ABSTRACT. WANG, JIA. Battery Management Issues for Battery Pack in Electric Vehicle. (Under the direction of Dr. Alex Q. Huang and Dr. Iqbal Husain.

Transcription

1 ABSTRACT WANG, JIA. Battery Management Issues for Battery Pack in Electric Vehicle. (Under the direction of Dr. Alex Q. Huang and Dr. Iqbal Husain.) The portable storage unit of a battery is the preferred option for the energy storage system in the electrical vehicles (EV) industry. Battery management system (BMS) is necessary and important to help maintain the long series connected string of battery cells work within its safe operating area (SOA), maximizing the battery s performance and prolonging the battery cells lifetime. A low power battery monitoring and management system has been designed and developed in this thesis to help collect feedback information for balancing and protection. The system can monitor 4 to 8 battery cells connected in series, which has 8 channels for voltage signals, 6 channels for temperature signals and 1 channel for current signal. The voltage resolution can be as great as 1.3mV, current resolution can be up to 0.2A and the temperature can be up to 1 0 C. The sample rate can be customized to a speed of 500ksps. The system also has 1 isolated and non-isolated UART and 1 I2C communication channels, which can communicate with the second controller to do the online parameter and state of charge (SOC) co-estimation. Simple coulomb counting method can also be employed to estimate the SOC. It has a power management system that can help save up to 39% of energy in sleeping mode, and can also detect an unsafe operating state and send out the protection signal. A high current passive balancing circuit has been developed and integrated with the monitoring system. Different types of imbalance factors were analyzed such as the SOC with the terminal voltage difference among cells. A thermal analysis with different resistor size

2 and package is presented and the long side termination thick film chip resistor is chosen as the dissipate resistor for this passive balancing method with excellent thermal performance results. With the thermal mitigation and passive balancing current optimized, the passive balancing current can be as much as 1A with acceptable thermal situation for the developed BMS. A novel active cell balancing topology has also been designed and simulated in this thesis for rapid balancing. It is designed based on single active bridge topology. It has rapid balancing speed with the balancing current approximating 18A. This active cell balancing converter is designed modularly for 6 battery cells connected in series. Hysteresis current control (HCM) that has good dynamic response and also insensitive to parameter changes is easier for the implementation. It is employed here to help make the balancing current controllable. To help overcome the drawback of uncontrollable frequency of HCC, a limited frequency HCC controller is designed and implemented to enhance the novel converter. The simulation results of the new converter with limited frequency HCC has been proposed and analyzed. The comparison between the novel converter and the LTC flyback converter has been presented to show the lower parts count, easier control and relatively simple implementation of this novel active balancing converter.

3 Copyright 2013 by Jia Wang All Rights Reserved

4 Battery Management Issues for Battery Pack in Electric Vehicle by Jia Wang A thesis submitted to the Graduate Faculty of North Carolina State University in partial fulfillment of the requirements for the degree of Master of Science Electrical Engineering Raleigh, North Carolina 2014 APPROVED BY: Dr. Alex Q. Huang Committee Co-Chair Dr. Iqbal Husain Committee Co-Chair Dr. Mo-Yuen Chow

5 ii BIOGRAPHY Jia Wang was born on February 6, 1989 in Wuxi, Jiangsu, China. He spent his early childhood in Wuxi, his hometown. He then spent about four year finishing his Bachelor of Engineering degree in Microelectronics from Nankai University. In 2011, he joined North Carolina State University to pursue the degree of Master of Science in Electrical Engineering under the guidance of Dr. Alex Q. Huang and Dr. Iqbal Husain.

6 iii ACKNOWLEDGMENTS I would like to acknowledge and deepest thank my co-advisor, Dr. Alex Q. Huang and Dr. Iqbal Husain. They have extensive knowledge in all areas of electrical engineering and flawless judgment, which helped me find direction and inspiration for my research in the past two years. I thank them for the extra hours spent in discussions and most importantly for teaching me how to think as an engineer and what I need to consider in a project. Above all, I thank them for their support and for providing an ideal environment for learning and conducting research throughout the time I spent at NCSU. I am grateful to my committee members, Dr. Mo-Yuen Chow for his time, valuable suggestions and help. I appreciate the assistance from the staff members of FREEDM Systems Center, especially Dr.Wensong Yu. He gave me a great deal of excellent advices on my research. Many thank to the visiting scholar Dr. Il-Song Kim, who spent much time with me discussing and guiding my research topic. It was a great pleasure to work with the all of the students here at FREEDM center. I am grateful to the postdoctoral, Dr. Xijun Ni and Dr. Ruiyang Yu who helped me siginificantly. I also learned much from my friends Mehnaz Kahan, Rajib Makail, Yang Lei, Fei Xue, Rui Wang, Changjian Hu who taught me many things and gave me much of their time and many helpful suggestions for my thesis. My heartfelt thanks my parents Huaping Wang and Jufeng Ying, whose unconditional love has supported me throughout my life.

7 iv TABLE OF CONTENTS LIST OF TABLES... ix LIST OF FIGURES... x Chapter 1 Introduction Perspective Need for a battery management system (BMS) Review of Current Battery Technology Necessity and Functionalities of BMS for EV Applications Battery Management System System Diagram of an EV Battery Pack System Diagram of EV Battery Management System Summary Scope of the Thesis Glossary Chapter 2 Evaluation of Current Battery Management System Basic Battery Management System Review of Current Battery Balancing Method Passive Balancing Method... 20

8 v a) Fixed Shunt Resistor b) Controlled Shunting Resistor Active Cell Balancing Method a) Capacitor Shunting Balancing Method b) Inductor and Transformer based Balancing Methods c) Energy Converter based Balancing Methods Summary Chapter 3 Characteristics of a Low Power Battery Monitoring System Motivation-Why We Need the Monitoring System Charging and Discharging Test Bed for Battery Packs High Precision/ Resolution Measurement Voltage Temperature: Current Cell Protection System State of Charge and State of Health Calculation Power Save Management Communication Interface... 61

9 vi 3.8 Package and Integration The Experimental Verification Summary Chapter 4 Design and Evaluation of a Passive Balancing Method Introduction Different Causes of Imbalance Affection of Imbalance to Cell Terminal Voltage a) State of Charge Unbalance b) Capacity Unbalance c) Internal Impedance Unbalance Harms Performance to Battery Pack from Imbalance a) Cell Degradation b) Early Charge and Discharge Termination c) Safety Concern of Overcharge and Over discharge Summary of the Imbalance situation Design of the Passive Balancing Topology Switch Device Selection High Current Bypass Dissipate Resistor Selection... 82

10 vii Passive Balancing Circuit Controller Design Thermal Problem Mitigation High Current Bypass Resistor Package Selection Bypass Current Optimization Experimental Verification Summary of the Passive Balancing Method Chapter 5 Design and Evaluation a Novel Active Cell Balancing Converter The Power Stage Conventional Single Active Bridge DC-DC Converter A Novel Active Balancing DC-DC Converter Topology Design of the Novel Active Cell Balancing Converter MOSFET and Diode Selections Energy Storage Inductance Value Design Transformer Design Hysteresis Current Controller Design Conventional DCM Controller Design and Analysis Design of Hysteresis Current Control Hysteresis Current Control with Frequency limit Controller Design

11 viii 5.4 DC-DC Converter Performance Simulation Result Hysteresis Current Control Simulation Result and Analysis Frequency Limit Control Simulation Result and Analysis The Efficiency and Loss Distribution Analysis Comparison with the LTC Synchronous Flyback Converter Summary of the Active Balancing Method Chapter 6 Summary and Future Work Research Contributions Future Work REFERENCES

12 ix LIST OF TABLES Table 1.1 Rechargeable battery chemistries characteristics....2 Table 1.2 WB-LYP40AHA battery parameter datasheet...14 Table 3.1 Alert signal when the battery is working outside SOA Table 4.1 MOSFET absolute maximum ratings Table 4.2 Power dissipation of CHP Thick Film chip resistor Table 4.3 Power dissipation of long side termination thick film chip resistors Table 5.1 Calculated maximum value of the MOSFET and diode Table 5.2 Maximum rate of the MOSFET and diode Table 5.3 Balancing topology components numbers comparison Table 5.4 Comparison between the proposed topology and LTC Flyback converter

13 x LIST OF FIGURES Figure Comparison of energy density among the rechargeable batteries Figure Loss of capacity at high discharge rates at Peukert numbers Figure An EV battery module Figure Thundersky 40Ah battery pack (5 cells) Figure System diagram of EV battery management system Figure Shunting resistor balancing method Figure Capacitor shunting balancing method Figure Modularized switched capacitor (MSC) balancing Figure Inductor based balancing method Figure Single-Winding Transformer based balancing method Figure Multi-Windings Transformer based balancing method Figure Cuk converter based balancing method Figure Buck-boost converter based balancing method Figure Flyback converter based balancing method Figure Working status of Flyback converter Figure Ramp converter based balancing method Figure Full-bridge converter based balancing method Figure Quasi-Resonant converter based balancing method Figure System diagram of battery monitoring system Figure Test bed setup for charge and discharge battery pack

14 xi Figure Discrete and differential multiplexed cell voltage measurement methods Figure High precision reference voltage schematic Figure Temperature sensing: one per cell; one per pack Figure Temperature sensor drive circuit schematic Figure Current sensor and current sensing circuit topology Figure Alert signal drive circuit Figure Online parameter and SOC co-estimation Figure Power distribution on the board Figure Switch and its drive circuit for OpAmp Figure Non-isolated communication interface Figure Isolated communication interface Figure I2C communication interface Figure Specific package design for the battery module connection Figure Passive balancing integration on the monitoring board Figure A discharging voltage curve Figure A discharging temperature and current curve Figure A charging voltage curve Figure A charging temperature and current curve Figure Voltage difference relationship with 1% SOC imbalance Figure Individual cell voltage vs. capacity deficiency from nominal Figure Passive balancing circuit topology schematic

15 xii Figure Passive balancing controller diagram Figure Balancing current vs. balance time for 1% SOC imbalance Figure Relationship between the resistor temperature and balance current Figure Passive balancing procedure Figure Conventional SAB converter Figure Operating mode of the SAB Figure Proposed active cell balancing topology Figure Proposed active cell balancing circuit for one cell Figure Voltage difference vs. balance peak current for DCM Figure Hysteresis current control diagram Figure Hysteresis current control mode Figure Average current comparison between the DCM and HCC Figure Switch frequency varies with the voltage difference Figure Hysteresis current control with frequency limit controller Figure Balance current change when the voltage difference change (HCC) Figure Zoom in of the convertering moment Figure Balance current under 0.1V voltage imbalance of HCC Figure Balance current under 0.1V,0.11V voltage imbalance of HCC Figure Primary side current under 0.07V voltage imbalance Figure Primary side current under 0.05V voltage imbalance Figure System level simulation by balancing the super capacitors

16 Figure Efficiency relation with the voltage difference xiii

17 1 Chapter 1 Introduction 1.1 Perspective The energy problem is a major issue that is of grave concern for many countries around the globe. Given the limited availability of traditional fossil fuel based energy sources together with the environmental pollution caused by these types of energy, there is a large demand for mankind to find alternative energy sources, especially some form of renewable energy system. Battery system that helps storage can help sustain the supply of energy from renewable sources and can replace the fossil fuel based energy supply together with the renewable energy. Additionally, with the technological developments in the industry, many applications select the battery as the primary option for the energy supply. Worldwide, the sales for rechargeable batteries approached around US $36 billion in 2008, and this is expected to grow to US $51 billion by US demand for rechargeable batteries has increased by 2.5% annually to US $11 billion in Typically in the energy storage system, batteries are widely used, such as electric vehicles, uninterruptible power supplies, and distributed energy storage system. The focus of this work is to develop a battery management system to help maintain the safety of these batteries, improve the overall efficiency and extend the life of batteries for the energy storage system in electrical vehicles.

18 2 1.2 Need for a battery management system (BMS) Review of Current Battery Technology Several different types of rechargeable batteries are available for various applications. Table 1.1 Rechargeable battery chemistries characteristics. [1] Parameter unit Leadacid Nicd NiMH Li-Ion Li- Polymer LiFe Po4 Ave cell voltage Volts Internal Low Very Moderate High High Hig resistance Low h Self-discharge %mon 2%-4% 15%-5% 20%-25% 6%-10% 18%-20% Cycle life cycles Overcharge High Med Low Very Low tolerance Energy by volume Whr/lite r Energy by Whr/kg weight

19 3 Table1.1 shows the commonly used rechargeable battery types: lead-acid, nickelcadmium (NiCd), nickel metal hydride (NiMH), lithium-ion (Li-ion), lithium-polymer, and LiFePO4. The choice of a rechargeable battery technology is limited by size, weight, cycle life, energy density, power density, operating temperature range and cost. Lead-Acid Batteries now are the most widely used batteries in the automotive industry. They first appeared commercially in the early 1859s. The voltage of the battery can remain quite constant until most of its capacity is exhausted and then it drops off sharply. The battery capacity will remain relatively stable for most of its useful life under normal operating condition, but will suffer some degradation because of the age and use. NiCd batteries use nickel oxy-hydroxide as the material for the positive electrode and cadmium as the active material of the negative electrode. The discharge voltage characteristics of a NiCd battery are similar to those of the Lead-Acid battery, typically remaining constant until most of its capacity is exhausted and then dropping sharply. Nickel based batteries are easily charged by applying a controlled current. There is a memory effect on NiCd battery. When NiCd batteries are used after numerous partial discharge cycles and overcharging at a higher temperature, the cell voltage decreases below 1.05 V before the cell reaches 80% of the discharge. This is called the voltage depression effect. NiMH cells, introduced around 1989, have shown a nearly 170% increase in energy density compared to NiCds The cadmium negative electrode is replaced by an alloy in these cells. This chemistry is used extensively in hybrid electric vehicles. It was also the early

20 4 choice for electric vehicles (EVs) until about In addition, NiMHs do not exhibit the notorious memory effect. Li-ion batteries with a 3.6V+ operating voltage, have three times the terminal voltage of nickel-based chemistries. The Li-ion battery packs are of much higher energy density compared to other battery types. With about 25% of the mass of lead-acid cells, Li-ion cell battery is the choice for electrical vehicles today. The Li-ion batteries have high energy density of about twice that of nickel-based chemistries. The charging and discharging efficiency is always high and it can accept a high charge rate and the deep discharge can be as much as 80%. Li-Ion battery has a low self-discharge rate, allowing for a shelf life of 5 to 10 years, which will thus increase the efficiency of the battery usage. The first noticeable difference between lithium-based chemistries and nickel-based chemistries is the higher internal impedance of the lithium-based batteries. More significantly, the prices of lithium batteries are much higher than other chemistries. The costs are expected to decrease as development continues. Since Li-ion batteries use the high energy materials, safety issues must be adequately addressed. Hence, these battery cells require an effective battery management system to keep the batteries working in a safe mode, to maximize the use of the battery energy and to prolong the life of the batteries.

21 5 Figure Comparison of energy density among the rechargeable batteries. [18] Necessity and Functionalities of BMS for EV Applications Energy storage systems, usually batteries, are essential for electric drive vehicles, such as hybrid electric vehicles (HEVs), plug-in hybrid electric vehicles (PHEVs), and all-electric vehicles (EVs). Transitioning to HEVs and PEVs could help reduce U.S. foreign oil dependence by 30-60% and greenhouse gas emissions by 30-45% [28]. All-electric vehicles (EVs) use a battery to store the electrical energy that powers the motor. They are categorized as zero-emission vehicles by U.S. Environmental Protection Agency because they produce no direct exhaust or emissions. EVs are sometimes referred to as battery electric vehicles (BEVs). EV batteries are charged by plugging the vehicle into an electric power source. Because EVs use no other fuel, widespread use of these vehicles could dramatically reduce petroleum consumption. Lithium-ion batteries, because of their high energy per unit mass relative to other electrical energy storage systems, a high power-to-weight ratio, high energy

22 6 efficiency, good high-temperature performance, and low self-discharge, are used by most of today s plug-in hybrid electric vehicles and all-electric vehicles. Research and development is ongoing to reduce cost and extend their useful life cycle. Most components of lithium-ion batteries can be recycled. To overcome the problem of long distance travel, where fast charging is neither available nor convenient, Tesla Motors s battery swapping idea is proposed to be a solution. Tesla Motors has developed battery swap infrastructure where drivers can go to a station and exchange a depleted battery with a fully charged one. The time required for the exchange is less than it takes to fuel a conventional vehicle. A specific swappable battery pack must be available for use at the battery swap stations. Battery performance for electric vehicle is the key issue for these vehicles' economic, social, and environmental sustainability. Therefore improving battery performance is an important issue for the electric vehicle. The battery packs used in EVs are a large number of battery cells connected in series or parallel in a long string to provide sufficient energy to meet the demand of the EVs working in various operating conditions. However, the long battery string will cause many problems if it doesn t work in its proper state. Many safety problems have been reported and there has been some dangerous accidents because the battery string in the EVs had not been managed properly, which result in the battery to be damaged and occasionally to catch fire. Safe Operating Area (SOA) [20]: The Li-Ion batteries perform excellently, but as mentioned above, it may result in dangerous situations if operated outside the Safe Operating

23 7 Area. The SOA of Li-Ion battery is bound by its current, temperature and voltage. The operating characteristics of the Li-ion battery cells are: The battery cell will quickly exhibit damage and may burst into flames if charged above a certain voltage. Most Li-Ion battery will be damaged if discharged below a certain voltage. The lifetime of the Li-Ion battery will be drastically reduced if discharged outside a certain temperature range, or charged outside a temperature range. The Li-Ion battery will experience a thermal runaway and ignite if operation exceeds a safe temperature. Even those cells that are not prone to thermal runaway may contain organic electrolytes, which will fuel flames Li-Ion battery lifetime will be reduced when charged too quickly or discharged at exceedingly high currents. Li-Ion battery may be damaged when charged at a high pulse current for more than a few seconds. As we see above, the batteries can get out of the SOA which will cause serious problems such as thermal runaway and which will at least shorten the life of the battery. The SOA limits can vary in different applications and is determined by the battery chemistry of each cell. For example the LiC 2 battery without any other protection mechanisms will go into thermal runaway at a very low temperature, while the LiFePO 4 s are immune to thermal runaway. In a normal city car the SOA are defined strictly to help keep the battery working

24 8 in a safe state and extend the battery life. However in the sports vehicles the battery is used near its limit. The SOA is less severe and this can also reduce the performance of the battery. Therefore a battery management system which can monitor the battery state and make sure it is working under SOA is really necessary and useful. Peukert's Law [19]: Batteries will have a reduction in usable capacity when discharged at higher rates, using a simple equation Cp k I t C p is the capacity of the battery; I is the discharge current; k is the Peukert constant; t is time. Peukert's law describes an exponential relationship between the discharge current (normalized to some base rated current) and delivered capacity. Figure1.2.2 shows the loss of capacity at high discharge rates. Rapid discharging or charging events reduce the lifetime and performance of batteries.

25 9 Figure Loss of capacity at high discharge rates at Peukert numbers. [28] Maintaining the battery current under the proper limit is necessary for the maximum use of the battery. Here the BMS system is necessary to help detect the current and thus maximize the use of batteries. Moreover, this capacity loss will also cause SOC imbalance where battery cell balancing is needed from BMS. Capacity Fade [21]: The capacity of a lithium ion battery decreases during cycling. This capacity loss or fade occurs due to several different mechanisms resulting from or associated with unwanted side reactions that occur in these batteries. These reactions occur during overcharge or over discharge and cause electrolyte decomposition, passive film formation, active material dissolution, and other phenomena.

26 10 The battery string in EVs has large numbers of batteries connected in series to provide sufficient energy. The capacity fade and production restriction may cause the difference among these battery cells. The difference will expand gradually with the usage and age. Take capacity for example, because the charging and discharging current is the same and those high capacity cells will under shallow charge and shallow discharge, tend to slow capacity fading and have a long life, while those with small capacity cells under deep charge and discharge, sometimes overcharge and over discharge, which will cause the capacity to be degraded. Then these differences will become larger and larger with the increasing charging and discharging cycles. The overcharge and over discharge will not only affect battery life, but also may cause dangerous situation, such as explosion. Therefore, it is necessary to employ effective balancing methods to help balance the difference among the cells and prolong the life of the entire battery pack. To help eliminate the unbalance situation mentioned above, which will cause to shorten battery cells cycle lifetime and potentially cause dangerous accidents because of overcharge and over discharge, a battery management system which can help balance the battery string is needed. Above all, with the inherent battery cell problems and those caused by long battery pack, a BMS is essential to help maintain the battery operation in a safe state and to extend the life cycle of batteries.

27 Battery Management System The battery management system is part of a battery pack which monitors the cell variables, controls the charge and discharge rates, estimates the battery state, protects against overvoltage, maximizes the battery s performance and communicates with external devices System Diagram of an EV Battery Pack Since the battery packs in EVs are a collection of battery cells connected in series or parallel, different terms are usedto describe the package of the battery pack in an EV, which help understand the various components of the battery pack. Cell: The basic components of a battery, which is 3V to 4V for a Li-Ion battery Block: Several cells connected in parallel to provide 3V to 4V, whichever has more capacity. In our project, there are 11 cells connected in parallel. Module: A collection of cells or the blocks connected in series, which can provide a higher voltage. Here we have 8 blocks connected in series for a single module. This module is also the monitoring board which we design. Pack: Several modules connected in series to form a collection of batteries in series or parallel to provide energy and power for an EV.

28 12 Figure An EV battery module This is the battery pack from an EV and is the board we designed for it. While the battery block has 11 cells in parallel requiring a great deal of time to charge and discharge, another battery pack is used for test purpose. 40Ah Thundersky battery cells are used here. These battery cells are used in a distributed energy storage system that also requires a BMS. Another reason for using the battery cells is that the complete datasheet is available.

29 13 Figure Thundersky 40Ah battery pack (5 cells). The parameters of this 40Ah battery are shown below. The capacity is 40Ah. Normally, the charge and discharge is 0.5C. The overcharge voltage for this battery cell is 4.2V and the over discharge voltage is 2.75V. These two voltage levels are the parameters that require greater attention.

30 14 MODEL: WB-LYP40AHA Table 1.2 WB-LYP40AHA battery parameter datasheet NOMINAL CAPACITY 40Ah OPERATION VOLTAGE CHARGE 4.2V DISCHAGE 2.75V MAX CHARGE CURRENT <3CA MAX DISCHARGE CURRENT CONSTANT CURRENT <3CA IMPULSE CURRENT <20CA STANDARD CHARGE/DISCHARGE 0.5C CURRENT CYCLE LIFE 80DOD% >3000TIMES 70DOD% >5000TIMES OPERATING TEMPERATURE CHAGRE DISCHARGE SELF-DISCHARGE RATE <3%(MONTH)

31 System Diagram of EV Battery Management System The battery pack in EVs is formed by several series connected battery modules. The BMS is designed with modularly to help decrease the cost and facilitate implementation. Figure below shows the system diagram of the EV BMS has been designed. Figure System diagram of EV battery management system. The battery module is shown above in figure Each module has 8 battery blocks and the BMS can monitor the state of these 8 blocks, including the voltage, temperature and series charging or discharging current. The BMS is designed to maintain 4 to 8 series connected battery cells or blocks working in a safe state. Then each BMS module will

32 16 communicate with the second level controller to transfer data and control signal for future data analysis and control. 1.4 Summary As discussed above, a BMS is essential to help the battery string work under it safe operating conditions, to maximize the use of the battery capacity and to prolong the lifetime of the batteries. 1.5 Scope of the Thesis The thesis is organized as follows. In chapter 2, conventional BMS function is presented and the evaluation and comparison of the existing balancing method is shown. In chapter 3, the performance and characteristics of a low power battery monitoring system are developed and presented. In chapter 4, different types of imbalance factors are analyzed and a passive balancing circuit with thermal mitigation is implemented. In chapter 5, a novel active cell balancing converter with limited frequency hysteresis current control is proposed. The simulation result and analysis are presented. 1.6 Glossary EV BMS SOA Electrical Vehicles battery management system safety operating area

33 17 DCM HCC CC CV OCV OpAmp discontinuous control mode hysteresis current control constant current constant voltage open circuit voltage operational amplifier

34 18 Chapter 2 Evaluation of Current Battery Management System 2.1 Basic Battery Management System The basic BMS should provide the functions of monitoring the battery, protecting the battery cells, estimating the battery state, maximizing the battery s performance and communication with the external devices. Currently, the BMS available in the market can be divided into two classes: the simple analog BMS and the sophisticated digital BMS. As for the analog BMS, each cell has its own board installed. A supervisor IC powered by the battery cell will be used as the controller. It will drive a balance shunt resistor when the cell voltage exceeds the threshold voltage. However, an analog BMS is unaware of the cell voltage. It only knows there is one cell voltage too low, but cannot identify which cell and how low it is. The Clean Power Auto s MiniBMS offers an analog BMS which can be considered as the minimum requirement for a BMS. It can handle overvoltage, under voltage and voltage balance. In the case of the digital BMS, it can detect each cell s working state, for example, cell voltage. It can be categorized as centralized, modular, master-slave or distributed based on how it is installed. Monitors, meters, balance and protector are considered most important in long Li-Ion battery strings. The BMS we designed and developed in this thesis is a digital

35 19 BMS. Many battery manufacturers sell BMS with their cells. However, many of them are custom designed for their own battery cells. Some other BMS manufactures will provide BMS product commercially. But in some applications, we do not need so many functions and the package of the case sometimes is also not suitable for applications. Many of the digital BMSs use wire to connect the battery cells with the BMS. It gets worse when the cell numbers increase. Therefore, we have designed a novel BMS which has the functions we require for an EV test bed and can be placed on the battery pack without any wire connection. 2.2 Review of Current Battery Balancing Method Cell balancing is necessary for large numbers of cells connected in series, and is subject to a higher rate of failure than the single cell. With the age and usage of batteries, their state will change. The capacity, internal resistor or the open circuit voltage will be slightly different among the batteries. The position or the environment of the battery may also cause the balance problems. With these differences over a long time, battery pack capacity will decrease deeply. This will cause the total failure of the battery system. The balancing methods can be classified as the passive balancing and active balancing. The passive balancing method removes the charge from the fully charged cells or higher voltage cells through the dissipated resistors to match the charge in those cells that have lower charging current. Active cell balancing transfers the charge from the higher cells to lower cells or delivers more charge to those lower charge cells [23]. It has different

36 20 topologies according to the component it chooses. It can be simply separated as capacitor based, inductor/ transformer based and converter based. For the capacitor based, it can be classified by the switched capacitor, single switched capacitor, and double-tiered switched capacitor based on the switches it uses. In the case of the inductor and transformer based active balancing method, it can be divided as single/ multi inductor, single winding transformer and multiple windings transformer. For the converter based method, it can be divided as cuk converter, buck-boost converter, flyback converter, full bridge converter and Quasi-Resonant converter. Some of them may have some overlap, just like the flyback converter active balancing method may also be treated as the transformer based method Passive Balancing Method Passive balancing method is the most straightforward and simple method. Because it is simple, low cost and easy to implement, it is the most popular and widely used method in industry currently. The shunting resistor will dissipate charge or energy from those cells which have a higher charge to accomplish cell balancing. The passive balancing method is realized by bypassing the current of the highest cells and wait to until all cells to be the same. There are two different ways to shunt the resister: fixed shunt resistor and controlled shunting resistor. a) Fixed Shunt Resistor As shown in Figure 2.2.1, the resisters are fixed on all of the cells and the bypass current is continuous. It is determined only by the value of the cell voltage and the resistor value it

37 21 chose. The fixed shunt resistor can be used for Lead-acid and Nickel based batteries because they have little damage under overcharge conditions. The fixed shunt is simple and cheap but it results in continuous energy dissipation as heat from all cells. b) Controlled Shunting Resistor As shown in Figure2.2.1, the charge removed from the higher charge cell is controlled by the switches. This will help reduce energy dissipation. Figure Shunting resistor balancing method. [2] There are two control algorithms: Continuous mode: all switches are controlled by the same on and off signal.

38 22 Detecting mode: in this case there is one controller. When the imbalance conditions are detected, the controller decides which switch should be connected and the resistor will be shunted. This method is more efficient, simple and reliable and is used for this project. Passive balancing has a higher energy loss, a larger radiator and thus is not effective. Thermal management is required in most applications. It also takes a great deal of time to achieve equalization, since the current is kept fairly low. Several hours are required to correct even a moderate imbalance. Because it is dissipating the energy form the battery, it is not suitable for balancing voltage during discharge, which will shorten the battery life. It cannot be used for high power application and rapid charging. Nevertheless, passive balancing is still widely used because of its reliability and simplicity. The control method is also very simple. It is cheaper when it is used for low power application, and is also simple to implement with a small size. It is suitable for both HEV and EV Active Cell Balancing Method a) Capacitor Shunting Balancing Method The capacity shuttling method is to use the capacitor as an energy storage device to transfer the energy among the batteries. It has three different types: the basic switched capacitor, single switched capacitor and double-tiered capacitor topologies.

39 23 Basic switched capacitor: As shown in the figure2.2.2, it requires n-1 capacitors and 2n switches to balance n cells. Its control method is simple because it has only two states. The energy will transfer between the two adjacent batteries, from the higher to the lower. It does not require intelligent control and it can work in both charging and discharging operation. But it needs a relatively long equalization time if the charge difference is not between two adjacent batteries. Figure Capacitor shunting balancing method. [3]. Single switched capacitor: It uses only one capacitor but n+5 bidirectional switches as shown Figure The control strategy is simple and the energy is transferred between the highest and the lowest cells by turning on the corresponding switches. A more advanced

40 24 control algorithm can be implemented on this topology and the balancing speed can be improved significantly. Double-tiered capacitor topologies: This balancing method needs 2n-2 capacitor and 3n switches to balance n cells. This method uses two capacitors tiers for energy shuttling as shown Figure The advantage of double-tiered switched capacitor over the switched capacitor method is that it can reduce the balancing time one quarter time. In addition, all switched capacitor topologies, single switched capacitors and doubletiered switched capacitors can work in the recharging and discharging operation. Super capacitor has the advantage of rapid charge (<20min) and discharge, large power density, and long cycle life [4]. It can be used both in charging and discharging mode. The voltage stress is low and the control strategy is simple. There is no need for close loop control and can be used in both EV and HEV.

41 25 Figure Modularized switched capacitor (MSC) balancing. [4][5] However, the uncontrolled balancing current and high peak current will be a problem. Average balance current cannot be high. This will result in too much time for balancing. It is possible to balance only two adjacent cells when using a basic switched capacitor, which also requires a great deal of time for balancing. When cost is considered, a large numbers of switches is a problem. Intelligent control is needed for fast equalization. b) Inductor and Transformer based Balancing Methods. This method uses inductors or transformers to transfer energy from one cell or group of cells to another cell. This method can tolerate high balance current which will require a small

42 26 amount of balance time. However, since the frequency is quite high, there may be a need for some filters, the cost for which is prohibitive and efficiency is still a problem. Single/Multi switched Inductor Figure Inductor based balancing method. [6] The single switched inductor (SSI): it needs one inductor and 2n switched and diode for balancing n cells. Then it will transfer the energy from the cell which has the highest energy to the lowest.

43 27 The multi switch inductor (MSI): it needs n-1 inductor, 2n-2 switches for balancing n cells. It transfers the energy between the two neighboring cells. Both can transfer a high energy and have the advantage of fast equalization. But it takes a long time for the MSI to transfer energy from the top cell to the bottom one. SSI is suited to solve this problem. The advantage of SSI topology is fast equalization speed and good efficiency. The complex control is required and MSI needs accurate voltage sensing. The switched current stress and voltage stress on the diodes is high. Filtering capacitors are needed for high switching frequency Single-Winding Transformer: Figure Single-Winding Transformer based balancing method. [7] As shown in figure 2.2.5, it needs n+6 switches, 1 diodes and 1 transformer.

44 28 This topology transfers the energy between the total pack and a single cell. This is similar to a converter. It can transfer the energy form a high energy cell to the whole battery pack through the transformer and it also can transfer the energy from the battery pack to the weakest cell if needed. This method can achieve fast equalization speed with low magnetic losses. The disadvantage is the high complexity of control and the expensive implementation for the bidirectional switches. Also, to add one or more cells, the core must be changed Multi-Windings Transformer: The multi-windings transformer: used here has one primary winding and multiple secondary windings. The energy is transferred from the battery pack to each cells and the energy will be distributed based on the voltage of each cell. The higher voltage cell will receive the lower power. It has two circuit topologies: [8] flyback and forward configurations. This can also be treated as the converter based active balancing method. The flyback structure: needs 1 transformer which has 1 magnetic core with 1 primary winding and n secondary windings, n diodes and 1 switch. When the voltage difference is detected, the switch at the primary side is switched on. The energy is stored in the transformer and when the switch is off, the energy is transferred to the batteries at the secondary of the transformer. Most of the current goes to the cell with the lowest voltage based on the voltage second balance of the inductor.

45 29 The forward structure: needs 1 transformer which has 1 magnetic core with 1 primary winding and n secondary windings, 1 diode and n switches. Figure Multi-Windings Transformer based balancing method. [8][9]

46 30 When the voltage difference is detected, the switch connected to the cell with highest voltage is turned on, and energy is stored in the transformer. When the switches are turned off the energy will goes from the transformer to the whole pack. The circuit is complex and expensive. There is also a saturation problem with the transformer. The advantage of Multi-Windings Transformer based balancing method is the rapidly balancing speed and no close loop control needed it can be used for EV and HEV applications The disadvantage is its high complexity control and expensive implementation. It is limited by the parasitic parameter of the transformer. To add one or more cells the core must be changed. c) Energy Converter based Balancing Methods. By using energy converters, we can have full control over the entire balancing process. It can be divided by the converter type: Cuk, Buck or boost. Flyback, Ramp, Full-bridge and Qusi-Resonant converter.

47 31 Ćuk converter: Figure Cuk converter based balancing method. [10][11] The cuk converter needs 2n+2 switches, 2n inductor and n capacitors for balancing n battery cells connect in series. It transfers energy between adjacent cells just as figure2.27 shows. The greatest disadvantage to this converter is very high cost and the great deal of time required for balancing if the battery string is long. The advantage is the efficient equalization system that it is suitable for both EV and HEV. However, it needs complex control and accurate voltage sensing. The equalization speed is slow.

48 32 Buck or/and Boost converter: Figure Buck-boost converter based balancing method. [12][13] Buck, Boost or the buck-boost converters are widely used for balancing part. As the converter shows in the above figure, it needs 2n+5 bidirectional switches and n inductor for balancing n cells. It can transfer the energy from an external source or the whole battery pack to the weakest cell or remove the energy from a single cell having the most energy to the whole pack. When there is one cell lower than the others, the controller will turn on the corresponding switch and the external source will transfer energy to this cell. When there is one cell having more energy than the others, the controller will turn on the corresponding switch and the energy will transfer from this cell to the external source.

49 33 This method needs intelligent control for the battery selection and the voltage sensor is mandatory. The cost of this method is quite expensive because of the large use of switches, which is also the main concern of the efficiency problem. It is highly complex when the battery string is long. The advantage of this topology is the good equalization speed and the ability to transfer energy bidirectional. It is easy to use for modular design. But it has a high cost due to the large amount of switches and it needs intelligent control. Flyback converter: Flyback converter is an isolation structure and it can transfer energy in bidirectional. It is widely used in industry. The LTC has launched a new controller IC, LTC3300, for this kind of transformer based bidirectional active balancing of battery packs, as shown in figure2.2.9

50 34 Figure Flyback converter based balancing method. [22] Each LTC3300 can balance up to six series connected batteries. Multiple LTC3300 devices can be connected in series for long strings. Each cell will be connected to the whole battery pack with the transformer and switches as a flyback converter. Each cell here can be operated individually, when the cell is weaker, it can get energy from the whole battery pack by operating in charge mode. Likewise, when the cell is stronger, it can operate in discharge mode to transfer energy to the whole battery pack.

51 35 Figure Working status of Flyback converter. [22] During the discharge mode, the primary NMOS is turned on first and the energy is stored in the transformer, when the current reaches 2A, which is sensed by the high precision resistor, the current signal ramps up to 50mV, the energy begins to transfer to the secondary side, the NMOS on the secondary side is turned on. It remains on until the current in the secondary side goes to 0. The cycle repeats until the LTC3300 gives a signal to stop or there is some fault such as overcharge or over discharge. During the charge mode, the NMOS on the secondary side is turned on first and remains on until the current reaches 2A. Then the primary NMOS is turned on, the energy goes from the whole pack to the single cell. The average current is determined by the value of the current sensing resistor, the turn ratio of the transformer, the number of the cells in the secondary side and the transfer efficiency.

52 36 I I discharge charg e 50mV S 2R S T sensepri 50mV ST 2R S T sensesce S is the number of the cells connected in the secondary side is the transfer efficiency T is the turn ratio of the transformer. The advantage of this topology is its suitability for modular design. It integrates the gate drive and the serial interface with built in protection. The disadvantage is the need for voltage sensing and its high cost because of the large amount of component it uses. Ramp converter: Ramp converter (RC) cell balancing topology is shown on the figure It needs 1 transformer which has the same primary winding and each pair of cells shares one secondary winding, 2n diodes and capacitors.

53 37 Figure Ramp converter based balancing method. [14] The operation principle is on the first half the current goes to the lowest cells of the odd number and on the other half, the current goes to the even cells which have the lowest voltage. The advantage is the fewer switches and a simple transformer is needed and soft switching can be easily implemented. The disadvantage is its complex control and low balancing speed.

54 38 Full-bridge converter: Figure Full-bridge converter based balancing method. [15] The full-bridge PWM energy converter needs 4n+4 switches and n capacitor, which can be treated as the fully controlled converter in the balancing method. It can be used as the AC/DC or DC/DC. The energy can be distributed by the control of the switches by the duty ratio or the phase shift to make the lower battery cell have the higher balancing current. The advantage is it can be fully controlled and the power rating can be high. In this way it can achieve quick equalization speed. However, the disadvantage is that it needs intelligent and complex control, and it has a relatively high cost.

55 39 Quasi-Resonant converter: Figure Quasi-Resonant converter based balancing method. [16] The quasi-resonant converter is shown in Figure , it needs2n switches, 2n inductors and n capacitors, it can be either zero-current quasi-resonant (ZCQR) or zerovoltage quasi-resonant (ZVQR) converters. It does not need to use controller to generate the PWM and it can drive the switch with only the resonance circuit. The energy is transferred between the battery cells by the resonant circuit alone. Lr and Cr are the resonant tank to achieve the zero current switching function for the converter. The advantage of the Quasi-Resonant converters is they can reduce the switching losses and then increase the balancing system efficiency.

56 40 The disadvantage is that resonant converters have a very complex control and difficult implementation. High cost is also a concern of that topology because each adjacent cell needs one converter to transfer energy. 2.3 Summary The basic BMS concept with monitor, meter, protection and estimation is described. A number of passive and active balancing topologies are compared. The advantages and disadvantages of these balancing methods are analyzed. This information can serve as a guide for the design of a novel monitoring system with passive balancing and a new active cell balancing converter.

57 41 Chapter 3 Characteristics of a Low Power Battery Monitoring System This chapter will show a novel battery management system which has the function of monitoring, protection, preparation for the estimation of SOC and SOH, power save management and communication interfaces. Moreover, a test bed for the CC or CV charging and discharging series connected battery pack is also shown here which allows us to do experiments on battery pack rather than only single batteries. 3.1 Motivation Why We Need the Monitoring System The figure shows the system diagram of our monitoring system. Our monitoring board is designed modularly for the battery module. Each monitoring board can monitor 4 to 8 battery blocks connected in series. The battery pack is composed of battery modules. The modules are shown in figure When these modules are connected in series, the battery pack is formed. The state of the battery cells will be collected by the monitoring board modularly and then transferred it to the second level controller for future use. Each monitoring board has communication interfaces which can be used with wire or wireless communication

58 42 Figure System diagram of battery monitoring system. The monitoring system is really necessary and helpful for battery packs experiments and tests, especially for protection from and prevention of some dangerous situations. Most equipment on the market for battery testing is limited to the number of cells. They can only test individual cells, not the battery packs. In additional, the price is very expensive for this kind of equipment. The battery pack, unlike the battery cells, will cause many different issues when we use them. Because many cells are connected in series, we need to know each cells state to make sure it is working inside its SOA. As for the battery packs in EVs, the terminal voltage of the whole pack is very high, which is one reason why there is few equipments for

59 43 testing of these battery packs. Our monitoring board can collect the voltage, current and temperature of each cell, which can be used as the feedback information for protection and can indicate if the battery cells are under SOA. The modular design can help eliminate the problem of high terminal voltage for the pack. Another reason for the necessity of this monitoring system is for the test of novel SOC and SOH algorithm. Our monitoring board has its own controller and it can do some simple SOC estimations. It can also be connected to the second level controller to perform sophisticated SOC estimation methods, similar to the online parameter and SOC estimation developed in the FREEDM center. As for the battery pack, it has balancing problems which does not occur in single cell test and use. Concerning the balancing method, the state of the battery cells is necessary and important. High precision measurement is important for balancing, especially for the passive balancing method. Our monitoring system can provide 1.3mV voltage resolution to implement balancing method, which is higher accuracy better than most of the products on the market. What's more, our monitoring board is integrated with the passive balancing circuit. Our monitoring system is designed in a specific modular package for the battery modules, which can eliminate many wire connections. Because we need to know each cells state, we need to test each cells voltage and temperature. Most products in the market need a wire connection to each cell electrode. Our monitoring system is designed without wire connection to the electrode of each cell. Consider the battery pack, we can save 16 wires

60 44 connection for each module. If we use 8 modules to form the battery pack. We can save 128 wires connection. This is significant number of wires which will make the system look mess. 3.2 Charging and Discharging Test Bed for Battery Packs This is the basic diagram of the test bed for the battery pack experiment. It can charge the battery pack in two modes: constant current or constant voltage. Also it can discharge the battery pack in two modes: constant current or constant voltage. The charge current can be as large as you want based on your choice of power supply. Here I choose the LAMBDA ZUP60-14 as the charge source and the current can be as large as 14A. Figure Test bed setup for charge and discharge battery pack.

61 45 LAMBDA ZUP60-14, which is the Zero up programmable power supplies, has the feature below: Constant Voltage / Constant Current Last Setting Memory Digital Meters Built-in RS232 & RS485 Interface w/ GPIB optional Embedded Microprocessor Controller This charger can provide up to 60V voltage and 14A current. This can be used to charge up to two battery modules. Here we choose 10A (0.25C) to charge the battery. The electric load is Chroma6312, whose feature is shown below: Max Power: 600W Wide range 0~500V operating voltage Parallel load modules up to 1200W for high current and power applications Synchronization with multiple loads Flexible CC, CR, CP and CV operation modes Dynamic loading with speeds up to 20kHz Fast response of 0.32mA/µs ~ 10A/µs slew rate Minimum input resistance allows the load to sink high current at low voltages Real time power supply load transient response simulation and output measurements

62 46 The switch UL 1077 which controls the on and off of the charger and discharge can be controlled manually and also has the function of protection. The over current control is automatic: Complete range of UL 1077 listed DIN rail mounted miniature circuit breakers up to 63 ampere current rating Standard ratings of 10 ka IC at 277 / 480 Vac Current limiting design provides fast short-circuit interruption that reduces the let-through energy, which can damage the circuit Suitable for supplementary protection Thermal-magnetic over current protection In addition to the protection from the automatic switches, we can turn it off using the alert signal from the LED or beep when overcharge and over discharge is detected. The detail will be shown in later section. 3.3 High Precision/ Resolution Measurement The measurement of voltage, current and temperature is the basic function of a sophisticated digital BMS. This novel low power battery monitoring system can monitor at most 8 channels of voltage, 6 channels of temperature and 1 channel of current for the whole series connected battery pack.

63 Voltage A distributed BMS may measure the voltage on each cell. The measuring board is always on each cell and self-powed, as figure 3.2 shows. However, we use a multiplexer and calculate the voltage difference between the two taps on each side of the battery. In this way, we can save many I/O ports on the microcontroller. Rate: The rate of the AD converter here can be as fast as 500ksps. The read rate requirement depends on the need different applications. In EV, it is rapid and should be approximately one sample per second. For backup power, one sample per 10 second period is acceptable. For EV, when it is driving, the sample should be fast to make sure the battery pack is working in a perfect state and can provide sufficient energy to support acceleration. In case of long time charging, for example all night charging, where the state of the battery does not change rapidly, the sample rate can be slow to help save energy. The sample rate of this new monitoring system can be determined by the customers.

64 48 Figure Discrete and differential multiplexed cell voltage measurement methods. Accuracy: The accuracy of this novel BMS is pretty high. It can be as much as 1.3mV. First the AD converter used here has 10bits and we also use a high precision reference voltage source as the reference voltage for the AD converter. Thus the accuracy can be calculated: V ref Vref 1.3mV 10 2 Below is the information about the AD converter and the high reference voltage source. 10-bit A/D Converter has the following key features: Successive Approximation (SAR) conversion Conversion speeds of up to 500 ksps

65 49 Up to 13 analog input pins External voltage reference input pins Automatic Channel Scan mode Selectable conversion trigger source 16-word conversion result buffer Selectable Buffer Fill modes Four result alignment options The high precision reference voltage source we choose the Analog Device ADR4520 and ADR4533 here. Figure High precision reference voltage schematic. The ADR4520 and ADR4533 devices are high precision, low power, low noise voltage references featuring ±0.02% maximum initial error, excellent temperature stability, and low output noise. These voltage references use an innovative core topology to achieve high

66 50 accuracy while offering industry-leading temperature stability and noise performance. The low, thermally induced output voltage hysteresis and low long-term output voltage drift of the devices also improve system accuracy over time and temperature variations Temperature: Temperature monitoring is really important and useful to battery packs. Li-Ion battery must not be discharged if it is outside at a certain temperature. Also it should not be charged outside a range of temperature. Eliminates many problems which will cause the battery to overheat and even burst into flame, similar to the internal ones where the cell is abused or bad and the external one which may have connection problems or other problems caused by the external environmental issues. All of these problems will cause battery damage and may also be dangerous. Therefore, the temperature sensor is very important.

67 51 Figure Temperature sensing: one per cell; one per pack. With the distributed BMS, it is easy to include a sensor on each cell board. Most normal digital BMS may or may not sense temperature. Only a small number of them has one sensor for the entire battery pack. Here our novel BMS has a sensor for each cell. We have 6 channels temperature sensing. The battery pack has two layers and in each layers two batterie electrodes are connected together. So it only needs 6 sensors for all 8 battery cells.

68 52 Figure Temperature sensor drive circuit schematic Current The measurement of the current allows BMS to have some additional function [20]. Prevent battery pack from working outside the SOA which will cause damage to battery packs Use the current as part of internal SOC, SOH and DOD calculation. Simply report the battery current Use the current with the internal resistance to calculate the terminal voltage of the battery cells Here we use a current sensor to monitor the battery pack current, Allegro ACS758. This sensor provides economical and precise solutions for AC or DC current sensing. Typical applications include motor control, load detection and management, power supply and DCto-DC converter control, inverter control, and over current fault detection.

69 53 Figure Current sensor and current sensing circuit topology The device consists of a precision, low-offset linear Hall circuit with a copper conduction path located near the die. Applied current flowing through this copper conduction path generates a magnetic field which the Hall IC converts into a proportional voltage. Device accuracy is optimized through the close proximity of the magnetic signal to the Hall transducer. A precise, proportional output voltage is provided by the low-offset, chopperstabilized BiCMOS Hall IC, which is programmed for accuracy at the factory. 3.4 Cell Protection System A good BMS will protect the battery pack from working outside the SOA based on the parameters it monitoring: pack current; cell voltage; cell temperature. If it detects something abnormal, it will give out signal to interrupt the current or disconnect the battery pack with

70 54 the charger or load. There are two types of signal on board to tell us about the abnormal state of the battery: The alarm and the LED signal. The frequency of the noise produced by the alarm and the flash of the LED will change with the abnormal degree. Table 3.1 Alert signal when the battery is working outside SOA. Outside the SOA Over current Over voltage Thermal problem Alert Signal LED Flash Buzzer Screaming Both Buzzer and LED Warning Figure Alert signal drive circuit.

71 55 These two signal control PIN is designed for the future use as the control signal PIN for isolate circuit. When battery pack is implemented it the system, these two PINs can send the command to decide if disconnecting the battery pack with motors. If there is some abnormal situation detected, it can turn off the connection to isolate the battery with the whole system. The switches connected to the charger and the load also has over current protection at the test process. The UL1077 working current range is 0-64A and the short circuit rating is 10KA. It will disconnect the charger or load with the battery when the current is over its limit, just like short circuit, which we called it a backup protection. 3.5 State of Charge and State of Health Calculation Here we have collect the voltage, current and temperature. We can do some simple estimation of the SOC and SOH. There are lots of ways to do this. The most accurate way is use the electrochemical models to estimate SOC. But it is hard to be implemented in the hardware. Coulomb counting, which is easy for implementation, is a good way to do estimation. Our software has enough sources to do this kind of estimation. However, it needs the initial SOC, the details will be discussed in the following. Open circuit voltage measurement is also a good way to predict the SOC, because the OCV has a direct relationship with SOC. It does not need any algorithm to be implemented. However, the OCV of the battery need the battery to be in the resting mode for a long time, which is not suitable in most applications. The extended Kalman filter model is also a popular way to

72 56 estimate and it deals with the white noise. But complicated algorithm needs large memory and long time to compute. Our FREEDM center also has the Online parameters/soc co-estimation [25][26], which is able to increase the performance of the estimation with identifying and updating the parameters in the observer structure. This is only need the voltage current and temperature as the parameters for calculation. The monitoring board has the communication port which can transfer the data to the online estimation just like the figure shows. Figure Online parameter and SOC co-estimation. The online adaptive parameter identification and SOC estimation has a high accuracy for the SOC estimation because it update parameters of the battery model continuously. An RC equivalent model was built and the V oc and SOC should have a nonlinear relationship. But

73 57 here in this estimation method, because the SOC has small change at the ordinary charging and discharging current, it is separated into piecewise linear part with varying parameters, which can be useful for easy implementation. The online adaptive identification algorithm will update the parameter of the battery model and the coefficient of the V oc and SOC at each time step. A reduced order observer was designed to estimate the SOC of the battery model. The function of V oc and SOC is used in this observer. With a linear identification method and a linear observer, it can be easy implemented and the continuous parameter update, the accuracy of the battery model and SOC estimation also is improved. Also in this monitoring board itself, it has enough sources to do some simple estimation of SOC, just like the coulomb counting. Integrating the current in or out of the battery can show the charge of the battery. Because we have current sensor so we can do the calculation and estimate the SOC of each cell, if the initial charge of the cell is known. However, there are three limitation of this coulomb counting method: Leakage current of the battery does not go through the current sensor and then it is not taken into account. This will limit the accuracy of the SOC estimation Offset in the measure will result in the SOC drifting. So it is not suitable for fast charging and discharging It needs the initial SOC of the battery, which is not easy to get in some application. As for the Li-Ion battery, they have low leakage current. So this method will has less limitation in this part. The drift problem is the only problem it needs concern. It is suitable for those application not charge and discharge fast.

74 58 So the most efficient and accuracy way is use the online estimation and transfer the data to the second control board for the calculation. 3.6 Power Save Management This monitoring system has two modes of operation: sleep mode and the measure mode. Because in some applications, we do not need to measure the voltage all the time. We can shut down some of the chips on board to help save power. Volt sense OpAmp Current sensor Multiplexer Balancing drive Reference chip Communication Temp sensor chip controller Figure Power distribution on the board.

75 59 Measure mode: During the measure mode, all devices will be turn on and each part will be working on monitoring the voltage, current and temperature. The protection, balance or communication will be on if necessary. Sleep mode: In this mode, if the state of the battery does not need to monitor in a high rate. We can shut down the voltage monitoring part by shut down the Operational Amplifiers of the voltage measurement and shut down most of the chip which are supported by the 5V power supply. Figure Switch and its drive circuit for OpAmp. This one is the switch part and the drive circuit of the switch. It helps shut down the OpAmp which is used to calculate the cell voltage powered from the battery pack.

76 60 To calculate the power saving amount by the sleeping mode, we assume each chip is working on its nominal state to do some approximated estimation, because it is hard to test the power of each chip. Microcontroller (PIC24FJ64GA004): Temperature sensor (MCP9700): Voltage Operational Amplifier (OP490): Reference chip (ADR4520, ADR4533): P 3.3 V*250mA 0.825W 3.3 P 3.3 V*6uA 0.02mW P 30 V*80uA 2.4mW P V *10mA 20.5mW P 3.3 V *10mA 33mW Balancing drive (PS2801): Multiplexer (74HCT4067): P 5 V *5mA 25mW P 5 V*50mA 250mW Current sensor (ACS758ECB-200B-PFF-T): Communication (MAX3221): power save at the sleep mode: P 5 V *4mA 20mW P 5 V *1uA 5uW

77 61 2.4* * % P 39% *6 2.4* * So this can help save the power by 39% if every chip is working on its nominal state. 3.7 Communication Interface There are three type of communication interface: 1 UART non-isolated, for direct data logging to the PC and for configuration. 1 UART isolated, for chaining multiple BMS boards. 1 I2C for multi-board systems or the Zigbee communication. The UART is asynchronous communication protocol. Its maximum bit rate is 500 kbit/s. It can be used in the long distance with the improved noise immunity. Typically application is for the personal computers and other data acquisition system. But it requires accurate clock frequency. This communication interface can be used between the microcontroller and child Zigbee communication to transfer data. I2C is a synchronous communication protocol. The maximum can be 1Mbit/s. it allows multiple masters. But it is limited by the slowest and short distance. The I2C interface, for example, uses two wires (and therefore two pins of the microcontroller), one for the clock (referred to as SCL) and one (bidirectional) for the data (SDA). It is simple and easy for implemented

78 62 Figure Non-isolated communication interface. The MAX3221 transceivers are 3Vpowered EIA/TIA-232 and V.28/V.24 communications interfaces intended for notebook computer applications. A proprietary, highefficiency, dual charge-pump power supply and a low-dropout transmitter combine to deliver true RS-232 performance from a single +3.0V to +5.5V supply. A guaranteed data rate of 120kbps provides compatibility with popular software for communicating with personal computers. Figure Isolated communication interface.

79 63 Figure I2C communication interface. 3.8 Package and Integration To help eliminate lots of wire connection, a special design for the battery module is developed. The figure3.8.1 below shows the details.

80 64 Figure Specific package design for the battery module connection The ten round black holes are80 the connection with the battery electrodes. This will help eliminate the 128 wires connection, which will cause a mess of the whole system. If not connected properly, the wires will easy be short circuit or reverse connected, which will cause dangerous or damage of the monitoring system. Our monitoring board is also integrated with the passive balancing circuit. The detail of the passive balancing method will be discussed in the next chapter.

81 65 Figure Passive balancing integration on the monitoring board. The passive balancing circuit is integrated on the monitoring board. It is controlled by the MCU to decide which battery cell needs dissipation. It can be connected or disconnected with the battery pack. In this way we can choose use passive or active balancing method for balancing. This monitoring board also has enough PINs for the control signal for the active cell balancing. It also has an interface for the connection with the active cell balancing board. 3.9 The Experimental Verification Here we use the 40Ah Thundersky Li-Ion battery to do the test. 5 of them are connected in series. We charge and discharge it at 10A (0.25C) and monitoring the voltage, current and temperature. The following is test result. One most important thing I should mention here is

82 66 the batteries we used here is not in the same condition, which means they are old batteries and some of them are used charge and discharge many cycles, others may only be used few cycles. This is the difference compared with the batteries used in EVs, because the EVs batteries are new or even manufacture in a same batch. So the difference of the internal imbalance or sometime the capacity imbalance for the batteries used here is much worse than those in EVs. Figure A discharging voltage curve. We discharge the battery pack for about 1 hour. From the curve, we can see two of them voltage drop more rapidly than the others, which directly indicates the SOC of them is lower

83 67 than others. But the almost equal voltage at first is probably caused by the internal resistance imbalance. This shows the impedance of these two batteries is lower than the others, which indicates these two batter cells are newer. Figure A discharging temperature and current curve. The temperature and current is stable during the charging process. Because the outside temperature of these cells is the stable, the temperature of the cells does not change under the 0.25C discharge. The only different temperature is happened for the blue cell. Because we put the thermal sensor at the electrode of the batteries which also contact the connector which

84 68 connect two battery positive and negative electrodes, the difference of the connector factor will may be the reason why the temperature is higher. We detect the temperature of the 4 connectors. Only the connector of the blue curve has a higher temperature. This is the same with the temperature sensor we get. Figure A charging voltage curve. We charge the same battery pack for about 2 hours until the highest cell voltage reached the threshold value under 0.25C. From the curve, we can see one of the blue cell has higher SOC than the others. This can also be caused by the higher internal resistance of that cell. We

85 69 can compare it with the discharge curve. During discharging, the blue curve cell has the lowest voltage. This indicates the internal resistance is the factor that causes the situation. This cell is the oldest cell which has higher impedance and the initial SOC is also higher. Figure A charging temperature and current curve. The temperature and current is also stable during the charging process. Because we test it under the same outside situation and the charge and discharge current is not exceed or even

86 70 close to its threshold value, the temperature is decided most by the external environment. Then the current for the charge and discharge is almost same here Summary A novel low power monitoring system is presented here.. It can monitor 4 to 8 battery cells connected in series, which has 8 channels of voltage signal with resolution of 1.3 mv, 6 channels of temperature signal with resolution of 1 0 C and 1 channels of current signal with resolution of 1A. It has 1 isolated and non-isolated UART and 1 I2C protocol, which can communicated with the second controller to do the online parameter/ state of charge (SOC) co-estimation. It has the power management system which can help save the 39% energy when working in sleeping mode, as well as detect the unsafe operating state and give out the protection signal.

87 71 Chapter 4 Design and Evaluation of a Passive Balancing Method 4.1 Introduction Different Causes of Imbalance The cells imbalance for battery pack is about 3 types of unbalance: SOC unbalance; capacity unbalance and internal impedance unbalance. These unbalance situation can be caused by many reasons. We separate it simply into two kinds: internal and external sources. Internal sources are caused mostly by manufacturing variability. It will cause the difference in charge storage capacity, variations in internal impedance and differences in self-discharge rate even in the same batch. For example, new cells may have up to ±15% difference in the value of internal impedance at low frequencies. External sources are caused by the temperature difference, rate of charge or discharge and some multi-rank pack protection ICs, which drain charge unequally from the different series ranks in the pack. As for the temperature gradient, the battery cells can be exposed in different situations where the component dissipate different heat, because the battery pack is a large and long pack in electrical Vehicles. The capacitor, impedance and voltage are thermal dependent parameter. This will cause the difference in the SOC, capacity and the internal impedance. In addition the thermal difference will also cause the different self

88 72 discharge rates of the cells. For the rate of charge or discharge, those cells which have low capacity will reach full charge first and will easy be overcharged. The same is for the over discharged. This will cause the reduction of the capacity. Then this vicious loop will continued until the whole pack damaged. For the protection ICs, they gain power from each cell and this is common on those distributed BMS. Each IC on the battery cell is powered by the cells itself. The difference of the working state of the IC will gain difference power from each cell. This will cause the difference of the SOC of the cells Affection of Imbalance to Cell Terminal Voltage a) State of Charge Unbalance SOC is one of the most important parameter we concern about. This is also the imbalance we need to equalize. Just like the battery we use, the fully charge is 40Ah. Assume we discharge one of them 2Ah and the others 3Ah. The first state of charge will be 95% and the others will be 92.5%. The imbalance will be 2.5%. The difference of the SOC will cause the difference of the open circuit voltage (OCV), because the OCV has a direct relationship with the SOC. The figure shows the difference of OCV between two cells with 1% SOC unbalance.

89 73 Figure Voltage difference relationship with 1% SOC imbalance. [27] The terminal voltage under load can be approximately got from the OCV and SOC. V OCV( SOC) I * R Because the current for the series connected current is the same and we can assume the impedance difference is not significant because we consider the SOC difference first caused by other influence. The difference of the SOC will have a direct correlation with the terminal voltage, which we can detect. It is significant at the low SOC. So if we can balance at the beginning of the charge when the SOC is low, we can directly balance the terminal voltage of the battery cells rather than the SOC. This is good for the active cell balancing method where the converters working based on the condition of the terminal voltage.

90 74 b) Capacity Unbalance Because of the manufacture issue, the initial capacity of the battery cells will have a small difference to start with. Just take the 40Ah battery we use for an example again. If one of the batteries initial capacity is 39Ah, then we discharge all of them by 2Ah. The 39Ah battery SOC is 94.9% and the others are all 95%. The SOC difference is only 0.1%. We can see 1Ah difference in capacity has a much less impact on the SOC difference than the 1Ah difference in SOC. Because the SOC correlates to the terminal voltage, this shows the capacity difference has less voltage difference than the SOC. With the manufacture technology improves, the difference among the initial capacity is less and is tolerant for the battery pack, which will not be a major concern if the imbalance is not becoming worse. In EVs, the battery cells are almost new and sometimes are made from the same batch. They are charge and discharge in the same condition where the capacity can be little different. Then the capacity imbalance only has little impact on the SOC imbalance where we can ignore this capacity influence to the battery cell unbalance. c) Internal Impedance Unbalance The impedance we mention here is the DC series resistance, not the AC impedance the manufactures provides. Because when we use the battery cells, DC current is flow in and out, only the DC impedance has impact to the terminal voltage. Therefore, the impedance mention in this thesis is the DC series resistance of the battery.

91 75 The impedance imbalance will occur when the battery is charge and discharge lots of cycles, because many factors will affect the value of the impedance, such as the thermal problem and the operating current. The new battery internal impedance difference can be approximately 15% on the same manufacture batch. The impedance for Li-Ion battery is really low because the battery needs working under large current condition, where the manufacture difference for the impedance can be ignored. The impedance unbalance does not have affect on the OCV but has influence on the cell terminal voltage. V OCV I* R During the charging part, the voltage will be higher with those cells which have higher internal impedance. During discharging part, the terminal voltage will be lower with those which have higher impedance. No balance method can help eliminate the impedance imbalance, because it is caused by the internal chemistry difference. What we can do is to make the battery cells working condition the same especially the temperature to make sure there is no outside affect to make the impedance imbalance. However, the impedance imbalance will distort the SOC balance if we ignore it. If we just look at the terminal voltage and use it to determine which one needs to balance, we may not know if the voltage difference is caused by the SOC difference or the internal impedance difference. If it is caused by the impedance difference, bypass more current or charging more current will cause the imbalance situation worse.

92 76 This problem can be reduced if we do the balance at the end of charge when the current is reduced. Then the voltage drop in the impedance will drop. Because the impedance value is not very large and if the current is low, the terminal voltage difference will be determined by the SOC difference Harms Performance to Battery Pack from Imbalance a) Cell Degradation Figure Individual cell voltage vs. capacity deficiency from nominal. [27] For the SOC or capacity imbalance, the cell with higher SOC will result in higher voltages. Typically the charging ends with the battery pack voltage reaches the nominal working condition to make sure the battery pack can support enough power. For example, if we use 6*4.2V cells connected in series, the charger will stop charging when the pack

93 77 voltage reach at 25.2V. However, the individual cells voltage will not be the same. The lower capacity cell will have a higher voltage, just like the Figure shows. The normal cells will have a lower voltage. When the capacity deficiency large enough, the battery cells will be overcharged, which will cause dangerous for Li-ion battery. To make things worse, the degradation of the cell will accelerate if no balance method is applied. The one which has lower capacity will be exposed to a higher voltage, which will makes it degrade faster so that the capacity will become even less. So cell balancing is necessary for decrease of the cell degradation. b) Early Charge and Discharge Termination To consider the safety issue, the charge and discharge limit can be set at the point not to overcharge and over discharge. Just like at the charging duration, the charger will stop charging when any of them reach their threshold overcharge voltage. If the imbalance situation occurs, those imbalance cells will not be fully charged. This will first reduce the cycle life of the batteries, because most of them are not fully charged and its SOC is lower and cannot be used such a long time and provide so much power. Another serious situation is the EV maybe cannot work normally because the voltage of the battery pack may be less than its nominal voltage. The same situation is for the discharge termination. So the cell balance is significant important for the increase of the battery pack cycle life. c) Safety Concern of Overcharge and Over discharge Li-ion batteries are in small volume which has high energy. It cannot be overcharged or overheating. Thermal runaway may occur which will cause danger, which has been

94 78 mentioned at the beginning of the thesis. The imbalance of the SOC will result in the voltage difference and the higher voltage cells will easy reach the overcharged. So the cell balance is really important to help all the cells are fully charged and no cell overcharge occur Summary of the Imbalance situation There are three types of imbalance: SOC, capacity and internal impedance. We need to balance the state of charge for every battery cells. The other two imbalance factor will result in the SOC imbalance. As for the capacity imbalance, it is not significant and can be ignored with the development of the manufacture technique. The impedance imbalance can be reduced if we balance at the end of the charge when the current is not very high. The SOC has a direct correlation with the cell voltage, which we show above. So in most situations, we can use the cell voltage as the reference to judge if the SOC has balanced. For the passive balancing, we can balance the cells voltage at the beginning of the charge and the end of charge because the voltage will be obviously different with the SOC imbalance. If we can get the SOC of each cell, we can use the SOC as the control parameter to control each switch turn on or off. Then the internal impedance imbalance will have no influence on this kind of balance method. Here, because we do not communicate with the online SOC estimate, we only use the voltage as the control parameter and balance the voltage at the beginning and end of charge. For the active cell balancing, because the balance current is really high and the balance time is not long. So we can use the battery voltage as the reference parameter to estimate the

95 79 SOC difference. We can balance the battery pack at the end of charge when the internal impedance impact reduced and the voltage difference can better reflect the SOC unbalance. Another important reason is that the number of switches will increase if we use the SOC as the control parameter. We need to control each channel of the balance circuit and the control algorithm will be much more difficult and complicated. So in this thesis, the active cell balance converter is to balance the voltage at the end of charge. 4.2 Design of the Passive Balancing Topology The passive balancing method is the simplest and cheapest way to balance cells. It dissipates energy from those higher SOC cells by turn on the corresponding switches. It can be operated continuously with the switched turning on and off all the time if required. The efficiency and effect is determined by the resistor and the control algorithm we choose. However, this kind of method is dissipating energy into heat which has a low efficiency. But this one is simple and stable. The cost is also low, which makes it a very appealing balancing method in EVs.

96 80 Figure Passive balancing circuit topology schematic. We used 8 balancing circuit for 8 batteries. The cell voltages are sensed by a 10-bit AD converter. The microcontroller (MCU) acquires the cell voltage from the ADC and check if there is any difference of among the cell voltage. If there is any cell voltage difference is large than 5mV, the MCU will turn on the MOSFET through a gate driver IC. Then the bypass shunt resistor will begin to bypass the current. When all the cell voltages differences are less than 5mV, the balance part will be turned off Switch Device Selection Here the MOSFET voltage stress is limit by the battery itself. So max( V ) 4.2V The current stress limit is the largest balancing current we set Consider the safety margin Q max( I ) 1A Q

97 81 max( V ) 6.3V Q max( I ) 1.5A Here we choose the RZR040P01TLCT-ND, Silicon P-channel MOSFET. It has Low onresistance, high power package, and Low voltage drive. (1.5V). Its drain current can be as large as 4A. This can help us have a large margin to do the bypass current design. The block voltage is 12V which is also enough in this application. Q Table 4.1 MOSFET absolute maximum ratings. parameter limits unit Drain-source voltage -12 V Gate-source voltage 10 V Drain continuous 4 A current pulsed 16 A Total power dissipation 1 W Channel temperature C

98 High Current Bypass Dissipate Resistor Selection There are two concerns about the resistor selection: the limit value of the balancing current and the power rating of the resistor. Here we need the bypass current be as much as 1A. So the resistor value and power rating should be max( I) 1A V R 4ohm I max( P) V * I 4W The resistor should be 4ohm to help make the bypass current as much as 1A in some specific situation when the battery is fully charged. The power rating of the resistor should be no less than 4W. Here to consider about the thermal management and the separation of the power, we choose two 8ohm resistor connected in parallel. In this way the heat dissipation square will double and the power will be separate into two parts which will lower the selection standard of the resistor. Here we choose the RCL12258R06FKEG-ND, Long Side Termination Thick Film Chip Resistors. FEATURES: Enhanced power rating Long side terminations Enhanced thermo cycling performance in 1225 size Pure tin solder contacts on Ni barrier layer, provides compatibility with lead (Pb)-free and lead containing soldering processes

99 83 The power rating of this resistor can be as large as 2W. We choose two 8ohm resistor connected by parallel as the dissipate part. Then the bypass current can be as large as 1A. And the 1A is also suitable for the MOSFET drain current limit. The power can be as much as 2W only if we balance it when it fully charged. If we balance it at the beginning of the charge, it will not reach its rated power limit. So this kind of resistor is suitable in this application Passive Balancing Circuit Controller Design ADC read Calculate the average flase Value>4V Valueaverage>limit true true Turn on: MOSFET Red LED on false Turn on: yellow LED Turn off: MOSFET Red LED off Exit Figure Passive balancing controller diagram.

100 84 If we can get the value of SOC of the battery cell, we can use it to set our control command. This will be the best and will be done after we communicate with the second level controller. We can calculate out the difference of the SOC and decide which cell needs to balance and turn on the switched. But this method also has some drawbacks. When the balance circuit is turned on, we need to monitor the balance current for the calculation of the SOC. This will increase much more component just like the current sensor or high precision resistor. Because the SOC has a direct correlation with the terminal voltage of the cells, the simplest way is do balance based on the voltage difference among cells. Just like the control diagram shows, if the voltage exceeds the threshold, the bypass circuit will be enable. To help resolve some problem, we can do more to help improve the control of the passive balancing. Balancing during the beginning of charge: this can be used to help save energy because the voltage at that time is lower. Another important reason is the SOC difference can be obviously reflected by the voltage difference at the beginning of the charging. This can help balance the cells SOC much better Balancing at high states of charge only: This is used to decrease the effect of impedance unbalance, because at the end of charge, it will go into constant voltage or

101 85 floating charge where the current is low, the impedance imbalance has little effect on the voltage difference. The difference of the voltage is only caused by the SOC imbalance. So this passive balancing is working at the beginning of the charging. At the end of charge, it will turn on to see if there is still some unbalancing caused by the impedance unbalance to make sure the whole pack has been balanced. 4.3 Thermal Problem Mitigation High Current Bypass Resistor Package Selection Thermal problem is the most serious problem which eliminates the application of the passive balancing. The bypass resistor is the main component dissipates heat. This is also the part needs thermal management. So the choose of the type of the resistor has a significant influence on the thermal problem of the passive balance method. The resistor component is small compared to the PCB. The heat removal from the radiation can be ignored so the thermal model can be simply defined: Where T j is the temperature of the resistive layer T a is the ambient temperature T T Rth * P T ( Rth Rth )* P j a ja a jsp spa

102 86 Rth ja is the thermal resistor between the resistive layer and the ambient Rthjsp is the thermal resistor between the resistive layer and the solder joint Rthspa is the thermal resistor between the solder joint and the ambient. P is the power dissipation of the resistor The key thermal resistance value for component manufacturers is affected by the choice of material, the resistor pattern, and the terminations. The normal package resistor parameter is show in table 4.2. Rth jsp, which is Table 4.2 Power dissipation of CHP Thick Film chip resistor. Size Rth jsp ( 0 C / W ) Rth spa ( 0 C / W ) Rth ja ( 0 C / W ) These kinds of package resistors are not suitable for the application for this passive method. The power here is: P P min max / / W W The temperature can be as large as C at the worst case which is not acceptable.

103 87 relieved. By using the Long Side Termination Thick Film Chip Resistors, this situation can be Table 4.3 Power dissipation of long side termination thick film chip resistors. Size Rth jsp ( 0 C / W ) Rth spa ( 0 C / W ) Rth ja ( 0 C / W ) Here we choose the RCL1225 which has the best thermal condition. We test it at about 1W dissipate power, the temperature is stable at 60 0 C, which is acceptable Bypass Current Optimization The bypass current is one of the most important parameters in the cell passive balance method. There are two factors limit its value: balance speed, thermal limitation caused by power dissipation. Balance speed: cell balance speed is a significant feature for the choice of the balance method. Passive balance method speed is not very fast. It is the limitation for passive balance method widely used, especially for the fast charge. However, if the balance current is too low, it will not be able to balance the whole battery pack during the charge.

104 88 The relation between the balance current and the balance speed can be show in the figure below. It shows the relation between different balance current and the balance time needed for balance 1% imbalance of the 20Ah battery pack. Figure Balancing current vs. balance time for 1% SOC imbalance. The larger the balance current is, the faster the balancing speed will be. However, the current cannot be too large. It is limit by the power dissipation and the thermal problem. Thermal limitation: As we can see, if the balance current is too large, the power dissipation will be high, the efficiency of the circuit will be low. What is worse, the power dissipate on the resistor will

105 89 dissipate as heat, which will cause the thermal problem. The relations ship between temperature of the resistor and the balance current is show below. Figure Relationship between the resistor temperature and balance current. We can see when the current increase, the temperature will exceed unlimited. So we cannot choose the current as large as we need. Here, the resistor operating temperature range is form C to C, we choose the working temperature no larger than100 considering the safety margin and some unexpected working condition occur. From the figure, we find if the current larger than 1.2A the temperature will exceed than C. Here we set the balance current no large than 1A to make sure the temperature is no larger than C and the balancing speed is also fast enough.

106 Experimental Verification We set the voltage difference as V1=3.364; V2=3.366; V3=3.360; V4=3.364; V5= Because the passive current is low and it needs a long time to balance. So we charge or discharge the cells individually to make the voltage difference not too large, which can help us to see the balance procedure much more clearly. The charging current is 1 A to eliminate the impedance imbalance and the balance current here under this voltage type can be as much as 0.7A. Figure Passive balancing procedure.

107 91 From the figure 4.4.1, we can see the balance process. The 3.36 battery (blue curve) has higher internal impedance as the analysis in chapter 2, so the SOC is lower than the others if they have same voltage if we consider the impedance imbalance. It may need more charge current to help balance. The voltage of the 3.36 battery is lower than the others. Then we can see the SOC of the 3.36V cell is much lower than the others despite we use a low current charge to help decrease the impact of the internal resistance imbalance. From the result, we can find when the other 4 battery cells balance, the 3.36V is still unbalance. But we can see the voltage difference is decrease and the SOC difference is then decrease. It only needs more time and then all the cells will be balanced. 4.5 Summary of the Passive Balancing Method A passive balancing circuit is designed and developed here. The passive current can be as much as 1A. Thermal problem is acceptable by the choice of the long side termination RCL thick film chip resistors and the passive balancing current optimization. Analysis of the thermal problem caused by the package of dissipate resistor is presented and the experiment verification shows this passive balancing circuit works well.

108 92 Chapter 5 Design and Evaluation a Novel Active Cell Balancing Converter This chapter will show a novel active cell balancing converter designed for 6 battery cells connected in series as a module. The converter is designed based on the single active bridge. Some improvements have been done to help the converter have a high efficiency and fast balance. The details will show in this chapter. Design details are presented in the second section. The comparison between the conventional DCM and the control algorithm HCM we used here is doing in the third section. At last, simulation results and analysis are proposed. 5.1 The Power Stage Conventional Single Active Bridge DC DC Converter As for the conventional active bridge, a half bridge is on the primary side and a full bridge diode rectifier is on the secondary side.

109 93 Figure Conventional SAB converter. [17] The inductor, as the main energy storage element, is a triangular waveform. It is determined by the voltage across it. The voltage applied on the primary side of the transformer is a square waveform. The parasitic parameters of the transformer will have a significance influence on the converter, especially the winding capacitance and the magnetizing inductance. There may be some oscillations in some interval. The leakage inductance, which will also be part of the energy storage element, is just like the inductor. In some applications, the inductor will be integrated in the transformer as the leakage inductance of it. The higher the leakage inductor value will cause the lower the duty cycle. At light load, the SAB converter can work in DCM. Just in our application, the power is no very high, it can work in the DCM. But at nominal load or the heavy load, it needs work on the CCM.

110 94 Figure Operating mode of the SAB. [17] As illustrated in the figure 5.1.2, the voltage across the inductance is positive. The current is flowing across DA. The voltage across the TA is zero. The transistor can switch under zero voltage condition. And then the current reverses. When the TA is turned off. The energy in the inductance can be transferred into the snubber capacitors by resonance. The snubber capacitor across the TA will take over the current and the transistor can be turned off under zero voltage condition. At the same time, the DA across the TA will be forward bias by the discharge of the snubber capacitor across TA. Similarly, the TA will turned on ZVS like TA. Then the commutation will repeat. The steady state analysis of the SAB is on CCM. The DCM is similar, which will be analysis in the control mode later.

111 95 In subinterval one, the current is flowing across the diode and is negative in the inductance. The voltage across the inductance is positive. V V I d T L dc o neg TA s V o here is the voltage in the primary of the transformer. In subinterval two, the current is flowing across the MOSFET and is positive in the inductance. The voltage across the inductance is positive. V V I d T L dc o pos DA s The inductor should be voltage second balance. So M is the conversion ratio Vdc Vo Vdc Vo d T d T A A L L 1 M dd d A TA 1 M D s T s M V V o dc In order to achieve the soft-switching, the element should be satisfied the following equation [17] 1 1 LI (2 C )(2 V ) 2 2 2V dc I L Z Z o 2 2 L s dc o L 2C s

112 96 Combine the equation d D A 2L Z T (1 M) o s Our novel active balancing converter is designed based on the SAB, because this full bridge converter can be fully controlled and the power rating can be high. In this way it can equalize fast because of the balancing current can be as much possible. The unidirectional power flow can distribute the power from the whole pack to each cell based on its voltage value to achieve cell balance. Here, instead of using close loop control, a novel hysteresis current control with frequency limit can be implemented in the SAB to help simplify the control algorithm A Novel Active Balancing DC DC Converter Topology Here the power flow is unidirectional. It transfers power from the battery pack to the battery cells. Then the energy will redistributed to each cell. The cell which has the lower voltage will get more power. The one whose voltage is high enough will get no balance power. In this way, the cell will be balanced very fast without any complicated control. Because the power flow is unidirectional, we can eliminate the small loop between two individual cells. For example, if we use the switches to replace the diode, the power can flow a high voltage cell to the low voltage cell just through the transformer, cell to cell. This will cause many uncontrollable elements to this circuit. The unlimited loop current will also cause the

113 97 transformer saturation. We use the diode to force the power flow unidirectional, which will help eliminate many unrespectable situation and make the current and power controllable D1 ieq1 i1 D2 ieq2 D3 S1 S3 i2 D4 ieq3 D5 S2 S4 i3 D6 ieq4 L i4 T3 D7 D8 B4 ieq5 L i5 D9 D10 B5 ieq6 D11 i6 D12 Figure Proposed active cell balancing topology.

114 98 Here, we can assume the 6 cell voltage V1, V2, V3, V4, V5, V 6 V1 V2 V3 V4 V5 V6 to help better analyze the work state of this converter. There are m cells need to be balanced, V... 1 V m, which means V... V V 1 m V... V V 6 m 1 ave ave Where V ave V V V V V V The balance current of those cells whose voltage is larger than the V ave will be block by the diode and the balancing current will be distributed to low voltage cells by the voltage amount in the following way. We can get the voltage difference between the two sides of each inductor of those cells which have balancing current. During subinterval 1, when S1 and S3 is on and the current is negative in inductor. V V V V V V V n* V V V V V V V V n* V m m If we do not consider the dissipate part form the diode and the MOSFET, we can get the inductor voltage above and because of the current of the inductor has a direct relationship with the voltage, then V i t L L1 1

115 99 Because the voltage is positive across the inductor, then the current will increase. When the current reaches zero, then it will be in subinterval 2. The current is positive in inductor. V V V V V V V n* V V V V V V V V n* V m m then V i t L L2 2 The working state is the same during subinterval 3 and 4 when S2 and S4 are on and S1 and S3 are off. Because the working state is symmetrically, so the positive peak current is equal to the negative peak current, which means I i i Lpeak L1 L2 We control the current of the primary full bridge side and set it no more than 3A. I I A B max Lpeak 3 So this 3A balancing current will be distributed to each low voltage cell which needs balancing. Ideally, the main influence of the amount of the balance current is the voltage difference of the inductor. So the average balancing current of each balance channel are: I I b1 bm IB V1 * n* 2 V... V I B Vm * n* 2 V... V If we consider the dissipation part, the difference is the 1 1 m m V.

116 100 V V V V V V V 2* I * R n*( V V ) ' B on 1 F V V V V V V V 2* I * R n*( V V ) ' m B on m F We calculate the average balancing current, so we just consider the working state when the primary side is on its slide current 3A. The equation above is an approximate equation for calculation of the average balancing current. By choosing a proper value of the turn ratio of the transformer, we can set the blocking voltage of the balancing channel the average voltage of the whole pack. I I I I ' ' B 1 b1 * * ' ' 2 L' 1 V 1... V m ' bm ' bm 1 ' b6 I ' IB n V m * * 2 L' V... V 0 0 n ' 2 1 V ' m In this way, we can find that the balancing current amount is determined by the voltage difference across the inductor, which indicates the voltage difference of each cell. The cell whose voltage is lower, will have a larger which is the way we need. V. Then it will have a larger balancing current, To help reduce the loss and increase the efficiency, we change the full diode bridge into the center-tap transformer connected diode bridge. In this way the efficiency of this converter can be improve significantly, because the battery is working from 2.8V to 4V, while the diode forward voltage drop is 0.53V or sometimes 0.7V, which is significant to the battery cell voltage. In this simulation file, the efficiency can be increase from 77.72% to 87.16%

117 101 We set the voltage on the secondary part of the transformer as the average voltage of this battery pack. In this way, those cells whose voltage is higher than the average value will be cut off by the diode, which will receive no balance current. This will help increase the balance speed. Because the redistributed power is mostly going to those cells have lower voltage. Moreover, in this way, when the cell voltage is balancing, there will be no balancing current in the secondary part of the transformers. This acts as automatically turning off the balance. However, there will still be some current transfer between the battery cells and transformers. So we need turn off all the switches when detected all the cells are balance. Here, we use the transformer connected in parallel in the primary side, instead of the use a multi-output transformer. In this way, it is easier for repair and replace. If one of the balance circuit broken, we just need to replace one of them, but not replace the muti-output transformer.

118 Design of the Novel Active Cell Balancing Converter MOSFET and Diode Selections Figure Proposed active cell balancing circuit for one cell. Each channel of the converter can be simplified as the figure5.2.1 above. The primary side is the power by the whole pack battery voltage. Due to the symmetry conditions, the working mode can be simplified into two. The voltage in the two side of the inductor V, V is either the same polarity or the different polarity. As in mode one, the V, V have the same polarity, just assume they are both positive. in o When they are both negative, it is the same. The only difference is voltage stress is on the S2 and S3, but not S1 and S4. Then in o

119 103 max( V, V ) V 4*6 24V S1 s4 B _ pack The same is on S2 and S3. Then, max( V ) V 24V S B_ pack The current limit is based on the balancing current summary. Because we set the largest balance current 10A, the peak current should be 20A. The worst case is there is only one cell lower than the others. Then the whole balance current will go to this channel. So the current limit should be se as Here n is the turn ratio of the transformer I L I B n The current limit for the MOSFT is about 3.33A. Here we use the hysteresis current control, which will be discussed in the control part. So we set the current limit as 3A. In mode two, the limit for the MOSFET is the same. As for the diode, the largest voltage stress is when it has no forward current. For example, when D3 is on, the voltage stress on D4 is max( V ) 2*max( V ) 8 D4 B2 The same voltage stress is on the D3. As for the current limit on the diode, we set the balance average current limit as 20A. So, I D _ ave 10A

120 104 All of the voltage and current should consider a safety margin of at least 1.5. So the limitation of the diodes and transistors are in the table below: Table 5.1 Calculated maximum value of the MOSFET and diode. Rated voltage Rated current Voltage_margin Current_ma rgin MOSFET(Q 24V 3A 36V 4.5A 1-Q4) DIODE(D1-8V 10A 12V 15A D12) We select the diode and transistor: Table 5.2 Maximum rate of the MOSFET and diode. Vds Rds(on) Ion MOSFET(Q1-Q4) 55V 0.035ohm 30A IRZ34NS/L I F V R V F DIODE(D1-D12) 18A 35V 0.53V 18TQ035

121 Energy Storage Inductance Value Design The inductor, which is the key energy storage element of the converter, has a significant role in this converter. We set the hysteresis slide control current at 3A. Then the inductor will have an effect on the frequency of the current. The less the inductor value is, the higher frequency the current is. To help decrease the shape of the converter, we set the frequency at hundred kilohertz when its balance current is 10A. So the inductor of the transformer is Here T s is the cycle of the balance current. Vin V I 0 B 0.5T L V L 0.5Ts I Then we set the inductance at 2 u. If the voltage difference is 100 mv, the balance current frequency is about 100K. When the voltage difference decrease and the frequency decrease, the frequency control will be applied and the frequency will be limited at 100K. When the voltage difference increases, the frequency will increase. B s Transformer Design Turn ratio of the transformer is the key effect of this converter, because we want the voltage in the secondary of the transformer to be the average voltage of the whole battery pack. But there is one diode on the secondary side. The forward voltage will have a significant effect on the whole battery pack. If we just set the turn ratio as the 6:1, the converter will have no balance current when it is not totally balanced. For example, the 6

122 106 cells voltage is 3.9, 3.9, 3.9, 3.9, 3.9, The average value is The forward voltage across the diode is only 0.06V, which is not enough for the diode to turn on. So we need to do some correction on the turn ratio of the transformer to help it is more sensitive to the voltage difference. As consider the active balance is used for the fast charging and the balance current is really high. So it needs not very long time for balance. We decrease the turn ratio of the transformer to help offset the diode forward voltage. The lower the transformer turn ratio is, the more sensitive to the voltage difference. But if the transformer turn ratio is too low, the converter balance current cannot be block by the diode when all the cell voltage is balance. So we set the turn ratio as when the battery is fully charged and it will have no balance current. The turn ratio is 4*6 n In this way, the balance current will be blocked by the diode when the battery is fully charged if the cell voltage is balance at that time.

123 Hysteresis Current Controller Design Conventional DCM Controller Design and Analysis The discontinuous mode is a commonly control method for the SAB converter in case of the current of the converter is always return to zero. In the discontinuous mode, the balance current is only determined by the duty ratio if we choose a proper inductor. During the subinterval 1, when the current flows across the inductor is positive and the voltage across the inductor is also positive because the current is increasing. I V V L dt dc o max 1 s During the subinterval two the current begins to decrease, because the MOSFET is turned off and the voltage across the inductor is negative. I V V L dt dc o max 2 s During the subinterval three, all the transistor and diodes are turned off. There will be no current in the inductor. d 1 d2 1 Here, Then d1 0.44

124 108 However, the DCM has a disadvantage when it is used here for the balancing. Because the voltage unbalancing is not predictable in many different situations, it can vary from 1mV to 0.2V, sometimes even worse. The balancing current is uncontrollable here. We also set the average balancing current 10A when only one cell has 0.1V unbalance problem. From the figure below we can find the balance current can be too high for the battery cells when the voltage imbalance is too high and the balance current will be low, sometimes even lower than passive balancing current, when the voltage difference is small. The peak current can be as much as 40A. This is too much for battery cells, almost 1C charge. Another issue is the choice of the component because of the high balance current. The limit value of MOSFET and diode will be increase. Figure Voltage difference vs. balance peak current for DCM.

125 109 At the design procedure, the DCM will need component which has high voltage and current tolerance than HCC. This will make us to choose expensive device to conduct. If we set the maximum average balance current at 10A, the worst case for DCM is one of the cells has highest voltage, while the others voltage are low. Then the current stress for the MOSFT in the primary side is max( I ) 50A It is 5 times than the one we choose. This will increase the cost of the whole design. But in HCC the total current flow into the MOSFET will be controlled and limit at 10A. This is the limit current we do not consider the voltage difference. If we put the voltage difference into consideration, the current limit will be 100A, which is 10 times than the one we use HCC. To help overcome these problems, an HCC control method is designed. S Design of Hysteresis Current Control Hysteresis control mode has a good dynamic response, is insensitive for parameters change and easier in implementation. This make this control method extensively used for the DC-DC converter. A typical hysteresis control has two control modes for converter: voltage control mode and current control mode. Here, we care more about the balance current, so the current mode is employed. The current at the primary full bridge side is the parameter to be controlled.

126 110 Figure Hysteresis current control diagram. In hysteresis current control mode, the input signal of the switches is decided by a sliding surface of the full bridge MOSFET current we set. The switching status u (for S1) which indicates the turning on and off of the converter switches is decided by the sliding line we set. 1 ON, when I Ihys u 0 OFF, when I Ihys previous state, otherwise Ihys 3A Ihys 3A Ihys and Ihys are the current limit for the I, when I reach it up limit, the switch of S1 and S4 will turn off, S2 and s3 will correspondingly turn on, the current I will become small because the voltage across the inductor is negative. When I reach it down limit, the switch of S1 and S4 will turn on, S2 and s3 will correspondingly turn off, the current I will

127 111 become small because the voltage across the inductor is negative. When the current is between the up and down limit, the switches will remain at its previous state. In this way, the current will be controlled between a fixed range. Then the largest balance current will be controlled. I max( Ihys, Ihys ) ref The worst case will be occurred when only one battery cell voltage lower than the others. Then all the balance current will go to charge this battery cell. So the maximum balance current is I bmax n* Iref We set the maximum average balance current 10A. So the maximum balance reference current is Iref 3.7A Here we set the Ihys 3A Ihys 3A Then the controller is designed like the diagram below.

128 112 Figure Hysteresis current control mode. By using the hysteresis current control, the primary side current is controlled and then the maximum current of the second side is also controlled. In this way, we can make the converter working at the safety condition. The balance current can be stable at 8A with voltage difference varies if we do not consider the frequency problem. This is better than the DCM control. Figure Average current comparison between the DCM and HCC.

129 113 The most important feature of HCC is it is easier to be implemented and is tolerant about parameter changes. It also has no stable problems compared with close loop control Hysteresis Current Control with Frequency limit Controller Design The hysteresis controller is widely used in low voltage application due to its advantages. The most significant merits is its fast respond and robust with simple design and easy implementation. It has fast transient performance since it respond to the change of current or load after the transient happens. Compared with feedback loop control, it reduces the design effort for calculating and theoretical analysis. The elimination of compensation network also helps the fast transient. However, the Hysteresis controller has a major drawback: uncontrollable switching frequency. To help eliminate this problem, a frequency limit controller is designed. If there is no frequency control for this balancing converter, the frequency can variable range can be calculated approximately base on the voltage difference across the inductor. f 6* V L* I /6 B Here we assume only one battery cell is different from the others and the V is the voltage difference. Because we need to calculate the voltage across the inductor, then the voltage difference can be approximately as n* V if we do not consider the dissipation of the MOSFET and diode. Then according the working state analysis we can get the frequency

130 114 equation. Here we select L to set the frequency 100KHz when V is 100mV. So the relationship between the frequency and the voltage difference is show the figure below. Figure Switch frequency varies with the voltage difference. From the figure we can see the frequency will vary from 10KHZ to 200KHZ, which is too much for the design. The transformer core choice and the switching components requirements will increase. So a frequency limit control is really needed. Here, we set the switching frequency at hundred KiloHz during the balancing process. When the current hysteresis controller is working, the voltage difference will decrease and the switching frequency will be lower than 100KHz. Then the frequency limit control will work and it will limit the frequency at 100KHz. In this way, two parameters are under control at the same time, the primary side current and the switching frequency. Any of these two

131 115 parameters is working outside the limit, the controller will work and force it working under its limit. The current is between -3A to 3A and the frequency is no less than 100KHz. Because the voltage cannot be too large, if so we can consider the battery pack is damaged, there is no need to set the up limit of the frequency controller. Figure Hysteresis current control with frequency limit controller. As the figure shows, we do a frequency limit to the switching signal. We use a counter to calculate the switching signal. When it decreases to 100K, then it will change the on or off state to help limit the frequency. The result is shown in the next section.

132 DC DC Converter Performance Simulation Result In this section the different simulation results are discussed and shown. First, the hysteresis current control is presented in different voltage imbalance situation. Then the frequency limit control result is proposed. The system level simulation is done here by using the sup-capacitor to replace the battery cells to show a fast response. Finally, the efficiency and loss analysis of this converter is also presented here in this chapter. The battery cell imbalance situation is changing all the time and unpredicted. So to make simulation easy to implement and to help show the converter features directly. Some simplification and assumption is set here. In most cases, we just set on battery cell voltage lower than the others, which can help exclude the other difference influence and get the result we need. And for the system level control, we use the capacitor to replace the battery cells to show its charging procedure, which is not accurate but can shows the function of the converter Hysteresis Current Control Simulation Result and Analysis Here conditions for the voltage difference of series connected batteries are protein and unpredictable. To help solve this problem and make it intuitive and clear for us to understand the converter characteristic, we do some simplification here. We just set one of the battery cells lower than the others. The other 5 cell voltages are the same at 4V, which is almost fully charged.

133 117 Figure Balance current change when the voltage difference change (HCC). Figure Zoom in of the convertering moment. From the figure we can see the primary full bridge MOSFET peak current of HCC will be stable at about 3A when the imbalance voltage changes from 0.2V to 0.1V. When the

134 118 voltage difference changes, the controller will regulate the frequency of the control signal and turn on and off the MOSFET to make the current slide in the fix area. The average balance current will not change if there is only one cell unbalance. This is how hysteresis current control works when one cell imbalance condition changes. The average balance current will be stable and the frequency will be regulated based on the imbalance voltage difference. When the imbalance situation is worse, two or more cells have imbalance problem, the situation will be more complicated. To help simplify the analysis of the converter working principle, we firstly analyze the condition where only one battery cell voltage is lower than the others. We set the imbalance situation as 0.1V. The first battery voltage is 3.9V and the others are 4V. The balance current is

135 119 Figure Balance current under 0.1V voltage imbalance of HCC. We can find the battery whose voltage (SOC) is lower will have the balance current. The average balance current value is about 8A. The other will have no balance current because their voltages (SOC) are higher. This is the result we expect. If the imbalance situation is worse, just like two or three of them are imbalance, this active cell balancing converter will also work. The difference is that the balance current will distribute to those cells whose SOC is low. The lower the SOC is, the more balance current it will get. The amount of the balance will be distributed based on the amount of the voltage difference. For example, two battery cells voltage is lower than the other, one is 3.89V and another is 3.9V. The balance current will only charge theses two cells and the 0.11V cell will get more balance current.

136 120 The simulation result is shown below: Figure Balance current under 0.1V,0.11V voltage imbalance of HCC. We can find that the balancing current is almost separated. But the one whose imbalance is 0.11 is a little bit higher than the other one. The difference of 0.11 average balancing current is about 4.14A, the 0.1V one is about3.86. If the unbalancing situation becomes even worse, three or more cells need to balance. We can consider the converter will work in this way. Firstly, those cells whose voltage is higher than the others will receive no balance current. The others whose voltages are lower than the average one will get balance voltage based on its own voltage value. The one who

137 121 has lower voltage will get the higher balance current. The maximum balance current will not exceed 8A Frequency Limit Control Simulation Result and Analysis When the voltage imbalance becomes better and better, the frequency will decrease and be lower than 100KHz. Then the frequency limit controller will work. It will limit the frequency at the fix value at 100KHz. The simulation result is shown below. Figure Primary side current under 0.07V voltage imbalance.

138 122 Figure Primary side current under 0.05V voltage imbalance. Because we set the frequency limit at 100KHz and the frequency is around 100KHz when the voltage imbalance is 100mV, the frequency limit will work when imbalance voltage is less than 100mV. From the figure and figure 5.4.6, we can find the frequency is fixed when the voltage imbalance at 0.07V and 0.05V. They are both 100KHz. The only difference is the balance current value. The 0.07V imbalance balance current is a little bit higher than the 0.05V one. The frequency is limit, so if the voltage difference is higher, the inductor will transfer more energy and then the balance current is higher. This is reasonable. The frequency limit controller will work when the voltage imbalance is less than about 0.1V. The balance current will become smaller when the voltage imbalance is lower.

139 123 Above all, for the hysteresis current control with frequency limit, when the voltage imbalance is large enough, the controller will limit the balance current to set the average balance current no more than about 8A. When the imbalance situation becomes better, the balance current will be smaller and the controller will fix the switching frequency at 100KHz. When all the cells are balanced, there will be no balance current and the converter will be turned off. To help us understand the converter well, we use the capacitor to replace the batteries to do the system level simulation. To help get a fast response, we just set the capacitor 1000u. The initial voltage of these cells is set at 3.95V, 3.9V, 3.85V, 3.8V, 3.78V, 3.6V. The voltage difference of the battery cells in EVs will not be so much difference. We just want to see the converter working principle better to set the difference a little bit high. The simulation result is shown below Figure System level simulation by balancing the super capacitors.

140 124 From the result, we can see that this converter can balance the voltage difference very well by giving the lower voltage cell higher current and block off the balancing current for the high voltage cells The Efficiency and Loss Distribution Analysis. The main dissipate component here in this topology is MOSFET and the diodes. The loss for the MOFET is: P I R 2 loss rms * on And the loss of the diode is: P I * V loss ave F Figure Efficiency relation with the voltage difference.