POWER CONVERTER CIRCUITS FOR VOLTAGE BOOSTING, BALANCING AND RELIABLE OPERATION OF ENERGY SYSTEMS By Wei Qian A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY ELECTRICAL ENGINEERING 2011 ABSTRACT POWER CONVERTER CIRCUITS FOR VOLTAGE BOOSTING, BALANCING AND RELIABLE OPERATION OF ENERGY SYSTEMS By Wei Qian Hybrid electric vehicle (HEV) and alternative energy have received growing attention nowadays, due to their contribution to the sustainable development. In these systems, power electronic circuits play a key role in bridging over the differences between sources and loads. The thesis investigates power converters to interface the energy sources and to prolong the life of the critical energy storage devices. The boost converter in commercial HEV powertrains meets obstacles to upgrading the rating while downsizing the converter. A four-level dc-dc converter, and its special case―a 3X dc-dc converter that operates at three discrete voltage ratios, can overcome the drawbacks by dramatically reducing the inductance requirement. The operating principles, the current ripple, the power loss analysis and a clamping circuit are introduced. The concept is verified by the experimental results from two 30-kW and 55-kW prototypes. Yet, for other applications like FC, PV and TEG, it is beyond the ability of the 3X dc-dc converter to attain high boost gain. Thus, a new switched-capacitor dc-dc converter is proposed. The component cost comparison and the power loss analysis demonstrate its many features such as low component power rating and count, low capacitance requirement, light weight and high efficiency. The features are validated by a 450-W prototype. For the needs of converting the dc power to ac and vice versa, a family of transformer based impedance source (trans-Z-source) inverters is proposed. They possess the buck-boost functionality and the reliable operation against shoot-through/open-circuit fault, which are unavailable in traditional voltage source and current source inverters. Compared to the existing Z-source inverters, they exhibit the increased voltage gain and reduced voltage stress in the voltage-fed trans-Z-source inverters, and the expanded motoring operation range in the currentfed trans-Z-source inverters. Two 3-kW prototypes prove the analysis of the voltage-fed and current-fed trans-Z-source inverters respectively. To reduce the number of the serially connected battery cells/modules that are used in the energy systems, an improvement of a battery balance circuit is presented via phase-shift control. Each pair of battery cells are paralleled with a magnetically coupled half-bridge balance circuit. This cost-effective solution is able to achieve cell balancing and defective cell tolerance with more evenly distributed current stress, less component count and lower device ratings than its counterparts. Experimental results on Li-ion batteries verify the concept. ACKNOWLEDGEMENTS First of all, I would like to give my sincere thanks to my advisor, Dr. Fang Zheng Peng, not only for his enlightening guidance and generous support, but also for all his kind care, help and encouragement. His amazing insight, enthusiasm in research and his kindness in life has set a great model for me to follow in the years to go. I am grateful to my committee members, Dr. Robert Schlueter, Dr. Elias Strangas and Dr. Rajan Mukherjee, for their inspiring classes and valuable suggestion. In addition, their care and help for my career is very much appreciated. My gratitude also goes to my master advisor Dr. Dichen Liu in China, whose help, care and encouragement has helped my dream of PhD become true. I would like to thank all the members in the Power Electronics and Motor Drive Lab and all my other friends at MSU. I am very grateful to Dr. Junming Zhang for the impressive knowledge and technical experiences that he has kindly shared with me. My thanks also go to Dr. Baoming Ge, for the methodologies and advice he bestowed upon me in research. It is my great fortune to have learnt a lot from Dr. Honnyong Cha. Perhaps no word can I find in the world to express my gratitude to him for his selfless help all the time. I appreciate the help from Dr. Miaosen Shen on the 3X dc-dc converter, particularly for his innovation of the clamping circuit. I would also like to thank Dr. Lihua Chen, Mr. Qingsong Tang, Dr. Haiping Xu and Dr. Tawon Lee for helping me build the first 30-kW 3X converter prototype. Besides that, they are like elder brothers to me ever since I came to MSU. For the research afterward, I appreciate the help from Ms. Xi Lu, Mr. Sangmin Han, Mr. iv Jorge Cintron, Mr. Matthew Gebben, and Mr. Dennis Wey on the prototype fabrication and experiments. In addition, I would like to thank Dr. Yuan Li, Dr. Shuitao Yang, Dr. Dan Wang, Dr. Jianfeng Liu, Mr. Dong Cao, Ms. Qin Lei, Mr. Shuai Jiang, Dr. Uthane Supatti, Mr. Craig Rogers, Mr. Joel Anderson, Mr. Alan Joseph and Mr. Yantao Song for all the insightful discussion, inspiration and help. Great thanks go to my colleagues at Magana E-Car Systems. The last but not the least, I am indebted to my parents for their unconditional love, understanding and support. Their encouragement has always been there whenever I need support, whereas my company could not be there whenever they long to see their only kid. To this end, I would further like to thank all my relatives and friends who have been giving their generous care and help to my parents and to me for all these years. v TABLE OF CONTENTS LIST OF TABLES ....................................................................................................................... viii  LIST OF FIGURES ....................................................................................................................... ix  Chapter 1 Introduction .................................................................................................................... 1  1.1. Background and Motivation ................................................................................................. 1  1.2. Typical Electrical Characteristics and Issues of Some DC Energy Sources ........................ 3  1.3. Electrical Characteristics and Issues of Batteries ................................................................. 7  1.4. Power Converter Interfaces ................................................................................................ 11  1.5. Approaches to Improve the Energy Storage System .......................................................... 19  1.6. Scope of the Thesis ............................................................................................................. 23  Chapter 2 A Variable 3X DC-DC Converter for Plug-in Hybrid Electric Vehicles ..................... 26  2.1. Introduction ....................................................................................................................... 26  2.2. Four-Level Flying-Capacitor DC-DC Converter ............................................................... 28  2.3. Variable 3X DC-DC Converter and its Operation Principle .............................................. 39  2.4. Transient Current Control during Voltage Transitions ...................................................... 41  2.5. Minimum Requirement of Parasitic Inductance ................................................................. 46  2.6. Power Loss Analysis .......................................................................................................... 48  2.7. Experimental Results of a 55-kW 3X DC-DC Converter .................................................. 52  2.8. Auxiliary Clamping Circuit ................................................................................................ 56  2.8.1. Operating Principle ...................................................................................................... 56  2.8.2. Experimental Results of the Clamping Circuit ............................................................ 61  2.9. Summary............................................................................................................................. 61  Chapter 3 A Switched-Capacitor DC-DC Converter with High Voltage Gain and Reduced Component Rating and Count ....................................................................................................... 63  3.1. Introduction ........................................................................................................................ 63  3.2. A Brief Review of the MMCCC Structure ......................................................................... 66  3.3. The Proposed Switched-Capacitor DC-DC Converter and Operation Principle ................ 66  3.4. Characteristics of the Proposed Switched-Capacitor DC-DC Converter ........................... 71  3.5. Component Cost Comparison with Other Topologies ....................................................... 74  3.5.1. Total Device Power Rating .......................................................................................... 74  3.5.2. Capacitor Voltage, Current and Capacitance Requirement ......................................... 76  3.6. Power Loss Analysis .......................................................................................................... 78  3.7. Simulation and Experimental Verification ......................................................................... 84  3.8. Application Examples ........................................................................................................ 91  3.9. Summary............................................................................................................................. 92  Chapter 4 Trans-Z-Source Inverters ............................................................................................. 94  4.1. Introduction ........................................................................................................................ 94  4.2. The Voltage-fed Trans-Z-Source Inverters ........................................................................ 97  vi 4.3. The Current-fed Trans-Z-Source Inverters ....................................................................... 104  4.4. Comparison with the Z-Source and Quasi Z-Source Inverters......................................... 112  4.5. Simulation and Experimental Results .............................................................................. 116  4.6. Application Examples ...................................................................................................... 125  4.7. Summary........................................................................................................................... 127  Chapter 5 Balance and Long-life Operation of Battery Cells ..................................................... 128  5.1. Introduction ...................................................................................................................... 128  5.2. The Defective Cell Bypass Functionality in the Non-Coupled Buck-Boost Balance Circuit ........................................................................................................................................... 135  5.3. The Battery Balance Circuit with Phase-Shift Control .................................................... 142  5.4. Experimental Results ........................................................................................................ 147  5.5. Summary........................................................................................................................... 149  Chapter 6 Conclusion and Recommendation .............................................................................. 150  6.1. Conclusion ........................................................................................................................ 150  6.2. Recommendation for Future Work ................................................................................... 151  BIBLIOGRAPHY ....................................................................................................................... 157  vii LIST OF TABLES Table 1.1. Nominal voltage per cell in general for typical batteries [12] ........................................9  Table 1.2. General characteristics of Li-ion batteries using various chemistries [13] .....................9  Table 1.3. Comparison between supercapcitors and batteries [77] ................................................22  Table 3.1. Capacitor charge paths in two switching states.............................................................70  Table 3.2. Comparison of capacitor voltage rating, current and capacitance ................................77  Table 4.1. Comparison of the voltage-fed Trans-ZSI/-qZSI and ZSI/qZSI ................................. 114  Table 4.2. Comparison of the current-fed Trans-ZSI/-qZSI and ZSI/qZSI.................................. 115  viii LIST OF FIGURES Figure 1.1. V-I characteristics of a basic fuel cell. ...........................................................................4  Figure 1.2. V-I and V-P characteristics of a PV cell*.......................................................................5  Figure 1.3. A TE module with two electrodes and ceramic plates. ..................................................6  Figure 1.4. V-I characteristics of a commercial TEG module assembly from Tellurex...................6  Figure 1.5. Block diagram of a typical power electronic interface for energy systems.................12  Figure 1.6. A power converter example for PHEV. .......................................................................13  Figure 1.7. Power converter for fuel cell vehicle...........................................................................14  Figure 1.8. A centralized PV converter with battery energy storage. ............................................16  Figure 1.9. The configurations with modular dc-dc converters. ....................................................16  Figure 1.10. PV micro inverters. ....................................................................................................17  Figure 2.1. Powertrain configuration of conventional series-parallel HEV. ..................................28  Figure 2.2. Four-level flying-capacitor dc-dc converter. ...............................................................28  Figure 2.3. PWM signals and key waveforms of the four-level operation. ...................................31  Figure 2.4. Switching states of the four-level operation in boost mode. .......................................33  Figure 2.5. PWM signals and key waveforms of the three-level operation. ..................................36  Figure 2.6. Switching states of the three-level operation in boost mode. ......................................36  Figure 2.7. Comparison of normalized current ripples. .................................................................37  Figure 2.8. 3X dc-dc converter. .....................................................................................................39  Figure 2.9. Switching state for 1X mode. ......................................................................................40  Figure 2.10. Switching states for 2X mode....................................................................................41  Figure 2.11. Switching states for 3X mode. ...................................................................................42  Figure 2.12. PWM signals of the active switches during the 1X to 2X transition showing gradual duty cycle increase from 0 to 1/2. ..............................................................................44  Figure 2.13. PWM signals of the active switches during 2X to 3X transition showing gradual increase of duty cycle from 1/2 to 2/3. ......................................................................45  Figure 2.14. PWM signals of the active switches in the 3X to 2X transition. ...............................45  Figure 2.15. PWM signals of the active switches in the 2X to 1X transition. ...............................46  Figure 2.16. Equivalent circuit for the 3X mode. ..........................................................................49  Figure 2.17. Power loss vs. switching frequency in 3X mode. ......................................................51  Figure 2.18. 55-kW dc-dc multiplier/divider prototype.................................................................53  ix Figure 2.19. Experimental waveforms in the steady-state operation. ............................................54  Figure 2.20. Experimental waveforms of the nX mode transition. ................................................55  Figure 2.21. Measured efficiency of the 55-kW prototype. ...........................................................56  Figure 2.22. Power loss breakdown of the 55-kW prototype in 3X mode. ...................................56  Figure 2.23. Clamping circuit ........................................................................................................58  Figure 2.24. Typical discharge loops. ............................................................................................59  Figure 2.25. Experimental results of the voltage spike with the clamping circuit.........................61  Figure 3.1. A flying-capacitor dc-dc converter with the voltage conversion ratio of four. ............64  Figure 3.2. The original MMCCC with a voltage conversion ratio of four. ..................................65  Figure 3.3. MMCCC in the similar form as the flying-capacitor circuit. ......................................67  Figure 3.4. The proposed 6X switched-capacitor dc-dc converter. ...............................................67  Figure 3.5. Switching states of the 6X switched-capacitor dc-dc converter. .................................69  Figure 3.6. Complementary PWM signals .....................................................................................70  Figure 3.7. Charge/-discharge loops for two switching states. ......................................................70  Figure 3.8. A generalized NX switched-capacitor dc-dc converter constructed from basic cells. 73  Figure 3.9. Normalized total device power rating vs. voltage boost gain .....................................76  Figure 3.10. Normalized total capacitor voltage rating vs. voltage boost gain. ............................77  Figure 3.11. Equivalent circuits of Switching State I counting in the ESL and ESR. ...................79  Figure 3.12. Transient state after the Switching States I................................................................83  Figure 3.13. Simulation and experimental configuration. .............................................................84  Figure 3.14. Simulation results of gate drive signals and typical switch voltages. .......................86  Figure 3.15. Simulation results of capacitor voltages. ...................................................................87  Figure 3.16. Simulation results of input/output voltages and currents. .........................................88  Figure 3.17. Experimental results of complementary gate drive signals and the corresponding switches. .....................................................................................................................89  Figure 3.18. Experimental results of capacitor voltage VC1a(b) and VC2a(b). ...................................89  Figure 3.19. Experimental results of capacitor voltage VC3a(b) and switch waveform VS3a(b). .......90  Figure 3.20. Experimental results of input/output voltage and current .........................................90  Figure 3.21. Comparison of calculated and tested efficiency ........................................................91  Figure 3.22. Modular NX dc-dc converter for the voltage boosting in TEG. ................................92  Figure 4.1. Voltage source inverter ................................................................................................95  Figure 4.2. Current source inverter ................................................................................................95  Figure 4.3. Z-source inverter .........................................................................................................95  x Figure 4.4. Voltage-fed quasi Z-source inverter with coupled inductors. ......................................98  Figure 4.5. Voltage-fed trans-quasi-Z-source inverter. ..................................................................99  Figure 4.6. The equivalent circuits of the voltage-fed trans-qZSI viewed from the dc link. .......100  Figure 4.7. Voltage gain (MB) versus modulation index of the voltage-fed trans-ZSI/-qZSI (n=2) and ZSI/qZSI. ...........................................................................................................103  Figure 4.8. Active switch voltage stress of voltage-fed trans-ZSI/-qZSI (n=2) and ZSI/qZSI....103  Figure 4.9. Voltage-fed trans-Z-source inverter. ..........................................................................104  Figure 4.10. Current-fed quasi Z-source inverter with two inductors coupled. ...........................105  Figure 4.11. Current-fed trans-quasi-Z-source inverter. ..............................................................105  Figure 4.12. The equivalent circuits of the current-fed trans-qZSI viewed from the dc link. .....106  Figure 4.13. Dc-ac voltage gain and operation region in the current-fed ZSI/qZSI and trans-ZSI/qZSI.......................................................................................................................... 111  Figure 4.14. Current-fed trans-Z-source inverter. ........................................................................ 112  Figure 4.15 Comparison of normalized magnetizing currents in trans-ZSI and trans-qZSI. ...... 116  Figure 4.16. Simulation and experimental system configuration of the voltage-fed trans-qZSI. 118  Figure 4.17. Simulation results of the voltage-fed trans-qZSI considering the parasitic capacitance of the bifilar winding (constant boost control, M=0.93, Dsh=0.2). ...... 119  Figure 4.18. Experimental waveforms of the voltage-fed trans-qZSI (constant boost control, M=0.93, Dsh=0.2). ....................................................................................................120  Figure 4.19. Simulation results of the voltage-fed trans-qZSI without boost (M=1). .................121  Figure 4.20. Experimental waveforms of the voltage-fed trans-qZSI without boost (M=1). ......121  Figure 4.21. Simulation and experimental system configuration of the current-fed trans-qZSI. 122  Figure 4.22 Simulation results of the current-fed trans-qZSI (M=0.6)........................................122  Figure 4.23. Experimental waveforms of the current-fed trans-qZSI (M=0.6). ..........................124  Figure 4.24. An example of PV micro inverter using NX dc-dc converter plus trans-ZSI..........126  Figure 4.25. An example of the current-fed trans-qZSI for PHEV. .............................................126  Figure 4.26. Voltage source inverter in HEV traction drives producing pulse voltage ...............127  Figure 5.1. Dissipative method ....................................................................................................130  Figure 5.2. An example of the charge type battery balance circuit..............................................131  Figure 5.3. An example of the discharge type battery balance circuit .........................................131  Figure 5.4. An example of charge-discharge type battery balance circuit ...................................131  Figure 5.5. Centralized forward balance circuit with reduced number of secondary windings. .133  Figure 5.6. A magnetically coupled buck-boost balance circuit. .................................................133  Figure 5.7. Modular non-coupled buck-boost balance circuit .....................................................134  xi Figure 5.8. The switched-capacitor balance circuit. ....................................................................134  Figure 5.9. The resonant switched-capacitor balance circuit with phase-shift control ................134  Figure 5.10. The modular unit in the buck-boost balance circuit. ...............................................136  Figure 5.11. The switching states of the 5-cell balance circuit ....................................................137  Figure 5.12 The equivalent circuit of the 5-cell balance circuit ..................................................138  Figure 5.13. PWM signals and waveforms when cell #1 is overcharged. ...................................143  Figure 5.14. Operation modes in phase-shift control. ..................................................................144  Figure 5.15. Equivalent circuits. ..................................................................................................146  Figure 5.16. Transferred current vs. the phase-shift angle. ..........................................................147  Figure 5.17. Experimental data of charging without balance circuit ...........................................148  Figure 5.18. Experimental data of charging with the balance circuit and phase-shift control.....148  Figure 6.1. Modular dc-dc converter configuration .....................................................................155  xii Chapter 1 Introduction 1.1. Background and Motivation As the concerns about energy crisis and climate change grow, governments and businesses are increasingly investing technologies of alternative energy sources and more environmentally friendly vehicles worldwide. Alternative energy sources encompass a host of types and applications, such as fuel cell (FC), photovoltaics (PV), wind, tide, microturbine, biomass, geothermal resources, and so on. Unlike traditional energy sources, they are green and offer sustainable solutions to the current energy needs. Another solution adopted worldwide is to mitigate the dependence and consumption of fossil fuel via hybrid electric vehicles (HEVs). The economical and environmental benefits from commercial HEVs now stimulate the global interest in developing plug-in hybrid electric vehicles (PHEVs) and pure electric vehicles (EVs). In addition, the research on thermal electric generation (TEG) recently emerges to reduce exhaust heat from internal combustion (IC) engines by converting waste heat into electricity to charge the 12-V battery and to power auxiliary loads in vehicles. Yet, alternative energy sources like solar, wind and tidal wave are not dispatchable by their own. These are only available when there is adequate sun, wind and tide. Voltage swells and sags can occur by just injecting the power into the system when they are available. To accommodate the intermittent and variable behavior of these resources, energy storage devices can be engaged in a complete energy system. Energy storage devices include, but not limited to, batteries, super capacitors, superconducting magnetic energy storage (SMES), flywheels, compressed air energy storage, and pumped hydro. Among them, batteries have found the widest applications ranging from electronic devices like cellular phones and laptops to HEVs in addition 1 to their applications in alternative energy systems. With the aid of batteries, it is possible to continuously utilize the renewable energy, such that the power system capacity for high demand period is increased. It has already been well-known that batteries play a crucial role of storing energy from regenerative breaking and providing energy when it is needed HEVs and EVs. Besides the fact that batteries are employed in both systems, there are many things in common between the applications of power electronics circuits for transportations and for renewable energy systems. Power electronic circuits interface energy sources and a variety of loads, as one means of fully utilizing the maximum energy available in the energy system, and meanwhile prolonging the cycle and calendar life of batteries. For instance, dc-dc converters and dc-ac inverters are employed both in a HEV and a PV generator. The dc power from the source needs to be boosted and then converted into the useful ac power that can drive a vehicle or that can be directly interconnected with the utility grid. Furthermore, a successful marriage has been made between a hybrid electric vehicle and a PV generator in a newly developed thirdgeneration Toyota Prius [1]. Because of similar functions of these power electronics interfaces, the thesis will focus on power converters to address issues in HEVs/EVs and in alternative energy systems. The following paragraphs will begin with a briefly introduction of the common electrical characteristics of some typical alternative dc sources; afterwards it will review the characteristics and issues of the batteries; the reviews will disclose several common issues to be addressed; thereby, a comprehensive solution from the power electronics point of view will be given to improve the entire energy system from both system level and cell/module level. 2 1.2. Typical Electrical Characteristics and Issues of Some DC Energy Sources Many alternative dc energy sources like FC, PV, TEG and etc are made up of serial and/or parallel connections of basic low-voltage cells. For instance, fuel cells are stacked in series to provide sufficient voltages, say, 240~380 V in a FC vehicle [2]. There are many types of fuel cells, among which the Proton Exchange Membrane (PEM) and Solid Oxide Fuel Cells (SOFCs) are mainly considered for automotive. FCs is sometimes also called as FC battery since it is also electrochemical device. In fact, FC creates electricity as long as the reactants (hydrogen and oxygen) are refueled externally, unlike a battery will either run down or need recharge. The energy density of a typical fuel cell stack is 200 Wh/L, which is nearly ten times of a battery. It has energy conversion efficiency of around 40~60%, and can be even higher with heat recovery [3]. It has been considered the highest among a variety of alternative energy sources. The V-I characteristics of a basic fuel cell is shown in Figure 1.1 [3]. The theoretical EMF of a cell at zero current and 80 °C and 1 atm gas pressure is V0 = 1.16V. As the current density increases, the actual voltage at the electrical terminals drops dramatically. It indicates that, when directly fed by a FC stack in a vehicle, the inverter sees a wide variation from the dc bus. If not properly compensated for this variation in the motor control system, this may create stability problems in the drive system [4]. To this end, a unidirectional dc-dc boost converter is usually preferred. Another reason for the voltage boosting is that a large number of series cells are otherwise needed, which can lead to reduced reliability. That is, if a single cell fails open, the entire stack stops functioning because the current flow is interrupted [3]. Additionally, the FC stack is not fast enough to produce the needed hydrogen to meet the sudden load changes. Hence, battery and 3 supercapacitor have been employed as an energy buffer for it [5]. Cell voltage (V) 1.0 0.5 Region of activation polarization (reaction rate loss) Region of concentration polarization (gas transport loss) Region of Ohmic polarization Operation Voltage curve 0 Current density (mA/cm2 ) Figure 1.1. V-I characteristics of a basic fuel cell. Similarly, a typical PV panel (or module) usually consists of 36~72 cells connected in series [6]. As can be seen from the V-I characteristics in Figure 1.2, each PV cell only has an opencircuit voltage of less than 1 V. A 160-W PV panel has an output voltage of only 24~38 V in the defined maximum power point range [7]. Depending on the power rating, an array of PV panels may be serially connected for enough voltage and power to fulfill the load demand. Indeed, PV panels in an array are never exactly identical and the scenario is deteriorated when some panels are shaded. The resultant current out of the whole array and the efficiency are thereby determined by the least efficient panel and indeed the cell. The present PV module light-toelectricity efficiency is still very low (<15%), and the power generation capability may be reduced to 75~80% of nominal value due to aging [6]. 4 5 Cell current (A) 15C 40C 75C 1000W/m2 4 600W/m2 3 2 200W/m2 1 0 0 0.1 0.2 0.3 0.4 0.5 Cell voltage (V) 0.6 0.7 Figure 1.2. V-I and V-P characteristics of a PV cell*. (For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this dissertation.) *Model based on the British Petroleum BP5170 crystalline silicon PV module. Power at standard test condition (1000 W/m irradiation, and a cell temperature of 25 °C): 170 W @ 36.0 V. Legend: solid at 15 °C, dotted at 40 °C, and dashdot at 75 °C. Not as well-known as FC and PV, TEG was used for the applications for space and military purposes in the past. Nowadays it has drawn more and more attention in automotive, as there are more concerns over the efficiency and the environmental impact of the IC engines. As its name implies, a thermoelectric (TE) module can generate electricity from a temperature gradient. Actually, when electricity is applied across it in turn, the TE module can generate heat on one side and behave as a refrigerant on the other side. A TE module is normally comprised of an array of TE couples that are joined thermally in parallel and electrically in series, and each TE couple is a pair of pellets made up of dissimilar p and n type semiconductor materials, as illustrated in Figure 1.3 [8]. The voltage that one TE module can produce depends on the number 5 of thermal couples in series and the temperature gradient across it. According to a test result in [9] as shown in Figure 1.4 on a commercial TE module, one module only produces less than 5 V at the maximum power of under 5 W at 150 °C. Therefore, several modules have to be connected in series and in parallel to supply enough voltage and power. Again, too many modules in series are vulnerable to open-circuit failure. Cold side Conductors Ceramic plates Warm side Outputvoltage(V)Outputpower(W) Figure 1.3. A TE module with two electrodes and ceramic plates. 10 Voltage 150 C Voltage 115 C Voltage 79 C Power 150 C Power 115 C Power 79 C 9 8 7 6 5 4 3 2 1 0 0 0.5 1.5 1 Outputcurrent(A) 2 2.5 Figure 1.4. V-I characteristics of a commercial TEG module assembly from Tellurex. 6 1.3. Electrical Characteristics and Issues of Batteries Before embarking on the issues associated with batteries, it is helpful to briefly review some of the important terminologies and characteristics. Viewed as a single unit, like in the HEV powertrain, a battery pack usually has many rechargeable battery cells connected in series to make sufficient voltage. One way to describe the battery capacity is Ampere-hour (Ah). It is the amount of current that a battery can deliver, at a constant rate (in fractions or multiples of “C”) over a specified duration of time. For instance, a C/2 rate discharge of a 42-Ah battery can supply 21 A for two hours before the battery voltage reaches the end-of-discharge cut-off. Another property of most concern, especially for HEV designers, are the energy density (or specific energy) and power density (or specific power). The energy density is the amount of energy stored in terms of watt-hours per kilograms of a battery pack or cell; analogously, the power density is a measure of how much power can be extracted per weight at a specific rate of charge. While the battery is charged or discharged, the battery capacity is normally measured as state-of-charge (SOC). Although there are some different definitions of SOC, one common definition is defined as the percentage of energy currently stored in a battery compared to the battery’s maximum capacity at the last full charge. The difference between the battery’s maximum and minimum SOC is called the depth of discharge (DoD). The number of charge/discharge cycles is defined as cycle life when commonly referring to the battery life. When used over time, battery performance can substantially degrade in terms of capacity, power density, energy density and safety. Even when a battery is not in use, it is still aging by selfdischarge [10]. Therefore, calendar life (service life/shelf life) is also specified to represent the ability of the battery to withstand degradation over time. Battery lifetime depends on battery chemistry, the depth of discharge, charge current and voltage profile, battery temperature and 7 battery capacity termination level. There are a plenty of rechargeable battery chemistries, such as the Lead-acid, NickelCadmium (NiCd), Nickel-Metal Hydride (NiMH), reusable Alkaline and so on. Their nominal cell voltages are quite low, as listed in Table 1.1. Among them, Lead-acid battery is widely used for the 12-V starting-lighting-ignition (SLI) battery in vehicles. They also find applications in hospital equipment, wheelchairs, emergency lighting and UPS systems. NiMH batteries are used in most HEVs currently sold in the US. It has the proven longevity in cycle life and calendar life, fast discharge capability and an overall history of safety. Plus, it is environmental friendly. However, its primary demerits are the limitation in energy density and power density, and low prospects for future cost reduction [11]. Recently, a class of Lithium-ion (Li-ion) and Li-ion-polymer batteries are expected to be more suited than NiMH for the more demanding HEVs, such as PHEV with all-electric operation in charge depleting mode. The main difference between the Li-ion and Li-ion-polymer is that a solid ion conductive polymer replaces the liquid electrolyte used in a standard Li-ion battery, although most polymer batteries also contain an electrolyte paste to lower the internal cell resistance [10]. There are various kinds of Li-ion batteries, but the most popular chemistries in production can be narrowed down to three according to the cathode materials: Lithium-cobaltoxide, Lithium-manganese-oxide and Lithium-phosphate. The general characteristics of some Liion batteries are listed in Table 1.2. Their characteristics vary a lot from different electrode and electrolyte chemistries and manufacture technologies. Nevertheless, compared to other rechargeable batteries, they have relatively high cell voltage, high energy density, high power density, very good cycle life (from hundreds to a few thousand cycles [12, 13]), and low selfdischarge (3~5% [10, 12]). They are also environmentally friendly. These properties make Li-ion 8 batteries have great potential for automotive applications. However, the high chemical reactivity of lithium brings in higher risk to longevity and safety, such as simultaneous Li-plating, oxygen gas evolution and the flammability of the electrolyte solvent. Therefore, Li-ion batteries call for more careful control over the cell voltage and operation temperature than other battery chemistries do. Overcharge and over-temperature will degrade the capacity quickly and even cause permanent damage. Table 1.1. Nominal voltage per cell in general for typical batteries [12] Battery type Lead-Acid NiCd NiMH Reusable Alkaline Nominal cell voltage (V) 2 1.2 1.2 1.5 Table 1.2. General characteristics of Li-ion batteries using various chemistries [13] Cell voltage Energy Density Cycle life max/nom. (V) (Wh/kg) (deep) Graphite / NiCoMnO2 4.2 / 3.6 100-170 2000-3000 Graphite / Mn spinel 4.0 / 3.6 100-120 1000 Graphite / NiCoAlO2 4.2 / 3.6 100-150 2000-3000 Graphite / Iron phosphate 3.65 / 3.25 90-115 >3000 Lithium titanate / Mn spinel 2.8 / 2.4 60-75 >5000 Chemistry (Anode / Cathode) The Ragone plot in [14] compares battery cells/packs vs. several PHEV goals. It represents trade-offs between power density and energy density for a given battery chemistry. The present battery technologies still fall short of the ever-growing demand especially for PHEVs and EVs 9 on power density and energy density, never mention the lifetime, safety, and cost. No matter what kind of battery is employed and what inherent trade-offs are among its attributes, the expected attributes of a battery pack can be summarized as: - High capacity with small size and light weight; - Safety and long-term reliability; - Long cycle life and calendar life; - Low cost; - Environmental friendly. When the battery cells are connected in series for the high voltage demand and have experienced hundreds of high-rate charge/discharge cycles, another problem comes out. Individual cells may age differently due to the variance in chemical and electrical characteristics and thermal condition. As time goes by, the capacity of some cells will deviate from the others over multiple charge/discharge cycles. If they are not periodically balanced with the rest of the cells, some could gradually be driven into over-charged or under-discharged, either of which is prone to damage and eventually completes the failure of the whole battery pack. As a result, without cell balancing, a battery may typically be used in the range of 20% to 80% SOC, providing a battery usage of only 60% [15]. If the SOC range can be enlarged, a smaller battery pack is allowed for a desired output capacity. This can result in a large saving in overall weight, even when the weight of the balancer is counted in. Furthermore, its environmental impact involves fewer chemical materials and the released recycle stress to some extent. In commercial products, the battery management system (BMS) monitors every cell or module, but can only discharge the surplus cells in a lossy manner. 10 1.4. Power Converter Interfaces The above renewable energy sources and batteries face the common issues of low-voltage and uneven characteristics in cells. For the sake of reliability and life time, it is better to have a smaller number of cells or modules in series. However, the loads demand high voltage (e.g. 200~650-V dc bus in HEVs), high charge/discharge rates, high tolerance to faulty cells. Besides, there are a number of applications that need to convert the generated dc power into a regulated ac, and vice versa, depending on the applications. A power converter interface has to bridge over the differences between energy system and load. From the viewpoint of system level, an interface circuit should be capable of voltage boosting, dc-ac conversion and regulation, so as to properly rate the energy system. From the viewpoint of the energy storage device, another issue is how to boost the effective utilization, run-time and calendar life for uneven cells/modules, and thereby enhance the reliability, which will be discussed in the next section. A general block diagram of typical dc-ac interfaces is shown in Figure 1.5. It is divided into two energy paths for two energy sources. In the first path, the upper dc-dc converter stage takes in the uncontrolled input dc voltage and boosts it to a higher voltage, with or without galvanic isolation. The followed dc-ac inverter produces the required voltage and frequency from the boosted dc voltage. The inverter is a generic circuit in renewable energy systems, motor drives, uninterruptible power supplies, active power filter and etc. Note that the general dc-ac inverter stage here includes the ac filter for the grid interconnection, but not for the vehicle applications. The second energy path involves the battery energy storage, or sometimes supercapacitors. The battery can be directly paralleled to the inverter dc bus, or can go through a dedicated bidirectional dc-dc boost converter as shown in Figure 1.5. When charging the battery, the inverter rectifies the ac back to dc. 11 Energy source(e.g. ac-dc rectifier fossil fuel via IC or engine, FC or PV) dc-dc converter dc-dc boost converter Bidirectional Single-/threedc-ac 2or3 phase load/source inverter (e.g. grid, motor) Inversion/ rectification Figure 1.5. Block diagram of a typical power electronic interface for energy systems More specifically, an example of the above system, a PHEV powertrain, is given in Figure 1.6. PHEV is like a conventional HEV, whereas it uses a larger battery pack that can be charged from the utility via an on-board charger. It has been reported that it can be driven up to 40~50 miles without using a drop of gasoline. When the electric charge of the battery is depleted to certain level, the downsized gasoline engine kicks in to either charge battery as the car moves, or it is the primary source of propulsion until re-plugging the charger. The power converter interface consists of a dc-dc boost converter for the high voltage (HV) battery, a dc-ac inverter for the traction motor to drive the vehicle, and a second dc-ac inverter for the generator that is linked to the engine. There are also configurations with no dc-dc converter between the battery pack and the inverter. Nonetheless, the configuration in Figure 1.6 provides more flexibility and improves system performance. The reason is well justified. If the battery pack feeds inverters directly, due to limited cell voltage, more cells are required for high voltage high speed operation. Plus, more cell monitor, balance and protection units have to get involved. Conversely, at the low speed operation, a low-voltage dc bus is expected. A less flexible dc bus voltage leads to degraded efficiency. Admittedly, the conventional dc-dc boost converter has its limits in terms of size, weight and efficiency for PHEVs, especially for high temperature operation. 12 Traction motor On-board charger for HV battery M/G2 PFC+dc-dc IC engine Starter/ generator M/G1 Figure 1.6. A power converter example for PHEV. Replacing the energy path for the IC engine with a FC plus a dc-dc boost converter yields a powertrain configuration for a FCV, as shown in Figure 1.7 [16]. The power converter (dc-dc converter plus dc-ac inverter) matches the low-voltage and relatively slow output of fuel cell stack to the high voltage requirement and the rapid load change of the propulsion drive system. To start the motor, the power is sourced by the 12-V SLI battery, whose power electronic circuitry is not included in the powertrain configuration. When the vehicle is in the steady-state speed range, only the fuel cell stack provides the propulsion power, and it may also charge the battery. During rapid acceleration, both the fuel cell and the battery power the vehicle. When the break pedal is pressed, the battery gets charged, in which its bidirectional dc-dc converter operates in buck mode. For the sake of cost, there are also different configurations with either the fuel cell or the battery directly connecting to the inverter dc bus [4, 13]. Nevertheless, the addition of dc-dc converters allows for the use of low-voltage dc sources and the optimal design of the propulsion system irrespective of the fuel cell voltage. Alternatively, a new converter topology called Z-source inverter has been proposed in [17-20] to combine both the voltage boost and the dc-ac inversion. 13 FC M/G HV Battery Figure 1.7. Power converter for fuel cell vehicle. As for grid-connected PV systems, there have been many power converters, since they have less constraint than FCVs do. Most of them are transformer-isolated for high voltage gain and isolation ([6, 7, 17-23]). Meanwhile, non-isolated converters are adopted in Japan and European countries where galvanic isolation is not required [7, 24-26]. One of the PV inverters is shown in Figure 1.8, which embraces battery energy storage as well, such that it becomes possible to offset the daily and seasonal intermittency of the primary energy source, to smooth out load fluctuations, to damp out utility transients, and to facilitate islanding operation. In this configuration, several PV panels are connected in series and in parallel, and they feed a centralized two-stage power converter. In the first stage, the dc-dc converter boosts the PV voltage and tracks the maximum power point; in the second stage, the dc-ac inverter produces a regulated ac output to the local load and/or to the grid. This two-stage power converter is preferable to the one that has only a dc-ac inverter. The reasons are that the voltage from the PV panel is low and that the voltage-source inverter is essentially a buck-type converter. The shortcoming of the centralized configuration is that it cannot fully extract the maximum power point of individual PV panels due to the aforementioned unbalance from panel to panel. Hence, it has been reported in [6, 22, 27, 28] to equip each half-panel (or panel substring) with a modular dc–dc converter. These modular dc-dc converters can be in series (Figure 1.9(a)) or in parallel (Figure 1.9(b)) before connecting to a central dc-ac inverter. Recently, there is a trend to also modularize the dc-ac inverter and integrate it with the dc14 dc converter into a plug-and-play device called as micro inverter [29, 30]. An example is shown in Figure 1.10. Its rating usually ranges from 150 to 300 W. Micro inverters attract much attention mainly because: 1) they are able to harvest the maximum capacity of each individual panels; 2) they are easy to install and are flexible to expand for any power rating; 3) the modular fabrication can yield high economical profits in massive production. Yet, the application of micro inverters presents some challenges, especially for the dc-dc converter. When fed by a 160W PV panel with the voltage of 24 V at the maximum power point, the maximum voltage conversion ratio for the dc-dc converter has to be more than 12 times, so that the subsequent dcac inverter can produce a 110-V ac output for single-phase residential utility in US. The voltage ratio is more than doubled in Europe and in China. Transformers or coupled inductors [31, 32] have to be used for the dc-dc converter, since traditional non-isolated dc-dc boost converter has constrained voltage gain due to the parasitic resistive loss. Their demerit is the degraded efficiency and the high temperature issues on magnetic and electrolytic capacitors that precludes placing a micro inverter behind a PV panel. Even if this buck-type dc-ac inverter is replaced by the aforementioned Z-source inverter (or the later developed quasi-Z-source inverter) that have been proposed for PV generation in [33, 34], the attainable voltage boost gain is not high enough in practice. 15 dc-dcconverter PV panels dc-acinverter Grid Local load dc-dc Figure 1.8. A centralized PV converter with battery energy storage. PV panel Modulardc-dcconverter Centralinverter Grid Local load (a) Series connection of modular dc-dc boost converters. Figure 1.9. The configurations with modular dc-dc converters. 16 Figure 1.9 (cont’d) Modulardc-dcconverter dcbus PV panel Centralinverter Grid Local load (b) Parallel connection of modular dc-dc full-bridge converters. microinverter PV panel acbus Grid Local load Figure 1.10. PV micro inverters. 17 Different from FC and PV, the present power converter interface for TEG is mainly a dc-dc converter [35-37]. Besides load interface, the dc-dc converter has to perform the maximum power point tracking or the power matching [38], since the produced voltage as well as the load varies. It is expected that the development of scalable, modular, low-cost, highly reliable power electronic interfaces will improve the overall cost and durability of the above energy systems, because of similar functions of these power electronics interfaces [39]. It would be best if one circuit could fit into all the applications. Unfortunately, in the real world, no circuit is perfect to address all the issues in a variety of applications. Several dc-dc converters and dc-ac inverters are examined in this thesis to improve the power conversion systems. Firstly, a four-level flyingcapacitor dc-dc converter is explored as an alternate to overcome the drawbacks of the conventional dc-dc boost converter interfacing the battery and the inverter in PHEVs. When it operates at three discrete output/input voltage ratios, the converter reduces the inductance requirement to a minimal value (almost zero). Yet, for other applications like FC, PV and TEG, it is beyond the ability of the flying-capacitor dc-dc converter to attain high boost gain. Therefore, secondly, a new switched-capacitor dc-dc converter is proposed. It has much lower current stress and lower total device power rating than the flying-capacitor dc-dc converter, which are fairly desirable features for dc-dc converters with high voltage gain. Thirdly, for the dc-ac power conversion, a class of transformer based impedance source (trans-Z-source) inverters is proposed. They possess the buck-boost functionality, and the reliable operation due to the immunity to shoot-through/open-circuit. None of these are unavailable in traditional voltage source and current source inverters. Moreover, their new features, thanks to their transformer windings, lend themselves to HEVs, FC and PV applications. 18 1.5. Approaches to Improve the Energy Storage System When we take a closer look at the energy storage device itself besides the system-level circuitry, the problems of the over-charge and deep-discharge concerns, the restricted charge/discharge rate, low cell voltage and etc call for the micro-management of the cells connected in series. The success of PHEVs and EVs greatly depends on the performance and lifetime of the battery. The state-of-the-art electric vehicle can not go thus far with a full charge cycle as the internal combustion engine vehicles do with a tank of fuel. Moreover, it takes just a couple of minutes to refuel a 15-Gallon tank in the internal combustion engine vehicle, whereas it takes hours to recharge a PHEV. The battery pack made up of low-voltage cells is large, heavy, and costly. Even the Li-ion batteries, which are poised to be a preferable choice for the next generation of electric vehicles, face several technical challenges to be addressed as in the other batteries. Their sensitive nature leaves many concerns in terms of longevity and safety. The development of new materials and structures in the device level has been improving the performance of batteries. At the same time, the advanced circuitry and energy management are indispensable in boosting the performance and propelling the applications. Hence, this section summarizes some approaches to boost the effective capacity and the reliable lifetime of the batteries. A. Cell balance and defective cell bypass As posed to the passive balance method used in the commercial BMS, many active cell balance circuits were proposed in literature [40-69]. They are essentially small dc-dc converters transferring energy to equalize each cell actively and efficiently. In this way, each cell approaches the same float voltage at the end of the charging process, or the same cutoff voltage at the end of the discharging process. Consequently, a wider charge/discharge range can be 19 achieved, thus the effective battery pack capacity will be expanded. However, many existing solutions are either low-performance or high-cost, which preclude their implementation. Furthermore, it is more desirable that a battery pack can have some fault tolerance, so that as much juice can be squeezed out of the pack as possible. Otherwise, the capacity of the rest cells is wasted even though only one or very few cells are malfunctioning. A cost-effective solution to handle defective cells/modules will give rise to significant profit to the battery applications. Therefore, the battery balance circuits and its functionality of defective cell bypass will be investigated in the thesis. B. Battery fault protection and status prognosis Safety is a crucial criterion for battery usage. Batteries are very sensitive to voltage and current profiles, charging speed, depth of discharge and battery temperature. The overcharge, over-discharge, over-current, over-temperature and etc are taken care of by the battery monitor and protection circuits. These circuits are integrated in a single IC [70-72], with small profile, very low power consumption and strong immunity to EMI noise. Besides that, many ICs have been developed in some companies like Texas Instrument for not only monitoring the battery status but also providing the remaining SOC. Some researches have also been reported in [73-75] to diagnose the health status of batteries in advance. They aim to predict the battery usable capacity and the discharge termination level. Basically, the data from manufacturer or from periodic tests were first collected after necessary data processing. Battery model or empirical equation was built using regression methods off line, according to the data like open-circuit voltage, current, temperature. During operation, the pertinent data are sensed for calculating the present and remaining SOC, and/or impedance which are closely related to battery capacity. Compared to historic data, the life time or aging is forecasted. The health status prognosis can 20 actively enhance the battery reliability and offer information for maintenance and fault diagnosis. The onset of problems can be aware of, and actions can be taken before they develop to a cascade failure. This approach not just prevents permanent capacity loss, but also reflects the true end of battery life. Otherwise, a battery is considered to have come to the end of life when the battery capacity drops below 80% of its rated capacity. In fact, batteries can still deliver usable power, although they have shorter run-time [10]. As a mater of fact, very accurate estimation is a challenge for HEVs, due to the difficulty in correct battery model and state of health estimation. The methods so far are based on a series of user-prescribed charge/discharge profiles, and specified average current and duration, whereas the depth of cycle, load condition, and thermal environment may differ with various drive cycles and environmental conditions. C. Combination of batteries and super-capacitors Batteries naturally have high energy density yet low power density. It is favorable to size the battery in PHEVs and EVs by the energy demand for a desired all-electric range. At the same time, a battery pack also has to meet the frequently irregular and rapid acceleration and regenerative braking in vehicles. Unfortunately, batteries that are tailored for a little higher power density usually have to sacrifice the energy density and are more costly, which ends up with an oversized and expensive design [76, 77]. Nowadays super-capacitors (or so-called ultra capacitor, electrochemical double layer capacitor) are gaining popularity for their high pulse power, fast charge/discharge response, high efficiency and long calendar life. A general comparison with batteries is listed in Table 1.3. As shown in the table, supercapacitor is good at handling large impulse power with high efficiency due to its small time constant and very low ESR, whereas slow power exchange rate are preferential for batteries. Supercapacitors with longer cycle life are more suitable for frequent charge/discharge than batteries. Besides, 21 supercapacitors have long shelf life and excellent performance over a wide temperature range including both hot and extreme cold conditions [77, 78]. Table 1.3. Comparison between supercapcitors and batteries [77] Characteristics State-of-the-art Li-ion battery Supercapacitor *Charge time ~3-5 minutes ~1 second *Discharge time ~3-5 minutes ~1 second *Cycle life <5,000 @ 1C rate >500,000 Specific Energy (Wh/kg) 50-100 5 Specific power (kW/kg) **1-2 5-10 Cycle efficiency (%) From less than 50% to more than From less than 75% to 90% more than 95% Cost/Wh $0.5-1/Wh $10-20/Wh Cost/kW $50-150/kW $15-30/kW * Time for discharge and charge of the useable total energy stored in the devices **Power capacity of the battery for short multi-second pulses at 90% efficiency If batteries are designed for the maximum energy density and supercapacitors for the maximum power density, combining them together can gain the maximum capacity, fast response, long cycle life and high efficiency. Supercapacitors act as a temporary reservoir, it can be fully charged before acceleration and can be empty before regeneration [79]. Otherwise, in current HEVs, the SOC for the batteries is managed with specific margin for regeneration. Plus, batteries will suffer from less impulse power that affects its lifetime. It is a significant thermal 22 burden shift from batteries to super-capacitors, which has the potential to reduce the overall thermal loading and enhance the safe operation of the vehicle regenerative energy storage system. Since the terminal voltage of a battery pack is relatively constant, direct paralleling supercapacitors with the battery pack will limit the charge transfer capability. Instead, they can be actively coupled via dc-dc converters. Yet, despite the above advantages, the addition of supercapacitors and dc-dc converter(s) to the energy storage system would not be PHEV and EV designers’ first selection, unless it can deliver superior cycle performance with long-term reliability, lower total size, less total mass, lower cost. In addition to the above approaches, researchers have also studied the charging methods to improve battery thermal performance and safety [80, 81] or to reduce charging time [82]. It has been found important to choose the proper charge methods and termination levels for different types of batteries. Besides that, another interesting innovation is to use TEG to adjust the battery thermal environment [83]. Battery life could be extended if battery temperature is maintained in the optimum range for that battery chemistry. 1.6. Scope of the Thesis The thesis is divided into two parts: system level circuits for the power conversion, and the cell/module level solutions for the battery energy storage itself. Chapters 2 to 4 propose several novel dc-dc converters and dc-ac inverters for voltage boost and dc-ac conversion. Chapter 5 deals with solutions for voltage balancing and reliable operation of one of the most costly devices ― the rechargeable batteries. Chapter 2 first explores the operating principle and current ripple of a four-level flyingcapacitor dc-dc converter. A special case of the four-level converter, the 3X dc-dc converter is 23 presented. The transient control to limit the inrush current and the power loss analysis are provided. Additionally, a clamping circuit is proposed to mitigate the voltage spike at the diode reverse recovery. The analysis is verified experimentally. Chapter 3 proposes another novel dc-dc converter that is derived from the flying-capacitor dc-dc converter. It features low component (switching device and capacitor) power rating, small switching device count and low output capacitance requirement. In addition to its low current stress, the combination of two short symmetric paths of charge pumps further lowers power loss. Therefore, a small and light converter with high voltage gain and high efficiency can be achieved. The operating principle, the features, the component cost comparison with its counterparts and the power loss analysis are introduced. Simulation and experimental results validate the principle and features of this topology. Chapter 4 proposes a class of trans-Z-Source inverters which have buck-boost capability with a single power stage. They employ an impedance network of two transformer windings and one capacitor to couple the dc source and the inverter bridge. While inheriting the main features of its origin ― the Z-source inverters, the new networks exhibit some unique advantages, such as the increased voltage gain and reduced voltage stress in the voltage-fed trans-ZSIs and the expanded motoring operation range in the current-fed trans-ZSIs, when the turns-ratio of the transformer windings is over one. Simulation and experimental results are provided to prove the analysis of the voltage-fed and current-fed trans-ZSIs respectively. Chapter 5 first reviews the battery balance circuits and discusses the functionality of defective cell bypass for the reliable operation. A battery balance circuit is introduced that employs phase-shift control for the purposes of reducing the component count and rating while improving the balancing performance. Each pair of battery cells are paralleled with a 24 magnetically coupled half-bridge balance circuit. Charge is transferred directly from the surplus cells to the weak ones. This cost-effective solution is able to achieve cell balancing and defective cell tolerance with more evenly distributed current stress, less component count and lower device ratings than its counterparts. The principle is demonstrated by experimental results on four Li-ion battery cells. Chapter 6 summarizes the work and suggests the prospective future work. 25 Chapter 2 A Variable 3X DC-DC Converter for Plug-in Hybrid Electric Vehicles 2.1. Introduction Many commercial hybrid electric vehicle (HEV) systems, such as Prius, Camry, and Fusion, use a traditional bidirectional dc-dc converter to interface the battery and the inverter dc bus. For example, the powertrain configuration of the Camry hybrid is shown in Figure 2.1. In this system, the dc-dc converter boosts the battery voltage, from 244 V to three optimized discrete voltage levels up to 650 V according to the motor/generator speeds [84]. Combined with the continuous adjustment of modulation indices of the inverters, then both the motor and generator and their individual inverters are able to operate in the most efficient region to accommodate the wide speed and power demand of the vehicle. In this traditional dc-dc boost converter, an inductor is employed. Besides its large size and heavy weight, the lossy inductor limits the operating temperature. Because of the inductor, the efficiency of the dc-dc converter is relatively low compared to the inverter. Nowadays, the economical and environmental benefits from commercial HEVs have spurred the global interest in further developing plug-in hybrid electric vehicles (PHEVs). For the pure electric drive operation of PHEVs, the dc-dc boost converter has to deliver the full power needed by the traction drive, not just the difference between MG1 and MG2 as would be the case in blended operation mode with the engine running (Figure 2.1). For example, the Prius’ present dc-dc converter is rated at 20-kW peak power (10-kW continuous power) [85], which is not enough for all-electric operation at higher speeds and has to be upgraded from 20-kW to 55kW peak power (30-kW continuous) for PHEV’s pure electric drive. It is not viable to simply 26 resort to increasing the switching frequency to limit the size, weight and cost of the converter since the core and copper loss of the inductor will go up as a result of the increased switching frequency and the power rating. In addition, the semiconductor heat dissipation in the converter limits the switching frequency. Thereby, multi-phase dc-dc converters [86-88] were developed for high power HEVs and fuel cell vehicles (FCVs). The inductor design is still a challenging issue for increasing the converter efficiency and power density. Multilevel dc-dc converter topologies [89-92] have been proposed for many applications. By contrast to the traditional dc-dc converter, a three-level flying-capacitor dc-dc converter greatly reduces the inductance requirement in addition to other benefits such as much lower device voltage rating and fast dynamic response [90, 91]. In this paper, further investigation and theoretical analysis are performed to utilize the multilevel structure to achieve high voltage boost and high power for PHEVs. A four-level flying-capacitor dc-dc converter is presented as shown in Figure 2.2 to replace the traditional dc-dc converter. This four-level structure can further reduce the input current ripple. More interestingly, a closer and further look into the input current ripple will reveal that the inductor can be eliminated or minimized when the converter operates at three discrete voltage ratios: 1X, 2X, and 3X. Power loss induced by the bulky inductor is diminished. Therefore, the operation of the four-level dc-dc converter at three discrete voltage ratios (thus named a variable 3X dc-dc converter) will be proposed later in this paper. The transition to achieve the three variable voltage ratios will be analyzed, followed by the experimental verification on a 55-kW 3X dc-dc converter prototype. It should be noted that a fixed 1:3 voltage ratio dc-dc converter has been introduced in [93, 94], however, they cannot be used for the above-mentioned HEV and PHEV applications that require variable voltage levels. The main new contribution of this paper resides in the introduction of the variable 3X operation 27 and smooth transition between voltage levels. Engine S/G Starter/ generator inverter Traction Traction inverter motor MG1 MG2 Dc-dc boost converter Battery Figure 2.1. Powertrain configuration of conventional series-parallel HEV. S3p DCbusto theinverter S2p L Vin Battery + S1p Vout C1 C2 Cout S1n S2n _ S3n Figure 2.2. Four-level flying-capacitor dc-dc converter. 2.2. Four-Level Flying-Capacitor DC-DC Converter In this section, the general operation and features of the four-level flying-capacitor dc-dc converter are explained and discussed first. The relationship between the input current ripple and voltage ratio is derived for the four-level converter and extended to the three- and two-level converters for comparison purposes. The current ripple comparison indicates a dramatic reduction of the inductance requirement. Analysis reveals three operation modes and the relationships between the different operation modes. In addition, the analytical results will be further used in a later section for transitions from one voltage level to another of the variable 3X 28 operation. As shown in Figure 2.2, each pair of switches Sjp and Sjn (j=1, 2, 3) conduct complementarily and are clamped by capacitors C1, C2, and Cout. An input inductor (L) on the battery side plays the same role as in the traditional boost converter. It should be noted that this structure has the capability of bidirectional power flow. This paper focuses on the boost mode operation, that is, power flows from the battery (low voltage side, Vin) to the high voltage dc bus (Vout) to inverter(s). Under the boost mode, each lower-side switch Sjn is controlled as the active switch, and the anti-parallel diode of the high-side switch Sjp conducts as the complementary switch. This four-level converter can be operated as a three-level one as well, based on a switching pattern proposed in this paper. First, the four-level operation can be divided into three operation ranges. The PWM signals and key waveforms in these three ranges are depicted in Figure 2.3(a), (b) and (c). The PWM signals for both upper and lower switches Sjp and Sjn are shown in the figures, thus valid for both boost and buck operations. For example, the PWM signals for Sjp as sketched in blue dash lines indicate the conduction intervals of the anti-parallel diodes in the boost mode, whereas they represent PWM signals for the switches in the buck mode. The switching states are shown in Figure 2.4. The voltages of capacitors C1 and C2 are controlled to be Vout/3 and 2Vout/3, respectively. A duty cycle D is defined as the ratio of the duration that the switch Sjn conducts over one switching cycle, T. For 0 ≤ D ≤ 1/3, as can be inferred from Figure 2.3(a), the converter operates in a sequence of Figure 2.4(a)-(d)-(b)-(d)-(c)-(d) over one switching cycle. The input 29 voltage Vin equals the average value of the leg voltage vm in steady state according to the voltage-second balance of the inductor. The leg voltage vm is switched between two potentials, 2Vout/3 and Vout, as sketched in Figure 2.3(a). For instance, when only S3n is turned ON as shown in the switching state of Figure 2.4(a), C2 gets charged, making vm = 2Vout/3; when S3n is OFF as shown in Figure 2.4(d), the current freewheels through the anti-parallel diodes of S1p, S2p and S3p, making vm = Vout. Similarly for other switching states, when only one of the switches Sjn is ON as shown in Figure 2.4(a), (b) and (c), vm = 2Vout/3; otherwise vm = Vout in the freewheeling states in Figure 2.4(d). Thus, one can get 2V 1  Vin  Vm  3   D  Vout  3D  out  1  D  Vout . 3 3  (2.1) During this operation range, each inner capacitor (C1 and C2) gets charged and discharged equally for a duration of DT over one switching cycle as shown in Figure 2.3(a). However, due to gate delays and device tolerance, the voltage of each capacitor C1 and C2 may settle down to a value slightly deviated from their theoretical values, Vout/3 and 2Vout/3, respectively, as described in [95]. Therefore, in order to maintain their voltages as desired, an accurate duty cycle control is needed. This capacitor voltage balancing issue has been addressed and analyzed in the literature [91, 95-98]. 30 S3p 1 0 StateI IV IV II III IV t S2p 1 0 t S1p 1 0 S1n 1 0 S2n 1 0 S3n 1 0 vm 0 t DT DischargeC1 ChargeC1,dischargeC2 ChargeC2 t t III t t S1n 1 0 DT S2n 1 0 t i 0 t (a) 0 ≤ D ≤ 1/3. t t S3n 1 0 0 2Vout / 3 T /3 2Vout / 3 ( D  1 / 3)T (2 / 3  D )T T (b) 1/3 ≤ D ≤ 2/3. Figure 2.3. PWM signals and key waveforms of the four-level operation. 31 V t S1p 1 0 Vout 2Vout / 3 T VII II S2p 1 0 vm i 0 t S3p 1 0 StateI VI t t t Figure 2.3 (cont’d) StateV VIII VI VIII VII VIII S3p 1 0 t S2p 1 0 t S1p 1 0 t S1n 1 0 DT S2n 1 0 t S3n 1 0 vm t t Vout / 3 t 0 i 0 T t (c) 2/3 ≤ D ≤ 1. When the duty cycle increases and enters the range of 1/3 ≤ D ≤ 2/3, three new switching states as shown in Figure 2.4(e), (f) and (g) replace the previous freewheeling state. As a result, vm is switched between two potentials: Vout/3 and 2Vout/3, as shown in Figure 2.3(b). The operating sequence follows Figure 2.4(a)-(f)-(b)-(g)-(c)-(e) over one cycle. Likewise, when 2/3 ≤ D ≤ 1, a new state comes into play. All the switches S1n, S2n and S3n conduct as shown in Figure 2.4(h). Consequently, vm presents two voltage levels: 0 and Vout/3, as sketched in Figure 2.3(c). The corresponding operating sequence is Figure 2.4(e)-(h)-(f)-(h)-(g)-(h) repetitively. In spite of the above three variation ranges for D, it can be proven that the input and output voltage 32 relationship expressed in (1) always holds true. Apparently, it is the same as the traditional boost converter. The buck mode can be analyzed similarly. It can be proven that the output/input voltage ratio for the buck mode is also the same as that for the buck converter. However, it can be seen from Figure 2.3 that not only the effective switching frequency is tripled but also the volt-second across the inductor is minimized when seen from the inductor in this four-level converter, thus having great potential to reduce inductor requirements. Given an input voltage Vin in (1), the duty cycle, D can be expressed as a function of Vin and Vout. According to the inductor current change in each state, the input current ripple can be expressed as a function of Vin and Vout: VinT   L V T  i   in  L V T  in  L  Vin  Vout 3  2 Vout   1  3 V  1  V   for  V  2 ; in   out  in  3 Vout  Vout   2 Vin  1  3V   3  V   for 2  V  3; in   out  in  (2.2) Vout  1 Vin   3  V   for V  3. out  in  S3p S2p L i + S1p C2 Cout + L i Vout + Vin S2p + S1n _ Vin vm _ S3n C1 vm _ (a) State I. (b) State II. Figure 2.4. Switching states of the four-level operation in boost mode. 33 Vout Cout _ Figure 2.4 (cont’d) S3p S3p S2p + L i S1p C1 C2 Cout + Vin L i + Vout vmS2n _ Vin Vout Cout _ vm _ _ (c) State III. C1 + (d) State IV. S2p + S1p L i Vin + S1p L i Vout Cout S2n + + S1n _ Vin vm S _ 3n Vout C1 C2 Cout _ vm _ S3n (e) State V. (f) State VI. S3p + L i Vin + S1n S vm 2n + L i Vout C2 Cout + S1n _ Vin vmS2n Vout Cout _ _ S3n _ (g) State VII. (h) State VIII. Second, the four-level converter can be operated as a three-level converter, when the 34 capacitors C2 and Cout are connected in parallel by continuously turning ON switch S3n. The PWM signals and key waveforms are shown in Figure 2.5. Note that the duty cycle, D is only the duty cycle for S1n and S2n in this case. Figure 2.6 shows the switching states for this three-level operation. The voltage of C1 is controlled to equal Vout/2. By applying the same analysis as before, the three-level operation can be divided into two operation ranges. When 0 ≤D ≤ 1/2 as shown in Figure 2.5(a), the operation sequence is Figure 2.6(a)-(c)-(b)-(c); otherwise when 1/2 ≤ D ≤ 1 as shown in Figure 2.5(b), the sequence is Figure 2.6(a)-(d)-(b)-(d). It was derived in [91] that the voltage relationship can be also expressed by (2.1). The current ripple can be obtained as VinT  Vout  Vin  Vout  1  2V 1  V   for  V  2; in  out  in  L  i   VinT  1  Vin  , forVout  2.  L  2 Vout  Vin    (2.3) For comparison purposes, the current ripple of the traditional two-level (2L) boost converter can be calculated as V T V i  in 1  in L  Vout  .  (2.4) The normalized inductor current ripples of the four-level and three-level converters are compared with that of the traditional boost converter in Figure 2.7. The base value is the current ripple of the traditional boost converter at a boost ratio of three. It is evident that the maximum current ripple of the four-level converter is almost one tenth of the traditional dc-dc boost converter’s over 1 ≤ Vout/Vin ≤ 4. In other words, given the same current ripple specification, a much smaller inductor is sufficient for this four-level converter. 35 StateI III II StateI IV III S3p 1 0 DT vm t t Vout / 2 vm Vout / 2 0 t S3n 1 0 t Vout DT S2n 1 0 t S3n 1 0 t S1n 1 0 t S2n 1 0 t S1p 1 0 t S1n 1 0 t S2p 1 0 t S1p 1 0 IV S3p 1 0 t S2p 1 0 II 0 t t i i 0 T 0 t (a) 0 ≤ D ≤ 1/2. T t (b) 1/2 ≤ D ≤ 1. Figure 2.5. PWM signals and key waveforms of the three-level operation. S3p S3p L i S1p + Vin S2p + L i Vout C1 C2 Cout vmS2n + + S1n _ Vin _ S3n C1 C2 Cout vm _ S3n (a) State I. (b) State II. Figure 2.6. Switching states of the three-level operation in boost mode. 36 Vout _ Figure 2.6 (cont’d) S3p S3p S2p L i L i Vout C2 Cout + Vin + + S1p + S1n _ vm Vin C2 Cout vm S2n _ _ S3n _ S3n (c) State III. (d) State IV. Traditional boost converter 4-level converter 3-level converter 1.2 Normalized current ripple Vout 1 0.8 0.6 0.4 0.2 0 1 1.5 2.5 3 2 Voltage boost ratio 3.5 4 Figure 2.7. Comparison of normalized current ripples. One concern about the four-level converter is that six semiconductor switching devices have to be used instead of two in the traditional two-level boost converter. The total switching device power rating (TDPR) or total device power stress (TDPS) of a converter circuit is an indication of how much the total silicon area is needed for the semiconductor devices. The following analysis shows that both two- and four-level converters have the same TDPR, thus requiring similar or the same amount of silicon areas. For the six switches in the four-level (4L) converter 37 operation, the TDPR is TDPR _ 4L  6  Vout    Iin  2  m  Po , (2.5) where m = Vout/Vin and output power, Po = Vin Iin = Vout Iout. For the two switches in the traditional two level bidirectional boost converter, the TDPR is TDPR _ 2 L  2 Vout  I in  2  m  Po . (2.6) As can be seen from (2.5) and (2.6), they have the same total device power rating. For the traditional boost converter, each switch has to sustain the full dc voltage, whereas the switches in the four-level converter only sustain 1/3 of the dc voltage. For example, Camry hybrid uses the traditional dc-dc boost converter, in which each switch employs four IGBTs and four diodes in parallel to reach the current (400 A) and voltage (650 V) ratings [99]. However, if the four-level converter is used, each switch needs to sustain only 1/3 of the output voltage, thus much lower voltage and higher current IGBT can be used without the need for paralleling. A different way to look at the four-level converter is that it uses three IGBTs in series to reach the required power rating, instead of parallel in the traditional boost converter. Therefore, the four-level converter is an attractive candidate to replace the traditional boost converter for PHEVs. More interestingly, when the converter operates at three discrete voltage ratios: one, two and three times the input voltage (3X), as marked with dots in Figure 2.7, the inductor current in the four-level dc-dc converter becomes ripple free. This implies that the inductance requirement is zero. The question left to be answered is how to produce three discrete voltage ratios with smooth transition from one level to another. As mentioned previously, three discrete voltage levels are enough and have been used in many HEVs such as Camry for optimum operation of the power train. 38 2.3. Variable 3X DC-DC Converter and its Operation Principle Figure 2.8 shows the proposed variable 3X dc-dc converter, in which no magnetic inductor is needed on the input side for three discrete voltage output levels: 1X, 2X, and 3X the input voltage. In the figure, Ls represents stray inductance of the circuit, which can be the equivalent series inductance (ESL) within the battery pack and the parasitic inductance of the cable connecting the converter to the battery in the PHEV. A fixed ratio of 1:3 was presented in [93, 94] to interface the 14/42 V accessory power supplies in vehicles. However, this fixed voltage ratio is not suited for HEV and PHEV applications, in which a variable dc voltage is needed. In this paper, we further explore the possible operation of variable voltage ratios and their smooth transitions. Compared with other switched-capacitor dc-dc converters [100-111], the following proposed variable 3X dc-dc converter has the simplest structure and finds its niche in the PHEV application where only discrete voltage ratios are desired from the battery to the traction drive dc bus voltage. S3p S2p LS S1n Vin + S1p Vout C1 C2 Cout S2n _ S3n Figure 2.8. 3X dc-dc converter. Because of its bi-directional nature, the 3X dc-dc converter can be viewed as a multiplier (step up) or a divider (step down) depending on power flow direction. The switching states as shown in Figure 2.9, Figure 2.10 and Figure 2.11 represent the boost mode operation for the 39 three voltage ratios, 1X, 2X, and 3X. 1X: To achieve the voltage ratio of 1, namely 1X, the converter operates in just one switching state as shown in Figure 2.9, with switches S2n and S3n always ON. Thus, all the capacitors are connected in parallel with the input voltage, which ensures the voltage ratio of 1X. Under this 1X operation, it is apparent that there is no voltage balancing problem. S3p S2p LS S1p Vin S2n + V C1 C2 Cout out _ S3n Figure 2.9. Switching state for 1X mode. 2X: While C2 and Cout are always paralleled by having S3n continuously ON, the converter alternates with 50-50% duty cycle between two switching states I and II as illustrated in Figure 2.10 (a) and (b). In the switching state I with S2n ON, the capacitor C1 is paralleled with the input battery and charged to the battery voltage, and in the switching state II with S1n ON, C1 is connected in series with the battery and discharged to the parallel of C2 and Cout. As a result, the voltage of the capacitor C1 is maintained close to the battery voltage and the output voltage becomes 2X input battery voltage. The capacitor voltages should be well clamped and balanced. Note that the anti-parallel diode (in bold dashed line) of S3n may conduct when the discharge current of C2 is greater than the input current charging C1. 40 S3p S3p S2p LS Vin + S1p S2n + LS V C1 C2 Cout out _ S1n Vin S3n (a) Switching state I. V C1 C2 Cout out _ S3n (b) Switching state II. Figure 2.10. Switching states for 2X mode. 3X: When the desired ratio is 3X, the converter circulates among switching states I, II and III as shown in Figure 2.11 (a), (b) and (c), with 1/3 duty cycle each state. In state I, VC1 = Vin; in state II, VC2 = VC1 + Vin; in state III, Vout = VC2 + Vin. After these three states, the capacitor voltages will all be balanced automatically, as demonstrated in [93, 94]. The output voltage is three times (3X) the input voltage. 2.4. Transient Current Control during Voltage Transitions In steady state, the capacitor voltages are well balanced and the voltage differences are very small, so the current through the switches is well limited as demonstrated in [93]. Theoretically, the 3X dc-dc converter does not need any input inductance in steady-state operation for any of the three output voltage levels. However, during a transition when the output voltage changes between Vin and 2Vin, or between 2Vin and 3Vin, the voltage differences are very large, which can lead to high transient current through the devices and capacitors. In order to limit this transient current, a variable duty-ratio PWM with a higher switching frequency is proposed for 41 the transitions. In addition, a minimum inductance indicated as Ls in Figure 2.8 is required for limiting the transient current, which will be discussed later. S2p + LS S1p C1 Vin Cout Vout S2n + LS V C1 C2 Cout out _ S1n _ Vin S3n (a) Switching state I. S3n (b) Switching state II. S3p + LS V C2 Cout out _ S1n Vin S2n S3n (c) Switching state III. Figure 2.11. Switching states for 3X mode. A. Changing the output voltage from 1Vin to 2Vin Before changing the output voltage from 1Vin to 2Vin, all three capacitors are initially charged to 1Vin in 1X mode. In the transition, C2 and Cout should be charged up to 2Vin gradually. To simplify analysis, assume that the stray inductance is large enough for the inductor 42 current to be continuous. (The transient input current does not have to be continuous, in principle). Therefore, the transient operation can be referred to the same sequence of Figure 2.6(a)-(c)-(b)-(c) for the voltage ratio of 1 ≤ Vout/Vin ≤ 2. The corresponding PWM signal in any switching cycle can be referred to Figure 2.5(a), except that the duty cycle, D has to be increased gradually over the transition to limit inrush current. The duty cycle is again defined in Figure 2.5(a) for S1n and S2n. As a result, the transition is divided into two active switching states and one freewheeling state between them. In the two active switching states, the switching signals are given to the switches as shown in Figure 2.6(a) and (b). The resultant leg voltage, vm, equals VC1 in Figure 2.6(a), and it equals Vout – VC1 in Figure 2.6(b) for the duration of DT, respectively. After each active state, the switches S1n and S2n are turned OFF. Therefore, the input current flows through the anti-parallel diodes in S1p and S2p as shown in Figure 2.6(c). The resultant leg voltage, vm, equals Vout for the duration of (1/2–D)T. According to the inductor voltage-second balance in steady state, the relationship of the input voltage and output voltage still satisfies (1) as follows: 1  Vin  DVc1  D Vout  Vc1   2   D  Vout  1  D  Vout . 2  (2.7) From the steady-state boundary condition of 1X and 2X modes, the duty cycle should vary from 0 to 1/2 to change the output smoothly to 2Vin. As illustrated in Fig. 12, the PWM duty cycle of the active switches S1n and S2n have to increase gradually from 0 to 1/2, with a higher transient switching frequency. 43 S1n D : 0  1/ 2 S2n S3n 1 Figure 2.12. PWM signals of the active switches during the 1X to 2X transition showing gradual duty cycle increase from 0 to 1/2. B. Changing the output voltage from 2Vin to 3Vin Applying the same principle to the transition from 2Vin to 3Vin, the switching states can be referred to the sequence of Figure 2.4(e)-(a)-(f)-(b)-(g)-(c). The corresponding PWM signals and their time interval are marked in Figure 2.3(b) over one switching cycle. Likewise, by averaging the leg voltage, vm, the same voltage relation can be derived as (1). For example, when S2n and S3n are turned on as shown in Figure 2.4(e), the leg voltage, vm equals VC1. After the duration of (D–1/3)T, S2n is turned off as in Figure 2.4(a), the freewheeling current forces vm to equal VC2 for the duration of (2/3–D)T until the next switching state (Figure 2.4(f)) starts. Likewise, when the converter stays in Figure 2.4(f) and (g), vm is equal to VC2 – VC1 and Vout – VC2 for (D– 1/3)T, respectively. In their corresponding freewheeling states, vm is equal to Vout – VC1 in Figure 2.4(b) and equal to Vout –VC2 + VC1 in Figure 2.4(c) for (2/3–D)T, respectively. Likewise, by averaging the leg voltage, vm, the same voltage relation can be derived as (2.1).The ideal duty cycle for the active switches varies gradually from 1/2 to 2/3 during this transition, as illustrated in Figure 2.13. 44 S1n D :1/ 2  2 / 3 S2n S3n Figure 2.13. PWM signals of the active switches during 2X to 3X transition showing gradual increase of duty cycle from 1/2 to 2/3. C. Changing the output voltage from 3Vin to 2Vin The control strategy for the transition from 3Vin to 2Vin is to reduce the initial three switching states (Figure 2.11(a), (b), and (c)) of the 3X operation to the first two switching states (Figure 2.11(a) and (b)) and to control the two switching states with 50-50% duty cycle. The output capacitor Cout is disconnected from the source and will be gradually discharged by the load. When the output voltage decreases to the voltage of C2, the anti-parallel diode of S3p conducts and clamps the output voltage to 2Vin. The PWM signals for this transition are shown in Figure 2.14. S1n S2n S3n 3X steadystate Transition 2X steadystate Figure 2.14. PWM signals of the active switches in the 3X to 2X transition. D. Changing the output voltage from 2Vin to 1Vin In a similar manner, a transition from 2X to 1X can be implemented by the gate signals illustrated in Figure 2.15 to keep S2n and S3n ON. When the three capacitor voltages decrease 45 and become equal, the converter settles down to the 1X mode steady-state operation as shown in Figure 2.9. S1n S2n S3n 1 2X steadystate Transition 1X steadystate Figure 2.15. PWM signals of the active switches in the 2X to 1X transition. 2.5. Minimum Requirement of Parasitic Inductance In this section, an analysis is made to determine how much parasitic inductance is needed to limit the transition current within the rated input current. The voltages across C1 and C2 are assumed constant in each short switching period, Ttr. Based on the above transient PWM control methods, the two worst cases for high transient currents are considered among all the transition modes. A. Vout: 1Vin→2Vin During this transition, the input current change (ripple) can be calculated according to the active switching state II, illustrated in Figure 2.6(b): V  DTtr i  Ls Ls   Vin  Vc1  Vout  DTtr Ls Vin 1  2 D  DTtr Ls 1  D   I max 46 (2.8) where Vc1  Vin ,Vout  Vin / 1  D for   D  1 / 2, and Imax is a given maximum current value allowed. The maximum i occurs at D = 0.29 in the above function, which should be limited to no larger than the specified maximum current allowed, Imax. B. Vout: 2Vin→3Vin Similarly, the current change in the 2X to 3X transition is computed according to the active switching state III in Figure 2.4(g): i   Vin  Vc 2  Vout  DTtr Ls Vin  2  3D  D  1/ 3 Ttr Ls 1  D  (2.9)  I max where Vc2  2Vin ,Vout  Vin / 1  D for    D  2 / 3. The maximum i occurs at D = 0.53 in the above function. In comparison of the maximum current changes in both cases, the minimum inductance requirement can be calculated as 4.1 µH for a 480-A allowable current change, assuming that the 55-kW 3X dc-dc converter is switched at 20-kHz transient switching frequency and 230-V input voltage. The requirement of the parasitic inductance can be further reduced by increasing the switching frequency during transitions. In practice, a further lower parasitic inductance than the above conservative calculated value is tolerable, because the capacitance of C1 and C2 is finite and the voltages are not constant, which further reduces transient inrush current. In practice, the total ESL of a battery pack and connection cable should be sufficient, especially when the battery is placed in the trunk area far away from the converter under hood in most HEVs and PHEVs. 47 Otherwise, a small air-core inductor can be added when the parasitic inductance is not enough. 2.6. Power Loss Analysis Power loss analysis is usually estimated for assessing the converter parameters. It can be calculated numerically based on the current in the charge/discharge loop. Thus, the following analysis starts with modeling the input current. Each nX steady-state operation can be modeled with its equivalent circuit. Take the 3X mode for instance. The equivalent circuits are illustrated in Figure 2.16. There is a lumped constant voltage drop Von resulting from two IGBTs and one diode in the serial loop. The voltage drop of one IGBT is modeled as a constant voltage plus a T current times its equivalent series resistance (ESR). In the first state of t  0 ~ s , C1 is charged 3 by the battery, as modeled in Figure 2.16(a). The input current can be solved as:     2  12 1I (0)   i  e 1t  I (0) cos 1t   C1V1 1  sin 1t   1 1        (2.10) where I (0) is the initial input current through the ESL in the charging loop, and V1  Vin  Von  VC1(0) is the initial voltage difference. i  Rsi / (2Ls ) in each state; 1, 2, 3 implies three states; the resonance frequency in each switching state can be explicitly found by the equivalent loop capacitance and ESL: i  Cloop1  C1 . 48 1 R  ( si ) 2 . In State I, Ls Cloopi 2 Ls i i Vin Ls I out Von Rs1 C1 C1 C2 Cout Rload RCout Vin Cout C2 Rload RCout Rs 2 (b) Switching state II. i Vin I out Von (a) Switching state I. C2 Ls Ls Von I out Cout Rs3 RCout C1 Rload (c) Switching state III. Figure 2.16. Equivalent circuit for the 3X mode. T 2T In the same way, for the second state of t  s ~ s as modeled in Figure 2.16 (b), the 3 3 input current is expressed as  Ts    T  2 (t  s )  Ts Ts  Ts   2 2  2 2  2  I ( 3 )  3  I ( )  cos  2 (t   )   Cloop 2 V2 ie  sin 2 (t  )  3 3  2 2  3   (2.11)       C1  C2 T T where V2  (Vin  Von )  VC1 ( s )  VC 2 ( s ) , Cloop 2  . C1  C2 3 3 For t  2Ts ~ Ts , the output capacitor gets charged as modeled in Figure 2.16(c). The load 3 49 current, Iout can be approximated as a constant dc, considering the very small output voltage ripple. The input current can be expressed as a function of both initial input current and load current. ie  3 (t  2Ts )   2Ts C I 3 I( )  2 out   3 2Ts   cos 3 (t  3 )  C2  Cout   C I  2Ts 2 2 )  2 out Cloop3 (V3  K )( 3  3 )   3  I ( 3 C2  Cout   2Ts  C2  1     sin 3 (t  3 )   C  C I out out 2  3  (2.12) where V3  (Vin  Von )  VC 2 ( C2 2Ts 2T , the )  VCout ( s ) , the constant K  Rs3 I out Cout  C2 3 3 Cout  C2 . loop capacitance Cloop3  Cout  C2 By equalizing the initial value of the current in the first state and the final value in the third state, the initial current 0 and the current expression can be solved. For a simplified iteration, the voltage difference in every state can be approximated from half of the capacitor voltage ripples. The total power loss includes conduction loss, switching loss, and gate drive loss. The conduction loss here refers to the loss dissipated in the semiconductor devices and in the ESR in the passive components. The latter is sometimes separated as the charge/discharge loss, although it is essentially consumed by the loop ESR. It is discussed in [94] that the charging loss is not relevant to the ESR when the switching period and the time constant satisfy and the stray inductance is negligible. However, it is not true when the stray inductance is not negligible. The conduction loss can be summed up for every 1/3 switching cycle: 50  3 Von  I avg , j  I 2 RMS , j  Rsj Pcon     3  j 1   2  3I out  I 2 RMS ,3   I avg ,  I out RCout   3 (2.13)  where Iavg, j and IRMS, j are the average and RMS values of the input current respectively. At the end of each switching state, two IGBTs and one diode are turned on/off. The IGBT switching loss is directly proportional to the switching current, given a fixed converter input voltage. The diode reverse recovery current can be expressed as a function of its snap-off current from the datasheet. Therefore, the switching loss is computed as the integration of the corresponding voltage and current: 3 1  1 jTs  Psw  Vce _ pk  f s   I ( )   ton  toff   I rr , j  t rr , j  6 3 2  j 1  0.04 Normalized Power loss 0.035 0.03 (2.14) Conduction loss Switching loss Total loss 0.025 0.02 0.015 0.01 0.005 0 0 10 1 10 Switching frequency (kHz) Figure 2.17. Power loss vs. switching frequency in 3X mode. As one of the guidelines for selecting the switching frequency, the estimated power loss is normalized and plotted in Figure 2.17 according to (2.13) and (2.14). The circuit parameters are the same as in the next section. The total loss also includes gate drive loss. The base value is 3051 kW continuous power at 220-V input voltage. As can be seen, the conduction loss is dominant when the switching frequency is low. As the switching frequency keeps rising, the switching loss becomes dominant. In terms of the overall efficiency, the optimum switching frequency falls in the range of 4~6 kHz. At the same time, the input current ripple to the battery is another factor to be taken into account for determining the switching frequency within the desirable efficiency range. 2.7. Experimental Results of a 55-kW 3X DC-DC Converter A 55-kW prototype as shown in Figure 2.18 was built to verify the functionality of the 3X dc-dc converter concept. The switching devices are commercially available 600-V IGBT/IPM modules, which is higher than what is required for the 3X dc-dc converter with 230-V input voltage. In fact, the efficiency of the converter would be higher if a lower (around 400 V) voltage rating module were available. The capacitors for C1 and C2 are 500 µF and 240 µF film capacitors, respectively. The output capacitor consists of 40-µF film capacitors in the 3X dc-dc converter and 820-µF capacitors at the inverter dc bus. In addition, a 10-kΩ bleeding resistor is paralleled across each IGBT. Besides draining the energy out of the capacitor at power-off, it helps balance the voltage across each IGBT under some particular conditions, for instance, the initial power-on stage. The duty cycle and frequency control is implemented by one CPLD XC95288XL, which can be readily integrated with the control board for the entire power converter system. 52 Figure 2.18. 55-kW dc-dc multiplier/divider prototype. Experimental results in the steady-state and transient operation are shown in Figure 2.19 and Figure 2.20. A total 5-µH ESL in the battery and cable was estimated. The converter operates at 8 kHz in the 3X mode and 12 kHz in the 2X mode. Figure 2.19(a) and (b) show the input/output voltage and current waveforms in 1X and 2X modes at their 30-kW peak power. Figure 2.19(c) shows the waveforms in 3X mode at the 55-kW peak power. Figure 2.20 shows voltage transition waveforms, in which the converter operates at 20 kHz. As can be seen, the transient current was successfully limited below 240 A, the rated current of the converter. Figure 2.21 shows the measured efficiency at different output powers in nX mode (n=1, 2, 3). The overall efficiency in the 30-kW continuous power range is over 97%. Figure 2.22 gives an estimation of the power loss breakdown in the full power range of the 3X mode, with a comparison to the experimental results. The analysis well predicts the characteristics of the loss with respect to the output power. Compared to the design of the fixed 3X 1-kW converter in [93], the utilization of parasitic inductance in the proposed variable 3X converter reduces pulse current, the conduction loss, and the capacitance requirement. As a result, a compact size (27.2 cm × 24.4 cm ×8.8 cm) and light weight (5.6 kg) were achieved. Furthermore, the power density of the proposed converter is 53 3 considerably increased, compared to the power density (30 kW / 3483.55 cm and 30 kW / 6.6 kg) of the dc-dc converter in a commercially mass-production vehicle [99]. The 3.5-kg 212-µH inductor alone contributes significantly to the volume and weight of the traditional boost converter. There is still great potential to make the converter even lighter by replacing the film capacitors with multilayer ceramic capacitors (MLCCs). Vin : 50 V/div Vin : 50 V/div Vout : 100 V/div Vout : 50 V/div Iout : 50 A/div Iin : 50 A/div Iout : 20 A/div Iin : 50 A/div t: 50 μs/div t: 100 μs/div (a) 1X boost mode at 30-kW output. (b) 2X boost mode at 30-kW output. Vout : 100 V/div Vin : 50 V/div Iin : 50 A/div Iout : 50 A/div t: 100 μs/div (c) 3X boost mode at 55-kW output. Figure 2.19. Experimental waveforms in the steady-state operation. 54 Vin : 100 V/div Vin : 100 V/div Vout : 100 V/div Iout : 10 A/div Vout : 100 V/div Iout : 10 A/div Iin : 50 A/div Iin : 50 A/div t: 10 ms/div t: 10 ms/div (a) Transition from 1X to 2X (b) Transition from 2X to 3X Vin : 100 V/div Vin : 100 V/div Vout : 100 V/div Vout : 100 V/div Iout : 10 A/div Iout : 10 A/div Iin : 50 A/div Iin : 50 A/div t: 10 ms/div t: 10 ms/div (d) Transition from 2X to 1X. (c) Transition from 3X to 2X Figure 2.20. Experimental waveforms of the nX mode transition. 55 100 3X boost mode Vin=220 V 2X boost mode Vin=220 V 1X boost mode Vin=220 V Efficiency (%) 99.5 99 98.5 98 97.5 97 96.5 96 0 10 20 30 40 Output power (kW) 50 60 Figure 2.21. Measured efficiency of the 55-kW prototype. 2500 Power loss (W) 2000 Switching loss Measured loss 1500 Diode conduction loss 1000 IGBT conduction loss 500 0 0 10 20 30 40 Output power (kW) 50 Other: Charge/ discharge loss on ESR + Gate drive loss 60 Figure 2.22. Power loss breakdown of the 55-kW prototype in 3X mode. 2.8. Auxiliary Clamping Circuit 2.8.1. Operating Principle For the power rating of over tens of kilowatt, one of the challenges for the multi-level dc-dc 56 converters in high power application is the high frequency spikes during the diode reverse recovery. The voltage spikes add significant stress and additional losses to the device, and also bring in EMI problems to the system. In these converters, each pair of complementary switching devices is naturally capacitor-clamped, and thus the IGBT switching spike is ideally well clamped. However, in practice, the capacitor-clamped loop contains considerable ESL, due to the constraint of the package of commercial IGBT modules and the entire main circuit layout. Hence, the clamping effect to the high frequency voltage spikes could not be sufficient in some real-world designs. Therefore, a RCD clamping circuit as shown in Figure 2.23 is proposed for this multilevel dc-dc topology. The basic operating principle is similar to that of the traditional RCD clamping circuit. When the switch is turned off, the current will be first diverted into the clamping capacitor through the clamping diode. Then the clamping capacitor regenerates the energy back to the main capacitor through a resistor. Such an example can be found in the energy regeneration loop for the clamping capacitors C1p and C1n. The clamping mechanism for all the switches is the same, but the discharge loops for other clamping cells are different, which innovatively make use of the switching devices in the main circuit to create the discharge loop. For instance, when the freewheeling diode in the switch S2n turns off, the stray energy from the diode reverse recovery will flow into the clamping capacitor C2n. When the switch S2n is on, as shown in Figure 2.24(a), the stored energy will be fed back to C1 through resistor R2n and diode Dd2n. Such diodes in the discharge loop as Dd2n are added to guarantee the unidirectional current. That is, the voltage ripples of the main capacitors C1, C2 and Cout cannot make the clamping capacitors 57 get charged backward. The reason of having resistors in the discharge loop is to limit the current and to maintain a reasonable clamping voltage. Nevertheless, power loss is still very low. In the same way, the clamping capacitor C3n will get a chance to recover the energy back to C1 when the switches S2n and S3n are both conducting, as shown in Figure 2.24(a). It is noteworthy that in the boost mode operation, the actual conducting devices are the upper three freewheeling diodes in Sxp and the lower three IGBTs in Sxn (x=1, 2, 3), and vice versa in the buck mode operation. Nevertheless, the discharge current can still pass through either the IGBT or its freewheeling diode as any of them is carrying the load current. S3p S2p S1p Dd3p DC3p C2p R2p DC2p Dd2p R1p Dd1p C1p DC1p LS S1n Vin C3p R3p S2n S3n DC1n R1n Dd1n C1 C2 Cout C1n DC2n R2n Dd2n C2n DC3n R3n C3n Dd3n Figure 2.23. Clamping circuit The discharge loop for the clamping capacitor C3p behaves in the same fashion, but there are two discharging states for C3p as shown in Figure 2.24(b) and (c). In the 1/3 switching period 58 in Figure 2.24(b), when the switches S2p and S3n are on, the captured energy in C3p will flow to Cout through C2, R3p and Dd3p. In another 1/3 period in Figure 2.24(c), the energy in C3p will flow to C2 through C1, R3p and Dd3p, when the switches S3p and S2n are on. Thus, the time constant of the discharge loop for C3p can be designed twice the time constant of the other discharge loops, leading to less power loss. S3p S2p LS S1p C1 S1n Vin S2n S3n R2n Dd2n C2n C3n R3n Dd3n (a)The discharge loop for C2n and C3n. Figure 2.24. Typical discharge loops. 59 Figure 2.24 (cont’d) S3p C3p R3p Dd3p S2p LS S1p S1n C2 S2n Vin S3n (b) The discharging state I for C3p. S3p C3p R3p Dd3p S2p LS S1p S1n Vin C1 S2n S3n (c) The discharging state II for S3p. 60 C2 Cout 2.8.2. Experimental Results of the Clamping Circuit The clamping circuit was designed for a 30-kW 3X dc-dc converter prototype. This converter was constructed by three 250-V 600-A CM600DU5F dual IGBT modules. The IGBT package imposes some layout restrict and difficulty. The capacitance for C1 and C2 are 500 µF and 160 µF. The capacitance at the output dc bus is 410 µF in total. The total input inductance is around 3.6 µH. The converter operated at 7 kHz. Without the clamping circuit, it could only handle the output power a little more than 10 kW. With the clamping circuit, the highest spike across switch is less than 15% at 30 kW in 3X boost mode, as shown in Figure 2.25. It demonstrates the effectiveness of the clamping circuit. t: 50  s/div vsw : 50 V/div Figure 2.25. Experimental results of the voltage spike with the clamping circuit. 2.9. Summary An alternative solution has been presented to overcome the demerits of the traditional boost converter for plug-in HEVs. Analysis has shown that a general four-level flying-capacitor dc-dc converter reduces the inductance requirement dramatically. Moreover, a variable 3X dc-dc converter has been proposed that was derived from the four-level dc-dc converter to further 61 minimize the inductance to null for HEVs and PHEVs that require three discrete voltage levels. The experimental results of a 55-kW prototype in steady-state and transient operation validated the operating principle and circuit analysis. Compared with the traditional low power switched capacitor dc-dc converters, the variable 3X converter achieves flexible voltage ratios without sacrificing efficiency (>97%) or the component count (only six switching devices and three capacitors) and with low voltage stress (Vin) across the switching devices. The transient current is well under control with the aid of the duty cycle control. By advantageously utilizing stray inductance, the variable 3X dc-dc converter achieves high efficiency with high power density. Therefore, it is a promising alternative to the existing boost converter for HEVs and PHEVs. 62 Chapter 3 A Switched-Capacitor DC-DC Converter with High Voltage Gain and Reduced Component Rating and Count 3.1. Introduction When no input inductor is engaged, the 3X dc-dc converter in the previous chapter can essentially be categorized into the switched-capacitor dc-dc converters. Switched-capacitor dc-dc converters are a network of switching devices and capacitors. They are sometimes also called charge pump. Switched-capacitor dc-dc converters have gained popularity in industrial switched mode power supplies due to their attractive features such as magnetic-less structure and high efficiency. Since they can be easily integrated without bulky magnetic components, the power density of dc-dc converters can be significantly boosted. They can achieve high efficiency even at very light load condition, and can maintain good no-load output voltage regulation [100]. Therefore, many switched-capacitor dc-dc converters and the derived resonant switchedcapacitor converters [102, 104, 107, 108, 112-120] have been developed. Yet, conventional circuits that are usually used in low power applications have some of the following drawbacks when a high voltage gain is desired: 1) quite diverse voltage/current stresses for switching devices in some circuits, which are not suitable for modular configuration or for high efficiency requirement; 2) a large number of switching devices in some other circuits; 3) pulsating input current and the resultant EMI; 4) unidirectional power flow. More importantly, their total device power ratings are unfavorable for a practical design to maintain high efficiency. On the contrary to low-power switched-capacitor converters, the magnetic-less flying63 capacitor dc-dc converters have the advantages of small component (switching device and capacitor) count, low voltage stress across the switching devices, and bidirectional power flow. However, when it comes to high voltage conversion ratios in applications like PV generation, thermal electric generation, a pure magnetic-less flying-capacitor structure has its practical limits to reach high voltage gain. For an output/input voltage boost ratio of N times (namely NX), the input current has to go through N switching devices. An example is given in Figure 3.1, with a conversion ratio of four. The resultant voltage drop and power loss overshadow their advantages presented at the low voltage ratios. S4p S3p + S2p S1p Vin   S1n S2n S3n S4n C2 C1 C3 V C4 out  = or Figure 3.1. A flying-capacitor dc-dc converter with the voltage conversion ratio of four. To address the above issues, a modular multilevel capacitor clamped dc-dc converter (MMCCC) was proposed in [121, 122]. A MMCCC is shown in Figure 3.2, with a voltage conversion ratio of four. Improved efficiency can be expected, because: 1) the current to charge a capacitor flows through at most three switching devices, regardless of the voltage ratio; 2) the currents through the switching devices and capacitors reduce to roughly 2/N times the corresponding currents in the original flying-capacitor structure for a voltage ratio of N, leading to the reduced total device power rating. However, the number of switching devices becomes 64 3N–2, rather than 2N for the conventional flying-capacitor structure, and the extra N–2 switching devices have to sustain the voltage stress of twice the input voltage (defined in boost mode). Additionally, as in many switched-capacitor dc-dc converters, while maintaining the device voltage (or current) stress low, it inevitably experiences the increased capacitor voltage with the increment of the voltage conversion ratio. The voltage diversity and the maximum voltage rating then affect the design, size and efficiency for a high voltage conversion ratio. S2a S1a Vin   S1p C1 S1n S2p S4a S3a C2 S2n S3p + C3 C4 Vout  S3n = or Figure 3.2. The original MMCCC with a voltage conversion ratio of four. This paper presents a switched-capacitor dc-dc converter that is very suitable for high voltage gain applications. It sums up the output of two symmetric charge pumps to reduce the device count, capacitor voltage rating and power loss. Moreover, it keeps very low total device power rating (TDPR). As will be discussed later, its many merits lead to the possibility of a compact, light and high efficient converter. The following sections will start with a brief review on the structure of the MMCCC, such that the features of the proposed new converter could easily be brought to light. The operation principle and features of the proposed converter will be introduced afterward, followed by a quantitative comparison with other counterparts. The concept and analysis will be validated by simulation and experimental results. 65 3.2. A Brief Review of the MMCCC Structure The MMCCC as shown in Figure 3.2 is composed of three basic cells, plus a switch S4a and a capacitor C4 connected to the output. In boost mode, it steps up the voltage from the low voltage input (defined as Vin) to the high voltage output (defined as Vout). From another point of view, the MMCCC can be reverted to the similar form as the flying-capacitor circuit and can be redrawn in Figure 3.3. The switch Sja (j=1, 2, 3, 4) creates a path for charging the capacitor Cj in one of two alternate switching states. The capacitor C1 is charged by the input, Vin, and the other capacitor Cj (j=2, 3) is charged via the addition of Cj-1 and Vin. A phase leg of complementary switches Sjp and Sjn (j=1, 2, 3) from each basic cell is in parallel with the input source Vin, in order that Cj can be directly connected to the positive (or negative) terminal of the input through just one switch Sjp (or Sjn). By the same token, the switching states are reduced from four to two, since the current path becomes independent. This explains why the MMCCC has shorter current paths and lower current stress than the flying-capacitor circuit as shown in Figure 3.1 does. 3.3. The Proposed Switched-Capacitor DC-DC Converter and Operation Principle Figure 3.4 shows the proposed switched-capacitor dc-dc converter with a voltage conversion ratio of six (named 6X). It can also function as a buck when the energy flows in the opposite direction. To explain the operating principle, its boost mode is taken as an example. The buck mode operation can be analyzed analogously. The charge pump splits into two symmetric 66 horizontal paths to build up the output voltage. In the upper path, the capacitor voltage is pumped up one by one to make the voltage of C3a equal to 3Vin; the lower path works in same manner, making the voltage of C3b equal to 3Vin. = or S4a S3a S2a + S1a S1p S2n C1 S3p S1n Vin S2p C2 C3 C4 Vout S3n    Figure 3.3. MMCCC in the similar form as the flying-capacitor circuit. S3a S2a S1a S1p Vin S2p C1a S3p C2a C3a   Vout S2n S1n S3n S1a = or C1b S2b C2b C3b S3b Figure 3.4. The proposed 6X switched-capacitor dc-dc converter. This converter alternates between two switching states as illustrated in Figure 3.5, with 50% duty ratio for each state as shown in Figure 3.6. The switching devices marked in solid line are 67 on-state devices and current paths; the remaining devices in dashed lines are off-state devices. Table 3.1 summarizes the switching states. The corresponding individual charge/discharge loops are shown in Figure 3.7. It is easy to infer that the switches Sjp and Sjn in the same phase leg are complementary; switches Sja and Sjb are complementary as well (i.e., if one is on, the other should be off and vice versa. j=1, 2, 3). In the switching state I as shown in Figure 3.5(a), in the upper path, the capacitor C1a is charged to Vin by the input through devices S1a and S1n, as simplified in Figure 3.7(a); the capacitor C2a is in series with the input to charge the capacitor C3a through the switches S2p, S3a and S3n, as simplified in Figure 3.7(b). In the lower path, the capacitor C1b is in series with the input to charge the capacitor C2b through the switches S1n, S2p and S2b, as simplified in Figure 3.7(c). C3b is discharged by the load current. In the similar way, in the switching state II as shown in Figure 3.5(b), the complementary switches are gated on, so that the capacitors C2a, C1b, C3b that are discharged in the first switching state become charged in the second switching state, while the capacitors C1a, C3a, C2b become discharged. In particular, C3a is discharged by the load current this time. Combining the voltage relations in the two switching states and neglecting the voltage drop, one can get the following voltage relations: VCja  VCjb  j  Vin ( j  1,2,3). (3.1) 68 S3a S1a Vin C2a C1a S2p   C3a Load Vout S3n S1n C1b C2b C3b S2b = or (a) Switching states I. + S2a S1p Vin C1a S3p C2a   C3a Load Vout S2n S1b = or C1b C2b C3b – S3b (b) Switching states II. Figure 3.5. Switching states of the 6X switched-capacitor dc-dc converter. Consequently, as the sum of the voltages across C3a and C3b, the output voltage is six times the input. In reality, the two capacitors, Cja and Cjb may be stabilized at a value slightly deviated from their theoretical values, because of the voltage drop, device tolerance and possibly diverse gate delays as analyzed and addressed in [95-97]. Nevertheless, the possible voltage difference 69 between a pair of capacitors, Cja and Cjb are tolerable during operation, since they are involved in two independent upper and lower charge paths, as can be seen from the equivalent circuits in Figure 3.7. 1 0 S1n , S2p , S3n , S1a , S2b , S3a t S1p , S2n , S3p , S1b , S2a ,S3b 1 0 T t Figure 3.6. Complementary PWM signals Table 3.1. Capacitor charge paths in two switching states Switching state I Switching state II Capacitor charge paths On-state switches Capacitor charge paths On-state switches Vin  C1a  S1a, S1n Vin  C1b  S1b, S1p C2a  Vin  C3a  S3a, S3n, S2p C 2b  Vin  C3b  S3b, S3p, S2n C1b  Vin  C 2b  S2b, S2p, S1n C1a  Vin  C2a  S2a, S2n, S1p S1a Vin   S1n S2p S3a C1a (a)Charging C1a. C2a S2p Vin   C3a S3n (b) Charging C3a. Iout Vin   S1n C1b S2b (c) Charging C2b. Figure 3.7. Charge/-discharge loops for two switching states. 70 C2b Figure 3.7 (cont’d) S1p Vin   S1b S3p S2a S1p C1b (d) Charging C1b. Vin   C1a C2a Vin S2n (e) Charging C3b.   S2n C2b S3b C3b Iout (f) Charging C2a. 3.4. Characteristics of the Proposed Switched-Capacitor DCDC Converter Evidently, the new circuit shares some of the salient features as the MMCCC: short current paths and low current stress, which are preferable for high voltage gain and high efficiency. Besides, it is much closer to a modular structure. A generalized NX dc-dc converter is depicted in Figure 3.8(a). It consists of N/2 basic cells (N=2k, k=1, 2,…) as shown in Figure 3.8(b). Inside each cell, a phase leg of two complementary switches Sjp and Sjn (j=1, 2, 3,…) and a pair of capacitors Cja and Cjb are connected together at their respective midpoints. Externally, the switch phase leg is in parallel with the input voltage source; two terminals PCja+ and PCjb– are connected to the next cell; two terminals PC(j-1)a+ and PC(j-1)b– from switches Sja and Sjb respectively are connected to the preceding cell, expect that the first cell is directly fed by the input. Only the capacitor voltages in different cells differ, like in the MMCCC. Yet, in the new converter, since the capacitor voltages range from Vin to N/2Vin in different cells, rather than from Vin to NVin as in MMCCC, it is easier for a modular design. 71 Moreover, the new converter has its unique features, compared to the original MMCCC for the same NX voltage ratio: 1) Two charge pump paths feed the load directly, leading to less power loss in the energy transfer. 2) Half of the capacitors reduced their voltages by N/2Vin. Switched-capacitor dc-dc converters rely on capacitors to transfer energy and to filter the output voltage. Normally the larger equivalent series resistance (ESR) associated with the higher capacitor voltage rating lowers the efficiency. Plus, capacitors contribute proportionally to the total volume and weight of the converter. High power density can be expected by employing low voltage capacitors. 3) There is lower capacitance and ripple current requirement for the two output capacitors. The output voltage ripples are reduced given the same capacitance as in the MMCCC, due to the interleaved charge/discharge of two output capacitors. It can be explained from the switching states in Figure 3.5: while one of the capacitors is being charged, the other one is being discharged. Consequently, the sum of two complementary voltage ripples makes the output voltage almost ripple-free. This feature is quite beneficial if the converter operates at zerocurrent switching (ZCS) as the ZCS-MMCCC does in [123], such that the output voltage ripples can be minimized. Besides, both output capacitors have smaller current ripples than the capacitors in the other basic cells, because they always supply load current while one of them gets charged alternatively. 4) The new converter employs fewer switches (2N versus 3N–2 in MMCCC) with no penalty of total device power rating, as will be calculated in the later section. The associated gate drive and accessory power supply are saved accordingly. 5) Each pair of complementary switching devices is truly capacitor-clamped. For instance, 72 when S2a in Figure 3.8(a) is gated off, the voltage across S2a is clamped by a natural clamp circuit formed by capacitors C1a, C1b, C2a, C2b and the diode in S2b. This is a virtue for designing a high power converter without assistance from extra clamping circuits. S1a Vin   SN/2,a S2a S1p C1a S2p C2a SN/2,p CN/2,a S1n C1b S2n C2b SN/2,n CN/2,b S1b + Vout  SN/2,b S2b = or (a) The generalized NX switched-capacitor dc-dc converter PVin+ S ja PC(j-1)a+ PCja+ S jp S jn PC(j-1)b- C ja C jb S jb PCjb- PVin- (b) Basic cell. Figure 3.8. A generalized NX switched-capacitor dc-dc converter constructed from basic cells. In sum, its many features allow further higher efficiency with more compact package and lighter weight for high voltage boost gain than the MMCCC. Apparently, when N equals two, this converter reduces to the dc-dc converter module with a voltage conversion ratio of two in [111], in which very high efficiency was already demonstrated on a 10-kW converter. 73 3.5. Component Cost Comparison with Other Topologies To compare the cost of the new converter with its counterparts, the total device power rating, the capacitor voltage stress, current rating and capacitance requirement are itemized. The total device power rating is based on the product of the maximum voltage imposed on the device and the average current flowing through it over the duration when the device conducts. Note that neither the peak current nor RMS current is used. In this way, the comparison is less dependent on the actual shape of the charge and discharge current, which is a function of the ESR and equivalent series inductance (ESL). 3.5.1. Total Device Power Rating 1) For the original flying-capacitor (FC) structure, all the 2N devices sustain the voltage equal to the input voltage and the input current. Its TDPR is the same as the traditional boost converter: TDPRFC  2 N  (Vin  I in )  2 N  Pin (3.2) where Vin is the input voltage and Iin is the input current. 2) The MMCCC also pumps charge from one capacitor to the next one, but its switching states reduce to two. Thus, the charge current into one capacitor is the discharge current from its preceding capacitor, except that the output capacitor has half the charge and discharge current. Also considering that the average charge current of one capacitor in half switching period equals its average discharge current in the other half switching period, the average current through each switching device is 2Iout in one of the two switching states. There are (N–2) switches sustain twice the input voltage, as stated earlier. Thus, the TDPR is 74 TDPRMMCCC  2 N Vin  2 I out    N   2Vin  2 I out  8N  8  Pin ( N  2,3, 4,...). N (3.3) 3) For the new converter, the (N–2) switches in the complementary phase leg convey the sum of the current in two charge pump paths, 4Iout, which is twice the current through the rest switches. It is not hard to find the voltage stress of each switch. Hence, the TDPR can be derived as: TDPRnew  ( N  2)Vin  4 I out    Vin   I out  N  2 Vin   I out 8N  8  Pin ( N  2, 4, 6,...). N (3.4) The above equations clearly demonstrate that unlike the conventional flying-capacitor structure, the new converter has no penalty of total device power rating even with fewer devices than the MMCCC. The ratio of the TDPR and the input power is plotted with respect to the voltage gain in Figure 3.9. It is quite interesting that this ratio for the new converter will get saturated as N approaches infinite. This property is different from the TDPR in the FC as well as in the conventional two-level boost converter. It can be physically explained by the aforementioned fact that the charge/discharge current will only go through three switching devices at most, and that the maximum switching device voltage stress is no more than 2Vin, in spite of the increment of voltage gain. 75 20 Flying-capacitor MMCCC The new NX converter 18 TDPR / Pin 16 14 12 10 8 6 4 2 4 6 Boost gain, N 8 10 Figure 3.9. Normalized total device power rating vs. voltage boost gain 3.5.2. Capacitor Voltage, Current and Capacitance Requirement Table 3.2 compares the total capacitor voltage, the current and the capacitance of the new converter with that of the FC and MMCCC. While the voltage ratings for the FC and the MMCCC are the same, the voltage rating for the new converter is reduced nearly by half, as plotted in Figure 3.10. The RMS current is related to the parasitic parameters in the circuit, but the average charge (/discharge) current of the internal capacitors in the new converter (and in the MMCCC) can be calculated as 2/N times of the current in the FC dc-dc converter, as stated before; the average charge (/discharge) current of the two output capacitors is 1/(N-1) times of that in the FC dc-dc converter. Thereby, the capacitance requirement can be obtained accordingly. The quantitative comparison supports the statement that the ripple current and capacitance requirement of the two output capacitors in new converter can be much lower. 76 Table 3.2. Comparison of capacitor voltage rating, current and capacitance Total capacitor voltage ratings MMCCC NX Capacitance requirement  N ) N Vin 2  I out  f V  j  1,2, N  1)  NI out ,( j  1,2,...,N  1)  s j Cj  Ij  ( N  1) I out ,( j  N )   N  1) I out  j  N )  Nf s V j   N ) N Vin 2  I out  f V  j  1,2, N  1)  s j Cj   I out / 2  j  N )  f s V j   N / 2) N Vin 2  N  2,4,6,...) 60 Total capacitor voltage rating / Vin FC Average charge/discharge current 50 2 I out ( j  1,2,..., N  1) Ij   I out ,( j  N ) C ja  C jb I ja  I jb  I out N 1  f V ( j  1,2,..., 2 )  s j   I out / 2( j  N ) 2  f s V j   2 I out , ( j  1,2,..., N 1)  2  N)  I out ( j  2  Flying-capacitor and MMCCC The new NX converter 40 30 20 10 0 2 4 6 Boost gain, N 8 10 Figure 3.10. Normalized total capacitor voltage rating vs. voltage boost gain. 77 3.6. Power Loss Analysis The power loss is broken down into the conduction loss, switching loss and gate drive loss. The conduction loss, which includes the charge/discharge loss due to the ESR of the capacitors, is the dominant loss in this design. Thereby, the problem turns to identifying the current in the individual charge/discharge loops in two switching states. Since the two states are essentially symmetric, the total loss is calculated based one of the switching states as shown in Figure 3.5(a) and multiplying it by two. This switching state is divided into three charge/discharge loops, as illustrated in Figure 3.11. In the modular design, each module is fed by two parallel electrolytic capacitors in a distributed way, and the ESR introduced in each module is defined as Rin. An MOSFET is modeled as an equivalent series resistance (ESR). To decouple the two loops sharing the same MOSFET Sjp(n) (j=1, 2), the resistor RSjp(n) is split into two equivalent resistors 2RSjp(n) in parallel, based on the fact that the average is ideally the same. In fact, they are equal only if the currents are exactly identical. Nonetheless, this approximation only makes the assumption a little more conservative. Stray inductance lumped as Ls is advantageously utilized in the circuit to reduce the capacitance requirement and to mitigate the impulse charge current present in the switchedcapacitor circuit where the capacitor and the ESR used to be the main players in the circuit. The presence of the ESL makes the initial current in each loop starts from zero. Define the resonant frequency in each loop as: j  1   2 ,  j  1, j Ls Cloopj 78 (3.5) where the decay rate  j  R j / (2 Ls ) , Cloopj is the equivalent loop capacitance as will be detailed later. Ls 2 R2 Ls1 C1b C2b C1a Vin Vin R1 (a) Ls3 C2a (b) R3 C3a C3b I out I out RC 3b Vin (c) (d) Figure 3.11. Equivalent circuits of Switching State I counting in the ESL and ESR. 1) The charge loop for C1a The capacitor C1a gets charged by the input. The equivalent circuit is reduced to Figure 3.11(a). The current loop can be expressed as 2  2  1 i1 (t )  e t (C1V1 1 sin 1t ) 1 (3.6) where C1  C1a  C1b , ESR in the loop is: R1  RSja  2 RSjn  Rin  RC1a  RLs1 , and the 79 voltage difference is V1  1 2 I out  T / 2 PoutT  . C1 C1(12Vin ) 2 2) The charge loop for C2b In this loop as shown in Figure 3.11(b), one more capacitor comes into the picture. The capacitors C1b and C2b are equivalently in series, making the loop capacitance Cloop2  C1C2 , where C2  C2a  C2b . C1  C2 The loop current can be computed as i2 (t )  e  t (Cloop 2 V2 2  2  2 2 2 sin  2t ) (3.7) where the ESR in the loop is: R2  RSjb  2(2 RSjn )  Rin  RC1b  RC 2b  RLs 2 and the voltage difference is V2  1 PoutT 1 1 (  ). 2 6Vin C1 C2 In general, for a conversion ratio of NX (N=6, 8, 10, …), the charge/discharge loops, except the two final output stages, comply with a general equation: i j (t )  e jt (Cloopj V j 2   2 j j j sin  j t ), for j  1, 2,...,( N / 2  1) (3.8) 3) The charge/discharge loop for the output stage The output stage is different from the others. The load current, Iout gets involved via discharging the output capacitors C3a(b), as shown in Figure 3.11(c). It can be approximated as a constant dc, due to the fact that the output voltage is almost ripple-free. The resultant loop current is: 80 i3 (t )  e 3t  k1  cos 3t  k 2  sin 3t   k1 (3.9) where the ESR in the loop is: R3  RSja  2(2 RSjn )  Rin  RC 2 a  RC 3a  RLs3 , the constant is defined as k1  C2 I out , k2  C2  C3 capacitance is Cloop3  V3  I 2 2 out 32  32  Cloop3  V3  CCRs3Cout   C32CIC3 2 3 C2C3 C2  C3 3 . The loop with C3  C2a  C2b , and the voltage difference is: 1 PoutT 1 1 (  ). 2 6Vin C2 2C3 A. Conduction loss With the loop current obtained, the conduction loss for three loops can be calculated: T 1 2 Pcon1   2 i1 R1dt 0 T 1  2T 2  T   e T  cos  T   sin  T   P T 2   2    1 1 1   .(3.10)  out 1 1  R   e    1   2 2  I out 1  2 2 1  1         T 1 2 Pcon 2   2 i2 R2 dt T 0 1  2T 2   2T   e  2T  cos  T   sin  T   P T 2   2   2 2 2 2 2   .(3.11) out 2 2  R   e    2   2 2  I out 2  2 2 2  2  2         For the output stage, the conduction loss counts in the loss of the output capacitor C3b being discharged by the load current: 81  T  1 2 2 T 2 Pcon3  i R3dt   I out RC 3   T  0 3 2      2 2  2 2 k1  k 2  e  T k1  k 2  3  e  T    cos 3T  3 sin 3T   . (3.12)      2   2 2 2T 2 2  3  3       The total conduction loss is twice the sum of the above itemized losses: Pcon  2( Pcon1  Pcon2  Pcon3 ) . (3.13) B. Switching loss For the purposes of estimating the current at the instant when the MOSFET is turned off, · substitute the value of the actual duty cycle times the switching cycle, · (3.9). If the dead time (<100 ns) is neglected, into (3.6), (3.7) and is simplified to /2, and the current at the end of this half switching cycle is: 2 2 1  1 T  T / 2 (C1V1 sin 1  T / 2) i1 ( )  e 2 1 . (3.14) When the resonant frequency of each charge/discharge loop is designed to be close to but below the switching frequency of the converter, the switching-off current is very small. More interesting, a natural zero-voltage turn-on and turn-off can occur for the MOSFETs Sjp(n). For example, during the transition from Switching State I to Switching State II, the ESL in the circuit forces the loop current to charge the junction capacitors of S1n, S2p and S3n and discharge the junction capacitors of their complementary switches. After S1n, S2p and S3n turn off, the inductive currents freewheel through the diodes during the short dead time, if only they have not decayed to zero. The freewheeling loops are shown in Figure 3.12. When the converter enters 82 into the Switching State II, the MOSFETs S1p, S2n and S3p can be turned on at zero-voltage. It is the same for the transition from Switching State II to Switching State I, due to the symmetry. Thus, their switching loss can be negligible. The total switching loss is mainly the loss of charging and discharging the junction capacitors of the MOSFETs Sja(b):  2  Psw  C1V 2  2  C2  2Vin   4  f s in . (3.15) C. Gate drive loss Turn-on and turn-off twice. Therefore, the gate drive loss is arrived at: Pgd  V gs  QSjp ( n )  QSja (b ) ) f s  N (3.16) where N=6 indicating N pairs of MOSFETs Sjp(n) and Sja(b), and V gs  12V ,QSjp ( n )  195nC ,QSja (b)  210nC . D. The total loss: Ploss  Pcon  Psw  Pgd . S2a S1a  Vin  (3.17) S3a S1p C1a S2p C2a S3p C3a S1n C1b S2n C2b S3n C3b S1b S2b S3b = Figure 3.12. Transient state after the Switching States I. 83 + Vout  3.7. Simulation and Experimental Verification The principle and analysis are verified by simulation and experiments of a 6X 450-W prototype. The modular configuration in the simulation and experiment is depicted in Figure 3.13. Ceramic capacitors C5750X7S2A106M are used. Since their capacitance is affected by the operating voltage, the equivalent capacitance at their corresponding voltages is: C1a(b)120 μF, C2a(b)60 μF, C3a(b)40 μF, according to the manufacture’s data. The switching frequency is 100 kHz. The MOSFET IRF1324S-7PPbF is employed for the devices Sjp(n), and IRFS30047PPbF is for Sja(b). The minimum stray inductance needs to be more than 21 nH in the first module, and to be more than 63 nH and 84 nH for the second and third module respectively, based on the critical zero current condition and the equivalent loop capacitance. Considering the parameter tolerance and variation, the stray inductance is intentionally designed to make the converter switch off at relatively small yet non-zero current. Iin Vin   S1a LS S2a S3a Iout S1p C1a S2p C2a S3p C3a S1n C1b S2n C2b S3n C3b S1b S2b + Vout  S3b = Figure 3.13. Simulation and experimental configuration. Simulation results were shown from Figure 3.14 and Figure 3.16. In Figure 3.14, VGS1a and VGS1b represent two complementary gate drive signals, and voltages VS1a (/VS1b) and VS3a 84 (/VS3b) represent two kinds of voltage stress across the switching devices. The capacitor voltages are shown in Figure 3.15. The input/output voltages and currents are shown in Figure 3.16. The corresponding experimental results are given from Figure 3.17 to Figure 3.20. The output voltage is boosted from a 12-V input to 68.2 V. Note that the input current, Iin is referred to the current before the capacitors across the input dc bus as shown in Figure 3.13. As can be seen from Figure 3.18 and Figure 3.19, the two symmetric capacitors in one basic cell have the same very low average voltage and nearly complementary voltage ripples. Hence, the output voltage ripple is quite small as shown in Figure 3.20. The experimental results agree with the analysis and simulation. The calculated efficiency and the experimental result are given in Figure 3.21. The input voltage is 12 V and the gate drive power loss of 2.9 W is included. The power loss analysis well predicts the trend of the efficiency variation vs. the power. The calculated efficiency is a bit more above the experimental result at the light-load condition because the loop current becomes insufficient to make the soft switching assumption valid. 85 VGS1b VGS1a (V): t(s)  20.0 10.0 0.0 10.0 (V) VS1a (V): t(s)  20.0 10.0 0.0 10.0 (V) VS1b (V): t(s)  40.0 20.0 0.0 20.0 (V) VS 3b VS 3a (V) 40.0 20.0 0.0 20.0 (V): t(s)  2.16m t(s) 2.18m 2.2m Figure 3.14. Simulation results of gate drive signals and typical switch voltages. 86 (V) : t(s) (V) 15.0 VC1a VC1b 10.0 5.0 (V) : t(s) 30.0 VC2a (V) 25.0 VC2b 20.0 15.0 10.0 (V) : t(s) 40.0 VC3a (V) 35.0 VC3b 30.0 25.0 20.0 2.16m 2.18m t(s) Figure 3.15. Simulation results of capacitor voltages. 87 2.2m (V) : t(s) (V) 20.0 Vin 10.0 0.0 (A) : t(s) (A) 60.0 Iin 40.0 20.0 (V) : t(s) 80.0 (V) 0.0 Vout 60.0 40.0 (A) : t(s) (A) 10.0 Iout 5.0 0.0 2.16m 2.18m t(s) 2.2m Figure 3.16. Simulation results of input/output voltages and currents. 88 VGS1a : 10 V / div VGS1b : 10 V / div VS1b : 10 V / div VS1a : 10 V / div t : 2  s / div Figure 3.17. Experimental results of complementary gate drive signals and the corresponding switches. VC 2b : 5 V / div VC1a : 5 V / div VC 2a : 5 V / div VC1b : 5 V / div t : 5  s / div Figure 3.18. Experimental results of capacitor voltage VC1a(b) and VC2a(b). 89 VC 3b : 10 V / div VC 3a : 10 V / div VS 3b : 20 V / div VS 3a : 20 V / div t : 5  s / div Figure 3.19. Experimental results of capacitor voltage VC3a(b) and switch waveform VS3a(b). I in : 20 A / div I out : 5 A / div Vout : 20 V / div Vin : 10 V / div t : 5  s / div Figure 3.20. Experimental results of input/output voltage and current 90 97 Efficiency (%) 96 95 94 93 92 91 Calculation Experiment 50 100 150 200 250 300 350 Output power (W) 400 450 Figure 3.21. Comparison of calculated and tested efficiency 3.8. Application Examples One of the applications of the new converter is the high-gain voltage boosting for TEG. Isolated dc-dc converters are usually applied to achieve high voltage gain, although no isolation is needed. The associated transformer degrades efficiency. A multi-phase ZCS MMCCC has been proposed in [124], which involves many switching devices, which affects reliability. A configuration example using the NX dc-dc converter is shown in Figure 3.22. The original Sja(b) can be either diode or MOSFET for the unidirectional operation. The NX converter produces an unregulated voltage of clearly N times the generated voltage from TE module(s). It can be followed by a boost- or buck-type dc-dc regulator, such as a traditional boost or buck converter, three-level flying capacitor dc-dc [91], resonant switched-capacitor [125] and so on. Since the regulator is only dedicated to voltage regulation, and the maximum power point tracking (MPPT) or load matching, their current stress and the power rating is dramatically mitigated. As a result, a high efficiency can still be expected. 91 Modular NX dc-dc converter Boost- or buck-type regulator TE module(s) SLI battery = Figure 3.22. Modular NX dc-dc converter for the voltage boosting in TEG. Another prospective application is PV micro inverters. As briefly introduced in Chapter 1, a micro inverter is comprised of a dc-dc converter and a dc-ac inverter. The NX dc-dc converter is employed to boost the low voltage from the PV panel to a decent voltage for the dc-ac inversion stage afterwards. It can be directly mounted underneath the PV panel. The next question is: what about the dc-ac inverter in the micro converter? It will be answered in the next chapter. 3.9. Summary A novel switched-capacitor dc-dc converter with the potential of high voltage gain is proposed. Compared to its switched-capacitor dc-dc converter counterpart, it has the main advantages: - Less power loss due to the two symmetric short paths of charge pumps; - Substantially reduced total capacitor voltage ratings; - Lower capacitance and ripple current requirement of the output capacitors; - Reduced switching device count, low device current stress and low total device power rating; 92 - Improving efficiency; - Lower cost; - Able to have bidirectional operation; - Quasi modular structure. The experimental results verified the operation principle and features. 93 Chapter 4 Trans-Z-Source Inverters 4.1. Introduction Generally, there exist two types of traditional inverters: voltage source inverters (VSI) and current source inverters (CSI) as shown in Figure 4.1 and Figure 4.2. Traditional VSIs and CSIs have similar limitations and problems. For VSIs: 1) The obtainable ac output voltage can not exceed the dc source voltage. So a dc-dc boost converter is needed in the applications, for instance, with limited available dc voltage or with the demand of higher output voltage. 2) Dead time is required to prevent the shoot-through of the upper and lower switching devices of each phase leg. However, it induces waveform distortion. For CSIs: 1) Their output voltage can not be lower than the dc input voltage. 2) Overlap time between phase legs is required to avoid the open-circuit of all the upper switching devices or all the lower devices. A Z-source inverter (ZSI) [126] as shown in Figure 4.3, as well as the derived quasi Z-source inverters (qZSI) [127, 128], has been proposed to overcome the above problems. They advantageously utilize the shootthrough of the inverter bridge to boost voltage in the VSIs (or open-circuit in the CSIs to buck voltage). Thus, buck-boost functionality is achieved with a single stage power conversion. They also increase the immunity of the inverters to the EMI noise [129], which may cause mis-gating and shoot-through (or open-circuit) to destroy the conventional VSIs and CSIs. 94 DC current source or load DC voltage source or load To ac load or source To ac load or source  Vdc  = , , Figure 4.1. Voltage source inverter DC (voltage or current) Source or load OR , etc. Figure 4.2. Current source inverter Z Source L1 C1 Converter or Inverter C2 To (DC or AC) Load or Source L2 A switch or a combination of switching device(s) and/or diode(s) Figure 4.3. Z-source inverter The voltage-fed (VF) Z-source inverter can have theoretically infinite voltage boost gain. However, the higher the voltage boost gain is, the smaller modulation index has to be used. In applications such as grid-connected photovoltaic (PV) generation and fuel cell power conversion, a low voltage dc source has to be boosted to a desirable ac output voltage. A small modulation index results in a high voltage stress imposed on the inverter bridge. Several pulse width modulation (PWM) methods [130, 131] have been developed with the attempt of obtaining as much voltage gain as possible and thus limiting the voltage stress across the switching devices. The maximum boost control [130] achieves the maximum voltage gain through turning all the zero states in the traditional VSIs to shoot-through zero states. Nevertheless, it brings in low 95 frequency ripples associated with the ac side fundamental frequency. So the constant boost control [131] has been proposed to eliminate those ripples and thus reduce the L and C requirement in the Z-source network, with slightly less voltage gain, compared to the maximum boost control. These PWM methods still have limits to further extend the voltage gain without sacrificing the device cost. Recently, some modified impedance source networks were proposed in [132-135] for the sake of increasing the output voltage gain. Among them, a T-source inverter [132] has the possibility of increasing voltage gain with the minimum component count. As will be discussed in the next section, it can be grouped into a general class of transformer based Zsource inverters (trans-ZSIs) presented in this paper, which employ two transformer windings in the impedance network. The voltage-fed Z-source/quasi Z-source inverters can not have bidirectional operation unless replacing the diode with a bidirectional conducting, unidirectional blocking switch [136]. Neither can the traditional current source inverters ([137-139]) do unless they are fed by a phasecontrolled rectifier in front that can change the dc link voltage polarity. Interestingly, the currentfed (CF) Z-source/quasi Z-source inverters [127, 140, 141] can have voltage buck-boost and bidirectional power flow only with a diode in the impedance network. With the newly developed reverse blocking IGBTs [142], this single stage power converter becomes a promising topology. Yet, their dc–ac voltage gain can not exceed two in the boost mode operation. In other words, the dc input voltage cannot be lower than half of the peak output line-to-line voltage. Hence, this paper further proposes two current-fed trans-Z-source inverters that are capable of reaching wider voltage boost range and bidirectional power flow with a single diode. As a result, a wider output voltage range can be obtained, which is essential to some applications such as HEV/EV motor drives. 96 The trans-Z-source inverters can be derived from the voltage-/current-fed quasi Z-source inverters or the voltage-/current-fed Z-source inverters. The trans-Z-source-inverters inherit their unique features, and they can be controlled using the PWM methods applicable to the Z-source inverters. This paper will begin with the derivation of two voltage-fed trans-Z-source inverters from one of the quasi Z-source inverters. Next, the same idea is extended to the development of two current-fed trans-Z-source inverters. Then the followed comparative analysis, simulation and experimental results will demonstrate their new properties different from the Z-source/quasi Zsource inverters. 4.2. The Voltage-fed Trans-Z-Source Inverters In the voltage-fed quasi Z-source inverter with continuous input current, two dc inductors can be separated or coupled. When the two inductors are coupled as shown in Figure 4.4, the voltage across the inductor L1 is reflected to the inductor L2 through magnetic coupling. Then one of the two capacitors, for instance, C2 can be removed from the circuit. The rearrangement of the circuit yields the structure as shown in Figure 4.5. Furthermore, the voltage across L2 can be made proportional to the voltage across L1 by changing the turns ratio n2/n1. As the voltage constraint associated with one of the capacitors is released, the two windings, to some extent, behave more like a flyback transformer rather than the original coupled inductors [143], except that the currents flow simultaneously through both windings in some of the operation states. Therefore, it is named as the voltage-fed trans-quasi-Z-source inverter (trans-qZSI). 97 C1 1:1 Vdc   L1 D L2 C2 Toacload Figure 4.4. Voltage-fed quasi Z-source inverter with coupled inductors. Like the Z-source inverter, the trans-quasi-Z-source inverter has an extra shoot-through zero state besides the six active states and two traditional zero states. The shoot-through zero state can be realized by short-circuiting both the upper and lower switching devices of any one phase leg, any two phase legs, or all three phase legs. The shoot-through zero state contributes to the unique buck-boost feature of the inverter. Otherwise, when the dc voltage is sufficient to produce the desirable ac output voltage, a traditional PWM without shoot-through zero state is used. For the purpose of analyzing the characteristics of the trans-Z-source inverters, this paper will focus on the two general continuous current modes as in the Z-source inverter: the shoot-through zero state and the non-shoot-through states [126, 144]. By replacing the two windings with an ideal transformer and a mutual inductance (Lm) (a model used in [145]), the overall equivalent circuit viewed from the inverter dc side can be obtained as shown in Figure 4.6(a). 98 C1 n :1  Vdc  D L2 L1 Toacload Figure 4.5. Voltage-fed trans-quasi-Z-source inverter. In the shoot-through zero state, the inverter is equivalent to a short circuit as shown in Figure 4.6(b). Given that the inverter is in the shoot-through zero state for an interval of DshT during a switching cycle, T, the voltages across L1 and L2 are: vL1  Vdc  VC1 (4.1) n vL 2  2 vL1 n1 (4.2) Thus, the diode is reversed biased. Note that the symbol Dsh is used here for shoot-through duty ratio in voltage-fed Z-source inverters. 99 VC1   iC1 idiode n :1 idc vdiode  iL 2 vL 2   Vdc  iL1  Lm im v vL1 i  ii (a) With the transformer equivalent circuit. VC1   iC1 n :1 idiode idc vdiode  iL 2 vL 2   Vdc iL1  im Lm v vL1 i  ii (b) Shoot-through zero states. VC1   iC1 idiode n :1 idc vdiode  iL 2 vL 2   Vdc iL1  im Lm v vL1 i  ii (c) Non-shoot-through states. Figure 4.6. The equivalent circuits of the voltage-fed trans-qZSI viewed from the dc link. In any of the non-shoot-through states for an interval of (1―Dsh)T, the inverter bridge can be modeled as an equivalent current source as shown in Figure 4.6(c). The non-shoot-through 100 states include the six active states and two traditional zero states. For the traditional zero states, the current source has zero value (i.e. an open circuit). During one of the non-shoot-through states, one can get: vL2  VC1 (4.3) n n vL1  1 vL2   1 VC1 n2 n2 (4.4) The average voltage of both inductors should be zero over one switching period in the steady state. Thus, from (4.1) to (4.4), we have: n (Vdc  VC1 ) DshT  ( 1 VC1 )(1  Dsh )T n2 v L1  0 T (4.5) From the above equation, the capacitor voltage can be calculated as: VC1  n  Dsh Vdc 1  (1  n) Dsh (4.6) where n  n2 / n1  1 . From (4.4) and (4.6), the dc link voltage across the bridge in the non-shoot-through states can be boosted to: ˆ vi  1 Vdc  BVdc 1  (1  n) Dsh (4.7) 1 1  (1  n) Dsh (4.8) where the boost factor is: B The peak value of the phase voltage from the inverter output is: 101 ˆ ˆ V ph  M Vi / 2  M  B Vdc / 2 (4.9) where M is the modulation index. When the constant boost control [131] is used, the voltage gain (MB) as defined in [130] is: G  MB  M 1  (1  n)(1  3 M ) 2 (4.10) It can be seen that if the turns ratio is 1, the inverter dc link voltage boost gain is the same with that of the original Z-source/quasi Z-source inverters, but one capacitor is saved in the new trans-Z-source network. If the turns-ratio is over 1, the inverter dc link voltage boost gain can be higher given the same modulation index, M. In other words, it needs a smaller shoot-through duty ratio Dsh (accordingly a larger modulation index M) to produce the same ac output voltage than the Z-source/quasi Z-source inverters do. The voltage gain (MB) versus the modulation index for the voltage-fed trans-quasi-Z-source inverter (with a turns ratio n=2) is compared in Figure 4.7 with that for the Z-source/quasi Z-source inverters, using the constant boost control. The voltage stress, Vs across the switching devices can be assessed by comparing its peak dc link voltage against the minimum dc voltage (GVdc) [131] needed for the traditional VSI to produce the same ac output voltage at M=1. The ratio represents extra cost that the voltage-fed trans-quasi-Z-source inverter and Z-source/quasi Z-source inverters have to pay for the voltage boost in association with the higher voltage stress. The ratio of the voltage stress to the equivalent dc voltage for the trans-quasi-Z-source inverter is: Vs BVdc 3 1 1   (1  )  GVdc GVdc 2 n nG (4.11) When n=1, it is the same with the relative voltage stress in the Z-source inverter. When n>1, 102 as can be seen in Figure 4.8, the voltage-fed trans-Z-source inverter has less voltage stress across the inverter bridge for the same dc-ac output voltage gain. Hence, this circuit is beneficial to applications, in which a high voltage gain is required. Voltage gain (MB) 10 VFTrans-ZSI/-qZSI (n=2) VFZSI/qZSI 8 6 4 2 0 0 0.4 0.6 0.8 Modulationindex 0.2 1 1.2 Figure 4.7. Voltage gain (MB) versus modulation index of the voltage-fed trans-ZSI/-qZSI (n=2) Voltage stress/equivalent dc voltage and ZSI/qZSI. 1.8 VF Trans-ZSI/-qZSI, n=2 VF ZSI/qZSI 1.6 1.4 1.2 1 0.8 2 4 6 Voltage gain (MB) 8 10 Figure 4.8. Active switch voltage stress of voltage-fed trans-ZSI/-qZSI (n=2) and ZSI/qZSI. Similarly, another trans-Z-source inverter can be reconfigured as shown in Figure 4.9, if C1 103 is removed in Figure 4.4 instead of C2. This trans-Z-source inverter is the same as the T-source inverter proposed in [132]. It is obvious that it essentially has the same operation principle, voltage gain and voltage stress as the previously developed voltage-fed trans-quasi-Z-source inverter, except for different capacitor voltage stress and different input current drawn from the dc source. Therefore, Figure 4.4, Figure 4.5, and Figure 4.9 can be classified as a class of voltage-fed trans-Z-source inverters. In Figure 4.9, the capacitor voltage is: VC1  1  Dsh Vdc 1  (1  n) Dsh (4.12) When the turns-ratio, n of the trans-Z-source inverter in Figure 4.9 equals 1, (4.12) becomes the same equation for the capacitor voltage in the original Z-source inverter.  Vdc D L2  n :1 iL 2  iL1 L1 Toacload C1 Figure 4.9. Voltage-fed trans-Z-source inverter. 4.3. The Current-fed Trans-Z-Source Inverters Applying the similar concept to the current-fed Z-source and quasi Z-source inverters leads to the current-fed trans-Z-source and trans-quasi-Z-source inverters. The current-fed quasi Zsource inverter with continuous input current is shown in Figure 4.10, with inductors L1 and L2 coupled. The dc current source is provided by a dc inductor Ldc. Similarly, this circuit can be 104 further modified to a current-fed trans-quasi-Z-source inverter (trans-qZSI) as shown in Figure 4.11, where n  n2 / n1  1 . The current-fed trans-quasi-Z-source inverters have two unique open zero states besides six active states and three traditional short-circuit (or shoot-through) zero states. The open zero states can be realized by turning off either all the upper switches or all the lower switches. The open zero states are forbidden in the conventional CSI. The new Z-source inverters, however, utilize the open zero state so that the inverter can step up and down the voltage. I dc L2  Vdc C2 1:1 D C1 Toacload orsource L1 Figure 4.10. Current-fed quasi Z-source inverter with two inductors coupled. I dc  Vdc Ldc L2 C1 Toacload orsource n :1 L1 D Figure 4.11. Current-fed trans-quasi-Z-source inverter. Its equivalent circuit viewed from the inverter dc side is shown in Figure 4.12(a), assuming an ideal coupling for the transformer. When in an open zero state, the inverter bridge turns to an open circuit as shown in Figure 4.12(b). When in one of the non-open states (which includes the 105 traditional active states and short-circuit zero states), the inverter bridge can be modeled as an equivalent voltage source, vi, as shown in Figure 4.12(c). The equivalent voltage source, vi is the line-to-line voltage when in one of the six active states and is zero when in one of the shortcircuit zero states. I dc Ldc iL 2 iC1  VC1   Vdc   n :1 ii  iL1 iL1 idiode  Lm im vdiode  vi  (a) With the transformer equivalent circuit.  Vdc  I dc  Ldc vL 2  iL 2 iC1  VC1   n :1  iL1 iL1 idiode vi Lm im vL1  (b) Open zero state. Figure 4.12. The equivalent circuits of the current-fed trans-qZSI viewed from the dc link. 106 Figure 4.12 (cont’d)  Vdc  I dc Ldc v L2  iL 2 iC1  VC1  n :1   iL1 iL1 ii  v idiode  i Lm im vL1  (c) Non-open states. Using the similar circuit analysis with the above voltage-fed trans-Z-source inverters, the following current relation and the dc-ac voltage gain can be obtained for the current-fed trans-Zsource inverters. According to the KCL and the relationship of the winding currents, the following relations hold:  iC1  iL1  iL2 iL1  im  i 1iL1  niL2  0 L (4.13) In the open-zero states as shown in Figure 4.12(b) for a duration of DopT , the diode is conducting to carry the current through L1. From (4.13), the capacitor current can be expressed as iC1  (1  n)idc  im (4.14) In the non-open states for a duration of (1  Dop )T , the diode D is biased off; no current flows through L1. The capacitor current is iC1  iL2  im / n (4.15) With the assumption that the inverter operates in the continuous current mode, given small 107 magnetizing current ripple and input current ripple, both im and idc can be approximated by their dc components Im and Idc respectively. Application of the amp-second balance to the capacitor yields Im  n(1  n) Dop I dc 1  (1  n) Dop (4.16) Substituting Im to the above expressions for the non-open states, the peak value of the current flowing through the inverter dc link can be obtained as ˆ ii  1 I dc  B  I dc 1  (1  n) Dop , (4.17) where B is the boost ratio in terms of current. Correspondingly, the ac voltage can be bucked (stepped down). Based on the operation principle of SPWM for the CSI, the maximum duty ratio of two active states in one switching period for each phase can be calculated by 3  M sin t M sin(t  2 / 3)  D A max  max  M   2 2   2 (4.18) where M is the modulation index based on the current reference. Therefore, the peak value of the line current can be expressed as 3 ˆ ˆ il  ( M )  ii 2 (4.19) Assuming the energy conservation of the input and output, the peak value of the line-to-line voltage can be written as ˆ vll  4 Vdc [1  (1  n) Dop ] Vdc  I dc   ˆ 3M cos  3 il    cos  where cos is the power factor. 108 (4.20) Similar to the voltage-fed ZSI, there are corresponding control methods: simple boost, maximum boost and constant boost controls in terms of current (or buck in terms of voltage) for the current-fed Z-source inverters. The constant current boost is used here to compare the operation region of the current-fed trans-ZSIs and current-fed ZSIs. In order to keep the active state unchanged, the maximum available open zero duty ratio is: Dop max  1  D A max  1  3 M 2 (4.21) Hence, (4.20) can be rewritten as ˆ vll  3 M  1)  1] 2 3M cos  4  Vdc [(1  n)( (4.22) When n=1, (4.20) and (4.22) can be proven to be the same as in the Z-source and quasi-Zsource inverters. The ratios of the peak line-to-line voltage to the input voltage as a function of the modulation index are compared in Figure 4.13(a) and (b) for (quasi-)Z-source and trans(quasi-)Z-source inverters, where n equals 2 for the trans-Z-source inverters. The blue dash line is the minimum voltage envelope for the buck mode and the regeneration mode, with the constant current boost control. The shaded areas are operable regions with only one additional diode, D in Figure 4.10 and Figure 4.11. When no open state is employed (that is, Dop = 0), the red envelop line in Figure 4.13(a) and (b) that can be obtained from (4.20) depicts the maximum voltage boost gain for both the current-fed (quasi-)Z-source inverters and the trans-quasi-Zsource inverters. It is noticeable that, however, the trans-Z-source has a wider boost range in the motoring operation mode. In the current-fed quasi Z-source inverter [140], the dc link voltage vi cannot exceed the sum of capacitor voltages, because the diode will otherwise improperly 109 conduct in the active states and cause waveform distortion. For the same reason, in the transquasi-Z-source inverters, to maintain the diode reversed biased in the active state, the voltage across it should satisfy vdiode  1 1 1 (VC1  vi )  VC1  (1  )Vdc  vi  0 n n n (4.23) Hence, the peak value of the line-to-line voltage is constrained to ˆ ˆ vi  vll  4 Vdc  (1  n)Vdc 3M cos  (4.24) It implies that the minimum modulation index is related to the power factor seen from the inverter bridge. When n>1, the peak line-to-line voltage can reach (n+1)Vdc, as shown in Figure 4.13(b). Note that the dc-ac voltage gain can also be analyzed by calculating the voltage relation and active duty ratio in the same way as in [140], where the active and short-circuit zero states are depicted separately. Similarly, a current-fed trans-Z-source inverter (trans-ZSI) can be derived as shown in Figure 4.14. It has the same operation principle, voltage gain and operation regions as the transquasi-Z-source inverter, but with different current stress, which will be discussed in the next section. 110  4   6cos  ,2    v ˆ Dc-ac voltage gain ll Vdc 3 Dop  0 2 Motoring 1 region Dop  0 0 -1 Regeneration region -2 -3 0.2 0.4 0.6 0.8 1 1.2 Modulation index M (a) The current-fed ZSI/qZSI.  4   9cos  ,3    4 v ˆ Dc-ac voltage gain ll Vdc 3 2 Dop  0 0 Dop  0 Motoring 1 region Regeneration -1 region -2 -3 0.2 0.4 0.6 0.8 1 1.2 Modulation index M (Turns ratio n=2) (b) The current-fed trans-ZSI/-qZSI. Figure 4.13. Dc-ac voltage gain and operation region in the current-fed ZSI/qZSI and trans-ZSI/qZSI. 111 1: n I dc Ldc  Vdc  L1 D L2 Toacload orsource C1 Figure 4.14. Current-fed trans-Z-source inverter. 4.4. Comparison with the Z-Source and Quasi Z-Source Inverters Table 4.1 compares the governing equations for the voltage-fed trans-Z-source, trans-quasiZ-source, Z-source and quasi Z-source inverters. Table 4.2 compares the class of current-fed Zsource inverters. Two tables also provide a guideline for determining the ratings of the components such as the active switch, the diode, the capacitor and the transformer. The voltage and current directions in the Z-source and quasi Z-source inverters follow the same definition as that shown in the equivalent circuits of the trans-Z-source and trans-quasi-Z-source inverters. All the magnetizing current is reflected to the primary winding. While SD is the shoot-through switching function in the voltage-fed inverters, it is the open-circuit switching function in the current-fed inverters due to the duality between the voltage-fed and current-fed inverters. SD is defined as 1 when the inverter is in the shoot-through zero states and 0 when it is in the nonshoot-through states in the voltage-fed inverters. It is 1 for the open zero states and 0 for the nonopen states in the current-fed inverters. In Table 4.1, when the turns-ratio is over 1, the voltage-fed trans-Z-source and trans-quasi112 Z-source inverters, with reduced component count, have higher dc link voltage gain than the voltage-fed Z-source and quasi Z-source inverters. When the turns-ratio is 1, some of the governing equations can be unified, including the relation for the magnetizing current of the two coupled windings. Comparing the voltage-fed trans-Z-source and trans-quasi-Z-source inverters, less capacitance is needed for an input filter capacitor in parallel with the low voltage dc source in the voltage-fed trans-Z-source inverter, whereas the capacitor voltage stress is reduced in the voltage-fed trans-quasi-Z-source inverter. In Table 4.2, when the turns-ratio is over 1, besides a wider motoring operating range, the current-fed trans-Z-source and trans-quasi-Z-source inverters feature more voltage buck ratio. This feature can also be found in Figure 4.13. When the turns-ratio is 1, some of the governing equations in the Z-source inverter and quasi Z-source inverter are consistent with those for the trans-Z-source inverter and trans-quasi-Z-source inverter, respectively. Similarly as the Z-source and quasi Z-source inverters and as shown in Figure 4.15, the trans-quasi-Z-source inverter has less magnetizing current than the trans-Z-source inverter in the motoring operation (Im>0) when the open zero duty ratio Dop is: 0