330µF

Switching

capacitors

(Ceramic)

Vishay K102K15X7RH5UH5 2 50V, 1000pF

TDK FG28X7R1H334KRT06 2 50V, 0.33µF

TDK FK26X7R1H225K 2 50V, 2.2µF

TDK FK20X7R1H335K 2 50V, 3.3µF

Murata

RDEC71H106K3K1H03B

2 50V, 10µF

(Unknown models) 11 3.3pF, 10pF, 33pF, 68pF,

100pF, 220pF, 680pF,

3300pF, 6800pF, 0.01µF,

0.022µF, 0.047µF, 0.1µF,

0.22µF, 0.47µF, 1µF

Ferrite cores Fair-Rite 2675821502 4 200kHz – 30MHz, 75

material

Fair-Rite 2675102002 6 200kHz – 30MHz, 75

material

Fair-Rite 2631101902 6 1MHz – 300MHz, 31

material

Page 141: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

120

6.1.3 Sensors

There are two voltage sensors, one each for measuring the battery and UC voltages (vbatt and

vuc respectively). They are custom-built 1.8kΩ//6.8kΩ resistor voltage dividers, with output

over the 1.8kΩ resistor. The signal is fed into a 3-pole active Sallen-Key topology low-pass

filter, built from an operational amplifier (op-amp). The cutoff frequency is set at 40Hz. Then

this signal is sent to the Arduino microcontroller, and to a personal computer (PC) over

Universal Serial Bus (USB) where it is logged at 10Hz. Originally, the op-amp was only to be

used as a voltage follower due to the high output impedance of the voltage divider

(recommended maximum input impedance of Arduino ADC is 10kΩ), so it was

underutilised. Later, it was reconfigured as a Sallen-Key topology low-pass filter to fully

utilise the op-amp.

There are three Hall-effect current sensors, for measuring the current consumed by the drive

cycle idr (labelled ACS1 in Figure 6-1), UC current on the low-side of the DC/DC converter

iuc (ACS2), and battery current ibatt (ACS3). The sensor for the high-side UC current iuc,H

(ACS4) is disused, but it can easily be computed by,

𝑖𝑢𝑐,𝐻 = 𝑖𝑑𝑟 − 𝑖𝑏𝑎𝑡𝑡 (6-1)

The signals from the current sensors are fed into a single pole passive RC low-pass filter with

a 40Hz cutoff frequency. As the hall-effect sensor already contains a built-in op-amp to

amplify the signal, no voltage follower was necessary. Therefore, no Sallen-Key topology

was implemented here. Then this signal was fed into the Arduino just like the voltage sensor.

These sensors and filters are powered by two 230V AC to 9V DC isolation transformer

adapters, and then further regulated to the required voltages by LM78xx series linear

regulators. The components used for the sensors are listed in Table 6-3. These components

Page 142: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

121

were soldered onto stripboards seen in Figure 6-4. The sensor electrical connection diagrams

are found in Appendix C.

Table 6-3 Sensors and supporting components.

Component Model Quantity Specifications

Current sensor Allegro ACS712ELCTR-30A-T 3 Hall-effect, ±30A

Voltage sensor Custom-built voltage divider 2 1:4.78 step down.

Max input of 23.9V

Temperature sensor Vishay NTCLE400E3103H 1 NTC Thermistor

DC/DC converter for

op-amp

TI DCP020515DP 2 5V input, unregulated

±15V output,

isolated, 2W

Op-amp Intersil CA3140E 2 -

Capacitors for op-amp AVX TAP105K035SCS 8 Tantalum, 35V, 1 µF

Precision micropower

shunt voltage

reference

TI LM4040A50I 1 5V

6.1.3.1 Sensor Noise Problems

As it was a self-made DC/DC converter, initially there was a lot of switching noise generated

and observed in the sensor readings. In the voltage and current measurements, ringing would

appear at the DC/DC converter switching frequency of 31.3kHz, with the ringing frequency

peaking at about 60MHz.

This was solved with three methods. First, star grounding was implemented to avoid ground

loops. Following the suggestions of [72], the ground circuits of analogue and digital signals

were separated and each star grounded. Also, the ground of the power supplies to the sensors

Page 143: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

122

and filters were floated by using isolation transformer adapters to avoid having too many

connections to ground.

Second, numerous ceramic capacitors were added in parallel to the circuit to avoid

differential mode (DM) noise, following the suggestions of [73] and [74]. DM noise occurs

between the converter’s output and the return line. Ideally, ceramic capacitors with an

impedance null (self resonance) that is the same as the ringing frequency should be added.

The impedance null is explained as follows. The impedance of an ideal capacitor ZC

decreases as frequency ω increases, as indicated by ZC = (jωC)-1. However, a real and non-

ideal capacitor also has some internal resistance and some inductance, given by Z = (jωC)-1 +

R + (jωL). The impedance null occurs at frequency ωn given when the impedance Z is

minimum. Below ωn, the capacitance term dominates, while above ωn, the inductance term

dominates.

However, this impedance null frequency ωn is usually not specified in the manufacturer’s

data. It requires performing experiments, where the capacitor is swept over a frequency range

while the impedance is recorded. Due to the lack of relevant equipment, 21 ceramic

capacitors with capacitances from 3.3pF to 10µF (approximately 2 for each order, listed in

Table 6-2) were installed instead for a quick fix, with one set of 21 for the DC/DC converter

high-side and another set of 21 for the low-side. This is an overkill solution. If a minimum

component or cost design is required in future, some optimisation can be performed here.

The third solution to overcome switching noise was to install chokes, in the form of ferrite

cores, to overcome common mode (CM) noise. CM noise occurs in both output and return

lines. Following the suggestions of [73], ferrite cores were installed. Two types of ferrite

cores were used – 75 material for rejecting noise in the 200kHz to 30MHz range, and 31

material for rejecting noise in the 1MHz to 300MHz range (listed in Table 6-2). The output

and return lines were wound round the ferrite cores with a 1:1 turns ratio on both the high and

Page 144: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

123

low-sides of the DC/DC converter. Further ferrite cores were installed on the voltage sensor

signal output lines, in case the cores on the output and return lines were unable to reject the

noise. Again, this is an overkill solution, and optimisation can be performed to minimise

components and costs in the future.

6.1.3.2 Sensor Calibration

Firstly, before any experiment was performed, the current and voltage sensors were calibrated

to two multimeters. Two multimeters of different models (Fluke 115 & Fluke 19) were used

as they produced slightly different results, so the sensors were calibrated to the average of the

two multimeters.

For voltage sensor calibration, a Thurlby Thandar Instruments (TTI) EX354T bench DC

power supply was connected to a small resistive load (to stabilise the voltage), and the two

voltage sensors and two multimeters were connected in parallel to the load to measure the

voltage drop across the load. Figure 6-6 shows the multimeter voltage versus the 10-bit data

logged on the Arduino. Using Microsoft Excel, a least squares best fit line was created. The

best fit line equations (V1 & V2, shown in the figure) were implemented in Arduino.

Figure 6-6 Voltage sensor calibration.

Page 145: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

124

For current sensor calibration, the TTI EX354T power supply was connected to the high-side

of the custom-built DC/DC converter. The three hall-effect current sensors and the two

multimeters were connected in series with a small resistive load at the low-side of the DC/DC

converter. The purpose of using the DC/DC converter was to step-up the current output to

15A as the EX354T power supply could only give a maximum of 4A output. Ideally, it

should be calibrated to 30A, but the multimeters could only measure up to 10A, with

overload up to ~15A for a few seconds. Figure 6-7 shows the results. Again, a best fit line

was created for the three sensors and the equations (shown in the figure) were loaded into

Arduino.

Figure 6-7 Current sensor calibration.

6.1.4 Safety Components

In addition, there are various safety components in the setup. There are two relays, one for

the entire setup and one for the UC. To turn on these relays, a physical switch must be turned

on, and the Arduino must give a turn-on command.

Page 146: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

125

Fast acting 20A fuses were also installed in multiple locations. Although the components can

withstand 30A, a 20A fuse was used instead as the 20A fuse does not blow immediately at

20A. According to the manufacturer’s datasheet [75], 20A passing through 20A fuses would

take more than 100s to blow. The time to blow decreases exponentially with increasing

current.

These fuses were installed after programming errors for the Arduino caused a short in the

DC/DC converter, destroying the MOSFETs and current sensors during an early testing stage.

A 12V fan was also installed above the DC/DC converter MOSFETs, sucking air away from

the MOSFETs to ensure it remained cool.

Table 6-4 Safety & miscellaneous components.

Component Model Quantity Specifications

Safety relay TE Connectivity T9AS1D22-5 2 240VAC, 30A

Safety fuse Cooper Bussmann AGC-20-R 3 20A, 32V, fast acting

Fuse holder Cooper Bussmann HKP 3 30A

Optoisolators

for relay

Vishay 4N35 2 -

Transistors for

relay

2N222A 2 -

Main wiring Pro-power 12AWG multiple 600V, 41A

Linear

Regulators

Fairchild LM7805ACT, Fairchild

LM7810ACT, ST LM7812

multiple -

LEDs for visual

reference

HP HLMP1700, HLMP1719,

HLMP1790

5 -

Page 147: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

126

In addition, a thermistor was installed to monitor the temperature of the battery. If the battery

temperature exceeded 50oC, the Arduino would stop the experiment by switching off the

relays. These components are listed in Table 6-4 and earlier Table 6-3 (for the thermistor).

Furthermore, the UC voltage was monitored closely. If it exceeded 16V (98.8% SOC), the

Arduino would trigger the relays to disconnect the UC to prevent overcharging. 100% UC

SOC (16.2V) was not used to give a small buffer in case of any programming or sensor

errors.

6.2 Scaling

In this section, scaling is discussed. First, a reduced-scale mechanical model should be

created. A concept known as similitude was applied, common in fluid dynamic reduced-scale

models. The reduced-scale model is said to have similitude with the full-scale setup if both

share geometric similarity, kinematic similarity and dynamic similarity. Geometric similarity

occurs when the model is the same shape as the full-scale setup, only scaled. Kinematic

similarity occurs when fluid flow of both the reduced-scale model and full-scale setup

undergo similar motions. Dynamic similarity occurs when ratios of all forces acting on

corresponding fluid particles and boundary surfaces in both systems are constant.

Dimensional analysis was performed and the following scaling law was derived,

𝐿𝑚𝐿𝑎

=𝑣𝑎𝑣𝑚

=𝜏𝑚𝜏𝑎= √

𝜔𝑎𝜔𝑚

=𝑃𝑎𝑃𝑚

=𝑚𝑚

𝑚𝑎= 𝑘

(6-2)

and the meaning of each symbol is shown in Table 6-5. The full dimensional analysis

workings can be found in Appendix B.

Page 148: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

127

Table 6-5 Symbols and their meanings for scaling derivation.

Symbol Meaning Symbol Meaning

k Scaling ratio. >1 for reduced-

scale

P Power

τ Torque m Mass

v Velocity Xa a subscript for full-size ‘actual’

L Length Xm m subscript for reduced-scale

‘model’

ω Angular velocity

A Maccor Series 4000 battery tester was used to simulate drive cycles. This machine has a

maximum output of 1.6kW, versus the 80kW motor of the mid-size EV.

To scale down the power while achieving similitude, k should be larger than 1 in equation

(6-2) (k = Pa/Pm > 1). This means the reduced-scale model has lower power, velocity and

angular velocity, but has larger dimensions, mass and torque generated. The force remains

constant (the same) for both cases.

6.3 Scale Factor k

In this section, the scale factor k is chosen.

6.3.1 Selecting k Based on Drive Cycle

First, the drive cycles are examined to select an appropriate scale factor. As the FTP-75 city

drive cycle showed the greatest improvement for the battery/UC system in the cycle life

simulations in section 5.5, the FTP-75 city drive was selected for the cycle life tests in

Experiment 2. The FTP-75 city drive cycle peaks at 42kW as shown in earlier Figure 5-14.

Page 149: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

128

However, the most aggressive official drive cycle is the LA92 drive cycle, which peaks at

60kW as shown earlier in Figure 5-5. The more aggressive LA92 is used here for scaling in

case it needs to be run for some reason in future. Since the circuit components can only

handle 30A, while the Maccor can only manage 20V, the maximum power which the

experiment can face is (30A)(20V) = 600W. This leads to k > Pa/Pm = 60000/600 = 100, i.e. k

cannot be smaller than 100, otherwise the components cannot handle the required power.

6.3.2 Selecting k Based on Simulation/Experiment Energy Ratio

Due to a limited budget, the UC was donated from a previous project in the university. The

battery was bought quite early while budget was still available, before the algorithm was

finalised. Therefore, the energy ratio of the actual battery and UC are different from what was

used in the HESS management strategy in the full-scale simulations.

Table 6-6 shows the energy specifications of the battery and UC used in the simulation and

the actual battery and UC for the experiments. The battery energy/UC energy ratio for the

simulation is 151, while that for the experiment is 68. So for the experiment setup, it can be

said either the battery is undersized, or the UC is oversized. For the former case, where the

battery is undersized in terms of energy and the UC is correctly sized, it means the battery

would be become flat faster, and given a specified fixed current, the C-rate would be higher

and the battery would be stressed more. This case would result in the scale factor k = 75 as

seen by the full-scale simulation vs. reduced-scale experiment UC energy ratio in Table 6-6.

While in the latter case, the battery is correctly sized, while the UC is oversized. This means

it is not necessary to use the entire UC SOC range. This case would result in the scale factor k

= 166.

Therefore, k should be picked in the range 75 < k < 166. However, k > 100 as limited by the

components. The appropriate values of k become 100 < k < 166.

Page 150: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

129

Table 6-6 Battery and UC energy specifications.

Full-scale simulation

specifications

Reduced-scale experiment

specifications

Battery

Capacity (Ah) 66 7.8

Nominal voltage (V) 362.6 18.5

Energy (kJ) (66)(362.6)(3600/1000) =

86153

(7.8)(18.5)(3600/1000) = 519

UC

Max voltage (V) 288 16.2

Capacitance (F) 13.8 58

Energy (kJ) (0.5)(13.8)(288)2 = 572 (0.5)(58)(16.2)2 = 7.61

Overall

Battery energy/UC energy

ratio

86153/572 = 151 519/7.61 = 68

Full-scale/reduced-scale

battery energy ratio

86153/519 = 166

Full-scale/reduced-scale UC

energy ratio

572/7.61 = 75

6.3.3 Selecting k Based on Battery Limitations

As an accelerated capacity drop for the cycle life tests is preferred, it is ideal to push the

battery harder, i.e. closer to the undersized battery k = 100 case. So the battery was cycled

with k = 100 (1/100th original mid-sized EV power) with three consecutive FTP-75 city drive

cycle. The battery was charged to 20.2V.

Page 151: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

130

However, it is unable to complete the drive cycles. This is because at demanding sections of

the drive cycles when the power drawn is too large, it causes the battery voltage to dip below

14V, leading to the built-in battery management system (BMS) to trip and disconnect the

battery (discussed further in section 6.7.7 Battery-only Undervoltage During Demanding

Sections). This is not surprising as the battery is undersized. Therefore, the scale was

increased further, until the drive cycles could successfully complete. This is summarised in

Table 6-7.

Table 6-7 Value of k which allows 3x FTP-75 city drive cycle to complete.

Scale factor k 100 130 140 150 160

Battery only x x x x √

Battery/UC x x x √ √

Both setups are able to complete the three FTP-75 drive cycles only when k = 160.

Interestingly, the battery/UC configuration can handle k = 150 while the battery-only

configuration cannot. This shows the UC is able to provide sufficient current during high

demands to prevent the battery voltage from dipping significantly.

From the table, k = 160 was selected, where both configurations are able to complete the

drive cycles. Since k = 160, this is closer to the case of UC being oversized. This means, only

part of the UC energy will be used.

6.3.4 Consequences of k = 160

From Table 6-8, after scaling the UC by k = 160, the experiment UC should only contain

3.58kJ. Therefore, the 7.61kJ experiment UC is oversized. Since in the full-scale simulation,

only 75% of the UC energy can be used, this results in only 2.68kJ of the 3.58kJ in the

experiment UC being used. This 2.68kJ corresponds to using only 35.2% energy or 19.5%

Page 152: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

131

charge. Therefore, to compensate for the oversized UC, the UC is restricted to using only a

19.5% Depth of Discharge (DoD) instead of 50%.

Table 6-8 UC scaling specifications.

Full-scale simulation

specifications

Reduced-scale experiment

specifications

UC energy (kJ) 572 7.61

Scaled UC energy (kJ) - 572/160 = 3.58

Usable Depth of Energy (DoE) 0.75 2.68/7.61 = 0.352

Usable energy (kJ) (0.75)(572) = 429 (0.75)(3.58) = 2.68

Usable Depth of Discharge

(DoD)

0.5 1-√(1-0.352) = 0.195

(computed with (6-3) & (6-4))

𝑈𝐶𝑆𝑜𝐸 =

𝐸

𝐸𝑚𝑎𝑥=

0.5 𝐶 𝑉2

0.5 𝐶 𝑉𝑚𝑎𝑥2 = (

𝑉

𝑉𝑚𝑎𝑥 )2

= 𝑈𝐶𝑆𝑜𝐶2

(6-3)

𝑈𝐶𝐷𝑜𝐷 = 1 − 𝑈𝐶𝑆𝑜𝐶 = 1 − √𝑈𝐶𝑆𝑜𝐸 = 1 − √1 − 𝑈𝐶𝐷𝑜𝐸 (6-4)

Next, the battery parameter scaling is examined. The battery is still slightly undersized (k =

160 instead of 166), therefore, the reduced-scale experiment battery will experience slightly

higher C-rates than the full-scale simulation battery. Table 6-9 summarises the scaling results.

From earlier section 4.1.2, the maximum power the car will consume at constant speed is

40.6kW. Scaling it would mean the reduced-scale experiment battery discharge power should

be limited to 0.254kW, which is 1.76C, slightly larger than 1.7C in the full-scale simulation.

Similarly, the charge power in the reduced-scale experiment is scaled to be 0.188kW, or

Page 153: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

132

1.30C, which is slightly larger than the original 1.25C. This is normal as the battery is slightly

undersized (k = 160 instead of 166).

Table 6-9 Battery scaling specifications.

Full-scale simulation

specifications

Reduced-scale experiment

specifications

Battery energy (kWh) (66)(362.6)/1000 = 23.9 (7.8)(18.5)/1000 = 0.144

Power at max speed (kW)

(self-limited)

40.6 40.6/160 = 0.254

Max discharge rate (C)

(self-limited)

40.6/23.9 = 1.7 0.254/0.144 = 1.76

Max regen power (kW) 30 30/160 = 0.188

Max charge rate (C) 30/23.9 = 1.25 0.188/0.144 = 1.30

6.4 Experiment 0

6.4.1 Objective

As mentioned in the previous section, experiments 1 and 2 are run at 1:160 scale. However,

the DC/DC converter power losses and inefficiencies are not able to scale perfectly.

Therefore, a preliminary experiment, Experiment 0, is run to investigate the efficiency of the

custom-built DC/DC converter. This is important as it allows the differences to be

compensated during scaling for more accurate results.

6.4.2 Procedure

Similar to section 3.9 DC/DC Converter Model, a plot of efficiency vs. high-side power vs.

low-side voltage is required. The buck and boost case are performed separately.

Page 154: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

133

For the buck case, the high-side of the DC/DC converter was connected to the Maccor, while

the low-side was connected to a bank of resistors as shown in Figure 6-8.

Figure 6-8 Experiment 0 setup.

Three values are varied in this experiment – the resistor bank value, the Maccor output

current, and the DC/DC converter duty cycle. Table 6-10 shows one data point as an

example.

Table 6-10 DC/DC converter efficiency test for one data point.

Variables Measured Computed

Resistance Duty

Cycle

Current Voltage Power Efficiency

ɳ IH

(Maccor)

IL VH VL PH

(input)

PL

(output)

0.5Ω 40% 1.69A 4.28A 5.98V 2.05V 10.11W 8.77W 86.8%

For that data point, a resistor bank value of 0.5Ω was selected. The Maccor was commanded

to output 1.69A to the DC/DC converter high-side, and then a DC/DC converter duty cycle of

40% was selected.

The voltages and currents on both the high and low-side were measured, and the powers

calculated. The input power (high-side) was 10.11W and the output power (low-side) was

8.77W, resulting in an efficiency of 86.8%.

The three variables were varied to get more data points.

Page 155: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

134

For the boost case, the same procedure was performed, except that the DC/DC converter was

connected in reverse, where the low-side of the DC/DC converter was connected to the

Maccor, and the high-side connected to a bank of resistors. In this case, IL corresponds to the

Maccor current, while IH, VL and VH are the measured values. Again, Maccor current IL, the

duty cycle, and the resistor bank values were varied to get data points.

6.4.3 Results

6.4.3.1 Missing Data Discussion

Figure 6-9 DC/DC converter boost efficiency.

Figure 6-9 shows the DC/DC converter efficiency for the boost case. Each blue circle

represents a data point, and the data points were interpolated. The efficiency values are

difficult to read right now; this will be addressed in a later figure. First, some concerns are

addressed. From the figure, the surface is triangular, and not square, this means that not all

Page 156: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

135

possible combinations of IH and VL were covered. This is because of limitations of the

experimental setup.

At large IH values, there is limited data. This is because large IH requires an even larger IL due

to the boost mode of operation. However, the setup is only designed for 30A, so it is not

possible to get data for large IH values.

At large VL values, there is also limited data. A large VL requires an even larger VH due to the

boost mode of operation. However, the setup is designed for 35V maximum. Also, the

Maccor can only operate at a maximum of VL = 20V.

At low VL values, there is also limited data. This is because the resistor bank needs to have a

smaller overall resistance (i.e. more resistors in parallel), based on the DC/DC converter

formulas, VL = DVH = DRIH, where R is the resistance and D is the duty cycle. However, all

the resistors of the required sizes in the laboratory were already in use (by the author). Also,

as the resistance decreases, the current increases (high IH values). The resistors do not have

sufficient power dissipation rating.

Figure 6-10 DC/DC converter buck efficiency.

Page 157: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

136

Similarly, Figure 6-10 shows the DC/DC converter buck efficiency. Again, there are some

missing data sections.

A large VL requires an even larger VH due to the buck mode of operation. However, the

Maccor is only able to operate at a maximum of VH = 20V. A low VL value, requires the

resistor bank to have a smaller overall resistance, based on the DC/DC converter formula VL

= RIL (A low VL value also corresponds to a large IL value due to the buck mode of

operation).

However, these missing sections are not a concern because these are outside the operating

range.

6.4.3.2 Flattened Results

Figure 6-11 shows a flattened 2D view of the combined boost and buck efficiency

experimental results, which is easier to read. The y-axis has been converted to power from

current by multiplying the high-side voltage.

Figure 6-11 Interpolated DC/DC converter efficiency from experiment.

Page 158: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

137

Reasonably good performance is observed for the self-made DC/DC converter, with at least

85% efficiency for most parts, dropping to 75% for low VL UC voltages (below 8V on the

left). This is not a concern because the UC will be operated above 12.2V due to the UC being

oversized as discussed in the earlier scaling section 6.3.4 and later section 6.5.1.

6.4.3.3 Matching Simulation to Experimental Values

Next, a model for the custom-built DC/DC converter was created, using the DC/DC converter

model discussed in section 3.9, parameters from the data sheets, and some tuning based on

the experimental data.

Table 6-11 shows the parameters used for the full-scale simulation, and the custom-built

DC/DC converter parameters, extracted from the data sheets of the MOSFET and inductor.

Table 6-11 Reduced-scale simulation parameters.

Parameter Full-scale

simulation

From data

sheets

Experimental Tuned

Vigbt (V) 0.857 0 N.A.

Rigbt (Ω) 0.00285 0.018

Vd (V) 0.88 0

Rd (Ω) 0.00184 0.018

RL (mΩ) 37 2.95

Turn on time (ns) 800 250 45 35

Turn off time (ns) 1000 250 45 35

As the data given in the Vishay IRFP048 MOSFET data sheet was for specific test conditions

which were not encountered during the experiment, some tuning was necessary for the

simulation model to match the experiment data.

The biggest difference to the test conditions in the experiment setup was the turn-on and turn-

off times, which was tested with 72A MOSFET drain current in the data sheet (the setup only

Page 159: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

138

reached 30A). So a further experiment was performed to observe the turn on and turn off

times.

A Tektronix TDS2012 oscilloscope was hooked to the DC/DC converter output before the

electrolytic capacitors, so the waveform seen on the oscilloscope was a square wave. The rise

and fall section of the square wave was zoomed in, and the rise and fall was observed to take

approximately 45ns each. Because of the ringing in the square wave, it was difficult to

determine the start and end of the rise or fall. So this parameter was further tuned to match

the experimental values as close as possible.

Figure 6-12 shows the efficiency of the reduced-scale DC/DC converter as produced from the

simulation model and tuned. Next, the simulation output was subtracted from the

experimental output (i.e. Figure 6-11 minus Figure 6-12) to see how well it was tuned,

resulting in Figure 6-13.

Eventually, the turn on and turn off time was tuned to 35ns to minimize the difference in

efficiency between the experiment and simulation. From Figure 6-13, there is less than 4%

difference in most parts. The difference hits 8% at low UC voltages (below 10V on the left).

Again, this is not a great concern as the UC will operate above 12.2V.

As a comparison, the full-scale DC/DC converter efficiency map was seen earlier in Figure

3-24. As the components used are different, it is expected the efficiency map would be

different.

Page 160: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

139

Figure 6-12 DC/DC converter efficiency from simulation.

Figure 6-13 DC/DC converter efficiency, experimental minus simulation output.

Page 161: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

140

6.4.4 Summary

In summary, the efficiency of the custom-built DC/DC converter has been investigated in

experiment 0. In general, it performs with an efficiency above 80%, which is reasonably good

for a self-made DC/DC converter. Then, a reduced-scale DC/DC converter model was

created to match the experimental values. This model is used later in the reduced-scale

simulations.

6.5 Software Implementation

As mentioned earlier, the microcontroller controls the duty cycle of the DC/DC converter.

This allows it to control the current flow or the voltage output. The Arduino microcontroller

is loaded with a reduced-scale algorithm, which is discussed in this section.

6.5.1 Algorithm Curve Fitting

To avoid the Arduino performing complex calculations on the fly to determine the battery

limits for the PMS and target UC bands for the EMS, polynomial curve-fitted equations have

been implemented instead. This reduces the processing power required. In this section, the

curve-fitting is explained.

In the Matlab simulation, k = 160 was implemented to create a reduced-scale simulation. The

DC/DC converter has been scaled separately as discussed in the previous section and is

incorporated into this simulation too.

For simplicity, the combined motor/inverter efficiency map profile is assumed to be scaled

perfectly as shown in Figure 6-14, stretched from the full-scale Figure 3-8.

Page 162: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

141

Figure 6-14 Reduced-scale combined motor/inverter efficiency.

Figure 6-15 Battery power to maintain EV at constant speed for k=160.

The reduced-scale simulation is run to get the battery power required to maintain the EV at a

constant speed, resulting in the curve as shown in Figure 6-15 (full-scale version earlier in

Figure 4-1). The data points are represented by the blue diamonds (very close together).

Page 163: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

142

Curve fitting was performed using Microsoft Excel’s least squares, shown by the thin black

line. The lowest order polynomial was chosen that fits the curve well, which is

𝑃𝑏𝑎𝑡𝑡 𝑎𝑡 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 𝑠𝑝𝑒𝑒𝑑 = 368.928𝑣2 − 72.106𝑣 + 8.382 (6-5)

As explained earlier in the PMS description, this battery power at constant speed is used to

determine the self-imposed battery power limits. Then the battery discharge power was

loosened by a factor of 3, and clipped to the maximum of 1.76C (254W). Similarly, the

battery charge power was clipped to -1.30C (-188W) according to earlier Table 6-9 Battery

scaling specifications. (Note that as given by equation (6-2), speed has also been reduced by a

factor of k = 160.)

Next, the target UC SOC bands for k = 160 are calculated and are shown in Figure 6-16

(Figure 4-10 for full-scale version). The band limits are defined by the red squares for the

upper limit and blue diamonds for the lower limit. Here, it is expressed in terms of UC

voltage instead of SOC or energy level as the sensor reads the UC voltage directly.

Figure 6-16 Target UC SOC band for k=160.

Page 164: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

143

Again, a curve fitting was performed with Microsoft Excel, shown by the thin black lines.

The best fit regenerative braking curve and best fit acceleration curve are given by (6-6) and

(6-7) respectively. Then, the maximum at each point of the two curves is taken as the target

UC voltage level, similar to what was discussed in the EMS section 4.2.4 earlier.

𝑣𝑢𝑐,𝑡𝑎𝑟,𝑟𝑒𝑔𝑒𝑛 = 10.391𝑣3 − 15.015𝑣2 + 1.0395𝑣 + 15.999 (6-6)

𝑣𝑢𝑐,𝑡𝑎𝑟,𝑎𝑐𝑐 = −195.15𝑣5 + 351.36𝑣4 − 194.14𝑣3 + 27.277𝑣2 + 1.8655𝑣

+ 15.881

(6-7)

In Figure 6-16, the minimum UC voltage value observed is 12.6V, corresponding to a DoD of

22.0%. However, it is mentioned earlier in the UC scaling of section 6.3.4 that the UC is

oversized and that only 19.5% DoD of the UC will be used. The reason for this difference is

that the DC/DC converter efficiencies of the full-scale and reduced-scale had not been

considered earlier when computing the 19.5% value. After inputting the interpolated DC/DC

converter efficiency from Experiment 0 into the simulation, the actual UC DoD should be

22.0%.

Therefore, to compensate for the differences in DC/DC converter efficiency scaling, more

UC energy is used, which is not a problem as the UC is oversized anyway.

In short, instead of the Arduino performing complex calculations on the fly to determine the

battery limits and target UC bands, the three simple polynomial equations (6-5) to (6-7) were

implemented in the Arudino. The three polynomial equations only require vehicle speed as

the input.

Page 165: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

144

6.5.2 Syncing Speed with Power

However, as there are no wheels in the reduced-scale experiment, there is no ‘speed’ value. In

real-life, the speed can be tapped from the speedometer. Therefore, for the experiments, the

FTP-75 city drive cycle’s reduced-scale speed vs. time data was loaded into the Arduino.

The proposed HESS management strategy requires the power drawn by the motor (in this

case, the Maccor) and the vehicle speed to function. The Maccor is loaded with a power vs.

time data, so it will automatically draw the specified power at the specified timing. Therefore,

the speed vs. time data loaded into the Arudino must be synced with the Maccor. For

simplicity, a manual sync at the start was used, where both Arduino and the Maccor are

turned on together.

6.5.3 Problems with Maccor

However, a problem was observed. When running highly transient drive cycles like the FTP-

75 city, the Maccor is unable to keep up and therefore runs slower. Three consecutive FTP-75

drive cycles were run on the Maccor. Instead of taking (1875s)(3) = 5625s, the Maccor was

timed to take 6156s, which is 9.44% slower. Therefore, the speed vs. time profile in the

Arduino was also slowed down by 9.5% to match the Maccor. The Arudino is only able to

handle milliseconds in integers, so 1000ms was not translated to 1094.4ms, but 1095ms

instead to avoid interpolation of the drive cycle. (1094ms was tested at first, but resulted in

the Arduino ending earlier than the Maccor (as expected), which was bad as the setup was

programmed to be turned off for safety reasons at the end, which was before the Maccor

drive cycles had actually ended).

The consequence of this is that the drive cycle is not perfectly scaled down. But since this

slowed drive cycle is applied to both battery-only and battery/UC setups, it is still a fair

comparison between the two.

Page 166: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

145

Interestingly, when running modal drive cycles such as the EUDC or ECE drive cycles, the

Maccor had no slowdowns. This showed the Maccor is unable to keep up only for highly

transient drive cycles.

Also, the current sensor and voltage sensor of the Maccor machine are out of calibration. It

costs an excessive amount to fix as a technician is required to be flown from the United

States, and his flight and accommodation have to be paid for, in addition to the calibration

services. Therefore, in the experiment, sensors separate from the Maccor have been used for

data logging as discussed in 6.1.3.

Intriguingly, the Maccor underreports voltages, while it overreports current. Since the Maccor

is programmed in power vs. time, the underreported voltage and overreported current cancel

each other out, resulting in the power calibration being reasonably accurate. This is shown in

more detail in the next section, 6.6 Experiment 1.

For safety reasons, the Arduino was also programmed to turn off the relays and stop the

experiment if the battery temperature exceeded 50oC, or if the UC was charged beyond 16.2V

(100% SOC), or when three consecutive FTP-75 city drive cycles were over.

6.5.4 Battery Power vs. Current Thresholds

To recall, the goal of the PMS is to ensure the UC voltage follows the target UC voltage and

to prevent the battery power thresholds from being exceeded. However, there were problems

implementing battery power thresholds. During demanding sections, due to high contact

resistance in the circuit, a substantial voltage drop was encountered between the Maccor and

battery (This problem is discussed in detail later in sections 6.7.7 and 6.7.8). Due to the

voltage drop, the Maccor attempts to pull more current to compensate, causing the battery

voltage to drop even further.

Page 167: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

146

If a battery power threshold is used, the Arduino is slow to respond because when the current

increases, the battery voltage also drops, so the battery power threshold may not be exceeded.

However, since the current is increased, the battery is experiencing a higher than expected C-

rate discharge, which is not good for the cycle life.

Therefore, for experiment implementation, a battery current threshold was used instead to

avoid the transient voltage dips from affecting the C-rate too heavily. The battery current

threshold is calculated by dividing the battery power threshold by a fixed voltage, 18.5V, the

nominal voltage of the battery.

Ideally, if the contact resistance is minimized in the experiment setup, there would not be a

significant difference whether battery current or battery power threshold was used, as the

battery voltage would not change much in the short-term. Even in the literature, this is not

standardised. For example works of [32] [45] [48] attempt to limit battery power, while

works of [25] [40] [47] attempt to limit battery current.

6.5.5 Integral Controller for UC Voltage Control

As mentioned earlier in section 3.3 Simulation Approach, a backward facing approach was

used for the simulations. In the backward facing approach, transients are not considered. So

in the simulations earlier, the algorithm is able to calculate the exact amount of power which

the UC should generate/absorb to relieve the battery at every instant. In order to do this, the

algorithm was programmed with the DC/DC converter efficiency map to compensate for the

actual UC power required.

However, physically, transients are unavoidable. Also, from Experiment 0, although the

DC/DC converter efficiency model was empirically fit based on data from the actual

efficiency, there are still some inaccuracies.

Page 168: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

147

Therefore in the physical setup, an integral controller was implemented, mathematically

expressed with the following equations. This controller ensures the battery current threshold

is not exceeded, even with transients, and without knowledge of the DC/DC converter

efficiency map.

𝐼𝑏𝑎𝑡𝑡,𝑜𝑣𝑒𝑟(𝑡)

=

𝐼𝑏𝑎𝑡𝑡,𝑚𝑎𝑥(𝑡) − 𝐼𝑏𝑎𝑡𝑡(𝑡) 𝑖𝑓 𝐼𝑏𝑎𝑡𝑡(𝑡) ≥ 0 𝑎𝑛𝑑 𝐼𝑏𝑎𝑡𝑡(𝑡) > 𝐼𝑏𝑎𝑡𝑡,𝑚𝑎𝑥(𝑡)

𝐼𝑏𝑎𝑡𝑡,𝑚𝑖𝑛(𝑡) − 𝐼𝑏𝑎𝑡𝑡(𝑡) 𝑖𝑓 𝐼𝑏𝑎𝑡𝑡(𝑡) < 0 𝑎𝑛𝑑 𝐼𝑏𝑎𝑡𝑡(𝑡) < 𝐼𝑏𝑎𝑡𝑡,𝑚𝑖𝑛(𝑡)

0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

(6-8)

𝑣𝑢𝑐,𝑡𝑎𝑟,𝑎𝑑𝑑(𝑡) =

𝐾𝑣 ∗ 𝐼𝑏𝑎𝑡𝑡,𝑜𝑣𝑒𝑟(𝑡) + 𝑣𝑢𝑐,𝑡𝑎𝑟,𝑎𝑑𝑑 (𝑡 − 1) 𝑖𝑓 𝐼𝑏𝑎𝑡𝑡,𝑜𝑣𝑒𝑟(𝑡) ≠ 0

(1 − 𝛾) 𝑣𝑢𝑐,𝑡𝑎𝑟,𝑎𝑑𝑑 (𝑡 − 1) 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

(6-9)

𝑣𝑢𝑐,𝑡𝑎𝑟,𝑎𝑑𝑗(𝑡) = 𝑣𝑢𝑐,𝑡𝑎𝑟(𝑡) + 𝑣𝑢𝑐,𝑡𝑎𝑟,𝑎𝑑𝑑(𝑡) (6-10)

The purpose of these equations are as follows. If the battery current (on current sensor ACS3)

exceeds its positive threshold (i.e. discharging more than the threshold), the UC should assist

by discharging as well. Therefore, the target UC voltage should be adjusted lower so the UC

will discharge more. The converse is true also. If the battery current is exceeding its negative

threshold (i.e. charging more than the threshold), the target UC voltage should be raised so

more current enters the UC instead.

However, if no battery limits have been exceeded, then the target UC voltage should be

followed as per normal and no adjustments are required.

Equation (6-8) checks if the battery limits (Ibatt,max or Ibatt,min) have been exceeded. If it has

been exceeded, Ibatt,over is the error term. Equation (6-9) calculates the target UC voltage

addition term (Kv is a tuning parameter), while equation (6-10) is the final adjusted target UC

voltage.

Page 169: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

148

As an example, by considering the case of exceeding the positive battery current limit (i.e.

discharging too much). If the discharge limit Ibatt,max has been exceeded, the error term Ibatt,over

is a negative number. This causes the vuc,tar,add term to be negative, leading to the target UC

voltage being adjusted downwards as vuc,tar,adj < vuc,tar. After a few iterations, the error

coefficient Kv was tuned to 0.2 for best performance.

The addition (integral) term allows the output to be held at the value which prevents the

battery currents from being exceeded. Once the battery limits are no longer exceeding the

limits, the addition term should be switched off and the original target UC voltage followed.

However, there are problems for a sudden switch off.

A sudden switch off is only suitable if the battery limit is no longer exceeded because the

drive cycle current demand has fallen. But if the battery limit is no longer exceeded because

the output is giving an appropriate addition term, causing the error term to fall to zero, a

sudden switch off would cause the addition term to fall to zero, leading to the battery currents

being exceeded again, and the addition term must ramp up again, leading to oscillations.

The solution is a decaying addition term, so the switch off is not sudden. The decay rate is set

by ɣ in equation (6-9). If ɣ = 0, there will be no decay and the addition term will never die. If

ɣ = 1, the addition term decays to zero in one instant, leading to oscillations as mentioned

above.

ɣ was tuned such that the effects of the decay term are non-observable within the 10Hz data-

logging. The addition term decayed within 0.1s if the drive cycle current demand fell. But

when a sustained control output was required to prevent the battery current limits from being

exceeded, no decay was observed at 0.1s intervals. Eventually, ɣ was tuned to 0.05 iteratively

in a manner similar to the bisection method to achieve the non-observable decay effect.

Page 170: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

149

Once the adjusted target UC voltage vuc,tar,adj has been computed from equation (6-10), a

Proportional-Integral (PI) controller is used to bring the actual UC voltage to the adjusted

target UC voltage. The PI controller outputs an 8-bit PWM, which controls the DC/DC

converter duty cycle, and the PI controller was tuned iteratively for best performance.

6.6 Experiment 1

6.6.1 Objective

The purpose of Experiment 1 is to demonstrate the algorithm works in a physical setup and

performs as intended. For the PMS, the target UC voltage should be followed, except when

the battery current exceeds the speed-dependent limit, where the UC should produce/absorb

the excess current. For the EMS, as long as the target UC voltage is followed, the two EMS

goals of sufficient space and sufficient energy are met. Battery 1 was used for this test.

6.6.2 EUDC Drive Cycle

Although the FTP-75 city drive cycle is used for the cycle life test (Experiment 2), the EUDC

is demonstrated here for easier observation as the EUDC is a modal cycle.

The EUDC power vs. time data were loaded into the Maccor, and the speed vs. time were

loaded into the Arduino, and the EUDC was run (at normal speed). The following figures

show the results.

Figure 6-17 EUDC speed profile after scaling with k=160.

Page 171: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

150

Figure 6-17 shows the scaled EUDC speed profile as loaded into the Arduino. Figure 6-18

shows the drive cycle, battery and UC currents as measured by the current sensors (high-side

UC current iuc,H is computed with (6-1)). It also shows the battery current limits as computed

by the Arduino using the curve fit equations discussed in section 6.5.1.

Figure 6-18 Battery & UC currents from experiment, EUDC.

Here, the UC takes over when the battery exceeds the current limits. At 320s to 340s, when

the vehicle is accelerating to high speeds, the total drive cycle current idr exceeds the upper

battery limit ibatt,max. So the algorithm limits the battery current ibatt to ibatt,max, and the UC

handles the rest of the current as shown by iuc,H (iuc is the UC current at the DC/DC converter

low-side, while iuc,H is the UC current at the DC/DC converter high-side). A similar case can

be seen during high regenerative braking from 120s to 130s.

Note: The Arduino loop runs at 100Hz, while the data is logged and plotted at 10Hz.

Figure 6-19 shows the battery and UC voltages. There are 3 UC voltages – the actual UC

voltage vuc, the target UC voltage (as computed by the curve fit equations in section 6.5.1)

Page 172: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

151

vuc,tar, and the adjusted target UC voltage to prevent excessive battery currents vuc,tar,adj. It is

observed that vuc follows vuc,tar,adj very closely, such that they overlap. This shows that the

voltage tracking of the PI controller is very good.

Figure 6-19 Battery & UC voltages from experiment, EUDC.

It can also be observed that vuc,tar,adj deviates from vuc,tar at 320s to 340s, at the same place

where the battery current was going to exceed its limits. With earlier equation (6-10), vuc,tar,add

= vuc,tar,adj – vuc,tar, the deviation (addition term vuc,tar,add) can be seen more carefully in Figure

6-20. To prevent the battery current from exceeding the limits, the algorithm lowers the UC

target voltage as designed, therefore causing the UC to discharge more, and assisting the

battery during high power phases.

Figure 6-20 Target UC voltage addition term vuc,tar,add, EUDC.

Page 173: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

152

The opposite case is seen from 120s to 130s, where high regenerative braking is encountered.

The algorithm raises the target UC voltage to ensure the UC is charged more.

During non-demanding sections, vuc follows vuc,tar quite closely. So it has been shown that the

two PMS goals have been achieved – to ensure the actual UC voltage follows the target UC

voltage, and to limit battery currents – with priority on the latter. When the actual UC voltage

follows the target UC voltage, the two EMS goals of sufficient space and energy are also

achieved.

Similar to the full-scale simulations, the EUDC also demonstrates the two goals of the speed-

dependent battery current limit. From 250s to 290s in Figure 6-18, it is observed that the

battery limits are not exceeded as idr < ibatt,max. However, the UC is still being utilised (iuc,H >

0), which satisfies the first goal of using the UC even during low demands to reduce battery

usage. From 290s to 320s in Figure 6-18, a constant speed is being maintained, and it can be

seen that the battery is supplying all the power at steady state, satisfying the second goal of

the speed-dependent battery limit.

As mentioned earlier in section 6.5.3, there are two problems with the Maccor machine – the

machine runs slow for transient drive cycles, and the sensors are out of calibration. As the

EUDC is a modal drive cycle, the Maccor is able to run at the correct timing, so the Arduino

was not purposely slowed to compensate.

As discussed earlier, although the Maccor sensors are out of calibration, the resulting power

output is reasonably accurate as the voltage is underreported and the current is overreported,

cancelling each other out. This is shown in Figure 6-21, where the blue curve is the actual

power drawn by the Maccor, measured by the setup’s separate current sensors and voltage

sensors and multiplied together. The red Pdc curve is the power vs. time profile programmed

Page 174: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

153

into the Maccor. It was added later in post-processing after the experiment was finished (i.e.

it was not logged). It can be seen that the two curves do not differ much from each other.

Figure 6-21 Power and time sync check, EUDC.

6.6.3 FTP-75 City Drive Cycle

The four HESS and two battery limits goals have already been demonstrated in the EUDC

run above. Here, the FTP-75 city drive cycle was loaded into the setup and also tested. This is

provided for reference as the cycle life tests in experiment 2 use the FTP-75 city drive cycle.

It was run at reduced speed as discussed in section 6.5.3. The following figures show the

results.

Figure 6-22 FTP-75 city speed profile after scaling with k=160.

Page 175: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

154

Figure 6-23 Battery & UC currents from experiment, FTP-75 city.

Figure 6-24 Battery & UC voltages from experiment, FTP-75 city.

Figure 6-22 shows the scaled FTP-75 city speed profile as loaded into the Arduino. Figure

6-23 shows the drive cycle, battery and UC currents. Figure 6-24 shows the battery and UC

voltages. Figure 6-25 shows the target UC voltage adjustment values.

Page 176: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

155

Figure 6-25 Target UC voltage addition term vuc,tar,add, FTP-75 city.

As it is transient drive cycle, it may look ‘messy’, so a zoomed-in section from 180 to 280s is

shown in the following figures, where high acceleration is required. Figure 6-26 shows the

zoomed in section of currents. During the section from 205s to 220s, it is observed the drive

cycle current idr exceeds the upper battery limit ibatt,max. The algorithm has correctly limited

the battery current ibatt to the limits, and the UC iuc,H handles the excess current, just like the

earlier EUDC case.

Figure 6-26 Battery & UC currents from experiment, FTP-75 city, zoomed 180-280s.

Page 177: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

156

Figure 6-27 shows the battery and UC voltages. Again, it can be seen that actual UC voltage

vuc follows the adjusted target UC voltage vuc,tar,adj very closely such that they overlap,

showing good control of UC voltage.

Figure 6-27 Battery & UC voltages from experiment, FTP-75 city, zoomed 180-280s.

Also, the adjusted target UC voltage vuc,tar,adj deviates from target UC voltage vuc,tar at 205s to

220s, at the same place where the battery current was going to exceed its limits. Figure 6-28

shows this target UC voltage addition term more clearly. Similar to the earlier EUDC case,

the algorithm lowers the UC target voltage to further discharge the UC, assisting the battery.

Figure 6-28 Target UC voltage addition term vuc,tar,add, FTP-75 city, zoomed 180-280s.

Page 178: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

157

As mentioned earlier in section 6.5.3, the Maccor is unable to keep up with transient drive

cycles. Figure 6-29 shows the actual power drawn vs. the expected power drawn.

Figure 6-29 Power and time sync check, FTP-75 city.

The Maccor is running slow, taking 2052s instead of 1875s for a single FTP-75 drive cycle.

Also, within the cycle, some sections run faster, while others run slower, as can be seen from

the blue actual power drawn curve sometimes leading and sometimes lagging the red

expected curve (The expected curve was added in post-processing. It is slowed by a fixed

9.5% throughout, just like the Arduino).

On the positive side, the Maccor was consistent between cycles. When the FTP-75 city cycle

was run again, the same sections would lead or lag, so it was repeatable.

Although some sections lead or lag, this was not a major concern. Ultimately, the battery-

only and battery/UC setup are both subject to the same leading/lagging drive cycle in the

cycle life tests in Experiment 2, so a fair comparison can still be made (though the battery/UC

setup is at a slight disadvantage due to the speed vs. time being slightly out of sync with the

power vs. time).

Page 179: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

158

Also as discussed earlier, the actual power drawn and expected power drawn are reasonably

similar in terms of magnitude, despite the Maccor sensors being out of calibration.

6.6.4 Average Currents

In this section, the average of absolute values of the battery current in the battery/UC

experiments above is computed and compared to that in a battery-only system. It will be

shown that the battery in the battery/UC HESS outputs a lower average current as compared

to a battery-only system. A lower average current (or lower C-rate) results in a longer battery

cycle life.

Figure 6-30 shows the logged current data from running a battery-only system over the FTP-

75 city drive cycle. It can be seen this battery-only system exceeds the battery current limits

numerous times as there is no UC to assist.

Figure 6-30 Battery-only setup. Currents from experiment, FTP-75 city run 1.

Table 6-12 shows the average of absolute battery current values from experiment data. Three

runs of the EUDC and FTP-75 city drive cycles were performed, and the results were

averaged.

Page 180: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

159

Table 6-12 Average of absolute battery currents for experiments

EUDC FTP-75 City

Run No. 1 2 3 Avg. 1 2 3 Avg.

Battery-only setup (A) - - - - 2.19 2.20 2.25 2.21

Battery/UC setup (A) 3.44 3.51 3.59 3.51 1.81 1.86 1.86 1.84

The EUDC battery-only setup was unable to complete for the selected scale factor of k = 160,

as the demanding sections of EUDC caused the battery voltage to dip below 14V and tripped

the BMS (see section 6.3.3 for a similar case). Interestingly, the battery/UC setup could

complete the EUDC. This shows the UC is able to provide sufficient current during high

demands to prevent the battery voltage from dipping too much.

From Table 6-12, the battery/UC system reduces the battery current (or C-rate) from 2.21A to

1.84A over the FTP-75 city drive cycle. Based on the works of Wang, et al. [7], a lower C-

rate leads to a longer battery cycle life. This is explained further in Experiment 2.

6.6.5 Summary

Therefore, from this section, it has been shown that the proposed HESS management strategy

works in a physical setup and performs as intended. Similar to the simulation results earlier,

the EUDC experiment has shown that the four HESS goals and two battery limit goals have

been achieved.

Specific to the experiment (i.e. not simulation), it was observed the target UC voltage is well

tracked. And when battery current exceeds the limit, the target UC voltage is adjusted and the

UC will absorb/generate the extra power.

Page 181: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

160

In addition, it was shown that the proposed HESS management strategy reduces the average

C-rate of the battery in a battery/UC setup as compared to that of the battery-only setup,

extending the battery cycle life.

Although there are some problems with the Maccor machine as discussed in 6.5.3, these have

already been compensated or are not a major concern.

In addition, when using a scale of k = 160, the EUDC is only able to run in a battery/UC

setup. During peak power demands, the battery-only setup is unable to supply enough current

and undervoltage occurs. In contrast, the UC in the battery/UC setup is able to assist the

battery in supplying current, so the undervoltage condition does not occur.

6.7 Experiment 2

6.7.1 Objective

The purpose of experiment 2 is to compare the cycle life of battery 2 in the battery-only setup

to battery 3 in the battery/UC setup. Each setup is cycled continuously for an accelerated

cycle life test, and the capacities of the batteries in the two setups are compared. This would

verify the ultimate HESS goal, which is to extend battery cycle life.

6.7.2 Description

As mentioned in section 5.5.2, three FTP-75 city drive cycles comprise one full day of

driving. To estimate how many cycles should be performed, existing works in the literature

are examined.

In [7], Wang, et al. performed constant charge and discharge accelerated cycling tests with

Sanyo UR18650W batteries. With 0.5C as the charge and discharge rate, approximately 5000

cycles were required for a 30% drop in battery capacity, which took about two years, while at

least 200 cycles were required to observe a noticeable drop of 5%.

Page 182: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

161

As the average C-rate for the battery/UC setup under the FTP-75 city drive cycle is only

0.24C (less than their 0.5C), it is expected to take even longer. Running thousands of cycles

for three or four years is not feasible within the authors’ PhD candidature. Instead, as many

cycles as possible were run within a six-month period (during office hours only due to safety

concerns) until the final version of this thesis was submitted.

6.7.3 Procedure

Two brand new 5S3P SuPower battery packs (from China) were used for this experiment –

battery 2 for the battery-only setup and battery 3 for the battery/UC setup. First, it was

required to find the two batteries’ initial capacity by performing a capacity test.

Secondly, each battery was cycled for 20 daily cycles (i.e. 60 FTP-75 city cycles), one for the

battery-only setup and one for the battery/UC setup. This was effectively repeating

Experiment 1 multiple times. Charging at 0.5C was performed when the battery fell below

17V or was unable to get through the demanding sections of the FTP-75 city cycle (more on

this in section 6.7.7 later).

After the 20 daily cycles, another capacity test was performed to determine the capacity loss.

This was repeated, where a capacity test would be performed every 20 daily cycles. The

value of 20 daily cycles was selected in order to perform about 6 to 8 capacity tests in over

the first three months. This would provide sufficient data points to observe a capacity drop

trend. In the subsequent three months, it was spaced to 30 cycles instead, to allow more drive

cycles to be run.

Page 183: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

162

6.7.4 Battery Discharge Capacity Test

6.7.4.1 Procedure

The battery capacity test is performed similar to that found in [7]. The battery is placed into

the green box as seen in the photo in Figure 6-2 for data logging, but with the UC and DC/DC

converter disconnected.

First the battery is charged at constant current (CC) to maximum voltage, then charged at

constant voltage (CV) to make sure it is fully charged. Then the battery is discharged at CC

and the data is logged to determine the discharge capacity. The discharge capacity procedure

is summarised in Table 6-13.

Table 6-13 Discharge capacity test procedure.

Step Procedure Run Until Data

1 4.7A CC charge Voltage reaches 20.2V Discarded

2 20.2V CV charge Current falls below 0.7A Discarded

3 Rest For 3 mins Discarded

4 4.7A CC discharge Voltage falls below 14V Logged

5 Post-processing to limit voltage range to

[18.9V, 15.3V]

- -

The maximum voltage of the battery is 21V, however, the Maccor is only rated for 20V.

Therefore, the upper charging limit was set at 20V on Maccor. As the Maccor sensor was out

of calibration, it was actually 20.2V. The lower discharge limit is set to 14V, which is when

the battery built-in BMS trips the battery due to undervoltage occuring.

Ideally, a small CC charging/discharging rate should be used to get the true maximum battery

capacity (to reduce losses) and to avoid the capacity test affecting the cycle life results.

Page 184: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

163

Wang, et al. in [7] use 0.5C. However, since the purpose of the experiment was to observe the

capacity fade, the true maximum capacity was not that important, but rather the relative

capacity between the two batteries. Therefore, the maximum discharge current given by the

manufacturer of 5A was used to speed up the experiment. The Maccor was commanded to

charge/discharge at 5A, but because its calibration was off, it produced 4.7A instead,

corresponding to 0.6C.

The CV charging section should be stopped when the current falls below a threshold. Ideally,

this threshold should be as small as possible. Wang, et al. in [7] used a threshold of 0.075A,

sometimes taking up to a maximum of two days to charge. Again, as only the relative

capacity is important, and to speed up the experiment, the threshold was set at 0.7A

(commanded to 1A on Maccor), resulting in the CV charging section taking slightly more

than an hour.

6.7.4.2 Post-processing Calculations

Once the discharge cycle had finished, there was further post-processing to calculate the

battery capacity. This is illustrated by the first discharge test of battery 3.

Figure 6-31(a) shows the CC discharge of 4.7A as logged by current sensor ACS3. The

charging starts at about 1700s and ends at 6600s (ignore the dashed lines for now). The curve

in Figure 6-31(a) is integrated over time (known as coulomb counting) to give Figure

6-31(b), which is the amount of charge which has been discharged from the battery so far.

Figure 6-32 shows the battery voltage over time as logged by the sensor. It decreases over

time to 14V, when the BMS trips the battery due to undervoltage, disconnecting and dropping

the voltage to 0V at the end.

Since the UC and DC/DC converter are disconnected (iuc,H = 0), from equation (6-1), idr = ibatt,

and there are two current sensors – ACS3 & ACS1 – logging the same data. However, there

Page 185: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

164

are some minor differences in the two readings (up to ~0.1A), so the average was taken from

both sensors. Figure 6-33 shows the battery capacity with respect to battery voltage, for each

of the sensors. Time is implicit in that plot.

Figure 6-31 (a) Discharging current from ACS3 (b) coulomb counting.

Figure 6-32 Discharging voltage.

Figure 6-33 Battery discharging capacity.

Page 186: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

165

The total battery discharge capacity can be determined by the right-most value on Figure

6-33, which is 6.44Ah on ACS3 and 6.31Ah on ACS1, giving an average of 6.37Ah.

In Table 6-13 earlier, the discharge capacity procedure was presented, where the first two

steps are for charging the battery to 20.2V. However, due to contact resistance and other

losses (discussed in section 6.7.8), some voltage drop is present, and once the battery stops

charging, the open circuit battery voltage is lower than 20.2V.

In the fourth step, the battery discharge is performed until the in-built BMS trips the battery

at 14V. However, it is not known how consistent the BMS is.

Since the upper 20.2V and lower 14V limits cannot be fully relied on, further post-processing

was done on the resultant data. Only the data between 18.9V and 15.3V was retained as

shown by the dashed lines in the previous three figures. The two values were selected to

ensure a small distance from the edges, to avoid the uncertainty at the limits.

Subtracting the capacity from the two ends, this results in 6.06Ah on ACS3 and 5.93Ah on

ACS1, giving an average of 6.00Ah. This is the discharge capacity test procedure result.

For completeness, the plots for the first two steps of Table 6-13, which is to charge the

battery, are shown in Appendix A.

6.7.5 Battery-only Setup

There are two independent channels in the Maccor, so it is possible to run both the battery-

only and battery/UC experiments at the same time.

To run the battery-only cycle life experiment, ideally, the green box setup in the photo in

Figure 6-2 should be duplicated, for example, all the sensors, data logging equipment and

safety switches, only without the DC/DC converter and UC. However, there was limited

budget and time to construct a new setup for the battery-only system, so the Maccor was

Page 187: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

166

directly connected to the battery through wires only, without any sensors, for running the

cycle life test.

But for the capacity test, the battery (battery 2) would be placed in the green box setup in

Figure 6-2 (with the UC and DC/DC converter disconnected) in order to log the current and

voltage.

6.7.6 Initial Results

Figure 6-34 Battery discharge capacity tests over 60 cycles.

Figure 6-34 shows the discharge capacity tests results over 60 cycles for battery 2 (battery-

only system) and battery 3 (battery/UC system). They are the averaged results from the two

current sensors.

As the number of cycles performed is small, the battery capacity is not expected to drop

much. The variation in the results above seem to be noise due to experiment variations.

However, it seemed that battery 2 was more consistent than battery 3, which has some drop in

battery capacity when comparing cycle 0 to cycle 60 (or even cycle 40). This is contrary to

the predictions, so further investigation was necessary.

Page 188: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

167

6.7.7 Battery-only Undervoltage During Demanding Sections

Before explaining why the battery/UC system seemed to perform worse, an observation from

the experiments is discussed first.

In section 6.3 Scale Factor k, and more specifically Table 6-7, it was shown that the battery-

only and battery/UC system were able to complete the FTP-75 city drive cycle only when k =

160. That was when the battery was charged to 20.2V. After going through a few more

cycles, naturally the battery would be more discharged and its voltage would drop.

As an example, the battery-only setup was run with the FTP-75 city drive cycle, where the

initial battery voltage was 19.1V (Nominal battery voltage is 18.5V).

Figure 6-35 Battery-only voltage for FTP-75 city, tripped, k=160.

Figure 6-36 Battery-only current for FTP-75 city, tripped, k=160.

Page 189: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

168

Figure 6-35 shows the battery voltage profile and Figure 6-36 shows the current profile. At

212s real-time (corresponding to 193s drive cycle time due to the slow Maccor), the FTP-75

city drive cycle experiences a peak in power. The Maccor attempts to draw higher current,

which causes the battery voltage to dip because of internal resistance and other unwanted

resistive losses in the circuit. As the battery voltage dips, the Maccor draws even higher

current to hit the required power, which causes the battery voltage to dip even further. From

Figure 6-36, the last recorded non-zero current was 30A (the limit of the current sensor),

which caused the battery voltage to dip excessively, leading to the BMS to trip the battery,

resulting in 0V at 212s in Figure 6-35.

In this example, the battery-only system is unable to complete the drive cycle not because it

is flat, but because it cannot meet the required power demands.

As a side-note, in the battery/UC system, the FTP-75 city drive cycle is able to complete

when the initial battery voltage was less than 19V. Figure 6-24 in section 6.6.3 of Experiment

1 showed this. This is because the UC is able to provide sufficient current during high

demands and prevents the battery voltage from dipping too low.

6.7.8 Battery-only Contact Resistance

To recap, from earlier section 6.7.6, it was found that the battery/UC system performed worse

in terms of capacity loss. The reason is explained in this section.

6.7.8.1 No. of Drive Cycles Completed

Some observations were noticed in the experiment and summarised in Table 6-14. If initially

charged to 20.2V, the battery/UC system could complete approximately 11 FTP-drive cycles

before battery undervoltage occurred as discussed in the previous section. On the other hand,

the battery-only setup could complete approximately 10 FTP-75 city drive cycles.

Page 190: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

169

Table 6-14 FTP-75 city drive cycles before battery undervoltage.

Direct connection (section

6.7.5 Battery-only Setup)

Placed into green box

(Figure 6-2)

Battery-only system

(battery 2)

10 3

Battery/UC system

(battery 3)

- 11

However, if battery 2 from the battery-only setup was placed in the green box setup (Figure

6-2) with the DC/DC converter and UC disconnected, it would last for only three FTP-75 city

drive cycles before undervoltage occured.

It was suspected there was high voltage drop due to contact resistance in the green box setup,

leading to a higher possibility of undervoltage occurring. From the Maccor to the battery in

the green box, the current had to pass through two connectors, two fuses, two current sensors

and a relay as shown earlier in the electrical diagram Figure 6-1. The current sensors were

connected via nuts and bolts, and the fuses and relays were connected via quick connect

crimp connectors, so there may be substantial contact resistance.

In contrast, for the directly connected battery-only setup, the current only had to pass through

two connectors, resulting in smaller voltage losses.

6.7.8.2 Contact Resistance Experiment

The voltage losses due to contact resistance (and others) were verified in an experiment as

follows. The two voltage sensors of the setup were disconnected. One was placed as close to

the battery as possible, by making cuts on the battery lead insulation. The other sensor was

placed as close to the Maccor as possible, at the connection between the Maccor and the

experiment setup. This is shown in Figure 6-37.

Page 191: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

170

Figure 6-37 Contact resistance experiment connections diagram.

A step current profile was programmed into the Maccor, from 10A to 0A (discharging), then

-10A to 0A (charging) as shown in Figure 6-38. As mentioned earlier, there are two sensors

measuring current, so there are two curves, one for each sensor (ACS1 and ACS3), almost

overlapping.

Figure 6-39(a) shows the voltage measurement of the two sensors, while Figure 6-39(b) is the

difference of the two curves in Figure 6-39(a), showing the voltage drop between the Maccor

and the battery.

Figure 6-40 shows the equivalent resistance between the Maccor and battery, due to contact

resistance and other losses. This was calculated by dividing the voltage drop (Figure 6-39(b))

by current (Figure 6-38).

The section from 67s to 77s is blank because current is zero, and voltage divided by zero

current is undefined. The sections from 57s to 67s and from 117s to 127s are very noisy and

have high peaks. This is because during that section, the current is small (0.7A), resulting in

only a small voltage drop. As the voltage drop is very small and close to zero (see Figure

6-39(b)), there is a low signal-to-noise ratio in that section, so it is ignored.

to μC

to μC

to μC

to μC

Page 192: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

171

Figure 6-38 Current profile to find voltage drop.

Figure 6-39(a) Measured voltages (b) voltage drop.

Figure 6-40 Equivalent resistance (contact resistance and other losses).

It is observed from this experiment, that the equivalent resistance in the battery/UC green box

setup is on average, approximately 0.07Ω, which is significant. Earlier Figure 6-26 shows the

Page 193: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

172

current demanded hits 15A at 212s in the FTP-75 city drive cycle. A 15A current over 0.07Ω

results in a 1.05V voltage drop, enough to push the battery to undervoltage conditions.

To summarise, although the battery stress is relieved by the UC in the battery/UC setup, it

also has to overcome an extra voltage drop due to contact resistance and other losses, which

increases battery stress again.

6.7.9 Battery-only Setup, Revised

Therefore, the battery-only setup was revised after the first 60 cycles. The additional voltage

drop was included to ensure the cycle life tests between the battery-only and battery/UC setup

are fair. After a few trials, 0.06Ω of resistance was added in series to the direct connection of

the battery-only setup as shown in Figure 6-41.

Figure 6-41 Photo of battery-only setup.

The voltage drop and equivalent resistance were tested in the new setup. The same current

profile in Figure 6-38 was cycled. Figure 6-42 shows the voltage measurements and the

voltage drop between the Maccor and the battery.

Figure 6-43 shows the equivalent resistance between the Maccor and battery. Again, the

noisy sections from 50s to 60s and from 110s to 120s were ignored. Although only 0.06Ω of

Battery

0.06Ω of

resistance

To Maccor

Connector

Page 194: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

173

resistors were used, it can be seen from the plot that the total resistance is approximately

0.07Ω. The extra 0.01Ω is likely due to actual contact resistance present.

Figure 6-42(a) Measured voltages (b) voltage drop.

Figure 6-43 Equivalent resistance (contact resistance and others).

Now, the contact resistance of the battery-only setup has been matched to the battery/UC

setup.

The FTP-75 city drive cycles were run again in this revised battery-only setup, and as

expected, the battery completed only three FTP-75 city drive cycles before undervoltage

occurred, matching the situation in the green box as shown earlier in Table 6-14.

Page 195: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

174

6.7.10 Final Results

Eventually, 220 cycles of the battery-only setup and 190 cycles of the battery/UC setup were

completed in the six-month period. Figure 6-44 shows the battery capacity versus cycle

number. Figure 6-45 shows the battery capacity as a percentage of the initial battery capacity.

Figure 6-44 Battery capacity of battery-only and battery/UC setup.

Figure 6-45 Relative capacity. Cycle capacity divided by initial capacity.

From Figure 6-45, both battery capacities are hovering around the 97-100% mark. As each

cycle is equivalent to a day of driving, only 60% of a year of driving has been completed. As

Page 196: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

175

only a small number of cycles have been completed, the results are too close to call and still

inconclusive at this point.

The variation in the data points seem to be more of noise due to experiment variations. For

example, the additional ‘contact’ resistance in the battery-only setup was introduced only

after 60 cycles. In the 100 to 130 cycles section, there was a rise in capacity. This was due to

an increase in contact resistance for both setups. The connectors were found to be corroding

due to the extra flux used in soldering the wire and connectors. The rise in capacity was due

to the smaller gradient at the start of the discharge versus the steeper gradient at the end as

shown in Figure 6-46. Therefore, a larger section of capacity was gained as compared to the

capacity lost. The connectors were eventually replaced after 130 cycles, and no extra flux was

used, only the solder rosin core.

Figure 6-46 Capacity “gained” due to high contact resistance

In short, the contact resistance causes some minor capacity gain to be logged. As a side note,

at first glance, it might seem there is some conflict, because earlier in section 6.7.6 Initial

Results, it was said that the contact resistance worsens the capacity loss for the battery/UC

system. There is no conflict, because in section 6.7.6, the extra contact resistance stresses the

battery in the battery/UC setup more during the drive cycle tests, leading to higher C-rates

High contact resistance

Low contact resistance

Capacity gained Capacity lost Capacity (Ah)

Volt

age

(V)

Page 197: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

176

and more capacity loss. But in this section, only the capacity test is discussed, not the drive

cycle tests, and it is applied to both setups.

As mentioned earlier in section 6.7.2, when Wang, et al. performed constant 0.5C charge and

discharge accelerated cycling tests, they took 200 cycles to observe a noticeable drop of 5%,

and 5000 cycles (two years in real-time) for a 30% drop. As the average C-rate for this

battery/UC setup under the FTP-75 city drive cycle is only 0.24C (half of their 0.5C), it

would take even longer for similar results.

6.7.11 Summary

In Experiment 2, the battery/UC setup and the battery-only setup was cycled to compare the

cycle life. The purpose was to verify the ultimate HESS goal of extended battery cycle life.

After 190 cycles, the results are still inconclusive. However, based on the works of Wang, et

al, they showed that a lower C-rate leads to a longer battery cycle life. From earlier

Experiment 1, section 6.6.4, it was shown that the proposed battery/UC setup indeed has a

lower averaged absolute C-rate for the battery as compared to that in the battery-only setup.

If the experiment was run over a longer period, the capacity drop of the battery-only system

should be more severe than that of the battery/UC HESS setup.

Page 198: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

177

7 HESS REDUCED-SCALED SIMULATIONS

Earlier in section 6.5 Software Implementation, the scale factor k = 160 for a reduced-scale

experimental implementation was discussed. In this section, k = 160 was implemented to

create a reduced-scale simulation. The purpose was to allow a comparison between the results

of Experiment 1 and the reduced-scale simulation. They are compared in terms of behaviour

over time and in terms of energy consumption for the EUDC and FTP-75 city drive cycles.

7.1 Differences between Full and Reduced-scale Simulation

There are some differences between the reduced-scale simulation and full-scale simulation.

First, the additional contact resistance was added to the reduced-scale simulation. Although

the measured resistance was 0.07Ω, 0.2Ω was used instead as it was a better fit to match the

experiment and simulation. In general, the larger the contact resistance in the simulation, the

more the battery voltage would dip during drive cycle peaks, leading to the current drawn

surging. So with some iterative tuning, 0.2Ω matched the current surge better.

Second, the battery power limits in the full-scale simulation was replaced by battery current

limits to match the experiment as explained in section 6.5.4.

Thirdly, in the full-scale battery-only simulation, the maximum regenerative braking was

clipped to -30kW to match the Nissan Leaf. However, no clipping was performed in the

experiment. Therefore, this clipping was removed in the reduced-scale simulation.

Lastly, the FTP-75 city drive cycle in the reduced-scale simulation was stretched by 1.095 to

match the slow running Maccor, as explained in section 6.5.3. However, as illustrated by

earlier Figure 6-29, the Maccor does not stretch the drive cycles linearly, some sections are

stretched more than others.

Page 199: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

178

7.2 EUDC

Figure 7-1 Battery & UC currents from simulation, EUDC (compare with experiment in

Figure 6-18).

Figure 7-2 Battery & UC voltages from simulation, EUDC (compare with experiment in

Figure 6-19).

Page 200: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

179

Figure 7-1 and Figure 7-2 show the reduced-scale simulation results for the EUDC. It can be

seen that they are quite similar to the experiment results shown earlier (Figure 6-18 and

Figure 6-19) in terms of behaviour (shape) with respect to time.

From the figures, the battery current has been clipped to the limits from 320s to 340s, and the

UC voltage deviates from the target voltage to handle the excessive power required, just like

what was seen in the experiment in section 6.6.2. Therefore, both the experiment and

reduced-scale simulation behave similarly time-wise.

For an additional reference, the full-scale simulation results were presented in Figure 5-12

and Figure 5-13 in section 5.4.2. They are also similar in terms of battery and UC behaviour.

7.3 FTP-75 City Drive Cycle

The following figures show the reduced-scale FTP-75 city drive cycle simulations, with a

zoomed-in section just like the experiments. Again, they are quite similar to the experiments

in terms of their behaviour with respect to time.

Figure 7-3 Battery & UC currents from simulation, FTP-75 city (compare with experiment in

Figure 6-23).

Page 201: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

180

Figure 7-4 Battery & UC voltages from simulation, FTP-75 city (compare with experiment in

Figure 6-24).

Figure 7-5 Battery & UC currents from simulation, FTP-75 city, zoomed 180-280s (compare

with experiment in Figure 6-26).

Page 202: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

181

Figure 7-6 Battery & UC voltages from simulation, FTP-75 city, zoomed 180-280s (compare

with experiment in Figure 6-27).

Again, the battery current is clipped to the limits from 205s to 220s, and the UC deviates

from the target voltage to handle the excessive power, just like what was seen in the

experiment in section 6.6.3. Therefore, both the experiment and reduced-scale simulation

behave similarly time-wise.

For an additional reference, the full-scale simulation results were presented in Figure 5-14 to

Figure 5-17 in section 5.4.3. They are also similar in terms of battery and UC behaviour.

7.4 Total Energy Use

Here, the energy consumption of the battery in the experiments is compared to that of the

reduced-scale simulations. The energy consumption was calculated by multiplying the logged

battery voltage and current data, and then integrating it over time. Three runs were performed

for both the EUDC and FTP-75 city drive cycles and the results were averaged as shown in

Table 7-1.

Page 203: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

182

Table 7-1 Experiment battery energy use.

EUDC FTP-75 City

Run no. 1 2 3 Avg. 1 2 3 Avg.

Battery-only setup (kJ) - - - - 38.88 38.78 39.12 38.93

Battery/UC setup (kJ) 21.17 21.37 21.52 21.35 38.31 38.89 39.38 38.86

Then the averaged results was compared to the reduced-scale simulation results in Table 7-2.

There is a percentage difference of at most 10.6% (seen in the battery-only FTP-75 city case),

so the simulation is a reasonable estimation of the energy consumption of the experimental

setup. Note that as mentioned earlier in section 6.6.4, the EUDC battery-only setup was

unable to complete due to undervoltage occurring.

Table 7-2 Comparison of battery energy use between experiments and reduced-scale

simulations.

EUDC FTP-75 City

Battery-only setup Experiment (Avg.) (kJ) - 38.93

Simulation (kJ) - 43.04

Battery/UC setup Experiment (Avg.) (kJ) 21.35 38.86

Simulation (kJ) 21.56 42.15

From the results, the FTP-75 city drive cycle exhibits a bigger difference between the

reduced-scale simulation and experiment results as compared to the EUDC. This is not

surprising as the FTP-75 city drive cycle suffers from the Maccor speed problems discussed

in 6.5.3 and 6.6.3 and illustrated in Figure 6-29 earlier, where the actual power drawn by the

Maccor drifts in and out of sync with the programmed power.

Page 204: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

183

As a side note, the battery energy consumption when comparing the battery-only and

battery/UC case are very similar for both the experiment and simulation (i.e. battery-only

simulation consumes 43.04kJ vs. 42.15kJ in the battery/UC simulation). This is because the

UC only has a slight contribution to the total energy use. The UC is fully charged at the start

of the drive cycle, varies during the drive cycle, and is fully charged again at the end of the

drive cycle. Ultimately, the battery provides the energy to move the car from one location to

another location, so the energy consumption should be similar.

There are three possible reasons why the experiment results differ from the reduced-scale

simulation results. The first is the Maccor sync problems mentioned above, and the second

may be due to the DC/DC converter. The simulation uses an interpolated DC/DC converter

efficiency chart as discussed in 6.4 Experiment 0, so there is still some difference between

the actual DC/DC converter efficiency and simulated efficiency. The third reason may be due

to the simulation being a backward approach, where transients of the electrical components

such as the MOSFETs, inductors, capacitors, etc. are not considered.

7.5 Summary

To summarise, the reduced-scale drive cycle simulations give a reasonably accurate

prediction of the behaviour of the actual battery and UC currents and voltages time-wise. The

proposed HESS management strategy in both the experiment and reduced-scale simulations

perform similar manoeuvres to utilise the UC to prevent the battery from exceeding its limits.

In addition, the reduced-scale simulation is a reasonably accurate representation of the

experiment in terms of battery energy usage, where only an 11% difference was observed.

Page 205: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

184

8 CONCLUSION & FUTURE WORKS

8.1 Conclusion

In this work, a novel HESS management strategy was proposed. The proposed EMS involved

a more rigorous method of setting the target UC energy level using a speed-dependent band,

which is the first contribution of this thesis. This allows the UC to achieve two goals – to

contain sufficient energy required for future accelerations, and to have sufficient space to

store energy captured from future regenerative braking. Such rigorous calculations and

justifications based on averaged worst case scenarios and real-life drive cycles are not seen in

existing literature. As the calculations were based on worst case scenarios, knowledge of the

future drive profile is not required.

The proposed PMS also has two goals – to ensure the EMS (target UC energy level) is

followed, and to ensure the battery charge and discharge rates do not exceed specified limits

in order to extend the battery cycle life. The second contribution of this thesis is that the

specified battery limits are speed-dependent. This allows it to achieve two goals – better

utilization of the UC and allowing the battery to supply the steady state constant speed power.

In addition, by using the proposed HESS design methodology, the UC can be appropriately

sized to reduce weight and costs.

From simulations, by running the mid-sized EV over the EUDC and LA92 drive cycles, it has

been demonstrated that the proposed strategy achieves the two EMS goals, the two PMS

goals, and the two battery limit goals mentioned above. Furthermore, despite the algorithm

being designed with averaged worst case scenarios, it was able to tolerate harsh drive cycles.

In addition, the simulations showed that existing works cannot always achieve the two EMS

goals simultaneously unless their UCs are sized twice as large (especially for those with fixed

target UC energy levels), increasing weight and costs. Similarly, it was shown the proposed

Page 206: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

185

speed-dependent battery power limit allows the battery to supply the steady state power and

achieves better UC utilization as compared to other rule-based deterministic PMS.

Subsequently, battery cycle life simulations were performed to observe the fall in battery

capacity for the proposed battery/UC HESS, and for a battery-only system. Almost 30%

reduction in capacity loss due to cycling was seen for the battery/UC HESS as compared to

the battery-only system when running three FTP-75 city drive cycles daily over 10 years.

This is because the battery/UC HESS is subject to a smaller C-rate as compared to the

battery-only system.

Afterwards, a reduced-scale experiment was created. A preliminary Experiment 0 was

conducted to evaluate the DC/DC converter efficiency. Then Experiment 1 verified the

proposed strategy could work physically and satisfy the four HESS goals and the two battery

limit goals. Also, the results showed that the battery/UC HESS has a lower average of

absolute current as compared to the battery-only system.

In addition, the experiment showed that the EUDC cycle is only able to run in the battery/UC

HESS setup for the selected scale. The battery-only setup was unable to run as it could not

supply the required current demand, demonstrating an extremely useful application of the

battery/UC HESS.

Subsequently, a reduced-scale simulation was created and compared to the experiments. The

results from both are reasonably similar in terms of energy consumption, where only an 11%

difference was observed. Also, they behaved similarly time-wise, for example, the battery

was about to exceed its limits in the same section for both the experiment and reduced-scale

simulations, and the algorithm in both the experiment and simulations performed similar

manoeuvres to utilise the UC to prevent the battery from exceeding its limits.

Page 207: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

186

Lastly, Experiment 2 was performed to compare the battery cycle life of the battery/UC

HESS to the battery-only system. Each setup was continuously cycled with the FTP-75 city

drive cycle. The results are too close to call due to the small number of cycles run. However,

based on the works of Wang, et al., a smaller charge/discharge current (which the battery/UC

HESS achieves) leads to a longer battery cycle life.

8.2 Future Works

There are still further research works which can be done to expand or improve the work.

8.2.1 Optimization to Extend Battery Cycle Life

In section 5.5, battery cycle life simulation results were presented for the proposed algorithm.

However, this may not be the best performance from the proposed algorithm. Optimization

can be performed to further improve the battery cycle life.

There are many variables which can be adjusted, for example in the PMS discussed in section

4.1.1 regarding the battery limits, the selected values of maximum battery power of 1.7C or

the battery power scaling of Pbatt,max,scale = 3 could be optimized to get the longest cycle life.

Similarly for the EMS, the selected brake torque of τbr,tar = 244 Nm and acceleration torque

ratio τacc ratio = 0.5 in section 4.2.3 could also be optimized.

This would be a multi-objective optimization, where the objectives are to extend battery cycle

life, and to ensure the two EMS and two PMS goals are met. Satisfying the EMS and PMS

goals would allow the algorithm to tolerate harsh drive cycles outside its design range, like

the LA92 as discussed in section 5.4.1. Also, multiple drive cycles would need to be

considered, as the optimal solution for one drive cycle may not be the optimal for another

drive cycle.

Page 208: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

187

8.2.2 Full-scale Implementation

In this work, only a reduced-scale experiment was performed. Further research can be done

with a full-scale implementation in an EV, such as a Nissan Leaf. For a full-scale

implementation, here are some suggestions by the author.

Sensor readings are expected to be noisy when implemented in a real EV, with many

disturbances. As the algorithm responds in real-time to the power demanded, for example,

limiting the battery power, care must be taken in selecting the filter cutoff values. A low

cutoff will lead to more stable readings, but will also cause excessive lag.

Also, as the battery power or current limiting is hugely dependent on the battery current

sensor, it might be wise to have a duplicate sensor for redundancy. In the experiment, wrong

sensor readings were occasionally encountered, resulting in the UC unable to assist the

battery at the appropriate times and blowing the circuit fuses. Furthermore, as the Hall-effect

current sensors used in the experiment were rather sensitive to disturbances, other techniques

of measuring current could be considered, for example, using shunt resistors.

In addition, effort should be spent studying contact resistance. Due to the vibrations

experienced by a real vehicle, contact resistance is expected to be a bigger concern as

compared to the stationary experiment performed in this work, as connections may become

loose over time.

Lastly, a gradient sensor and weight sensor are required for more accurate implementation. In

the simulations and experiments in this work, all drive cycles were for flat terrain. The

gradient sensor will feed the present gradient θ(j) into the vehicle model in section 3.6.

Similarly, the simulations and experiments were performed with constant weight. Weight

sensors, in the form of strain gauges, could be installed on the suspension to weigh the

loading on the EV. The sensor does not need to continually weigh the loading. A good

Page 209: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

188

suggestion would be to perform the weighing every time the doors close as the loading is

most likely to change then.

The different weight or gradient encountered by the EV would produce a different target UC

band. Therefore, each weight or gradient would have its own curve-fitting polynomial. The

author suggests dividing all the combinations into a few bands, e.g. a load of 50-80kg would

have its own curve-fitting polynomial, while 80-110kg would have another polynomial to

reduce the number of options.

8.2.3 SuPower Battery Cycle Life Curve Fitting

Earlier in section 3.11.2, assembling Sanyo UR18650W batteries to a 98S44P configuration

to match the Nissan Leaf battery capacity was discussed. The reason for doing this is because

an empirically fitted battery cycle life model was available for the Sanyo UR18650W

batteries from Wang, et al [7], but not for the Nissan Leaf batteries.

In future, a cycle life curve fitting can be performed on the SuPower batteries used in the

experiments. As only 190 cycles have been performed in Experiment 2, there is not enough

data to extract the required parameters to fit the model at this moment.

When an appropriate number of cycles have been performed and the battery capacity fade

curve shape is clearly visible, a battery cycle life model for the SuPower batteries could be

created, and its cycle life could be simulated. This could be used to predict the cycle life of

the SuPower batteries, and there could be less emphasis on running the cycle life experiments

since a simulation is available.

Also, the SuPower batteries are more modern, and may not suffer the severe calendar life

losses faced by the 2007 Sanyo UR18650W. Therefore, the mid-sized EV in the full-scale

simulation can be modified to use SuPower batteries instead of the Sanyo batteries for an

improved cycle life simulation.

Page 210: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

189

8.2.4 Cost-Benefit Analysis of UCs

After understanding the cycle life of modern batteries, a cost benefit analysis could be

performed to determine what circumstances would favour a battery/UC HESS as compared to

a battery-only setup in terms of costs in the long run.

From the author’s analysis, the most cost beneficial scenario is when the EV has heavy city

start-stop driving, for example, a public bus. From the cycle life simulations earlier, a city

drive cycle extends the battery cycle life the most. In addition, with heavy daily usage, the

battery capacity loss due to cycle life would dominate the capacity loss due to calendar life.

This is beneficial as the battery/UC HESS can only control battery capacity loss due to cycle

life.

8.2.5 Improvement to Experiments

Further improvements can be made to the experiment, which is specific to this work. To

make the experiment measurements more accurate, instead of adding resistors to the battery-

only experiment setup to match the contact resistance of the battery/UC HESS experiment

setup, effort should be spent on reducing the contact resistance in the green box battery/UC

setup. For example, quick connect crimp connectors were used for easy troubleshooting and

swapping out of parts. For convenience, existing quick connect crimp connectors in the

laboratory were used. However, some connectors have been idle for many years, such that the

surface is no longer shiny, but slightly oxidised. New connectors should have been bought

instead.

Page 211: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

190

BIBLIOGRAPHY

[1] J. Xue, “Electric vehicles ‘not economically feasible yet’,” TODAY, p. 2, 9 December

2014.

[2] Nissan USA, “Nissan Leaf Electric Car: 100% Electric. 100% Fun.,” 2014. [Online].

Available: http://www.nissanusa.com/electric-cars/leaf/. [Accessed 10 December 2014].

[3] Nissan USA, “2014 Nissan Sentra Sedan,” 2014. [Online]. Available:

http://www.nissanusa.com/cars/sentra. [Accessed 10 December 2014].

[4] C. Gaylord, “Hybrid cars 101: How long should batteries last?,” 6 March 2012.

[Online]. Available: http://www.csmonitor.com/Innovation/2012/0306/Hybrid-cars-101-

How-long-should-batteries-last. [Accessed 18 December 2017].

[5] T. Bradley, “Replacing A Dead Prius Hybrid Battery Doesn't Have To Cost Thousands

Of Dollars,” 9 April 2014. [Online]. Available:

http://www.forbes.com/sites/tonybradley/2014/04/09/replacing-a-dead-prius-hybrid-

battery-doesnt-have-to-cost-thousands-of-dollars/. [Accessed 18 December 2017].

[6] J. Voelcker, “Nissan Leaf New Battery Cost: $5,500 For Replacement With Heat-

Resistant Chemistry,” 28 June 2014. [Online]. Available:

http://www.greencarreports.com/news/1092983_nissan-leaf-battery-cost-5500-for-

replacement-with-heat-resistant-chemistry. [Accessed 18 December 2017].

[7] J. Wang, J. Purewal, P. Liu, J. Hicks-Garner, S. Soukazian, E. Sherman, A. Sorenson, L.

Vu, H. Tataria and M. W. Verbrugge, “Degradation of lithium ion batteries employing

graphite negatives and nickel-cobalt-manganese oxide + spinel manganese oxide

Page 212: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

191

positives: Part 1, aging mechanisms and life estimation,” Journal of Power Sources, vol.

269, pp. 937-948, 2014.

[8] J. C. Burns, D. A. Stevens and J. R. Dahn, “In-Situ Detection of Lithium Plating Using

High Precision Coulometry,” Journal of The Electrochemical Society, vol. 162, no. 6,

pp. A959-A964, 2015.

[9] Q. Cai, D. J. L. Brett, D. Browning and N. Brandon, “A sizing-design methodology for

hybrid fuel cell power systems and its application to an unmanned underwater vehicle,”

Journal of Power Sources, vol. 195, no. 19, pp. 6559-6569, 2010.

[10] Battery University, “BU-209: How does a Supercapacitor Work?,” 21 April 2017.

[Online]. Available:

http://batteryuniversity.com/learn/article/whats_the_role_of_the_supercapacitor.

[Accessed 18 December 2017].

[11] A. Khajepour, S. Fallah and A. Goodarzi, Electric and Hybrid Vehicles: Technologies,

Modeling and Control - A Mechatronic Approach, West Sussex: John Wiley & Sons,

2014.

[12] Maxwell Technologies, “48 Volt Module Ultracapacitor General Purpose Module,”

2014. [Online]. Available: http://www.maxwell.com/products/ultracapacitors/48v-

modules. [Accessed 18 December 2017].

[13] Mouser Electronics, “BMOD0083 P048 B01 Maxwell Technologies,” 2017. [Online].

Available: http://www.mouser.com/ProductDetail/Maxwell-Technologies/BMOD0083-

P048-B01/. [Accessed 18 December 2017].

[14] K. M. So, Y. S. Wong, G. S. Hong and W. F. Lu, “An Improved Energy Management

Page 213: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

192

Strategy for a Battery/Ultracapacitor Hybrid Energy Storage System in Electric

Vehicles,” in IEEE Transportation Electrification Conference and Expo, Detroit, MI,

2016.

[15] K. M. So, G. S. Hong, W. F. Lu and Y. S. Wong, “An Improved Speed-dependent

Battery/Ultracapacitor Hybrid Energy Storage System Management Strategy for Electric

Vehicles,” IEEE Transactions on Transportation Electrification, [Under Review].

[16] Q. Xu, X. Hu, P. Wang, J. Xiao, P. Tu, C. Wen and M. Y. Lee, “A Decentralized

Dynamic Power Sharing Strategy for Hybrid Energy Storage System in Autonomous

DC Microgrid,” IEEE Transactions on Industrial Electronics, vol. 64, no. 7, pp. 5930-

5941, 2017.

[17] S. K. Kollimalla, M. K. Mishra, A. Ukil and H. B. Gooi, “DC Grid Voltage Regulation

Using New HESS Control Strategy,” IEEE Transactions on Sustainable Energy, vol. 8,

no. 2, pp. 772-781, 2017.

[18] U. Manandhar, A. Ukil, H. B. Gooi, N. R. Tummuru, S. K. Kollimalla, B. Wang and K.

Chaudhari, “Energy Management and Control for Grid Connected Hybrid Energy

Storage System under Different Operating Modes,” IEEE Transactions on Smart Grid,

vol. PP, no. 99, 2017.

[19] M. M. S. Khan, M. O. Faruque and A. Newaz, “Fuzzy Logic Based Energy Storage

Management System for MVDC Power System of All Electric Ship,” IEEE

Transactions on Energy Conversion, vol. 32, no. 2, pp. 798-809, 2017.

[20] J. Hou, J. Sun and H. F. Hofmann, “Mitigating Power Fluctuations in Electric Ship

Propulsion With Hybrid Energy Storage System: Design and Analysis,” IEEE Journal of

Oceanic Engineering, vol. PP, no. 99, pp. 1-15, 2017.

Page 214: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

193

[21] J. P. Trovão, F. Machado and P. G. Pereirinha, “Hybrid electric excursion ships power

supply system based on a multiple energy storage system,” IET Electrical Systems in

Transportation, vol. 6, no. 3, pp. 190-201, 2016.

[22] J. P. Torreglosa, P. Garcia, L. M. Fernández and F. Jurado, “Predictive Control for the

Energy Management of a Fuel-Cell-Battery-Supercapacitor Tramway,” IEEE

Transactions on Industrial Informatics, vol. 10, no. 1, pp. 276-285, 2014.

[23] P. García, J. P. Torreglosa, L. M. Fernándeza and F. Jurado, “Control strategies for high-

power electric vehicles powered by hydrogen fuel cell, battery and supercapacitor,”

Expert Systems with Applications, vol. 40, no. 12, pp. 4791-4804, 2013.

[24] A. A. Ferreira, J. A. Pomilio, G. Spiazzi and L. de Araujo Silva, “Energy Management

Fuzzy Logic Supervisory for Electric Vehicle Power Supplies System,” IEEE

Transactions on Power Electronics, vol. 23, no. 1, pp. 107-115, 2008.

[25] P. Thounthong, S. Raël and B. Davat, “Energy management of fuel

cell/battery/supercapacitor hybrid power source for vehicle applications,” Journal of

Power Sources, vol. 193, no. 1, pp. 376-385, 2009.

[26] Amin, R. T. Bambang, A. S. Rohman, C. J. Dronkers, R. Ortega and A. Sasongko,

“Energy Management of Fuel Cell/Battery/Supercapacitor Hybrid Power Sources Using

Model Predictive Control,” IEEE Transactions on Industrial Informatics, vol. 10, no. 4,

pp. 1992-2002, 2014.

[27] P. Thounthong, V. Chunkag, P. Sethakul, S. Sikkabut, S. Pierfederici and B. Davat,

“Energy management of fuel cell/solar cell/supercapacitor hybrid power source,”

Journal of Power Sources, vol. 196, no. 1, pp. 313-324, 2011.

Page 215: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

194

[28] S. F. Tie and C. W. Tan, “A review of energy sources and energy management system in

electric vehicles,” Renewable and Sustainable Energy Reviews, vol. 20, pp. 82-102,

2013.

[29] S. Collins, “Flywheel hybrid systems (KERS),” 4 April 2011. [Online]. Available:

http://www.racecar-engineering.com/articles/f1/flywheel-hybrid-systems-kers/.

[Accessed 18 December 2017].

[30] M. I. Masouleh and D. J. N. Limebeer, “Fuel Minimization for a Vehicle Equipped With

a Flywheel and Battery on a Three-Dimensional Track,” IEEE Transactions on

Intelligent Vehicles, vol. 2, no. 3, pp. 161-174, 2017.

[31] F. A. Bender, M. Kaszynski and O. Sawodny, “Drive Cycle Prediction and Energy

Management Optimization for Hybrid Hydraulic Vehicles,” IEEE Transactions on

Vehicular Technology, vol. 62, no. 8, pp. 3581-3592, 2013.

[32] J. Cao and A. Emadi, “A New Battery/UltraCapacitor Hybrid Energy Storage System

for Electric, Hybrid, and Plug-In Hybrid Electric Vehicles,” IEEE Transactions on

Power Electronics, vol. 27, no. 1, pp. 122-132 , 2011.

[33] A. Ostadi, M. Kazerani and S.-K. Chen, “Hybrid Energy Storage System (HESS) in

vehicular applications: A review on interfacing battery and ultra-capacitor units,” in

IEEE Transportation Electrification Conference and Expo (ITEC), Detroit, MI, 2013.

[34] U.S. Department of Energy Vehicle Technologies Program, “Advanced Vehicle Testing

- Beginning-of-Test Battery Testing Results,” 2012. [Online]. Available:

http://media3.ev-tv.me/DOEleaftest.pdf. [Accessed 18 December 2017].

[35] S. Hu, Z. Liang and X. He, “Ultracapacitor-Battery Hybrid Energy Storage System

Page 216: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

195

Based on the Asymmetric Bidirectional Z-Source Topology for EV,” IEEE Transactions

on Power Electronics, vol. 31, no. 11, pp. 7489-7498, 2016.

[36] F. Naseri, E. Farjah and T. Ghanbari, “An Efficient Regenerative Braking System Based

on Battery/Supercapacitor for Electric, Hybrid, and Plug-In Hybrid Electric Vehicles

With BLDC Motor,” IEEE Transactions on Vehicular Technology, vol. 66, no. 5, pp.

3724-3738, 2017.

[37] W. O. Avelino, F. S. Garcia, A. A. Ferreira and J. A. Pomilio, “Electric go-kart with

battery-ultracapacitor hybrid energy storage system,” in IEEE Transportation

Electrification Conference and Expo, Detroit, MI, 2013.

[38] F. S. Garcia, A. A. Ferreira and J. A. Pomilio, “Control Strategy for Battery-

Ultracapacitor Hybrid Energy Storage System,” in Applied Power Electronics

Conference and Exposition (APEC), Washington, DC, USA , 2009.

[39] S. Dusmez and A. Khaligh, “A Supervisory Power-Splitting Approach for a New

Ultracapacitor-Battery Vehicle Deploying Two Propulsion Machines,” IEEE

Transactions on Industrial Informatics, vol. 10, no. 3, pp. 1960-1971 , 2014.

[40] B. Hredzak, V. G. Agelidis and G. D. Demetriades, “A Low Complexity Control System

for a Hybrid DC Power Source Based on Ultracapacitor-Lead-Acid Battery

Configuration,” IEEE Transactions on Power Electronics, vol. 29, no. 6, pp. 2882-2891,

2014.

[41] B. Hredzak, V. G. Agelidis and M. Jang, “A Model Predictive Control System for a

Hybrid Battery-Ultracapacitor Power Source,” IEEE Transactions on Power

Electronics, vol. 29, no. 3, pp. 1469-1479, 2014.

Page 217: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

196

[42] Z. Yu, D. Zinger and A. Bose, “An innovative optimal power allocation strategy for fuel

cell, battery and supercapacitor hybrid electric vehicle,” Journal of Power Sources, vol.

196, no. 4, pp. 2351-2359, 2011.

[43] C. Romaus, J. Böcker, K. Witting, A. Seifried and O. Znamenshchykov, “Optimal

energy management for a hybrid energy storage system combining batteries and double

layer capacitors,” in IEEE Energy Conversion Congress and Exposition, San Jose, CA,

2009.

[44] P. Golchoubian and N. L. Azad, “Real-Time Nonlinear Model Predictive Control of a

Battery–Supercapacitor Hybrid Energy Storage System in Electric Vehicles,” IEEE

Transactions on Vehicular Technology, vol. 66, no. 11, pp. 9678-9688, 2017.

[45] M.-E. Choi, J.-S. Lee and S.-W. Seo, “Real-Time Optimization for Power Management

Systems of a Battery/Supercapacitor Hybrid Energy Storage System in Electric

Vehicles,” IEEE Transactions on Vehicular Technology, vol. 63, no. 8, pp. 3600-3611,

2014.

[46] J. Moreno, M. E. Ortúzar and J. W. Dixon, “Energy-management system for a hybrid

electric vehicle, using ultracapacitors and neural networks,” IEEE Transactions on

Industrial Electronics, vol. 53, no. 2, pp. 614-623, 2006.

[47] J. W. Dixon and M. E. Ortúzar, “Ultracapacitors + DC-DC converters in regenerative

braking system,” IEEE Aerospace and Electronic Systems Magazine, vol. 17, no. 8, pp.

16-21, 2002.

[48] R. Carter, A. Cruden and P. J. Hall, “Optimizing for Efficiency or Battery Life in a

Battery/Supercapacitor Electric Vehicle,” IEEE Transactions on Vehicular Technology,

vol. 61, no. 4, pp. 1526-1533, 2012.

Page 218: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

197

[49] J. Armenta, C. Núñez, N. Visairo and I. Lázaro, “An advanced energy management

system for controlling the ultracapacitor discharge and improving the electric vehicle

range,” Journal of Power Sources, vol. 284, pp. 452-458, 2015.

[50] F. R. Salmasi, “Control Strategies for Hybrid Electric Vehicles: Evolution,

Classification, Comparison, and Future Trends,” IEEE Transactions on Vehicular

Technology, vol. 56, no. 5, pp. 2393-2404, 2007.

[51] K. Ç. Bayindir, M. A. Gözüküçük and Ahmet Teke, “A comprehensive overview of

hybrid electric vehicle: Powertrain configurations, powertrain control techniques and

electronic control units,” Energy Conversion and Management, vol. 52, no. 2, pp. 1305-

1313, 2011.

[52] E. D. Tate and S. P. Boyd, “Finding Ultimate Limits of Performance for Hybrid Electric

Vehicles,” in SAE Future Transportation Technology Conference, Costa Mesa, CA,

2000.

[53] J. Shen and A. Khaligh, “A Supervisory Energy Management Control Strategy in a

Battery/Ultracapacitor Hybrid Energy Storage System,” IEEE Transactions on

Transportation Electrification, vol. 1, no. 3, pp. 223-231, 2015.

[54] M. E. Ortúzar, J. Moreno and J. W. Dixon, “Ultracapacitor-Based Auxiliary Energy

System for an Electric Vehicle: Implementation and Evaluation,” IEEE Transactions on

Industrial Electronics, vol. 54, no. 4, pp. 2147-2156 , 2007.

[55] United States Environmental Protection Agency, “Dynamometer Drive Schedules,” 1

August 2013. [Online]. Available: https://www.epa.gov/vehicle-and-fuel-emissions-

testing/dynamometer-drive-schedules. [Accessed 18 December 2017].

Page 219: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

198

[56] T. Markel, A. Brooker, T. Hendricks, V. Johnson, K. Kelly, B. Kramer, M. O'Keefe, S.

Sprik and K. Wipke, “ADVISOR: a systems analysis tool for advanced vehicle

modeling,” Journal of Power Sources, vol. 110, no. 2, pp. 255-266, 2002.

[57] Nissan Newsroom Europe, “New Nissan Leaf Technical Specification,” 8 April 2013.

[Online]. Available: http://www.newsroom.nissan-europe.com/EU/en-

gb/NEW_LEAF/Product/TechnicalSpecs.aspx. [Accessed 10 December 2014].

[58] Z. Stević, New Generation of Electric Vehicles, Rijeka: InTech, 2012.

[59] SAE International, “Power from Within,” Vehicle Electrification, p. 17, 23 February

2011.

[60] M. Allen, “Real-world range ramifications: heating and air conditioning,” 22 January

2014. [Online]. Available: http://www.fleetcarma.com/electric-vehicle-heating-

chevrolet-volt-nissan-leaf/. [Accessed 18 December 2017].

[61] M. H. Rashid, Modern Electric, Hybrid Electric and Fuel Cell Vehicles - Fundamentals,

Theory and Design, Boca Raton: CRC Press LLC, 2005.

[62] Nissan Motor Corporation, “EDIB (Electric Driven Intelligent Brake),” n.d.. [Online].

Available: http://www.nissan-global.com/EN/TECHNOLOGY/OVERVIEW/edib.html.

[Accessed 18 December 2017].

[63] O. Tremblay, L.-A. Dessaint and A.-I. Dekkiche, “A Generic Battery Model for the

Dynamic Simulation of Hybrid Electric Vehicles,” in IEEE Vehicle Power and

Propulsion Conference, Arlington, TX, 2007.

[64] L. Shi and M. L. Crow, “Comparison of ultracapacitor electric circuit models,” in IEEE

Power and Energy Society General Meeting - Conversion and Delivery of Electrical

Page 220: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

199

Energy in the 21st Century, Pittsburgh, PA, 2008.

[65] R. W. Erickson and D. Maksimovic, Fundamentals of Power Electronics, New York:

Springer Science+Business Media, LLC, 2001.

[66] J. Shen, S. Dusmez and A. Khaligh, “Optimization of Sizing and Battery Cycle Life in

Battery/Ultracapacitor Hybrid Energy Storage Systems for Electric Vehicle

Applications,” IEEE Transactions on Industrial Informatics, vol. 10, no. 4, pp. 2112-

2121, 2014.

[67] J. Wang, P. Liu, J. Hicks-Garner, E. Sherman, S. Soukiazian, M. Verbrugge, H. Tataria,

J. Musser and P. Finamore, “Cycle-life model for graphite-LiFePO4 cells,” Journal of

Power Sources, vol. 196, no. 8, pp. 3942-3948, 2011.

[68] EcoModder, “Vehicle Coefficient of Drag List,” 12 September 2014. [Online].

Available: http://ecomodder.com/wiki/index.php/Vehicle_Coefficient_of_Drag_List.

[Accessed 18 December 2017].

[69] U.S. Department of Transportation, National Highway Traffic Safety Administration,

“Laboratory Test Procedure for Rollover Stability Measurement for NCAP: SSF

Measurement,” March 2013. [Online]. Available:

http://www.safercar.gov/staticfiles/safercar/NCAP/SSF_Test_Procedure-

March2013.pdf. [Accessed 18 December 2017].

[70] Powerex, “PM400DV1A060 Intellimod Module Data Sheet,” 2012. [Online]. Available:

http://www.pwrx.com/Product/PM400DV1A060. [Accessed 18 December 2017].

[71] S. Blanco, “Second Drive: 2011 Nissan Leaf - Some things you probably didn't know,”

22 October 2010. [Online]. Available: https://www.autoblog.com/2010/10/22/2011-

Page 221: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

200

nissan-leaf-review-drive-second/. [Accessed 18 December 2017].

[72] H. Zumbahlen, “Staying Well Grounded,” Analog Dialogue, vol. 46, no. 06, 2012.

[73] Crane Aerospace & Electronics Power Solution, “Measurement and Filtering of Output

Noise of DC-DC Converters,” 29 August 2016. [Online]. Available:

http://www.interpoint.com/product_documents/DC_DC_Converters_Output_Noise.pdf.

[Accessed 18 December 2017].

[74] A. Martin, M. Davis-Marsh, G. Pinto and I. Jorio, “Capacitor Selection for DC/DC

Converters,” 2012. [Online]. Available:

http://www.kemet.com/Lists/TechnicalArticles/Attachments/5/Avnet2012PowerForum_

CapacitorsSelection.pdf. [Accessed 18 December 2017].

[75] Cooper Bussmann, “1/4" x 1-1/4" Fuses, AGC Series, Fast Acting, Glass Tube,” May

2017. [Online]. Available:

http://www.cooperindustries.com/content/dam/public/bussmann/Electronics/Resources/p

roduct-datasheets/Bus_Elx_DS_OC-2543_AGC_Series.pdf. [Accessed 18 December

2017].

Page 222: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

201

APPENDIX

A Charging Procedure

This is the charging procedure for the first two steps of the battery capacity tests in section

6.7.4, which was summarised earlier in Table 6-13. The following figures show battery 3

charging to 20.2V as an example.

Figure A-1 Charging current (b) coulomb counting.

Figure A-1(a) shows the charging current as logged by sensor ACS3. CC 4.7A charging

occurs from 0s to about 3900s, while the remaining section is CV charging. Figure A-1 (b) is

the amount of charge which has been charged to the battery so far (The plot is in the negative

region because charging has a negative convention).

Figure A-2 Charging voltage.

Page 223: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

202

Figure A-2 shows the battery voltage over time. Again, the CC region is from 0s to about

3900s, and 20.2V CV charging takes up the remainder of the time.

B Scaling

In section 6.2, dimensional analysis for scaling the experiment was discussed. The full

dimensional analysis workings are found here.

The torque is dependent on 7 variables (n=7), with 3 base dimensions (m=3).

𝑇 = 𝑓(𝜌, 𝜇, 𝑣, 𝐿, 𝜔, 𝐹) (A-1)

The meaning of the symbols are shown in the following table,

Table A-1 Symbols and their meanings for scaling derivation.

Symbol Meaning Symbol Meaning

k Scaling ratio M Mass, basic dimension

τ Torque L Length, basic dimension

ρ Density T Time, basic dimension

µ Dynamic viscosity CD Drag Coefficient

v Velocity Re Reynolds Number

L Length Xa a subscript for ‘actual’

ω Angular velocity Xm m subscript for small-scale

‘model’

F Force

Page 224: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

203

The dimensional analysis is performed as follows,

Choosing ρ, v, L

Π1 = 𝜏𝜌𝑎1𝑣𝑏1𝐿𝑐1

Π2 = 𝜇𝜌𝑎2𝑣𝑏2𝐿𝑐2

Π3 = 𝜔𝜌𝑎3𝑣𝑏3𝐿𝑐3

Π4 = 𝐹𝜌𝑎4𝑣𝑏4𝐿𝑐4

Π1 = [𝑀𝐿2𝑇−2][𝑀𝐿−3]𝑎1[𝐿𝑇−1]𝑏1[𝐿]𝑐1 = 𝑀1+𝑎1𝐿2−3𝑎1+𝑏1+𝑐1𝑇−2−𝑏1

= 𝑀0𝐿0𝑇0

𝑆𝑜𝑙𝑣𝑖𝑛𝑔, 𝑎1 = −1, 𝑏1 = −2, 𝑐1 = −2 + 3𝑎1 − 𝑏1 = −3

Π1 =𝜏

𝜌𝑣2𝐿3

Π2 =𝜌𝑣𝐿

𝜇= 𝑅𝑒

Π3 = [𝑇−1][𝑀𝐿−3]𝑎3[𝐿𝑇−1]𝑏3[𝐿]𝑐3 = 𝑀𝑎3𝐿−3𝑎3+𝑏3+𝑐3𝑇−1−𝑏3 = 𝑀0𝐿0𝑇0

𝑎3 = 0, 𝑏3 = −1, 𝑐3 = 3𝑎3 − 𝑏3 = 1

Π3 =𝜔𝐿

𝑣

Π4 =2𝐹

𝜌𝑣2𝐿2= 𝐶𝐷

𝑇

𝜌𝑣2𝐿3= 𝑓 (

𝜌𝑣𝐿

𝜇,𝜔𝐿

𝑣,2𝐹

𝜌𝑣2𝐿2)

(A-2)

Page 225: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

204

For complete similarity,

𝜌𝑎 = 𝜌𝑚 , 𝜇𝑎 = 𝜇𝑚

(𝜌𝑣𝐿

𝜇)𝑎

= (𝜌𝑣𝐿

𝜇)𝑚

𝐿𝑚𝐿𝑎

=𝑣𝑎𝑣𝑚

(2𝐹

𝜌𝑣2𝐿2)𝑎

= (2𝐹

𝜌𝑣2𝐿2)𝑚

𝐹𝑎𝐹𝑚

=𝑣𝑎2𝐿𝑎2

𝑣𝑚2 𝐿𝑚2= (

𝐿𝑚𝐿𝑎)2

(𝐿𝑎𝐿𝑚)2

= 1 (𝐷𝑟𝑖𝑣𝑖𝑛𝑔 𝑓𝑜𝑟𝑐𝑒 𝑑𝑜𝑒𝑠 𝑛𝑜𝑡 𝑐ℎ𝑎𝑛𝑔𝑒)

(𝜔𝐿

𝑣)𝑎= (

𝜔𝐿

𝑣)𝑚

𝜔𝑎𝜔𝑚

=𝑣𝑎𝑣𝑚

𝐿𝑚𝐿𝑎

= (𝐿𝑚𝐿𝑎)2

(𝜏

𝜌𝑣2𝐿3)𝑎

= (𝜏

𝜌𝑣2𝐿3)𝑚

𝜏𝑎𝜏𝑚

=𝑣𝑎2𝐿𝑎3

𝑣𝑚2 𝐿𝑚3 = (

𝐿𝑚𝐿𝑎)2

(𝐿𝑎𝐿𝑚)3

=𝐿𝑎𝐿𝑚

This results in the scaling law,

𝐿𝑚𝐿𝑎

=𝑣𝑎𝑣𝑚

=𝜏𝑚𝜏𝑎= √

𝜔𝑎𝜔𝑚

= 𝑘

(A-3)

Further calculations are performed to get the power scaling into the equation,

𝜏𝑎 = 𝜏𝑚

√𝜔𝑚

√𝜔𝑎

(A-4)

Page 226: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

205

𝑃𝑎𝑃𝑚

=𝜏𝑎𝜔𝑎𝜏𝑚𝜔𝑚

=𝜏𝑚√𝜔𝑚 𝜔𝑎

𝜏𝑚𝜔𝑚√𝜔𝑎= 𝑘

(A-5)

It turns out that the force scaling is a 1:1 ratio as shown by,

𝐹𝑚𝐹𝑎=𝑚𝑚𝑎𝑚𝑚𝑎𝑎𝑎

=𝑚𝑚𝑣𝑚𝑚𝑎𝑣𝑎

=𝑚𝑚

𝑚𝑎

1

𝑘= 1

𝑚𝑚

𝑚𝑎= 𝑘

(A-6)

This results in the final scaling law as,

𝐿𝑚𝐿𝑎

=𝑣𝑎𝑣𝑚

=𝜏𝑚𝜏𝑎= √

𝜔𝑎𝜔𝑚

=𝑃𝑎𝑃𝑚

=𝑚𝑚

𝑚𝑎= 𝑘

(A-7)

Page 227: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

206

C Sensor Circuits

These are the sensor circuits used for the experiment as discussed in section 6.1.3

Figure A-3 Voltage sensors and filters.

to voltage measurement to voltage measurement

to μC

9V power

supply input

Voltage divider

1st order RC filter

2nd order Sallen-

key filter

1st order RC filter

Page 228: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

207

Figure A-4 Current sensor filters, precision voltage reference and thermistor.

to μC

to current

sensor

to μC

to

thermistor

1st order RC filter

Precision voltage

reference

9V power

supply input

Page 229: AN IMPROVED BATTERY/ULTRACAPACITOR … · STORAGE SYSTEM MANAGEMENT STRATEGY FOR ELECTRIC VEHICLES SO KAI MAN (B.Eng. (Hons.), NUS) ... Most existing works focus on PMS. The existing

208

Figure A-5 Relays.

to μC

to fan

Optoisolators

Main relay

to MOSEFT

drivers

15V power

supply input

UC relay

Switches for user