The Development of a 48V, 10kWh LiFePO4 Battery Management ...
Transcript of The Development of a 48V, 10kWh LiFePO4 Battery Management ...
The Development of a 48V, 10kWh LiFePO4Battery Management System for Low Voltage
Battery Storage Applications.
by
Bartholomeus van Wyk Horn
Thesis presented in partial fulfilment of the requirements forthe degree of Master of Engineering (Electrical) in the
Faculty of Engineering at Stellenbosch University
Department of Electrical and Electronical Engineering,University of Stellenbosch,
Private Bag X1, Matieland 7602, South Africa.
Supervisor: Dr. P.J. RandewijkDr. J.M. Strauss
March 2017
Declaration
By submitting this thesis electronically, I declare that the entirety of the work containedtherein is my own, original work, that I am the sole author thereof (save to the extentexplicitly otherwise stated), that reproduction and publication thereof by StellenboschUniversity will not infringe any third party rights and that I have not previously in itsentirety or in part submitted it for obtaining any qualification.
Date: March 2017
©Copyright c 2017 Stellenbosch University All rights reserved.
i
Stellenbosch University https://scholar.sun.ac.za
Abstract
The Development of a 48V, 10kWh LiFePO4 BatteryManagement System for Low Voltage Battery Storage
Applications.B.V. Horn
Department of Electrical and Electronical Engineering,University of Stellenbosch,
Private Bag X1, Matieland 7602, South Africa.
Thesis: MEng (Elec)December 2016
Renewable energy sources are a promising replacement for fossil fuels in future energygeneration. To fully replace fossil fuels some form of energy storage is required. Chemicalbatteries offer an energy storage solution that is flexible and scalable for applicationsranging from electric vehicles to residential and even commercial applications.
Lithium-ion (Li-ion) battery technologies offers the most promising performance interms of energy density, power density as well as cycle life. Unfortunately, Li-ion batteriesare very sensitive to usage outside of the specified operating range. These specifiedparameters include the battery operating temperature, over- and undervoltage thresholdsas well as the maximum charge and discharge current. A Battery Management System(BMS) is thus required to monitor all of the above mentioned parameters and to ensurethe battery is operated safely and within the specified range.
A BMS’s primary focus is on the safety and protection of the battery, to minimise therisk of sudden failure and to maximise the life of the battery. The secondary function ofthe BMS is to perform battery diagnostics which could be used for more effective energymanagement of the battery.
The objective of this project was to develop a BMS to be used within a Li-ion batterypack for a micro electric vehicle. The developed BMS was used for battery testing.The battery results were used to estimate the battery parameters off-line according toa specific battery model. The estimated battery parameters can be used as a basis forfuture energy management purposes. An on-line parameter estimation algorithm wasalso developed. The algorithm was proven to be successful with a simulation. Futurework is required in order to simplify the practical implementation of the algorithm.
Another objective of this project was to development a solid state contactor (SSC)that can be used to disconnect the battery from a load. Mechanical contactors, whichpresents some disadvantages, are typically used for high current applications. The proof
ii
Stellenbosch University https://scholar.sun.ac.za
ABSTRACT iii
of concept SSC was proven to be an efficient though costly substitute to replace themechanical contactor within the BMS design.
Stellenbosch University https://scholar.sun.ac.za
Uittreksel
Die Ontwikkeling van ’n 48V, 10kWh LiFePO4 Battery BestuurStelsel vir Lae Spanning Battery Stoor Toepassings.
B.V. HornDepartement Elektries en Elektroniese Ingenieurswese,
Universiteit van Stellenbosch,Privaatsak X1, Matieland 7602, Suid Afrika.
Tesis: MIng (Elek)Desember 2016
Hernubare energie bronne is belowende opsies om fossiel brandstof bronne mee te vervang.Om fossiel brandstowwe volledig te vervang is daar een of ander vorm van energie storingnodig. Chemiese batterye bied so ’n oplossing wat buigsaam en maklik skaleerbaar isvir toepassings wat strek van elektriese voertuie tot residensiële en selfs kommersiëleverbruik.
Lithium-ioon battery tegnologië bied belowende verrigting in terme van energie digt-heid, drywings digtheid en lewens siklusse. Ongelukking is hierdie tegnologië baie sensitiefom buite die vervaardiger se spesifikasies bedryf te word. Hierdie spesifikasies sluit in diebattery temperatuur, oor- en onderspannings drumpel asook die maksimum battery laaien ontlaai stroom. ’n Battery monitor stelsel (BMS) word gebruik, vir hierdie rede, omal hierdie spesifikasies te meet en te verseker dit is binne die veilige venster van gebruik.
’n BMS se primêre doel is die beveiliging en die beskerming van die battery om die lewevan die battery te maksimeer. Die sekondêre doel van die BMS is om battery diagnosete doen vir meer effektiewe energie bestuur.
Die doel van hierdie projek is om ’n BMS te ontwikkel vir ’n mikro elektriese voertuig.The ontwikkelde stelsel was gebruik om battery toetse te doen. Die resultate was gebruikom die battery parameters van ’n battery model af te skat. Hierdie afgeskatte parameterskan in die toekoms gebruik word vir energie bestuur doelwitte.
’n Aanlyn parameters afskatting algoritme word ook in hierdie projek ontwikkel. Diealgoritme word slegs bewys deur behulp van simulasies. Verdere werk sal gedoen moetword om die algoritme aan te pas om in ’n praktiese stelsel geïmplementeer te kan word.
Die projek ondersoek ook die onwikkeling van ’n “solid state contactor” wat gebruikkan word om die battery van die las te ontkoppel. Meganiese kontaktors, wat nadele het,word tipies gebruik vir hoë stroom toepassings. Die “solid state contactor” poog om ’nopsie te wees wat die meganiese kontaktor in die hoof BMS ontwerp kan verwag.
iv
Stellenbosch University https://scholar.sun.ac.za
Acknowledgements
The author gratefully acknowledges the contributions of the following individuals andinstitutions:• My Mother and Father who have supported the author both financially and men-
tally.
• My two supervisors, Dr Randewijk and Dr Strauss, for all the time spent helpingto solve problems.
• The NRF who invested in the project.
• Mr Arendse for helping with some of the soldering work.
• All the post graduate students sitting in the power electronics laboratory includingAdel Coetzer.
• The Mellowcabs team for all the support during the project.
v
Stellenbosch University https://scholar.sun.ac.za
Contents
List of Figures viii
List of Tables x
Nomenclature xi
1 Project Overview 11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Project Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.3 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.4 Thesis Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Literature Study 52.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.2 Batteries: A Short Overview . . . . . . . . . . . . . . . . . . . . . . . . . 52.3 Battery Management Systems . . . . . . . . . . . . . . . . . . . . . . . . 102.4 Battery Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.5 Off-line Parameter Identification of an ECM . . . . . . . . . . . . . . . . 172.6 State Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.7 Recursive Least Squares method . . . . . . . . . . . . . . . . . . . . . . . 232.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3 Hardware design 253.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.2 Proof of Concept Battery Management System . . . . . . . . . . . . . . . 253.3 Full Scale Battery Management System . . . . . . . . . . . . . . . . . . . 303.4 Prototype Solid State Contactor . . . . . . . . . . . . . . . . . . . . . . . 403.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4 Software Design 474.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474.2 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474.3 Main loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484.4 Battery Balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494.5 Current sense loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514.6 Master control loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
vi
Stellenbosch University https://scholar.sun.ac.za
CONTENTS vii
5 On-line Parameter Estimation 545.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545.2 Battery model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545.3 Recursive Least Squares Algorithm . . . . . . . . . . . . . . . . . . . . . 575.4 Simulation and results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6 Results 666.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 666.2 Battery Management System . . . . . . . . . . . . . . . . . . . . . . . . 666.3 Battery Off-line Parameter Estimation . . . . . . . . . . . . . . . . . . . 736.4 Solid State Contactor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 796.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
7 Conclusion 857.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 857.2 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 857.3 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Bibliography 88
Appendices 92
A Calculations 93A.1 LM5017 regulator design . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
B Code 96B.1 Python USB listener . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96B.2 MATLAB Symbolic solver . . . . . . . . . . . . . . . . . . . . . . . . . . 97B.3 Noise filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98B.4 Battery balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
C PCB Schematics 102C.1 Full Scale BMS Schematic . . . . . . . . . . . . . . . . . . . . . . . . . . 102
Stellenbosch University https://scholar.sun.ac.za
List of Figures
2.1 Ragone diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.2 Diagram illustrating the discharge of a LFP cell . . . . . . . . . . . . . . . . 82.3 Requirement fulfilment of various cathode materials . . . . . . . . . . . . . . 92.4 Possible battery pack cell configurations . . . . . . . . . . . . . . . . . . . . 102.5 Typical passive cell balancing topology . . . . . . . . . . . . . . . . . . . . . 132.6 Internal resistance equivalent circuit model . . . . . . . . . . . . . . . . . . . 152.7 Single polarisation Thévenin equivalent circuit model . . . . . . . . . . . . . 152.8 PNGV equivalent circuit model . . . . . . . . . . . . . . . . . . . . . . . . . 152.9 Dual polarisation Thévenin equivalent circuit model . . . . . . . . . . . . . . 162.10 RC equivalent circuit model . . . . . . . . . . . . . . . . . . . . . . . . . . . 162.11 Typical discharge OC voltage test . . . . . . . . . . . . . . . . . . . . . . . . 182.12 Typical open circuit hysteresis . . . . . . . . . . . . . . . . . . . . . . . . . . 192.13 Battery voltage measured during off-line test . . . . . . . . . . . . . . . . . . 202.14 Block diagram of the ECM SOC Estimation Methods . . . . . . . . . . . . . 23
3.1 Proof of concept BMS circuit diagram . . . . . . . . . . . . . . . . . . . . . 263.2 Proof of concept BMS with the SEM battery . . . . . . . . . . . . . . . . . . 273.3 Voltage ripple . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303.4 Full scale BMS circuit diagram . . . . . . . . . . . . . . . . . . . . . . . . . 313.5 Full scale battery pack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.6 Digital model of the battery pack . . . . . . . . . . . . . . . . . . . . . . . . 333.7 Main control BMS PCB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343.8 TPS54060 voltage ripple . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353.9 Balance circuit diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363.10 Balancing PCB connected to cell terminals . . . . . . . . . . . . . . . . . . . 373.11 Current sensor circuit diagram . . . . . . . . . . . . . . . . . . . . . . . . . . 393.12 Current sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.13 Solid state contactor circuit diagram . . . . . . . . . . . . . . . . . . . . . . 433.14 Dead time design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443.15 Manufactured solid state contactor . . . . . . . . . . . . . . . . . . . . . . . 45
4.1 Main flow diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484.2 Cell balancing flow diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . 504.3 Current sensing flow diagram . . . . . . . . . . . . . . . . . . . . . . . . . . 514.4 Master flow diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.1 Single polarisation (SP) Thévenin model . . . . . . . . . . . . . . . . . . . . 55
viii
Stellenbosch University https://scholar.sun.ac.za
LIST OF FIGURES ix
5.2 Dual polarisation (DP) Thévenin model . . . . . . . . . . . . . . . . . . . . 565.3 Simulink simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 605.4 Battery discharge profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 615.5 Estimated Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625.6 Voltage estimation error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625.7 Estimates of the RLS algorithm compared to the actual parameters . . . . . 635.8 τt2 quantization error result . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.1 ADC error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 676.2 Voltage measurement noise comparison . . . . . . . . . . . . . . . . . . . . . 686.3 Current sensor noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 696.4 Filtered current sensor noise . . . . . . . . . . . . . . . . . . . . . . . . . . . 696.5 Current sensor thermal performance . . . . . . . . . . . . . . . . . . . . . . 706.6 Weak terminal connections . . . . . . . . . . . . . . . . . . . . . . . . . . . . 716.7 Cell balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 726.8 Impact of protection fuse during cell balancing . . . . . . . . . . . . . . . . . 726.9 Dual polarization Thévenin equivalent circuit model . . . . . . . . . . . . . 736.10 Pulse discharge test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 746.11 SOC vs OC voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 756.12 Ohmic resistance characteristic curve of the battery . . . . . . . . . . . . . . 756.13 Curve fit of dynamic behaviour . . . . . . . . . . . . . . . . . . . . . . . . . 766.14 Dynamic resistance characteristic curve of the battery . . . . . . . . . . . . . 766.15 Total battery resistance characteristic curve . . . . . . . . . . . . . . . . . . 776.16 Time constant τt1 characteristic curve . . . . . . . . . . . . . . . . . . . . . . 786.17 Time constant τt2 characteristic curve . . . . . . . . . . . . . . . . . . . . . . 786.18 SSC test set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 796.19 SSC current at maximum load . . . . . . . . . . . . . . . . . . . . . . . . . . 806.20 Current sense and reference pin at maximum load . . . . . . . . . . . . . . . 806.21 SSC temperature at maximum load . . . . . . . . . . . . . . . . . . . . . . . 816.22 SSC trip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 826.23 Current sense and reference pin . . . . . . . . . . . . . . . . . . . . . . . . . 836.24 Dead time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 836.25 SSC during turn-off . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
B.1 Noise of prototype BMS compared to that of the up-scaled BMS . . . . . . . 99B.2 Battery balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100B.3 Battery balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100B.4 Battery balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
C.1 Balance schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106C.2 Current sense schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Stellenbosch University https://scholar.sun.ac.za
List of Tables
3.1 LP55100100 cell specifications . . . . . . . . . . . . . . . . . . . . . . . . . . 263.2 LY-100AH cell specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.3 CSD19535KTT Power MOSFET specifications . . . . . . . . . . . . . . . . . 42
5.1 Simulink model parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
x
Stellenbosch University https://scholar.sun.ac.za
Nomenclature
List of symbolsV VoltA AmpereAh Ampere hourQ CapacityR ResistanceC CapacitanceL InductanceW Wattω Frequency
Abbreviations & AcronymsADC Analogue to Digital ConverterBMS Battery Management SystemCAN Controller Area NetworkCC Constant CurrentCPU Central Processing UnitCRC Cyclic Redundancy CheckCV Constant VoltageDC Direct CurrentDP Dual polarisationECM Equivalent Circuit ModelEIS Electrochemical Impedance SpectroscopyEMI Electromagnetic InterferenceESD Electro Static DischargeEV Electric VehicleFET Field-Effect TransistorFF Forgetting FactorFTDI Future Technology Devices InternationalHPPC Hybrid Pulse Power CharacterisationIGBT Insulated-Gate Bipolar Transistor
xi
Stellenbosch University https://scholar.sun.ac.za
NOMENCLATURE xii
I2C Inter-Integrated CircuitIIR Infinite Impulse ResponseLCO Lithium Cobalt OxideLFP Lithium Iron PhosphateLi LithiumLMO Lithium Manganese OxideLNMC Lithium Nickel Manganese Cobalt OxideLNCA Lithium Nickel Cobalt Aluminium OxideLS Least SquaresLTO Lithium TitanateMCU Micro Controlling UnitMOSFET Metal-Oxide-Semiconductor Field-Effect TransistorNiMH Nickel-Metal HydrideNTC Negative Temperature CoefficientOC Open CircuitOV Over VoltagePNGV Partnership for a New Generation of VehiclesRLS Recursive Least SquaresSC Short CircuitSEM Shell-Eco MarathonSOC State Of ChargeSOH State Of HealthSOP State Of PowerSP Single PolarisationSPI Serial Peripheral InterfaceSSC Solid State ContactorTI Texas InstrumentsTVS Transient Voltage SuppressorUDDS Urban Dynamometer Driving ScheduleUSB Universal Serial BusUV Under Voltage
Stellenbosch University https://scholar.sun.ac.za
Chapter 1
Project Overview
1.1 IntroductionFossil fuels are currently used to generate electricity and to provide fuel for transportation.It has the advantage that the energy is stored in chemical bonds which can be utilisedwhenever necessary and has a very high energy density compared to other storage options.
Renewable energy sources are a promising replacement for fossil fuels, since no green-house gasses are released during the energy generation process. The cost of renewabletechnologies has decreased significantly [1] and the need for fossil fuel replacements areincreasing as the effects of global warming are becoming more and more apparent. Un-fortunately, these sources come with other difficulties.
One such difficulty is that renewable energy sources require the energy generated tobe utilised immediately after generation. A significant drawback with most renewabletechnologies is that they cannot continuously supply a baseload. Without large energystorage devices connected to renewable energy sources, the energy generated is eitherunpredictable (wind power) or it is periodical (solar power). Thus, some sort of storagedevice is required to supply power while power fluctuations in the renewable sources arepresent.
Energy storage options that are currently available for commercial implementationinclude hydro systems, thermal energy storage and chemical batteries. Unfortunately,hydro systems are depended upon the location of the system. Thermal energy storageis typically used at concentrated solar power plants where energy is stored thermallyin a storage medium. The thermal energy is converted at a later stage to electricalenergy by means of a turbine. Unfortunately, this technique is not ideal for renewableenergy sources that directly generates electrical power, due to the inefficient processof converting thermal energy into electrical energy. It is clear that none of the abovementioned technologies can be practically implemented for small scale energy storagesuch as required by electric vehicles or residences. Chemical batteries offer an energystorage solution that is flexible and scalable for applications ranging from electric vehiclesto residential and even commercial energy storage.
Currently battery storage options are still expensive, but the price is gradually de-creasing. This decrease can be attributed to more effective manufacturing processes aswell as economies of scale. The battery price is moving towards the point where it isbecoming a viable solution for either electric vehicles (EV) or peak power shaving. The
1
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 1. PROJECT OVERVIEW 2
latest research on battery technologies are promising. The current commercially availablebattery technologies, in terms of performance, are good enough to replace fossil fuels asan energy storage solution.
Lithium-ion (Li-ion) battery technologies offers the most promising results in terms ofenergy density, power density as well as cycle life. Li-ion battery technologies include awide range of different battery chemistries, all with the similarity that Li-ions are used tocarry the positive charge between the two electrodes of the battery. Li-ion batteries havemany advantages compared to batteries with other chemistries, which will be discussedin more detail in Chapter 2.
Unfortunately, Li-ion batteries are very sensitive to being used outside of its specifiedoperating range. These specifications include the battery operating temperature, over-and undervoltage thresholds as well as the maximum charge and discharge current. Thisis due to the fact that Lithium is a very reactive element. The Li-ion cells can potentiallyignite or even explode if it is used outside of its safe operating range specified by themanufacturer [2]. A Battery Management System (BMS) is required, for this reason,to monitor all of the above mentioned specifications and to ensure that the battery isoperated safely within the specified range.
The objective of this project is to develop a BMS for a Li-ion battery pack that willbe used on a micro electric vehicle (EV). This technology can also be applied to therenewable energy storage sector since the research of Li-ion battery storage for EVs aremore advanced than for renewable storage.
1.2 Project MotivationIn an attempt to reduce the high investment cost of a Li-ion battery it is crucial tomaximise the lifespan of the battery. One of the easiest ways to maximise the battery’scycle life is to use it within the specified operating range as indicated by the manufacturer.All the different aspects requires to be monitored continuously to ensure the battery isoperated within the specified range. This is achieved by using a BMS to monitor thebattery pack. The different cell voltages and temperatures, as well as the battery currentare measured by the BMS to ensure it is within the specified range of operation. TheBMS for instance disconnects the battery from the load if the resulting measurementsare not within the specified range of operation, effectively protecting the battery. Thisincreases battery performance by minimising the physical degradation of the battery.A BMS primary focus are therefore on the safety and the protection of the battery, tominimise the risk of sudden failure and to maximise the life cycle of the battery.
The secondary function of the BMS is to perform battery diagnostics, such as stateof charge (SOC) estimation, state of health (SOH) estimation and state of power (SOP)estimation. SOC is the amount of charge the battery has at a certain time, i.e. howmuch of the total energy capacity is available for usage. SOH refers to the current batterycapacity compared to the original capacity specified by the manufacturer. SOP refers tothe maximum amount of power that can be delivered by the battery at a specific time.These states cannot be measured directly, but can be estimated. Accurate estimationsof these states are very important for effective energy management. In order to estimatethese states accurately, a model is required to describe the dynamics of the battery.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 1. PROJECT OVERVIEW 3
A great variety of battery models exists. This project will investigate and select theoptimal model in terms of accuracy and complexity. Once the optimal model is selected,the parameters of the model will be estimated, which can in turn be used for energymanagement purposes.
1.3 Research ObjectivesThe motivations for this project as discussed above gives rise to the following objectives:
• Development of a proof of concept Li-ion BMS to demonstrate the basic workingof the BMS. This includes the design, manufacturing and testing of the proof ofconcept BMS.
• Development of a full scale Li-ion BMS for the purpose of a micro EV. This includesthe design, manufacturing and testing of the full scale BMS. The aim is to prolongthe life of the battery by ensuring the battery is operated in a safe and sustainableway.
• The successful implementation of a balancing algorithm within the BMS to max-imise the available capacity of the battery pack.
• Off-line parameter estimation using the BMS.
• The development of an on-line parameter estimation algorithm using the appropri-ate battery model. The algorithm will be proved by a simulation.
• Development of a proof of concept Solid State Contactor (SSC) to protect thebattery against overcurrent and short circuit conditions. This includes the design,manufacturing and testing of the proof of concept SSC.
1.4 Thesis Structure• Chapter 2: Literature Study
The relevant literature concepts and topics are discussed. The study describes,amongst others, related works, a short battery overview, BMSs, battery modellingand parameter estimation algorithms.
• Chapter 3: Hardware DesignThe hardware design of a proof of concept BMS is firstly discussed within thischapter. The design choices made to up-scale the system to a full scale BMS for anEV follows. Finally the design of a prototype solid state contactor are discussed inthis chapter.
• Chapter 4: Software DesignThe software design of the BMS is discussed in this chapter. This includes thedifferent monitoring loops and the balancing algorithm.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 1. PROJECT OVERVIEW 4
• Chapter 5: On-line Parameter EstimationAn on-line parameter estimation algorithm is investigated within this chapter. Asimulation is implemented to prove the accuracy of the algorithm.
• Chapter 6: ResultsThe test results of the designed proof of concept BMS, full scale BMS and proof ofconcept SSC are discussed in this chapter.
• Chapter 7: Conclusion and Future WorkA conclusion of the work presented within this thesis is discussed in this chapter aswell as some ideas for future work.
Stellenbosch University https://scholar.sun.ac.za
Chapter 2
Literature Study
2.1 IntroductionLi-ion batteries have the potential to significantly impact the energy storage sector. Thischapter discusses the history and operation of batteries in general but with a primaryfocus on Li-ion batteries. The surrounding concepts and principles used to monitorand manage batteries are also discussed within this chapter. It includes BMSs, batterymodelling, parameter estimation, state estimation and the recursive least squares method.
2.2 Batteries: A Short Overview
2.2.1 Introduction
The earliest forms of batteries are found in ancient civilisations, such as the Egyptiansand Parthians, where it was typically used for electroplating purposes, not storing energy.Batteries used for energy storing purposes were first investigated in the late 17th century.Alessandro Volta discovered in the year 1800 that two different metals joined together bya moist intermediary would generate a flow of electrical power when the two metals areconnected by a conductor. This discovery led to the invention of the first voltaic cell. Abattery is simply a set of cells in series.
In the two centuries that have pasted since, various new technologies have been devel-oped. The basic operation is however the same for all of these technologies, ranging fromnon-rechargeable to rechargeable batteries. Also, these technologies served an integralpart of society but because of the low specific energy and low specific power constraints,was never used as a mainstream energy storage medium. Specific energy (Wh/kg) is thenominal battery energy per unit mass, sometimes referred to as the gravimetric energydensity while specific power (W/kg) is the maximum available power per unit mass.
In 1912, experimentation started on Li-ion batteries. Only in 1970 was the first Li-ion batteries made available to the open market. It is clear that the development ofbatteries is slow and comes with a number of difficulties. In the past 40 years significantimprovements have been made to the initial Li-ion technology. Li-ion batteries nowadayshave the potential of high specific power and high specific energy.
The following section discusses the basic operation of batteries in more detail and alsoprovides a comparative study in terms of the performance of the commercially available
5
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 6
battery technologies.
2.2.2 Battery Operation
Batteries store electrical energy in chemical bonds. All batteries makes use of electro-chemical reactions referred to as reduction and oxidation reactions [3]. Reduction is thegain of electrons by a molecule, atom, or ion. Oxidation is the loss of electrons by amolecule, atom, or ion. A battery cell has three essential components: the anode, thecathode, and the electrolyte. The materials used for each of these components determinethe battery’s characteristics, but the basic working principle of a cell stays the same.
The anode and cathode materials are chosen in such a way that the anode donateselectrons, and the cathode accepts electrons. The tendency of a material to donate oraccept electrons is commonly expressed as the material’s reduction potential. Reductionpotential is measured in Volts (V). Each material has its own intrinsic reduction potential.The higher the potential, the greater the material’s tendency to be reduced.
The difference between the reduction potentials of the cathode and the anode deter-mines the nominal operating voltage of the cell. The anode and cathode are separatedby the electrolyte, which is typically a liquid or gel that conducts electricity. When theanode and cathode are connected to each other through a conductor, the anode undergoesa chemical reaction with the electrolyte in which it loses electrons, creating positive ions(oxidation). The positive ions flows through the electrolyte to reach the cathode. At thecathode, the positive ions and electrons reacts with the cathode (reduction). Togetherthe entire process is known as a redox reaction.
Lithium is the metal with the lowest density [4], the greatest reduction potential andthe highest energy-to-weight ratio. Therefore, it is at the forefront of battery technologiesand it will contribute extensively to the future of energy storage on a large scale. In reality,Li-ion batteries have a much higher energy density than other common rechargeable bat-teries, such as a nickel-metal hydride (NiMH) or lead-acid batteries. The Ragone diagramcharacterises the specific power and specific energy of the different battery chemistries.It is used to compare the different technologies and can be seen in Figure 2.1.
The Li-ion batteries are not displayed as a continuous range on the Ragone diagrambecause different cell technologies are used in energy storage cells compared to powerstorage cells. Some Li-ion technologies have the highest specific energy (Wh/kg) whileother Li-ion technologies have the highest specific power (W/kg). Other Li-ion technolo-gies tries to find a good balance between the two. It is important to note that typicallythere is a trade off between the specific energy and specific power, e.g. a lead-acid bat-tery’s capacity is greatly influenced by the speed at which the power is extracted from thebattery. This is one of the reasons why lead acid cells is not ideally suited for EV appli-cations. EV applications require a high specific energy, but also a relatively high specificpower. The combination of both these properties is one of Li-ion’s greatest advantages.The technology is, at the moment, expensive compared to other chemistries. This willchange in the future as the scale of production of Li-ion batteries increases which willlead to price decreases.
Li-ion batteries usually uses a mixture of different cathode and anode materials tocomplement the advantages and disadvantages of the respective materials. The redoxreaction for charging and discharging a Lithium Iron Phosphate (LFP) cell can be seen
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 7
Specificpo
wer,c
elllevel
[W/k
g]
Super-capacitor
Specific energy, cell level [Wh/kg]
1
10
100
1000
10000
100000
0 20 40 60 80 100 120 140 160 180 200
Lead-acidNi-Cd Ni-MH Na-NiCL
Lead-acidspirally wound
Li-ionvery high power
Li-ionhigh power
Li-ionhigh energy
Li-polymer
Figure 2.1: Ragone diagram [5]
below with a typical example of the operation of Li-ion batteries.
Discharge:
Anode: LiC6 −→ C6 + Li+ + e− (2.2.1)
Cathode: FePO4 + Li+ + e− −→ LiFePO4 (2.2.2)
Charge:
Anode: LiFePO4 −→ FePO4 + Li+ + e− (2.2.3)
Cathode: C6 + Li+ + e− −→ LiC6 (2.2.4)
A discharge diagram of a LFP cell can be seen in Figure 2.2. The oxidation andreduction reactions can be seen at the anode and cathode respectively. The positivedenoted terminal is at the cathode while the negative denoted terminal is at the anode.
During the charge cycle of a cell the redox reactions are reversed from that of thedischarge cycle. Oxidation takes place at the positive terminal and reduction at thenegative terminal. The positive terminal becomes the anode and the negative terminalbecomes the cathode.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 8
Electrolyte
CathodeFePO4 + Li+ + e− → LiFePO4
e−
Li+
AnodeLiC6 → C6 + Li+ + e−
- +
Figure 2.2: Diagram illustrating the discharge of a LFP cell [6]
The most common material used for the anode is some form of carbon. Carbonmaterials have the advantage that their mechanical and electrical properties are not sig-nificantly affected by accepting or donating large amounts of lithium. Carbon is typicallyused in some form of graphite, but other materials do exist for example Lithium Titanate(LTO).
The chosen cathode material varies greatly for the different cell technologies. It ischosen according to the requirements of the application. Typical examples of currentlyused positive electrode materials is shown in Figure 2.3. The main determining factorsof batteries are performance, lifetime, safety, cost, power and energy density. The deter-mining factors of various battery technologies is also shown in Figure 2.3. The differentbattery technologies are rated according to these different factors. Generally, the batterytechnology is named after its positive electrode’s composition.
Starting from the left, Lithium Iron Phosphate (LFP) is well balanced with high safetyand cycle life. LFP has a relatively low specific energy because of its low voltage plateau.
Lithium Cobalt Oxide (LCO) is a chemistry that is typically used in consumer elec-tronics. Drawbacks include low thermal stability and relatively low cycle life.
Lithium Nickel Manganese Cobalt Oxide (LNMC) is widely used in EVs because it isvery well balanced and has a high specific power.
Lithium Manganese Oxide (LMO) shows high safety performance at a relatively lowcost since no expensive metals are used. Unfortunately, it is very sensitive to high tem-peratures and it has a relatively low specific power.
Lithium Nickel Cobalt Aluminium Oxide (LNCA) has the highest specific energy andspecific power. Unfortunately, it lacks in terms of safety and cost.
The variations on all of these battery technologies can differ greatly to result in a
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 9
LMO
LFP
LNMC
LCO
LNCA
Energy density
Costs
Safety
Cycles
Powerdensity
SOC-operatingrange
Figure 2.3: Requirement fulfilment of various cathode materials [7]
battery with the needed requirements for a specific application. There is no clear optimalsolution for battery chemistries, only chemistries that suite certain applications betterthan others.
Despite all the advantages it has, Li-ion technologies do have some disadvantages. Li-ion batteries degrade over time that leads to an increase in the internal resistance, whichdecreases the battery’s ability to deliver power. It is also susceptible to a number of otherpotential problems including oxygen production due to overcharging at the cathode andoverheating of the anode. These potential problems can lead to battery degradation orin worst-case scenarios, ignite the battery. Thus, it is crucial for Li-ion batteries to havea BMS to prevent these problems.
This section, thus far, has given a short review of batteries by discussing batteryoperation and comparing the different battery technologies. The rest of this sectiondiscusses different battery configurations since this greatly influences the specificationsof the system monitoring the battery.
2.2.3 Battery Configuration
Battery packs are a combination of cells. The voltage of a single cell is typically lowcompared to the demanded voltage for various applications. Cells are stacked in series todeliver a higher voltage to a load. This reduces the current, which minimises losses. Thecells can be configured in various different configurations as shown in Figure 2.4.
The arrangement of cells is dependant upon the following factors:
• The mechanical layout of the battery pack for structural integrity.
• The wiring harness connecting the different cells.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 10
a) b) c) d)
Figure 2.4: Possible battery pack cell configurations: a) series, b) matrix, c) series-paralleland d) mixed
• Temperature management. How to best regulate the battery temperature.
• Robustness of the battery pack. By connecting the cells in a series-parallel configu-ration, if one cell fail, that series string could be disconnected by a contactor. Thiswould ensure that the rest of the pack could still be used.
• Li-ion cell voltages need to be operated within the specified range of operation forsafety purposes and to prolong the cell’s lifespan. To ensure this is the case, all thecell voltages needs to be monitored continuously. Connecting the cells in a matrixconfiguration reduces the amount of cells that needs to be monitored, effectivelysimplifying the BMS.
These various configurations have different advantages and disadvantages and wouldtypically depend upon the application of the battery. The battery configuration alsoinfluences the choice of BMS used to monitor it. The following section investigates BMSsin more detail.
2.3 Battery Management Systems
2.3.1 Introduction
The first priority of a BMS is to monitor the health of all the cells in the battery pack,while still being able to deliver the power required by the application. In order to prolongthe the life of the battery pack the BMS needs to maintain all the cells within the specifiedoperating range. In this section an overview of BMSs are presented. This includes BMSrequirements, architectures and battery balancing techniques.
2.3.2 Requirements
The BMS is an integral part of any large scale battery pack and is generally responsiblefor:
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 11
• Cell protection: Safety is the first priority of the BMS. Protecting the cells fromoperating outside of the manufacturer’s specified operating conditions are crucial.The rest of the system also needs to be protected from the battery in the event ofbattery failure.
• Data acquisition: Battery current, voltage and temperature measurements.
• Data analysis: State of power, state of health and state of charge estimation.
• Control: Charge and discharge current control to ensure it is within the manufac-turer’s specified operating conditions.
• Communication: Used to interact with other components in the system, e.g. thecharger or inverter. Typically also used to give users access to the data of thebattery.
• Battery balancing: Ensures that the maximum capacity of the battery is availablefor use.
2.3.3 Architectures
There are a variety of different BMS architectures. All of them have their advantagesand disadvantages. The architectures are typically set apart by the scalability and costof each system [8]. Some of the BMS architectures are presented next.
Centralised BMS Architecture: A centralised BMS architecture uses one main con-trol board that monitors all the different aspects of the battery. This architecture has theadvantage of low cost and easy implementation when used with a battery with a low cellcount. Disadvantages include large wiring harnesses as the size of the battery increases,since the main board needs to be connected to all the different cells and temperaturesensors. This also increases the complexity of the system as the cell count increases.Typically this architecture is ideally suited for battery packs with low cell counts.
Distributed BMS Architecture: A distributed BMS architecture has a node on eachcell which monitors the voltage and temperature of that specific cell. This node is a slavedevice which are connected to a master via a serial connection. All the different nodescommunicate its measurements via the serial connection to the master controller. Themaster interprets this data and controls the output of the battery accordingly. This archi-tecture has the advantage that it is very easily scalable and easy to install. Unfortunately,the cost of a distributed system can be relatively high compared to a centralised system.
Modular BMS Architecture: A modular BMS architecture is a combination of thecentralised and distributed architectures. It uses a set of slave devices that monitorsmore than one cell each. These slave devices are typically connected on top of each other,through a daisy chain communication interface, to monitor a large number of cells. Themodular BMS architecture is flexible and scalable. The slaves are controlled by a mastercontroller. The modular architecture delivers a good trade-off between the centralised
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 12
and distributed architectures. It is generally used for batteries containing a high numberof cells such as the batteries used for electric vehicles.
2.3.4 Battery Balancing
Cell balancing is used to ensure that the state of charge of each of the series connectedcells is even. The inconsistencies in the manufacturing of battery cells result in uniqueperformance characteristics for each individual cell in a battery pack. For this reasoncells accept and deliver charge at a slightly different efficiency and their overall capacitydiffers slightly. This small difference in efficiency and capacity results in one cell havinga higher SOC than another cell, even though the current through the series connectedcells are the same. These differences are aggravated during extensive use of a batterypack. The difference between the SOC of the individual cells continue to increase whichin turn lowers the overall capacity of the battery pack as a whole, since the battery packwill operate at the level of the weakest cell. The unbalanced state of the battery packcauses some of the cells to be operated in an overvoltage or undervoltage state which willsignificantly decrease the life cycle of the battery pack. It is therefore very importantthat the cells are balanced properly. There are two battery balancing techniques thatwill be discussed in the following subsections: passive and active cell balancing.
Passive Balancing: Passive balancing makes use of resistors to remove charge fromthe cells with higher cell voltages than the rest in the series string, until all the cellvoltages equal each other. The advantage is that the system has a relatively low costand complexity. The drawback of this method is that it is energy inefficient, since all theenergy removed from the higher cells are dissipated in resistors.
Passive balancing typically only balances the battery pack during charging. Duringdischarge, the whole battery pack’s capacity is constrained by the cell with the lowestcapacity in the series connected string. Balancing during discharge is not desirable sincethe energy is wasted.
A typical passive cell balancing topology is shown in Figure 2.5. Cell balancing ofCell1 can be achieved by closing switch S1 and S2. The resistance R1 determines thebalancing current and effectively the speed at which balancing is performed. ChoosingR1 depends on the cell chemistry (nominal voltage), the manufacturer’s tolerance of thecells and the application of the battery.
Active Balancing: Active balancing removes charge from a cell with a high state ofcharge and delivers it to a cell with a lower state of charge. Transformers, inductors orcapacitors are used as the active component to store the charge from the higher stateof charge cells and then deliver it to the lower state of charge cells. Switching circuitsare used in combination with these active components to effectively spread the chargefrom one cell to another. Active cell balancing has the advantage over passive cell bal-ancing that cells can be balanced during charging as well as during discharging. Thiseffectively increases the capacity of the battery pack. Another advantage of active cellbalancing, compared to passive cell balancing, is its high energy efficiency since energy isnot dissipated within the resistors. The active cell balancing circuits are more expensiveto build and more complex to control than passive cell balancing. The overall complexity
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 13
−Cell3+
S3
R3Voltagemonitor
−Cell2+
S2
R2Voltagemonitor
−Cell1+
S1
R1Voltagemonitor
Figure 2.5: Typical passive cell balancing topology
of active cell balancing also reduces the total reliability of the system. There are manydifferent active cell balancing topologies, but they will not be discussed in detail sincepassive balancing was chosen for this thesis due to the complexity of active balancing.
The following section discusses the different battery modelling techniques.
2.4 Battery Modelling
2.4.1 Introduction
Battery modelling is an effective way in which to manage the battery by estimating theSOH, SOP and the SOC. Accurate estimations of these states are very important sinceit cannot be measured directly. In order to estimate these states accurately a model isneeded to describe the dynamics of the battery. Research on electric vehicles has increaseddramatically and therefore a wide range of different battery models exists. These modelscan typically be split into three different categories, which include the equivalent circuitmodel, first principle model and the empirical model. These models are discussed in thefollowing section.
2.4.2 First Principle Model
The first principle model are usually the first choice for battery analysis due to its accuraterepresentation of a battery’s physical properties. It models the battery as simply acollection of atoms bound together by electrochemical reactions. These reactions are theinteractions between electrons, which can be described by the basic laws of physics. Thefirst principle model attempts to analyse the battery through the atomic number and massof the battery’s chemical elements. The different physical characteristics of the batteryare modelled though several electrochemical models. Although these models can achievehigh levels of accuracy, they are computationally complex and time consuming becauseof the involved partial differential equations. The first principle models are not suitablefor control-oriented or real-time applications. It is typically used to investigate and
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 14
understand the physical properties of batteries in order to design new battery materials.The first principle model is not, for these reasons, discussed in further detail in this thesis.
2.4.3 Empirical Model
The empirical model uses data-based methods to model the dynamic behaviour of abattery. These models rely entirely on experimental data sets and have no insight intothe physical electrochemical processes within the battery. A large amount of empiricalmodels has been proposed for various purposes. The Peukerts formulation being the mostcommonly used battery model. It was expanded to capture the non-linear relationshipbetween the battery’s capacity and the discharge current under different temperatures [9].Other more intelligent data-based models include, neural networks [10] and fuzzy logic[11], which is typically used to estimate the states of batteries. These models require largeoff-line training data to estimate the model’s parameters. These models have shown highlevels of accuracy. The downside is that training has to be acquired off-line. The trainingdata could also prove to be a constraint. New applications require new training data inorder to determine how the battery will react to a specific application. This makes themodel’s accuracy directly proportional to the quality of the training data [12]. In addition,the empirical models always rely on a large amount of experimental observations, whichrequire a large amount of experimental and modelling.
2.4.4 Equivalent Circuit Models
The Equivalent Circuit Model (ECM) is a good trade off between the empirical and firstprinciple models. It delivers some physical insight into the battery as well as a goodcontrol-orientated basis. A large variety of ECMs exist and will be discussed in thesections below. These models use electric circuits to mimic the dynamic electrochemicalprocesses of a battery. The accuracy of the model typically depends upon the number ofelectrical elements used. Parameters in the model in turn depend on many factors, suchas the SOC, the temperature and the SOH. The presented models disregard self-dischargeand other factors with long time constants.
Internal Resistance model: The internal resistance model is the most simplistic ECMavailable and is shown in Figure 2.6. It consists of an ideal voltage source VOC and anequivalent series resistor Rs. The voltage source is used to model the open circuit voltageand the resistor is used to model the ohmic resistance of the battery.
Single Polarisation Thévenin model: The Single Polarisation (SP) Thévenin modelis the most commonly used battery model and is shown in Figure 2.7. It consists of anideal voltage source VOC , series resistor Rs and a parallel resistor-capacitor Rt1-Ct1 pair.Again,the voltage source is used to model the open circuit voltage. The resistor Rs modelsthe ohmic resistance and the parallel RC pair models the polarization voltage.
Partnership for a new Generation of Vehicles linearised model: The Partner-ship for a New Generation of Vehicles (PNGV) linearised model is used in the "Free-domCAR Battery Test Manual" which is a result of the FreedomCAR program which
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 15
−VOC+
Rs
+
−
V (t)
I(t)
Figure 2.6: Internal resistance equivalent circuit model
−VOC+
Rs
Rt1
Ct1
Vt1+ −+
−
V (t)
I(t)
Figure 2.7: Single polarisation Thévenin equivalent circuit model
−VOC+
RsCpb
+Vpb
−
Rp
Cp
Vp+ −+
−
V (t)
I(t)
Figure 2.8: PNGV equivalent circuit model
ended in 2003. The program set out to lay the groundwork for more energy efficientand environmentally friendly highway transportation technologies and was funded by theAmerican government [13]. The PNGV model is based on the SP Thévenin model andis shown in Figure 2.8. Again, the voltage source VOC is used to model the open circuitvoltage, Rs models the ohmic resistance and the parallel RC (Rp-Cp) pair models thepolarization voltage. Capacitance Cpb models the cumulative open circuit voltage changewith respect to current I(t).
Dual Polarisation Thévenin model: The polarisation characteristic of a Li-ion bat-tery consists of the concentration polarisation and the electrochemical polarisation. Inthe SP Thévenin model both these are modelled by one parallel RC pair. Unfortunately,this one pair can only model it to a certain extent. The Dual Polarisation (DP) Thévenin
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 16
model adds an RC parallel pair to the SP Thévenin model as shown in Figure 2.9. Thisexpands the modelling of the polarisation characteristics of the battery, which increasesthe model accuracy.
The DP Thévenin model is an improved SP Thévenin model and consists of a voltagesource (VOC), three resistors (Rs, Rt1, Rt2) and two capacitors (Ct1, Ct2). Again,thevoltage source VOC is used to model the open circuit voltage and Rs models the ohmicresistance. The first parallel RC (Rt1-Ct1) pair models the electrochemical polarizationand the second parallel RC (Rt2-Ct2) pair models the concentration polarization.
−VOC+
Rs
Rt1 Rt2
Ct1
Vt1+ −
Ct2
Vt2+ −+
−
V (t)
I(t)
Figure 2.9: Dual polarisation Thévenin equivalent circuit model
RC model: The RC model shown in Figure 2.10 was developed by the battery manu-facturer SAFT [14]. It comprises of two capacitors (Cb, Cc) and three resistors(Re, Rc,Rt). The capacitor Cb models the capacity of the battery and is therefore extremelylarge. The capacitance Cc represents the dynamic behaviour and is typically relativelysmall. The resistance Rt is known as the terminal resistance, Re is the end resistance andRc is the capacitor resistance.
Cb −Vb+
Re
Rc
Cc+
Vc−
Rt
+
−
V (t)
I(t)
Figure 2.10: RC equivalent circuit model
The following section discusses the different methods of estimating the parameters foran ECM.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 17
2.5 Off-line Parameter Identification of an ECM
2.5.1 Introduction
ECMs are commonly used to model battery behaviour. Accurate estimations of thebattery model parameters have a direct correlation with the model’s performance. In thissection, an overview is given on the methods of off-line ECM parameter identification.
2.5.2 Open Circuit Voltage Identification
Determining an accurate estimation of the open circuit voltage is a crucial part of theimplementation of most ECMs. The open circuit (OC) voltage VOC is present in all butone ECM (RC) model. The OC voltage is typically measured as a function of the SOCand temperature of the battery. If the battery is not used for an extended period of time,the battery terminal voltage will equal the electrochemical potential of the battery, thusreaching equilibrium. This is the potential required to balance the difference betweenthe anode and cathode’s ability to gain or lose electrons. This state is also referred to asthe OC voltage. The OC voltage is determined at set intervals of the SOC and constanttemperature. The relationship between the SOC and OC voltage is not linear and is thusinterpolated by a polynomial.
A typical procedure to determine the OC voltage is as follows: the battery is fullycharged with a Constant Current-Constant Voltage (CC-CV) profile according to themanufacturer’s specification. The CC-CV charging profile is achieved by charging thebattery at constant current (CC) until the battery reaches its maximum voltage. Thebattery is then charged to maintain this maximum voltage until the charge current de-creases below the fully charged specification as detailed by the manufacturer. A restperiod follows to allow the battery to reach its equilibrium potential. The battery is thendischarged at constant current to 90% of its nominal capacity. Typically, the manufac-turer’s rated discharging current will be used as the reference for the amount of dischargecurrent. Another rest period is then inserted during which time the battery will againreach its equilibrium potential. This cycle of discharging (10%) and inserting a rest pe-riod continues until one of the cells reaches its minimum voltage which marks the end ofthe OC voltage test.
An example of the typical results obtained from this testing procedure is shown inFigure 2.11. The same process can also be followed to determine the OC voltage of thebattery whilst the battery is charging, however the battery must be drained beforehandto 0% capacity. The battery has to be charged with the same amount of current usedwith the discharging method. The battery is charged at constant current charging cycles,similar to the discharge cycles, where the battery is charged 10% and given a rest periodto reach its equilibrium potential. This is repeated until one of the cells reaches itsmaximum voltage and the charge OC voltage test is completed.
It is worthy to note that the charge and discharge test do not compare perfectly,since the equivalent potential of the battery is dependant upon the previous usage of thebattery [15]. An exaggerated example of the OC voltage hysteresis during charging anddischarging is shown in Figure 2.12. This non-linearity of the battery is caused by the
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 18
0 5000 10000 15000Time [s]
42
44
46
48
50
52
Volta
ge[V
]VoltageOC voltage
Figure 2.11: Typical discharge OC voltage test
electrochemical reactions within the battery. The magnitude of the hysteresis is typicallydependant upon the cell chemistry.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 19
0 20 40 60 80 100State of charge [%]
42
44
46
48
50
52
Ope
nci
rcui
tVol
tage
[V]
DischargeCharge
Figure 2.12: Typical open circuit hysteresis
2.5.3 Time Domain Parameter Identification
Simple time dependant testing can be used to characterise a battery’s behaviour. This isparticularly important to enable ECM parameter fitting. In addition to the OC voltagetest detailed in the previous section, there is the Hybrid Pulse Power Characterisationtest (HPPC) that is used to estimate the characteristics of the current dependent ECMparameters. A HPPC test consists of a high current pulse charging and discharging thebattery for for a short duration of time. Normally, pairs of equal magnitude dischargeand charge current pulses are applied at different SOC operating points of the battery.Applying a HPPC test at different SOC operating points leads to a dynamic ECM whichdescribes the battery behaviour over the full SOC range of the battery. Variations of thistest can be applied in order to suit the application of the battery.
The parameters and OC voltage of an ECM can be identified simultaneously as de-tailed in [16]. A typical example of the test is shown in Figure 2.13. The voltage Vs is thepeak value of the linear part of the curve while Vt is the voltage value of the exponentialpart of the curve.
The voltage Vs is used to calculate the ohmic resistance of the battery pack. It canbe calculated by
Rs =VsIbat
, (2.5.1)
where Ibat is the battery current just before the rest period starts.The voltage Vt is used to estimate the dynamic behaviour of the battery. Curve fitting
is typically used to estimate an appropriate time constant or time constants to obtainthe ECM parameters. Curve fitting is only applied to the exponential part of the curve.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 20
0 5000 10000 15000Time [s]
42
44
46
48
50
52
Volta
ge[V
]
Vs
Vt
Rest period
Figure 2.13: Battery voltage measured during off-line test
The curve fit in combination with the battery current, just before the rest period starts,contain enough information to estimate the polarisation characteristic of the battery.
2.5.4 Frequency Domain Parameter Identification
Electrochemical impedance spectroscopy (EIS) is a frequency domain testing procedurethat employs small amplitude signal perturbations to measure the impedance of thebattery at different frequencies. This test is typically done at different cell operatingpoints such as temperature, SOC and whether the battery is busy being charged ordischarged. The magnitude of the perturbations have to be small in order to assume thatat that specific operating point, the system is linear. During the test, the temperature ofthe battery is required to be regulated to ensure the results can be related to a specificoperating point. Batteries are inherently non-linear devices, so it is crucial that theseconditions are met to ensure that the results for one operating point do not overlap withother operating points.
EIS testing can be used to identify the parameters of an ECM. The response of thebattery, when tested at mid-frequency ranges, exhibits behaviour which can be describedusing RC circuit elements [17]. The low and high frequency testing ranges results inbehaviour which can be described by capacitive and inductive circuit elements.
The impedance of Li-ion batteries are very low (within the milliohm range) and can-not be easily and accurately measured using laboratory EIS systems. It is said thatcommercial EIS systems become less accurate when the impedance is below 0.1 Ω [18].Another possible problem is the mutual inductance of the cell cable and placement of theleads which can have a major effect on the EIS system’s performance. The duration ofthe test at low frequencies (1 mHz or less), can be a considerable drawback since it can
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 21
last more than several hours. It is for these reasons that frequency domain parameterestimation is not investigated further within this thesis.
2.6 State Estimation
2.6.1 Introduction
A battery is a complex non-linear electrochemical device. Only a small amount of param-eters can be measured such as the voltage, current and temperature. Energy managementis crucial to use the battery optimally and effectively. Some parameters that are neededfor accurate energy management, cannot be measured, but only estimated. These pa-rameters include the SOH, SOC, and SOP and will be discussed in more detail in thissection.
2.6.2 SOH Estimation
The degradation of a Li-ion battery is indicated by two factors. One, the battery’scapacity decreases, and two, the internal resistance increases over time. This changes thebattery’s behaviour and could cause failure or limit the battery’s ability such that it isnot able to function as designed for. It is for this reason that the health of the batteryshould be analysed. SOH can be defined by either comparing the remaining capacity ofthe battery to the original rated capacity or by the increase in the internal resistance ofthe battery. These two definitions are defined below by
SOHcap =Qactual
Qrated
x 100%, (2.6.1)
SOHres = 1 +Rrated −Ractual
Rrated
. (2.6.2)
In the first definition, Qrated refers to the manufacturer’s rated capacity and Qactual
refers to the measured remaining capacity of the battery. This can be calculated bydischarging the battery from 100% to 0% according to the manufacturer’s standard dis-charging method. This will typically determine the discharge rate as well as the nominaloperating temperature.
In the second definition, Rrated refers to the initial internal resistance of the battery.Typically, the manufacturer does not supply this information in the datasheet. Thereforeit is important that the parametrisation of the battery is done at the beginning of its lifecycle in order to estimate the internal resistance. Ractual refers to the determined internalresistance of the battery. Again, both these values are calculated at the specified nominaloperating temperature and discharge rate.
Currently, there is no definite consensus in the industry to how SOH is measured.It is for this reason that a SOH has to be chosen that signals the end of life (EOL)of a battery. The general definitions are that either the EOL of a Li-ion battery is atSOHcap < 80% [19] or when the SOHres <= 0 [20].
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 22
2.6.3 SOC Estimation
The SOC is determined by comparing the the remaining charge Q of a battery comparedto the nominal capacity Qrated of the battery. The SOH of the battery influences theSOC since the capacity decreases over time as can be seen in the following equations
SOCnominal =Q
Qrated
x 100%, (2.6.3)
SOCactual =Q
Qactual
x 100% =SOCnominalSOHcap
x 100%. (2.6.4)
The SOC serves as a measurement of the amount of energy that is still available for usewithin the battery. Compared to an internal combustion vehicle, the SOC can be viewedas the fuel gauge of the battery. It is important to note that as the battery degradesand the capacity reduces, the SOC needs to be updated accordingly. For example ina degraded battery, the SOCactual might differ significantly from the SOCnominal of anew battery and will consequently convey a very inaccurate estimate of the remainingcharge. Additionally, the temperature conditions and the nature of the load could alsosignificantly impact the SOCactual compared to the SOCnominal.
A wide range of SOC estimation methods exist, as stated earlier. The most arbitrarymethod for estimating the SOC, is by integrating the current of the battery such that
SOC(t) = SOC(t0) +1
Q
∫ t
t0
I(τ)dτ. (2.6.5)
This method is also known as coulomb counting and is widely used in the consumerelectronics market. Unfortunately, the method is very sensitive to the accuracy of thecurrent sensor as well as to the initial SOC of the battery since an error in the measure-ment will be integrated over time. Another problem is the limited bandwidth due to thesampling period. Sudden changes in current can therefore not be accurately measured,which further decreases the accuracy of this method.
Other SOC estimation methods are mostly based on the way in which the batteryis modelled, except empirical methods of estimation. Pure empirical models uses thecurrent, temperature and voltage measurements as inputs and delivers an estimate onthe SOP, SOH or SOC as outputs. Hybrid empirical and ECM methods have beenproposed as detailed in [11], where the equivalent circuit model is based on fuzzy logic. Ingeneral the selected model is combined with some form of estimation method to estimatethe states within the model. A first principle model in combination with an extendedKalman filter is an example of such a model and is described in more detail in [21].ECMs are the most predominantly used solution for SOC estimation in automotive andlarge-scale battery packs [7].
An ECM has the advantage that the algorithms and models can be reused easilyfor different cell chemistries, and it is also efficient and robust. Typically, an ECM isused as the basis for the estimation process. The model is given the measured currentof the battery as an input. A control system is used to change the states of the ECMin such a way that the difference between the output of the model and real systemis minimised (negative feedback), as shown in Figure 2.14. These estimation methodsincludes the Luenberger observer, sliding mode method, Kalman filter and Proportional
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 23
Batterymodel
−
+
Estimationmethod
Battery
MeasuredCurrent
Current
MeasuredVoltage
CalculatedVoltage
Figure 2.14: Block diagram of the ECM SOC Estimation Methods
integral observer [22]. One of the states that these estimators will estimate, typicallyinclude the open circuit voltage VOC of the battery. This can then be used to obtain anestimation of the SOC of the battery.
2.6.4 SOP Estimation
State of power (SOP) is the ability of the battery to deliver or accept power to or fromthe application. An accurate SOP estimate is useful in practical battery applicationssince it can be used to determine the available power and as such ensure the battery isnot overcharged or over-discharged. SOP is defined similarly as discussed in [23]. It caneasily be estimated by the use of an ECM. Using the SP Thévenin model as an example,the maximum discharge current can be calculated if all the parameters of the model isknown.
The SOP can be calculated with the maximum allowed discharge current, withoutdecreasing a cell’s terminal voltage to below the cell voltage as specified by the manufac-turer. Thus,
SOP =Vlimit(VOC − Vlimit)
Rs +Rt1
[W]. (2.6.6)
Equation 2.6.6 is derived using node analysis. The voltage Vlimit is the minimumallowed terminal voltage of the cell. The voltage VOC is the open-circuit voltage and(Rs + Rt1) is the total internal resistance of the cell. This method can also be usedto calculate the maximum allowed charging current of a cell. In this case Vlimit is themaximum cell voltage specified by the manufacturer.
2.7 Recursive Least Squares methodThe least squares (LS) method is a mathematical algorithm that minimises the sum of thesquares of the differences between observed values and estimated values. This method istypically used for curve-fitting and this can be also be applied to the dynamic behaviourof Li-ion batteries. The parameters of an ECM can be estimated by the LS methodin such a way that the error between the measured data and the model is minimised.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 2. LITERATURE STUDY 24
Unfortunately, the LS method can only be applied to a complete data set, which makesit unsuitable for real-time estimation purposes. This is where the recursive least squares(RLS) method comes in.
The RLS method is derived from the LS method and is usually used for real-timeestimation. The model is recursively updated as new data is processed using the leastsquares method. The RLS method for on-line parameter estimation will be discussed inmore detail in Chapter 5.
2.8 ConclusionIn this chapter, the different concepts pertaining to the objectives of this thesis were pre-sented. This includes a short overview of batteries, where the different technologies werecompared and different configurations were discussed. The needed BMS requirementsand architectures were also investigated. Battery modelling and parameter estimationmethodologies were discussed in detail. Definitive definitions for the different state esti-mations of batteries were also given.
Stellenbosch University https://scholar.sun.ac.za
Chapter 3
Hardware design
3.1 IntroductionThe hardware design of a proof of concept BMS for a small Li-ion battery pack is discussedwithin this chapter. This design serves as a prototype to prove the basic capabilities ofthe different subsystems within the BMS. This design is scaled up for the purpose of amicro EV and is also discussed in this chapter. This chapter also includes the designof a proof of concept solid state contactor that may be used as an alternative to themechanical contactor used in the full scale BMS design.
3.2 Proof of Concept Battery Management System
3.2.1 Introduction
In this section a proof of concept BMS is designed for a small Li-ion battery pack. Thisconcept design is inspired by a previous prototype BMS designed for the Shell Eco-Marathon (SEM) competition. The competition’s goal was to travel a certain distanceusing the least amount of energy possible. Energy efficiency was therefore extremelyimportant. The prototype EV was supposed to have competed in the Shell Eco-Marathonrace, but due to some technical difficulties with the electrical drive train, the EV wasunable to participate in the race.
The concept design is based on the battery pack used for the SEM competition.The design and specifications of the battery is described in more detail in the followingsubsections.
3.2.2 Design choices
The SEM battery pack consists of 13 EEMB LP55100100 Lithium polymer cells connectedin series. The cell specifications is shown in Table 3.1. The proof of concept BMS designfor this thesis is based on these specifications.
The nominal battery voltage of the complete battery pack is 48V. This voltage hasbecome the standard for low voltage energy storage systems since it is well within the safeoperating range which is limited to 60V. The human body has an internal resistance and
25
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 26
Table 3.1: LP55100100 cell specifications
Parameter RatingNominal voltage 3.7VMaximum voltage 4.2VMinimum voltage 2.75VNominal capacity 6.5Ah at 1.3A
Maximum discharge current 13AMaximum charge current 6.5A
as the magnitude of the voltage increases, so does the current in the case of an electricshock. An electric current too large can affect biological tissue in a hazardous manner.
A BMS consists of multiple components. This typically includes a battery monitoringchip that measures the cell voltages and pack temperature. A current sensor is used tomeasure the battery current. A Micro Controlling Unit (MCU) is used to process thedata and accordingly control the the output of the battery by means of some sort of aswitch.
A block diagram of the proof of concept BMS is shown in Figure 3.1. In the figure,B1 to B13 are balancing circuits that are controlled via the battery monitoring chip whichis programmed with the MCU. The design includes a shunt resistor as a current sensor.Back to back power field-effect transistors (FET) are used as a solid state switch. A MCUis the master controller which controls the monitoring chip. The MCU is powered by alow power buck regulator. The proof of concept BMS with the SEM battery is shown inFigure 3.2. The design choices for all the main circuit components will be discussed inthe rest of this section.
+
48 V
Currentsensor Switch−
B1
B2
B13
Monitorchip
I2C Micro-controller
Unit3.3 V Low
PowerDC-DC
Figure 3.1: Proof of concept BMS circuit diagram
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 27
Figure 3.2: Proof of concept BMS with the SEM battery
3.2.2.1 Cell Monitoring System
There is a large variety of battery monitoring chips that are used to measure the voltageand temperature of the different cells in a battery pack. Linear Technology, AnalogueDevices and Texas Instruments all provide possible solutions and will be discussed below.
AD7280A [24] is a modular battery monitoring system from Analog Devices. It canmonitor up to six cells per chip, but multiple AD7280A’s can be stacked on top of eachother through its daisy-chain interface to accommodate up to 48 cells connected in series.The AD7280A supports cell balancing, temperature and voltage measurements. It usesthe Serial Peripheral Interface (SPI) communication protocol to communicate with themaster controller.
LTC6802-1 [25] is a modular battery monitoring system from Linear technologies. Itcan monitor up to 12 cells per chip, but multiple LTC6802-1 can be stacked on top ofanother via its daisy-chain interface to accommodate a high battery voltage (>1000V).The LTC6802-1 supports cell balancing, temperature and voltage measurements. It alsouses the SPI communication protocol to communicate with the master controller and itis equipped with an open wire connection fault detection function.
BQ76940 [26] is a centralized battery monitoring system from Texas Instruments (TI).The chip periodically measures the cell voltage, battery current and battery temperature.It supports up to 15 cells and 3 temperature sensors. Multiple BQ76940’s cannot bestacked on top of each other to accommodate an increased battery voltage. It aims tobe a complete battery monitoring solution in itself. The BQ76940 has a current sensoras well as an output for a FET solid state switch. The solid state switch can be used tocontrol the output of the battery. It uses the Inter-integrated Circuit (I2C) communicationprotocol to communicate with the master controller.
Another important feature that is included within the BQ76940, is a built-in extralayer of analogue protection. As soon as one or more of the sensors sense a value out-side of the specified operating range, the solid state switch opens and thus protects thebattery. The maximum overvoltage (OV), undervoltage (UV), short-circuit (SC) andovercurrent values are programmed into the BQ76940 directly after the system start-up.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 28
The periodical measurements are continuously monitored and the solid state switch iscontrolled independently, thus not by the MCU but by the BQ76940 itself. This protectsthe battery if the digital control circuitry from the MCU should fail for some reason.
The BQ76940 was chosen as the battery monitoring chip for this project due toits ability to measure all the required variables, and also due the additional protectioncircuitry that it provides. The BMS needs to be operational over extended periods ofoperation. If the MCU should fail for some reason, the BQ76940 chip will keep monitoringthe voltage of all the cells and the current of the battery. If some fault occur the switchis opened to protect the battery from damage.
3.2.2.2 Cell Balancing
The BQ76940’s cell balancing feature is implemented within the proof of concept BMSdesign. Each cell is equipped with its own balancing circuit. All of the different cellbalancing circuits are situated on the proof of concept BMS printed circuit board (PCB)along with other components such as the BQ76940 and the MCU. Each cell balancingcircuit is controlled with a P-channel MOSFET which is switched on or off using theBQ76940 chip. The MCU controls which cells are balanced by programming the BQ76940chip.
A 20Ω balance resistor was selected to discharge the cell being balanced. This resultsin a balancing current of 200mA which is 3% of the LP55100100 cell capacity (Ah).Balancing currents of 1% of the battery capacity (Ah) are adequate for applications thatstay in standby for long periods, e.g. electric vehicles, according to [27]. The balancingcurrent is chosen in such a way that the proof of concept BMS is compatible with batterypacks with a larger capacity than that of the SEM battery pack.
3.2.2.3 Current Sensor
For the safety of the battery and the application it is extremely important to regularlymeasure the battery current to avoid continuous overcurrent conditions. The BQ76940chip is equipped with a 16-bit integrating analogue-to-digital converter (ADC) that pro-vides measurements of accumulated charge across a current sense resistor. This func-tionality is used since it reduces the overall system complexity and it also provides anaccurate measurement of the battery current.
Shunt resistors used for current sensing purposes are ideal for smaller battery packsdue to its low cost and simplicity. Larger battery packs, typically, do not use shuntresistors, because of the high power losses and the effect of temperature on the resistor’svalue. The value of the shunt resistor used in this design is chosen as 2mΩ. The valueof the resistor was chosen as such to minimise losses, since efficiency is of great concernfor this design. This does however slightly lower the measurement accuracy.
3.2.2.4 Solid State Switch
The BQ76940 chip provides two low side FET drivers. These may be used to controlback-to-back power FETs, which acts as a solid state switch. Two FETs are used, one tocontrol each direction of the battery current. This has the advantage for example thatif the battery is fully charged, the charge FET can be opened to only allow a discharge
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 29
current. This functionality does result in higher losses in the switch, but are rarely used.It is only used on the edge of operation, meaning at very low or high SOC. The FETschosen for the proof of concept BMS was the FS3207 power metal-oxide-semiconductorfield-effect transistor (MOSFET) from International Rectifier. It is chosen for its lowon-state resistance of 4.5mΩ and minimum breakdown voltage of 75V.
3.2.2.5 Micro Controlling Unit
The MCU used in this design is the Piccolo TMS320F28035 [28] from TI. TI MCUs arechosen for all the electronic development done within the electrical drive train. Thisensures compatibility and eases the integration process between the different workingparts of the overall project.
The TMS320F28035 was chosen for its Controller Area Network (CAN) functionality.It is the smallest available TI MCU with CAN functionality. The CAN interface will beused, at a later stage, to communicate with the dashboard unit of the prototype EV.
The MCU is used for the processing of data and receives the battery data via theI2C communication protocol. The MCU controls the complete BMS apart from partiallycontrolling the solid state switch. The MCU processes the data it received from themonitoring system such as the cell voltages, battery current and pack temperature. It isalso in charge of cell balancing and can control the output 48V battery supply via theBQ76940 chip.
3.2.2.6 Low Power Buck Regulator
A low power regulator is required to step down the battery voltage to power the MCUand future communication interfaces. The supply voltage is required to step down from48V to 5V and 3.3V. The battery voltage can fluctuate, depending on its state of charge,between a minimum voltage of 41.6V and a maximum voltage of 54.6V.
The LM5017 [29] buck regulator from TI is selected for this purpose. It can delivera maximum current of up to 600mA which is adequate since the MCU uses a maximumconstant current of 200mA. The LM5017 is primarily chosen for its 600mA currentsupplying capability, since additional current will be required at a later stage for theimplementation of the communication interfaces.
Secondly, the LM5017 is chosen for its flexibility in terms of its wide input voltagerange, adjustable switching frequency and under voltage lockout protection. It is alsoequipped with other protection features such as thermal shut down and a peak currentlimiting circuit which protects against overload conditions. This makes the LM5017 idealfor the purpose of this project.
The design to step down the battery voltage to 5V through the LM5017 is shown inAppendix A.1. The 5V supply will also be used in future versions to power the isolatedCAN transceiver. Isolated CAN is required since the solid state switch is connected to thenegative terminal of the battery. The 5V is stepped down to 3.3V using the TPS73633which is a linear regulator from TI. The 3.3V is used to power the MCU.
The voltage ripple on the 3.3V rail is shown in Figure 3.3. The switching frequencyof the LM5017 is 237.3 kHz which confirms the designed switching frequency detailed inAppendix A.1. The ripple voltage is 58.5mV. The large ripple voltage can be attributed
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 30
−0.000010 −0.000005 0.000000 0.000005 0.000010Time [s]
−0.02
−0.01
0.00
0.01
0.02
0.03
0.04
0.05
Volta
ge[V
]Ripple
Figure 3.3: Voltage ripple
to the fact that a ripple voltage of 25mV is required at the feedback pin to enable theLM5017 to react.
3.2.2.7 Temperature Sensor
The temperature of the battery pack is measured with three thermistors. The 103AT-2of Semitec is selected according to the BQ76940 datasheet. It specifies the use of 10 kΩnegative temperature coefficient (NTC) thermistors for measurements.
3.3 Full Scale Battery Management System
3.3.1 Introduction
The proof of concept BMS detailed in the previous section is scaled up for the purposeof a micro electric vehicle in this section. The required battery capacity is 10 kWh with amaximum continuous power output of 5 kW. The battery terminal voltage is selected as48V since this reduces the risk of fatal electric shocks. The required power specificationresults in a nominal current of 104A. By scaling the discussed BMS to meet all of thespecification for the EV, new design challenges are introduced. The design choices madeto address these challenges are discussed in this section.
3.3.2 System Overview
The full scale BMS is based on the proof of concept BMS that is expanded into multiplecomponents. The complete proof of concept BMS was manufactured onto a single PCB.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 31
Due to the scale of the micro EV battery, the full scale design is forced to be divided intoseparate PCBs, which include:
1. A main control board, which measures the different battery attributes and interpretsthem.
2. A current sensor board to deliver a voltage representation of the battery currentthat can be measured by the main control board. Currents of this magnitude couldpossibly induce Electromagnetic Interference (EMI) onto the cell voltage measure-ments if the sensor remained on the main board as with the proof of concept design.
3. Separate balancing boards are required and mounted on each parallel set of cells toincrease the battery balancing current.
The other circuit components within the system that was not developed include amechanical contactor and a fuse that serves as a back up to the contactor if it shouldfail. A condensed version of the complete BMS is shown in Figure 3.4. Starting fromthe left: the system has a local buck regulator to supply power to the MCU and variouscommunication hardware (CAN, USB and Bluetooth). The MCU is the master controllerand controls the 48V battery output by using a mechanical contactor. It also monitorsthe battery current through the current sensor board and does cell balancing duringcharging. The voltage and temperature measurements are done with a battery monitoringchip (BQ76940). Custom balancing PCBs are designed for cell balancing purposes inorder to enable high power cell balancing. The battery cells are connected in a matrixconfiguration to minimise the total required cell voltage measurements. The full scaleBMS installed along with the battery pack can be seen in Figure 3.5. The design choicesmade are discussed in detail in the following subsections.
B1
B2
B15
Monitorchip
LowPower
DC-DCI2CMicro-
controllerUnit
3.3 V
Communication250 A
F1 CurrentsensorContactor
+
48 V
M1
N1 −
Figure 3.4: Full scale BMS circuit diagram
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 32
Figure 3.5: Full scale battery pack
3.3.3 Battery Cell Configuration
The cells used inside the battery pack are 100Ah LFP (LY-100AH) cells from LiyuanBattery Corporation. The battery chemistry is chosen for its high cycle life and highsafety. The LFP cells have a nominal voltage of 3.2V each, resulting in nominal batteryvoltage of 48V when 15 of the cells are connected in series. The specifications of thecells are shown in Table 3.2. Two sets, of 15 cells connected in series, are used to achievethe required capacity of around 10 kWh. The resulting battery pack is a 48V - 200Ahbattery pack, which has a capacity of 9.6 kWh.
The two strings are connected in the matrix configuration, i.e. two cells are connectedin parallel, with 15 of these parallel connections connected in series. This configuration isselected to minimise the required cell voltage measurements. Only 15 parallel connectedcell voltages need to be measured. Another advantage, with careful selection, is that thecapacity tolerances of the cells can be used to cancel each other out by connecting a lowcapacity cell in parallel with a higher capacity cell. This will typically entail testing eachindividual cell’s capacity before constructing the battery pack.
A digital model of the full scale BMS installed along with the battery pack is shown
Table 3.2: LY-100AH cell specifications
Parameter RatingNominal voltage 3.2VMaximum voltage 3.6VMinimum voltage 2.5VNominal capacity 100Ah at 33A
Normal discharge current 33ANormal charge current 33A
Maximum discharge current 500AMaximum charge current 300A
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 33
Figure 3.6: Digital model of the battery pack
in Figure 3.6. The battery pack is divided into an upper and lower level due to sizeconstraints. All the control circuitry is packaged close together within the remainingspace allotted to the battery pack. This simplifies the installation of the BMS onto thebattery. It also results into a battery pack with a symmetric form as specified by the EVmanufacturer.
3.3.4 Fuse
The central fuse of the system is added to protect the battery pack from overcurrentconditions, either during charging or discharging. The fuse acts as the last resort currentprotection in situations where for example there is a short-circuit current, or the MCUor current sensor should fail for some reason.
The fuse selected is a 250A fast acting fuse. The battery pack can deliver up to600A and receive up to 300A without being damaged according to the datasheet. Themaximum continuous power that the system requires to deliver is 5 kW, which results ina maximum current of 104A. The maximum peak power is limited by the current sensor,since the current sensor is only rated to measure currents up to 150A. This is well belowthe 250A rating of the fuse.
3.3.5 Main Control Board
The main control board is responsible firstly to measure all the battery attributes and sec-ondly to regulate the battery output. The components used within the previous designedBMS that are also used within the full scale design include the MCU (TMS320F28035)and battery monitoring chip (BQ79640). The manufactured main control PCB is shownin Figure 3.7 and the schematic can be seen in Appendix C.1.1. Reusing as much aspossible of the previous design reduced the development time since many of the firmwarehas already been developed.
Similarly as the proof of concept BMS, the full scale BMS is centered around theBQ76940 battery monitoring chip from TI. Again the core functionality of the chip is
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 34
Figure 3.7: Main control BMS PCB
to monitor the 15 cells connected in series as well as to provide the additional analogueprotection circuitry. The BQ79640 is able to control the mechanical contactor using anAND gate and the cell voltage measurements. The current measurement is no longerperformed by the BQ79640 which results that the BQ79640’s analogue SC and overcur-rent protection are lost. The BQ79640’s analogue protection is active for OV and UVconditions.
Other changes that were implemented from the proof of concept BMS include theaddition of communication interfaces that can be used to communicate the data andstates of the battery to a driver interface. A different low power buck regulator is alsoused. These changes are discussed in the following subsections.
3.3.5.1 Bluetooth Communication
The LMX9839 [30] Bluetooth module from TI is added to the design to provide theBluetooth capability. This will typically be used for battery diagnostics and data logging.This module was chosen since it is a complete solution with a built-in antenna. Thishardware design is included but not tested in this project. The software development ofthe Bluetooth interface is outside of the scope of this project and will be added at a laterstage.
3.3.5.2 CAN Communication
The SN65HVD234 [31] transceiver from TI is added to the design for its CAN functional-ity. The CAN protocol delivers a robust solution for communication between the varioussystem components for example the BMS and the battery charger. It can however alsobe used for battery diagnostics and data logging. The SN65HVD234 transceiver actsas a buffer between the MCU and the CAN bus. The transceiver was chosen because itmeets the requirements of the system. This hardware design was succesfully implementedwithin this project. The software development of the CAN interface is discussed in detailin Chapter 4.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 35
3.3.5.3 Powerline Communication/ USB Communication
Originally, the PLC-stamp mini from I2SE was also added to the design for its power-linecommunication functionality. The PLC-stamp mini is a small breakout board. Unfortu-nately, it only became apparent at a later stage that using this technology is currentlytoo expensive to implement. This will however change over time as the price of the tech-nology decreases. Therefore, it is kept in the design for development and testing at alater stage. The PCB layout is done in such a way that other breakout boards can alsobe used instead of the PLC-stamp mini.
USB communication was added via Future Technology Devices International’s (FTDI)FT220 [32] breakout board which uses the same layout as that of the PLC-stamp mini.This is the primary interface used to send data from the BMS to a computer. USB com-munication is easy to set up on the computer’s side. A small Python listener applicationis used to read the incoming battery data into a database.
3.3.5.4 Low Power Buck Regulator
A low power regulator is used to power the MCU as well as all of the different communi-cation interfaces. The supply voltage is required to be 3.3V. The LM5017 buck regulatorwas used in the proof of concept design. Unfortunately, it has a very low feedback preci-sion which results in a voltage ripple higher than desired. It is for this reason that for thefull scale BMS design, the TPS54060 [33] buck regulator from TI is chosen. The designis done using TI’s SwitcherPro software package.
The voltage ripple of the regulator is shown in Figure 3.8. The voltage ripple hasdecreased to 10mV. The lower ripple minimises the regulator’s influence on the voltageand current measurements. The switching frequency of the device is 143.7 kHz whichconfirms the design.
−0.000015 −0.000010 −0.000005 0.000000 0.000005 0.000010 0.000015Time [s]
−0.008
−0.006
−0.004
−0.002
0.000
0.002
0.004
Volta
ge[V
]
Ripple
Figure 3.8: TPS54060 voltage ripple
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 36
3.3.6 Balance Circuit Design
The full scale BMS, similar to the proof of concept BMS, deploys passive cell balancing tomaximise the battery capacity. In large scale systems the magnitude of the cell balancingcurrent is large, thus a lot of heat is dissipated. The heat is substantially more than whatcan be dissipated on the BMS’s main control board. For this reason a small balance PCBis designed, one for each set of parallel connected cells, to distribute the heat across alarge area during cell balancing. The circuit diagram of the PCB is shown in Figure 3.9and the PCB schematic can be seen in Appendix C.1.2.
The size of the balance current is typically chosen according to the cell capacityand tolerance. A minimum balancing current of 1A was selected for this project, aswas suggested by Julian Gerber, an engineer involved in the Joule EV project. He alsoconsulted the University of Stellenbosch during this project.
Balancing is activated by the battery monitoring chip that is situated on the maincontrol board. The chip’s ADC pins are connected internally to enable a P-channelMOSFET (M1) externally, which enables cell balancing of that specific cell. A fuse (F1)is used to protect each cell connection from a short-circuit fault. A transient voltagesuppressor (TVS) (Dt) and two diodes (D1 and D2) are used to protect the MOSFETfrom electrostatic discharge (ESD). It is also disconnects the output by blowing the fusein case the PCB is connected to the cell the wrong way around. The resistor Rf is partof the BQ79640 hardware set-up and is used in combination with a capacitor, to act as afilter. The resistor Rg is used as a gate resistor to limit the inrush current which enablesthe balancing MOSFET. The resistor Rt is a 10 kΩ NTC thermistor and is selected assuggested by the BQ76940 datasheet. The thermistor is situated close to the batteryterminal on the PCB. This ensures an accurate cell temperature measurement.
The PCB is designed to bolt onto the terminals of a cell. It was decided to use a PCBtrace as the balance resistor Rb. This saves on some of the cost of the system since nohigh power resistors need to be bought and it reduces the height of the balancing PCBswhich was a constraint in the physical design of the battery pack. The required tracewidth for a specific current and temperature increase can be approximated [34] by
I = 0.048 ·∆T 0.44 · A0.725, (3.3.1)
where
I = maximum current [A],
∆T = temperature rise [C] ,
+Vcell −
Rb
M1
Rf
Rg
D1
D2
+F1
Dt
Rt+-
Figure 3.9: Balance circuit diagram
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 37
A = cross-sectional area [mil2].
The minimum balancing current was specified as 1A. This results, with a temperaturerise of 10C, in a PCB trace width of 0.3mm.
The total balance resistance is chosen as 2.5Ω. The cells are typically balanced ata voltage of 3.6V, limiting the balancing current to a maximum of 1.44A. The internalresistance of the MOSFET is 0.14Ω, which results in a required trace resistance of 2.36Ω.The resistance of the trace can be calculated by
R = ρ · L
T ·W · (1 + α · (temperature− 25C)), (3.3.2)
where
ρ = resistivity,
L = length,
W = trace width,
T = trace height,
α = temperature coefficient.
The values of ρcopper and αcopper are 3.9 · 10−3 Ω·cm and 1.7 · 10−6 1C respectively.
The length of the trace is calculated as 1.4m with a trace thickness of 35µm. Thisresults in a balancing current with a minimum value of 1A and a maximum value 1.44A.During practical testing the balancing current was measured to be 1.25A. A manufacturedbalance PCB connected to a cell’s terminals is shown in Figure 3.10.
Figure 3.10: Balancing PCB connected to cell terminals
3.3.7 Current Sense PCB Design
Accurate current measurements are required to successfully monitor the battery pack.This section details the design of the high current measurement board. The currentmeasurement board is separated from the main control board for several reasons.
The board is designed to measure a maximum steady state current of up to 120A. Asafety margin is added since only 104A is required to provide an output power of 5 kW.To achieve such a high current rating the PCB trace width and trace thickness need tobe maximised.
Increasing the trace thickness increases the overall cost of the PCB dramatically.One of the manufacturing constraints is that a PCB typically only have one PCB trace
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 38
thickness unless it is a specialised PCB which is expensive. The current sensor is thereforesplit from the main control board to minimise the PCB area that requires extra thicknesstracks.
Splitting the design also reduces EMI on the main control board. The high motorcontroller switching current needs to be as far away from the main control board aspossible to minimise the coupled noise. This noise will typically reduce measurementaccuracy.
3.3.7.1 Current Sensor Selection
There are two types of sensors that can be used to measure the current of the battery.The first is a shunt resistor, as was used in the proof of concept BMS design. The secondis the use of a Hall effect-based current sensor.
The shunt resistor method has some disadvantages when it comes to high currents.In extreme fault conditions the conductor connecting the battery and the shunt resistorcan induce voltage spikes which can damage the measuring device. This is due to thechange in battery current with respect to time ( δi
δt). The corresponding generated voltage
is calculated by: V=L δiδt. All conductors have some equivalent series inductance (L)
even though it is very small. This phenomenon decreases the overall robustness of thissolution.
Another drawback with using a shunt resistor, is the power dissipated within theresistor. It is important that adequate resistance is chosen. If the resistance is too large,too much power is dissipated which decreases the overall efficiency of the circuit. On theother hand if the resistance is too small, the accuracy of the measurement is decreased.The value of the shunt resistor is typically chosen extremely small to minimise the amountof power dissipated within the resistor. A shunt resistor with a small tolerance andtemperature coefficient is usually quite expensive. An operational amplifier is used toamplify the shunt resistor’s voltage measurement in order to increase the resolution ofthe measurement. This also adds to the complexity of the system.
A Hall effect current sensor is another option. It operates on the principle of using themagnetic field generated by the current to measure the size and direction of the current.The ACS758KCB-150B-PFF [35] sensor from Allegro is chosen for this design. The Halleffect sensor is much more suitable for high current applications and provides superiorperformance to that of the shunt resistor. The sensor chosen has a series resistance of100µΩ which is much smaller than that of a corresponding shunt resistor required for thisapplication. The Hall effect sensor provides galvanic isolation which isolates the sensor’soutput from voltage spikes and it allows high-side voltage current sensing. No additionalamplification of the measurement is required and as such it provides the complete currentmeasuring solution. The price of the Hall effect sensor also compares very well to that ofa high quality shunt resistor.
3.3.7.2 Sensor Design
A simplified diagram of the current sensing circuit is shown in Figure 3.11. The thickerlines represent high current paths. A low pass filter Rf and Cf , is shown at the outputof the current sensor. This is used for noise filtering.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 39
I in
I out
RfCf
CS out
Rp
PC out
D13.3 V
RPC in S1
Currentsensor
SSC
Figure 3.11: Current sensor circuit diagram
A pre-charge resistor Rp is also added to the current sensor design. The batteryis typically connected to loads that have some form of bus capacitance. Initially themechanical contactor is connected in open-circuit to allow the bus capacitance to chargeat a controlled rate in order to prevent high inrush currents that can harm the battery,the mechanical contactor or the load itself. A pre-charge resistor is used for this purpose.The mechanical contactor’s one terminal is connected to the output terminal (I out) ofthe current sense PCB. The pre-charge output (PC out) is connected to the mechanicalcontactor’s second terminal. A small solid state contactor (SSC) is used to control theoutput of the pre-charge circuit to the external bus capacitance. Once the voltage acrossthe capacitors equal that of the battery potential, the main mechanical contactor is closed.
The value of the pre-charge resistor Rp was chosen according to the motor controller’sdatasheet information as 470Ω. The SSC was chosen according to the pre-charge resistor’svalue and the maximum voltage of the battery which is 54.75V. A transient voltagesuppressor (TVS) D1 is used to protect the SSC from electro static discharge (ESD) aswell as from fault conditions where the voltage across the mechanical contactor couldspike. The SSC is controlled by the MCU by means of a MOSFET S1 and resistor R.The PCB schematic can be seen in Appendix C.1.3. The manufactured current sensePCB is shown in Figure 3.12.
3.3.7.3 Thermal Design
Accommodating a current of 120A on a PCB is a challenge. For this design a two layerboard was selected to minimise the cost of the PCB. The PCB layer thickness was chosenas 140µm on each side. This is four times the thickness of the standard of 35µm. Thetop and bottom layers are stitched together with via’s to ensure both layers are usedeffectively. This results in a combined layer thickness of 280µm. The required tracewidth is calculated using (3.3.1) and (3.3.2). The trace width is calculated using theworst case scenario in mind: an ambient temperature of 40 C, a current of 120A and a
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 40
Figure 3.12: Current sensor
temperature rise of 15 C. The resulting trace width is 22mm. The thermal results willbe discussed in more detail in Chapter 6 as part of the results discussion.
3.4 Prototype Solid State Contactor
3.4.1 Introduction
A SSC is designed, in this section, as a possible replacement for the mechanical contactor.Possible advantages a solid state contactor has over a mechanical contactor are a fasterreaction time, less wear as well as being more energy efficient.
The main focus of this design is on ensuring the SSC is more energy efficient thanit’s mechanical counter part. It is for this reason that the design is centered around thespecifications of the mechanical contactor. The basic operation of a mechanical contactoris therefore discussed in the following subsection. The rest of the subsections discuss thedesign of the SSC.
3.4.2 Mechanical Contactor
The mechanical contactor shown in Figure 3.4 is used as a switch to connect or disconnectthe battery from the load. In this subsection a mechanical contactor is investigated andis used to determine the different specifications required for a SSC.
Mechanical contactors use an electrical current as input, to generate a magnetic fieldin the control coil. This magnetic field closes the mechanical contact. Typically, themagnetisation current is the most energy inefficient part of the mechanical contactor.In order for the contactor to remain closed, magnetisation current is required. Thus,as long as the contactor is closed, power is being dissipated. The mechanical contactoralso dissipates power during use due to the contact resistance of the pole. This contactresistance typically increases with use due to the build-up of compounds on the contactswhich is caused by arcs when the contactor disconnects.
The mechanical contactor used in this project is the SW80B [36] from Albright In-ternational. It is a 100A rated single pole, normally open contactor. The continuouslyrated coil (on-state) power dissipation is between 7 – 13W according to the datasheet.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 41
The contactor was tested for the intended application under normal operating conditionsto more accurately determine the coil’s power dissipation. A coil dissipation of 11W wasmeasured. This value stays constant while the contactor is closed, irrespective of the loadpower usage.
The datasheet also specifies that the typical voltage drop per pole across new contactsat 100A, is 40mV (4W). Interpolated for 104A, this would result in a loss of 4.326W.The total power dissipation in the mechanical contactor at a current of 104A is thus15.326W. The SSC must thus dissipate less power in order to be more energy efficient.
3.4.3 Solid State Contactor Design
When designing a SSC, all the solid state switching technologies need to be investigated.The most commonly used solutions currently include a MOSFET or an insulated-gatebipolar transistor (IGBT). For this project’s application, the contactor is used at highcurrent and low voltage. This makes the use of a MOSFET an appropriate solution, sincelow voltage MOSFETs have very low on resistances. Thus, MOSFETs were chosen overother technologies such as IGBTs.
3.4.3.1 Design Overview
MOSFETs can only control current in one direction because of its freewheeling diode.Therefore, an extra MOSFET is added to ensure current control in both directions. Thisresults in two MOSFETs in series, connected back to back with a common source. TheMOSFETs need to be powered by an isolated supply to ensure operation within noisyenvironments. The solid state contactor must thus be isolated from the control circuitry.Opticalcouplers are used to deliver isolated signals to the gate drivers of the MOSFETs.MOSFETs do add some difficulties in terms of robustness.
3.4.3.2 MOSFET Considerations
It is critical that the SSC is robust and disconnects the battery from the load when desiredunder all circumstances. The worst case scenario is a short circuit condition across thebattery terminals. If the SSC is unable to disconnect the battery from the load in theevent of a short circuit condition, it could lead to the battery failing. A Li-ion batteryfailing, could lead to a fire or even an explosion.
MOSFETs are susceptible to being damaged in the case of overvoltage or overcurrentconditions. In the worst case scenario, the Miller capacitance can also lead to the MOS-FET being operated within the linear region. This would cause the MOSFET to dissipatetoo much power, which in turn could lead to MOSFET’s failure. MOSFETs also needshort circuit protection and thermal runaway protection to protect it from damage.
3.4.3.3 Detailed Design
MOSFET: A simplified schematic of the SSC design is shown in Figure 3.13. The de-sign is based on ultra-low resistance MOSFETs. The MOSFETs used is the CSD19535KTTPower MOSFET from TI as is shown in Table 3.3. To further reduce the internal resis-tance of the SSC, four MOSFETs were connected in parallel, which reduced the overall
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 42
Table 3.3: CSD19535KTT Power MOSFET specifications
Parameter RatingVDS maximum voltage 100VVGS maximum voltage ± 20V
RDS(on) 2.8 mΩ at 10VID maximum drain current 197A at 25 CQg maximum gate charge 98 nC at 10V
Junction-to-case thermal resistance 0.5 C/W
resistance to 0.7mΩ. This action was required to ensure the SSC is comparable to themechanical contactor in terms of efficiency. Two sets of parallel MOSFETs are connectedin series, back to back. This configuration minimises the required amount of DC-DCconverters to one, which reduces the overall cost of the system.
The two series sets allow the control of the battery current in both directions. Thisincreases the overall resistance to 1.4mΩ. A current of 104A results in a power dissipationof 15.14W, which is within the design specification.
It is important to note that the power dissipated inside the coil of the mechanicalcontactor, when it is within its closed state, is responsible for the majority of the powerbeing dissipated. This results in a very inefficient switch when only a small amountof power is being delivered to the load. The on-state power dissipation of the SSCon the other hand is very small. It is negligible compared to the series resistance powerdissipation. This is confirmed in the results section. Even though at maximum continuouspower the two contactor’s efficiencies are roughly equal, at low power usage the SSC’sefficiency declines quadratically with the load current.
Isolated Converter: The MOSFETs are controlled by MOSFET drivers which arepowered by an isolated DC-DC converter. This isolates the MOSFETs from any voltagespikes on the high current carrying conductors. It also allows the SSC to be connectedat the positive terminal as was the case with the mechanical contactor.
The supply voltage of the MOSFET driver was chosen as 12V. It is the midway voltagebetween the driver’s maximum voltage and the minimum on-voltage of the MOSFETs.The chosen DC-DC converter for this project is the MEU1S1212ZC [37] from MurataPower Solutions. Both the input and output voltages are rated to 12V. The converterwas chosen for its small size, cost and voltage rating.
Optical Isolator: The input signals to the SSC are also required to be isolated in orderto isolate the control of the SSC from the main control board. This can easily be achievedby means of optocouplers. The 4N35 [38] from Fairchild is selected for this project. Itdelivers a good trade-off between performance and cost.
Transient Voltage Suppressors: One of the problems with MOSFETs are its sensi-tivity to overvoltage. It is for this reason that transient voltage suppressors (TVS) are
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 43
High Current input
GND-ISO
High Current output
RG1 RG2 RG3 RG4
RG5 RG6 RG7 RG8
Mosfetdriver 1
12 V ISO
OUT1
R1CS
R2
GND-ISO
C1 D1
Mosfetdriver 2
12 V ISO
OUT2
R3CS
R4
GND-ISO
C2
D2
R5IN2
3.3 V ISOINPUT 2
R11
R6
IlimR7
D3
IN1R8
GND-ISO
3.3 V ISO
R9
IlimR10
D4
INPUT 1
R12
ISOSupply
12 V 12 VD5
Figure 3.13: Solid state contactor circuit diagram
added to the design. When a large current is disconnected by the SSC, the inductancewithin the conductors causes a voltage spike on the SSC terminals. The TVS clampsthe voltage at a set value. The TVS chosen for this project is the 5KP54CA-TP [39]from Micro Commercial Co. It can dissipate peak power pulses of up 5 kW at a maxi-mum clamping voltage 87.1V. The peak pulse current is 57A. Three of these TVS areconnected in parallel to accommodate the worst case battery current scenario of 120A.
MOSFET Driver: Each set of parallel MOSFETs are controlled by a MOSFET driver.The UCD7100 [40] from TI is selected for this purpose. Another MOSFET driver thatwas investigated is the TD352 from ST. The UCD7100 was chosen for its high sink andsource current of 4A. A large current is demanded in order to switch all of the parallelconnected MOSFETs on at the same time.
Another functionality that is crucial for this application is the current limit protection.With this functionality, the driver can turn off the power stage in the unlikely event thatthe digital system cannot respond to a failure situation in time. The UCD7100 has twoinputs to achieve this: a current sense (CS) pin and a current limit (Ilim) pin. Accordingto the datasheet these two inputs are compared and when the CS voltage level is greaterthan that of the ILIM voltage minus 25mV, the output of the driver is forced low. Areference voltage can be given to the Ilim pin to change the overcurrent threshold voltageof the MOSFET driver.
This current sense functionality is used to solve two possible problems: thermal run-away and short circuit conditions. The idea is to sense the voltage across the set ofMOSFETs while they are switched on. The voltage across a MOSFET can only increasedue to one of two reasons. Either the current increased through the MOSFET, or theMOSFET’s internal resistance increased due to a temperature rise. A dead-time circuit
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 44
is designed to delay voltage feedback from the CS pin as shown in Figure 3.14. Thisensures that the current limit protection is not activated during the time period requiredfor the MOSFETs to switch on.
R1
CS
R2
OUT
C1
D1
RDS(ON)
Figure 3.14: Dead time design
In Figure 3.14, OUT is the power output of the MOSFET driver, CS is the currentsense pin of the driver and RDS(ON) is the internal resistance when the MOSFET is turnedon. Before the MOSFET is switched on the diode D1 is typically reverse biased becauseof the high voltage across the SSC terminals. The diode is only forward biased whenthe MOSFET is switched on. Resistor R2 is used to limit the current flowing throughD1 during operation. Resistor R1 and capacitor C1 are added to introduce dead-timedead-time to the CS pin. This allows some time delay for the MOSFETs to switch onbefore normal operation of the UCD700 continues.
The derivation of the dead-time’s mathematical equation are done within the fre-quency domain. The assumption is made that the diode is reversed biased. A step inputat the OUT pin of the drive results in a voltage
VCS =
(1
C1R2
s+ R1+R2
C1R1R2
)(VOUTs
), (3.4.1)
at the CS pin. Calculating the inverse Laplace transform of (3.4.1) yields the time domainresponse
VCS =
(R1
R1 +R2
)(1− e−
R1+R2C1R1R2
t)VOUT . (3.4.2)
at the CS pin.A dead-time below 1µs is chosen. This minimises the risk of the SSC failing. The
increase in current, within the dead-time delay, is limited by the inductance of the con-ductors.
A check is required to ensure the MOSFETs turn on before the dead-time has passed.The MOSFET’s input gate pin is approximated by their total gate capacitance. The gatecapacitance can be estimated by Cg = Qg
Vgas 9.8 nF. The gate resistor, through which the
MOSFETs are controlled, is chosen as 15Ω. This yields a simple RC circuit. The timedomain function of the input gate pin can be approximated by
Vgate = VOUT (1− e −tRC ). (3.4.3)
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 45
According to the datasheet, the MOSFET is said to be fully switched on when Vgate =8V . The switch time for the gate voltage to reach 8V, is 161.5 ns. Adding the MOSFET’sturn-on delay of 33 ns yields a total time delay of 194.5 ns. This results in a large safetyfactor between the time required for the MOSFETs to be fully switched on and beforethe dead-time has passed.
Once the MOSFETs are switched on, the voltage present at the CS pin is the forwardbias voltage of the diode plus the voltage across the MOSFET. In order to enable bettercontrol of the output current of the gate driver, only the voltage across the MOSFET’sdrain and source terminals are required. This voltage is typically very small. The problemis that the diode voltage differs with current and temperature. It is for this reason that asimilar diode, D4, is placed in series with the reference resistors to set the current limit.When similar magnitude currents flows through both diodes, the two voltages cancel out.The manufactured SSC is shown in Figure 3.15.
3.4.3.4 Thermal Design
Ensuring that the SSC operate at the optimal temperature is crucial to making it morerobust. The thermal design of the SSC is discussed in the following section.
Power dissipation: The thermal design is conducted for a battery current of 120A.This adds an extra safety factor since the maximum continuous rated current is 104A.There are four MOSFETs in parallel and the assumption is made that the current aredivided equally between them. This results in a current of 30A per 2.8 mΩ MOSFET.The maximum power dissipated per MOSFET is 2.52 W. The switching losses are ignored,since the SSC is not design for high frequency switching.
Design: The maximum ambient temperature is selected as 40C. The maximum junc-tion temperature is chosen not to exceed 70C which results in a temperature rise of
Figure 3.15: Manufactured solid state contactor
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 3. HARDWARE DESIGN 46
30C. Therefore, the overall thermal resistance per MOSFET required for this design is11.9C/W. The MOSFET junction-to-case thermal resistance is 0.5C/W. The 7109DGfrom AAVID Thermalloy is a surface mount heat-sink with a thermal resistance of11C/W. It is chosen for this project for its small footprint. This results in a overallthermal resistance of 11.5C/W, which is within the required specifications. A heat-sinkis connected to every MOSFET on the PCB to dissipated the heat. The thermal resultswill be discussed in more detail in Chapter 6 as part of the results discussion.
3.5 ConclusionIn this chapter a proof of concept BMS was designed. The required design changes tothe proof of concept BMS were also implemented in order to scale the BMS for theapplication of a micro EV. This included splitting the designed circuitry into three parts,a main control circuit, balancing circuits and a current sensor circuit. This approachdelivered an adequate BMS in terms of cost and performance. A proof of concept designof a SSC was also investigated, which could possibly replace the mechanical contactor infuture iterations of the BMS.
Stellenbosch University https://scholar.sun.ac.za
Chapter 4
Software Design
4.1 IntroductionThe BMS is digitally controlled by means of a MCU. This ensures that the system isflexible and that it can be used with a wide range of different battery technologies withoutmaking hardware adjustments to the control circuitry. Only software changes have to beimplemented in order for the BMS to comply with different battery cells. The softwaredesign of the BMS is discussed within this chapter.
4.2 OverviewAn overview of the software design is discussed in this section. The MCU used for thisproject is the Piccolo TMS320F28035. The IDE used to program the MCU is Code Com-poser Studio from TI. The MCU is required to interact with a variety of components forthe BMS to function properly. The MCU controls the battery monitoring chip with theInter-Integrated Circuit (I2C) communications protocol. Battery data is communicatedvia Serial Peripheral Interface (SPI) to a Universal Serial Bus (USB) breakout boardwhich communicates with a computer. The battery data can also be requested via theCAN communication protocol. The CAN interrupts are used to respond to a measure-ment request. The analogue battery current is measured with an analogue-to-digitalconverter (ADC). Other peripherals used include Central Processing Unit (CPU) timersto gauge the time between samples taken.
The software design can be split into multiple sections and contains a main loop withan additional set of interrupts. The layout of the design is described in detail below.
1. The main loop is used to measure the cell voltages and battery temperature usingthe BQ79640 chip and thus forms the basis of the design. It sends the batterymeasurements via USB to a computer. Battery balancing is also performed withinthe main loop.
2. A current sensing loop is used to convert the current sensor’s analogue value intoa digital representation using the MCU’s ADC. The conversion triggers an ADCinterrupt after which the current sensing loop is executed.
47
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 4. SOFTWARE DESIGN 48
3. A master control loop, which is situated within a CPU timer interrupt, is used tocontrol the logic of the battery. The logic is controlled via flags that are set withinthe main loop and current sensing loop. These flags convey the status of the batteryat that specific time.
The design of the different software loops are discussed in the following sections.
4.3 Main loopA simplified flow diagram of the main loop is shown in Figure 4.1. The flow diagramis shown to have an initialisation phase. Within this phase all the different peripher-als are set up. It includes the set-up of the ADC, CPU timers, SPI, I2C and CAN.The BQ79640’s analogue protection and initialisation are also programmed within thisinitialisation phase.
Start
Initialise:BQ79640
CPU TimersI2CSPI
ADCCAN
Timer0interrupt
T
Read BQ79640measurements
CRC match
T
F
Compare toconstraints
T
Balance cells
Send datato USB
F Set flagsOpen contactor
F
Figure 4.1: Main flow diagram
An if statement in combination with the CPU timer 0 is used to execute the mainloop only once every second. Within this loop the battery measurements are read via theI2C communication protocol from the BQ79640. These measurements include the cellvoltage measurements and the battery temperature measurements. The data received isput through a cyclic redundancy check (CRC) method to ensure the validity of the data.This ensures a more robust system in a typically noisy environment.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 4. SOFTWARE DESIGN 49
The measurements taken are then compared to the operating specifications as givenby the manufacturer. When the battery is operated outside of the specified operatingrange, the appropriate error status flags are set. The contactor is also opened in the caseof a fault condition.
Once the measurements are taken, the balancing method is called to balance thebattery pack. The balancing function is only executed if the status flags indicate balancingis required. All the battery measurements are sent to the computer by the USB FT220breakout board. On the computer’s side, a python script is used to process the data andstore it in a database. This script is shown in appendix B.1.
4.4 Battery BalancingThe balancing algorithm is shown in Figure 4.2 and is located in the balancing functionwithin the main loop. The operation of the function can be described as follows: thebattery is charged with the traditional constant current - constant voltage (CC-CV)method.
As soon as one of the measured cell voltage is higher than 3.65V, charging is stoppedand balancing is started. All cells with a measured voltage of higher than 3.45V isbalanced through the balancing circuitry. Balancing is terminated for a particular cellonce the cell’s voltage drops below 3.45V. The cell balancing action is thus terminatedonce all the cell voltages are below the reference voltage. This algorithm was suggestedby James Verster, whom is the CEO of BlueNova, a local energy storage company.
This is not the general balancing algorithm used for Li-ion cells. Typically, cellsare balanced during charging. The reasons for not using the general algorithm are theapplication of the battery and the battery chemistry used in this project. The applicationof the battery allows the battery to be charged and balanced during nights when it it notused. The battery chemistry exhibits extreme non-linear behaviour at a high and lowSOC. This results in a large cell voltage difference between cells that have a small SOCdifference, when the battery pack is nearly fully charged. In most cases this results inthe battery not being charging in the CV mode for long enough periods of time in orderfor the charging current to decrease significantly. This results in the balancing currentbeing negligibly small compared to the charging current. Therefore, balancing is onlyactivated once charging has stopped. Another characteristic of the LFP cells is that theircell voltage do not stay constant at the charging voltage after being fully charged, insteadthe cell voltages settle at approximately 3.4V. This characteristic is combined within thebalancing algorithm as a threshold voltage where balancing is terminated.
The battery pack is balanced using the BQ76940 battery monitoring chip. One of theconstraints of the chip is that adjacent cells cannot be balanced simultaneously. Balancingadjacent cells simultaneously can possibly damage the integrated circuit. The balancingalgorithm is therefore adjusted to ensure that this does not happen. The even numberedcells are balanced inversely with the uneven numbered cells. This ensures that adjacentcells cannot be balanced simultaneously. Each set of cells are balanced for a time durationof 5 seconds. This minimises the amount of balancing commands sent from the MCUto the BQ76940 chip. This algorithm does have a negative impact on the time requiredfor cell balancing, since adjacent cells requiring balancing causes the average balancing
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 4. SOFTWARE DESIGN 50
current to be half of the maximum balancing current.The battery monitoring chip also impacts the average balancing current of the cells.
The total duty cycle devoted to cell balancing is approximately 70% of the 250ms period.This is due to a portion of the 250ms time duration is allocated to perform normal cellvoltage measurements using the BQ76940’s ADC. This ensures that the cell balancingdoes not affect the voltage measurements. It also ensures that the OV and UV protectionsdo not accidentally trigger, or that OV and UV conditions go undetected while cellbalancing is enabled.
Enterfunction
Initializevariables
Balancestatus
F
T
Counter==0
F
T
Resetbalancing
Balance unevencells above ref
Counter==period
F
T
Resetbalancing
Balance evencells above ref
Count ++
Counter >2 x period
F
T
All cellsbelow ref
F
T
Stopbalancing
Reset flagand counter
Stopbalancing
Exitfunction
Figure 4.2: Cell balancing flow diagram
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 4. SOFTWARE DESIGN 51
4.5 Current sense loopThe current sensing flow diagram is shown in Figure 4.3. It shows the ADC interruptwhich is triggered by Timer1 on the MCU. This interrupt has the highest priority toensure it is not interrupted while a fault condition exists. The mechanical contactorallows for a slower reaction time. The sampling rate is therefore chosen as 2 kHz, sincethis will ensure the MCU reacts fast enough to a fault condition.
Enter ADCInterrupt
Read: ADCvalue
Convert ADCvalue
Compare toconstraints
T
First orderdigital filter
Acknowledgeinterrupt
Exit ADCInterrupt
F
Set flagOpen contactor
Figure 4.3: Current sensing flow diagram
Once the interrupt is received, the ADC value is converted from a measured voltage toa corresponding digital current value. The output of the battery is disconnected from theload by means of the mechanical contactor if the current measurement is outside of themanufacturer’s specified operating range. A flag is also set to notify the master controlloop of the status of the battery. The current measurement is passed through a firstorder digital filter in order to filter all the excess noise. The interrupt is acknowledged toenable the ADC interrupt to read in new values.
4.6 Master control loopThe master control loop monitors the battery flags in order to determine the battery’smode. The battery output is controlled accordingly. The flags are for current, temper-ature, charged and discharged indications. The current flag is set in the current sensingloop if the battery current falls outside of the specified current boundaries. The temper-ature flag is set within the main loop if the measured battery temperature falls outside ofthe specified operating range. The charged flag is set when the battery is fully charged,
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 4. SOFTWARE DESIGN 52
when one of the cell voltages is higher than the maximum allowed voltage during charg-ing. The discharged flag is set when the battery is fully discharged, when one of the cellvoltages is lower than the minimum allowed voltage during discharging.
The master control loop, as shown in Figure 4.4, is executed within the CPU’s Timer1interrupt. This interrupt has a frequency of 100Hz. This allows the routine to quicklyrespond to the different external and internal inputs and interrupts.
There are three main modes that the battery can be in. In the normal mode all theflags are null and the charger is not connected. In combination with all these flags theBMS has some external inputs from the motor drive unit and the driver. These inputsare used to signal the BMS when to close the mechanical contactor if needed. One of theexternal inputs is the vehicle’s key switch used to signal the motor drive unit to start upand standby until the throttle is active. The key switch is also used to reset the currentflag after a fault condition.
Timer1Interrupt
Charger==0f_Discharge==0
F
T
Keyswitch==0
F
T
Keydrive==0
F
T
f_Current==0f_Temperature==0
F
T
Closecontactor
Opencontactor
Opencontactor
Setf_Current==0
Charger==1f_Charge==1
F
T
Charger==1
F
T
f_Temperature==0
F
T
Closecontactor
Opencontactor
Acknowledgeinterrupt
EndInterrupt
Figure 4.4: Master flow diagram
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 4. SOFTWARE DESIGN 53
The second mode the BMS can enter is the charged mode. This is when the chargeflag is high and the battery is still connected to the charger. The charge flag is set whenone of the cell voltages is higher than the maximum allowed voltage during charging.This results in the contactor opening to stop charging. Typically, this is when balancingstarts. The charged flag is reset to null when all the cell voltages decreases below aspecific reference level.
The third mode is the charging mode. The BMS enters this mode during batterycharging. More specifically, the mode is entered when the charger is connected and thecharge flag is null. The mechanical contactor is closed in order for the battery to charge.
4.7 ConclusionIn this chapter, the software layout of the BMS was discussed. This included the varioussoftware design choices made. Flow diagrams were used to explain the flow of the software.The main loop is used to monitor the cell voltages and battery temperature and is alsoused for battery balancing. The current sensing loop is used for current measurements.Flags are used to notify the master control loop which then controls the SOC and currentof the battery.
Stellenbosch University https://scholar.sun.ac.za
Chapter 5
On-line Parameter Estimation
5.1 IntroductionThis chapter investigates the on-line parameter estimation of an equivalent circuit model(ECM) of a battery. An ECM proves a good trade-off between model accuracy and modelcomplexity. Unfortunately, the battery parameters tend to change over time as the healthof the battery deteriorates (irreversible physical and chemical degradation). This makesthe on-line parameter estimation a more realistic approach to energy management.
A wide range of methods exist to accomplish on-line parameter estimation. Typically,a set of current and voltage measurements are used as the input and output of a systemrespectively. By minimizing the the error between the measured output and estimatedoutput, the parameters of the battery can be estimated with reasonable accuracy. In [41]the recursive least squares (RLS) algorithm is used to estimate the battery parameters.The RLS algorithm is easy to implement on-line and is computationally efficient. It istherefore chosen to determine the parameters of the ECMs.
The following chapter include the mathematical derivations of the single polarisation(SP)- and dual polarization (DP) Thévenin ECMs such that the RLS algorithm can beapplied to the models. A simulation, in combination with the RLS algorithm, is used toprove the derived mathematics.
5.2 Battery model
5.2.1 Introduction
In order to use the RLS algorithm to estimate the parameters of an ECM, the mathematicsdescribing the model is required to be written in the correct form. The SP Théveninmodel is evaluated first as it is the most common ECM. This model is then later inthe chapter extended to the DP Thévenin model which characterises the battery moreaccurately. The derivation of the models to the suitable RLS algorithm format is shownin the following subsections.
54
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 5. ON-LINE PARAMETER ESTIMATION 55
5.2.2 Single Polarisation Thévenin equivalent circuit model
The most commonly used ECM is the SP Thévenin model as shown in Figure 5.1. TheOpen Circuit (OC) voltage, VOC , is the equilibrium potential of the battery. For thepurpose of this thesis the OC voltage is assumed a function of the SOC of the battery.Also, the temperature is assumed to be constant. Thus, when the SOC decreases so willthe OC voltage. Typically this relationship is non-linear and will be determined off-line.The internal resistances include the ohmic resistance Rs and the polarization resistanceRt1. The equivalent capacitance Ct1 is used to describe the transient response.
−VOC+
Rs
Rt1
Ct1
Vt1+ −+
−
V (t)
I(t)
Figure 5.1: Single polarisation (SP) Thévenin model
The mathematical model is derived as
V (t) = VOC −RsI(t)− Vt1(t), (5.2.1)
Vt1(t) = Rt1I(t)−Rt1Ct1dVt1(t)
dt. (5.2.2)
The assumption is made that the open circuit voltage (VOC) remains constant overone sampling period [42]. By differentiating (5.2.1) with respect to time, a third equationis established. Solving these equations simultaneously, Vt1(t) can be eliminated in (5.2.1).The output V (t) is expressed in terms of I(t), dI(t)
dtand dV (t)
dtas shown in (5.2.3).
V (t) = VOC − [Rs +Rt1]I(t)−RsRt1Ct1dI(t)
dt−Rt1Ct1
dV (t)
dt(5.2.3)
This gives the same result as that in [43]. In the next section the more complex DPThévenin ECM is derived.
5.2.3 DP Thévenin equivalent circuit model
The DP Thévenin model is shown in Figure 5.2. This model adds an additional RCpair to more accurately match the dynamic behaviour of the battery voltage. The re-sistor Rt1 characterises the effective resistance of the electrochemical polarisation andRt2 characterises the effective resistance of the concentration polarisation of the battery.The capacitor Ct1 represents the electrochemical polarisation, whilst Ct2 represents the
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 5. ON-LINE PARAMETER ESTIMATION 56
−VOC+
Rs
Rt1 Rt2
Ct1
Vt1+ −
Ct2
Vt2+ −+
−
V (t)
I(t)
Figure 5.2: Dual polarisation (DP) Thévenin model
concentration polarisation, which is used to characterise the transient response of the bat-tery. According to [14], the DP Thévenin model presents the best dynamic performanceand offers a more accurate SOC estimation compared to other ECMs. In [44], a similarderivation was performed for the DP model, by transforming the state space model ofthe battery to the frequency domain. The derivation presented here is however muchsimpler, since all the equations are derived within the time domain, as shown below:
V (t) = VOC −RsI(t)− Vt1(t)− Vt2(t) (5.2.4)dV (t)
dt= −Rs
dI(t)
dt− dVt1(t)
dt− dVt2(t)
dt(5.2.5)
d2V (t)
dt2= −Rs
d2I(t)
dt2− d2Vt1(t)
dt2− d2Vt2(t)
dt2(5.2.6)
I(t) =Vt1(t)
Rt1
+ Ct1dVt1(t)
dt(5.2.7)
dI(t)
dt=
1
Rt1
dVt1(t)
dt+ Ct1
d2Vt1(t)
dt2(5.2.8)
I(t) =Vt2(t)
Rt2
+ Ct2dVt2(t)
dt(5.2.9)
dI(t)
dt=
1
Rt2
dVt2(t)
dt+ Ct2
d2Vt2(t)
dt2(5.2.10)
Again the assumption is made that the open circuit voltage, VOC , will remain constantover one sampling period. Equations (5.2.4) through (5.2.10) can be solved simultaneouslyusing a MATLAB solver function, as shown in Appendix B.2, to find V (t) in terms ofVOC and the other ECM parameters such that
V (t) = VOC − (Rs +Rt1 +Rt2)I(t)
− (Rs(Ct1Rt1 + Ct2Rt2) +Rt1Rt2(Ct1 + Ct2))dI(t)
dt
− Ct1Ct2Rt1Rt2RsdI2(t)
dt2− (Ct1Rt1 + Ct2Rt2)
dV (t)
dt
− Ct1Ct2Rt1Rt2dV 2(t)
dt2. (5.2.11)
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 5. ON-LINE PARAMETER ESTIMATION 57
In order to further simplify the above equation let Ct1Rt1=τt1 and Ct2Rt2=τt2 where τt1and τt2 are the time constants of the two RC pairs of the model, resulting in the simplifiedequation
V (t) = VOC − (Rs +Rt1 +Rt2)I(t)
− (Rs(τt1 + τt2) +Rt2τt1 +Rt1τt2)dI(t)
dt
−Rsτt1τt2dI2(t)
dt2− (τt1 + τt2)
dV (t)
dt− τt1τt2
dV 2(t)
dt2. (5.2.12)
5.3 Recursive Least Squares Algorithm
5.3.1 Introduction
The following section discusses the rewriting of the derived ECM mathematics into theform required by the RLS algorithm. The RLS algorithm is divided into 8 steps tosimplify the implementation thereof.
5.3.2 Recursive structure
The DP Thévenin ECM mathematics need to be rewritten into the required format forthe RLS algorithm in order to simplify the calculations. This is achieved by setting
V (t) = φT (t)θ, (5.3.1)
where the input vector is
φT (t) =[1 I(t) dI(t)
dtd2I(t)dt2
dV (t)dt
d2V (t)dt2
], (5.3.2)
and the parameters that needs to be estimated using the RLS algorithm is
θ =[θ1 θ2 θ3 θ4 θ5 θ6
]T
=
VOC
−(Rs +Rt1 +Rt2)−(Rs(τt1 + τt2) +Rt2τt1 +Rt1τt2)
−Rsτt1τt2−(τt1 + τt2)−τt1τt2
. (5.3.3)
In the case of an on-line application, the output V (t) and the input I(t) are discretisedat a constant sampling period T . Therefore, (5.3.1) requires to be converted into thediscrete form such that
V [k] = φT [k]θ. (5.3.4)
Within (5.3.2), I(t) and V (t) are discretised as I[k] and V [k] and represents thecurrent and voltage measured at a constant sample period T . The other variables can be
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 5. ON-LINE PARAMETER ESTIMATION 58
approximated by:
dI(t)
dt≈ I[k]− I[k − 1]
T, (5.3.5)
d2I(t)
dt2≈ I[k]− 2I[k − 1] + I[k − 2]
T 2, (5.3.6)
dV (t)
dt≈ V [k]− V [k − 1]
T, (5.3.7)
d2V (t)
dt2≈ V [k]− 2V [k − 1] + V [k − 2]
T 2. (5.3.8)
Once the vector θ is identified by the RLS algorithm, the parameters of the batterycan then be derived from (5.3.3) as follows:
VOC = θ1, (5.3.9)
Rs =θ4θ6, (5.3.10)
τt1 = −θ5 −√
(θ25 + 4θ6)
2, (5.3.11)
τt2 = −θ5 +√
(θ25 + 4θ6)
2, (5.3.12)
Rt1 =−1
θ25θ6 + 4θ26[2θ4θ6 + 2θ2θ
26 + θ3θ5θ6 + θ2θ
25θ6
+ τt2(2θ3θ6 − θ4θ5 + θ2θ5θ6)],
(5.3.13)
Rt2 =−1
θ25θ6 + 4θ26[2θ4θ6 + 2θ2θ
26 − θ3θ5θ6 + θ4θ
25
− τt2(2θ3θ6 − θ4θ5 + θ2θ5θ6)].
(5.3.14)
5.3.3 Recursive identification algorithm
As stated the RLS algorithm can be divided into 8 steps. By repeating these steps eachtime a new set of input and output data is measured, the RLS algorithm minimizes theestimated error. The algorithm should be initialised with the initial estimate of θ0 = 0and P0 = (106)I, where θ0 is the initial estimation, P0 is the initial covariance matrixand I is an identity matrix. This will place the emphasis on the incoming data insteadof on the input vector θ0.
In the 8 steps described below, K[k] is the recursive gain and P[k] is the covariancematrix, while λ is the forgetting factor (FF) of the RLS algorithm. The steps are asfollow:
1. Initialise parameters θ0 and P0.
2. Collect the input data I[k], I[k − 1]), I[k − 2] and the output data V [k], V [k − 1],V [k − 2].
3. Calculate φT [k].
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 5. ON-LINE PARAMETER ESTIMATION 59
4. Calculate the recursive gain matrix
K[k] =P[k − 1]φ[k]
λ+ φT [k]P[k − 1]φ[k]. (5.3.15)
5. Calculate the system output prediction
V [k] = φT [k]θ[k − 1]. (5.3.16)
6. Calculate the error between the measured output and the predicted output withthe previous estimated parameter vector values
E[k] = V [k]− V [k]. (5.3.17)
7. Calculate a new updated estimate of the system parameter vector
θ[k] = θ[k − 1] + E[k]K[k]. (5.3.18)
8. Calculate covariance matrix
P[k] =P[k − 1]−K[k]φ[k]P[k − 1]
λ. (5.3.19)
The battery parameters tend to change over time. Therefore, the dependence of theparameters on temperature, SOC and battery degradation is managed via the forgettingfactor as in (5.3.19). A small FF places more emphasis on new incoming data, whilea large FF places greater emphasis on older data. Typically the FF is chosen between0.95 and 1. If the value chosen for the FF is too small, the system will become unstable.Similarly, if λ = 1, the FF will not work at all, since older measurements are weighed thesame as new measurements.
5.4 Simulation and results
5.4.1 Introduction
In this section the suggested DP Thévenin ECM mathematics are proven using a simu-lation. The simulation can be split in two parts:
1. The ECM is simulated in Simulink to provide input (current) and output (voltage)data.
2. This data is then used, in an on-line manner by the RLS algorithm in order toestimate the parameters of the battery.
By comparing the parameters used inside the Simulink simulation with the estimatedparameters a conclusion can be drawn on whether the mathematical analysis is correct.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 5. ON-LINE PARAMETER ESTIMATION 60
5.4.2 Simulation
The Simulink simulation is discussed within this section. A block diagram of the Simulinksimulation is shown in Figure 5.3. The current is integrated to monitor the SOC of thebattery during discharge. A gain factor is used to scale this value to a percentage of thetotal battery capacity. This value is subtracted from the initial SOC in order to calculatethe remaining SOC of the battery. This SOC value is then fed into a polynomial function,which describes the relationship between the SOC and OC voltage, in order to supply areference OC voltage output to the voltage source. This results in a voltage, VOC(SOC),which is a voltage source controlled by a polynomial that describes the relationship be-tween the SOC and the OC voltage. The polynomial of the Shell Eco-marathon (SEM)battery pack is used in this simulation. The relationship was determined by [16] and willbe discussed in more detail in Chapter 6. The sampling period T is chosen as 1 s.
InitialSOC
+
− SOCpolynomial −
+VOC(SOC)
Gain
Integrator
currentRs Rt1 Rt2
Ct1
Vt1+ −
Ct2
Vt2+ −+
Rload
SidealPWM
In
-Ibat
Vbat Noise
Figure 5.3: Simulink simulation
The battery model parameters that were used is shown in Table 5.1.
Table 5.1: Simulink model parameters
Parameter ValueRs 1 ΩRt1 0.5 ΩRt2 0.5 ΩCt1 100 FCt2 20 F
The current profile at which the simulated battery model is discharged, is shown inFigure 5.4. The simulated battery is discharged from 95% to 15%. This ensures that thenon-linear relationship between the OC voltage and the SOC would be present in theoutput of the simulation. The current profile comprises of the sum of two components:
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 5. ON-LINE PARAMETER ESTIMATION 61
0 100 200 300 400 500 600 700
0
2
4
Time [s]
Current[A
]
Total battery I
Resistor I
Figure 5.4: Battery discharge profile
1. A discharge pulse through a resistor. The pulse has a period of 720 seconds and aduty cycle of 0.45.
2. A varying current to excite the dynamic behaviour of the battery. This varyingcurrent is added to the simulation by using a current source that is driven by noise.
In Figure 5.4 the resistor current is shown in red and the total battery current isshown in blue. The varying current is added to provide the system with some excitationduring periods when almost no dynamic behaviour is detected such as during constantcurrent periods. According to [45], poor excitation can lead to the exponential growth ofthe covariance matrix (or covariance wind-up). This results in the estimator becomingextremely sensitive and susceptible to numerical and computational errors.
5.4.3 Simulation results
The simulated measured output voltage (blue) and the estimated voltage (red) is presentin Figure 5.5. The error between the output voltage and estimated voltage, as a percent-age, can be seen in Figure 5.6. It is clear that the RLS algorithm effectively minimises theerror between output voltage and estimated voltage. The maximum error between thevoltage and estimated voltage is smaller than 1%. This proves that the RLS algorithmwas implemented successfully.
The estimated ECM parameters for the battery is shown in Figure 5.7. In each subfigure, the RLS algorithm’s estimation is indicated in red, while the Simulink simulationparameter value is indicated in blue. The estimated parameters compare closely withthe actual simulated parameters when the OC voltage is a constant value as shown inFigure 5.7a through 5.7f.
The parameters are unstable when the OC voltage of the battery changes too quickly.This is due to the RLS algorithm presented here inherently assume a model with a con-stant OC voltage. According to [17] the estimation response of the OC voltage is slow
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 5. ON-LINE PARAMETER ESTIMATION 62
0 100 200 300 400 500 600 700
45
50
55
Time [s]
Voltage[V
]
V
RLS estimated V
Figure 5.5: Estimated Voltage
0 1,000 2,000 3,000 4,000 5,000
−1
0
1
Time [s]
Estim
ationerror[%
]
Figure 5.6: Voltage estimation error
compared to that of the typical drive-cycle excitation dynamics. Typically, the RLS algo-rithm is combined with a current integration method since the two methods complementeach other. The purpose of the simulation was firstly to prove the mathematical modelwas correctly derived and secondly to test the implementation of the RLS algorithm.
The sampling period was also shown to have a significant impact on the accuracyof the estimation. There is a trade-off between the sampling time and the maximumvalue of the time constants of the ECM that can be estimated. The larger the samplingtime, the larger the time constants that can be estimated. This is due to the fact thatthe estimation rests on only three voltage and current measurements at a time. Forinstance, when the time constants of the model are very long and the sampling time isshort, the differences in the output are too small to measure. Under these circumstancesthe RLS algorithm cannot perform the estimation accurately. Increasing the sampling
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 5. ON-LINE PARAMETER ESTIMATION 63
0 1,000 2,000 3,000 4,000 5,00049
51
53
55
Time [s]
VOC
[V]
VOC
RLS estimated VOC
(a) VOC estimation
0 1,000 2,000 3,000 4,000 5,0000.6
0.7
0.8
0.9
1
1.1
1.2
Time [s]
Rs
[Ω]
RLS estimated Rs
Rs
(b) Rs estimation
0 1,000 2,000 3,000 4,000 5,0000
20
40
60
80
Time [s]
τ t1[s]
RLS estimated τt1τt1
(c) τt1 estimation
0 1,000 2,000 3,000 4,000 5,0000
0.25
0.5
0.75
1
Time [s]
Rt1
[Ω]
RLS estimated Rt1
Rt1
(d) Rt1 estimation
0 1,000 2,000 3,000 4,000 5,0000
2
4
6
8
10
Time [s]
τ t2[s]
RLS estimated τt2τt2
(e) τt2 estimation
0 1,000 2,000 3,000 4,000 5,0000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Time [s]
Rt2
[Ω]
RLS estimated Rt2
Rt2
(f) Rt2 estimation
Figure 5.7: Estimates of the RLS algorithm compared to the actual parameters
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 5. ON-LINE PARAMETER ESTIMATION 64
time too much on the other hand, impacts the assumption that the open circuit voltage,VOC , remains constant over one sampling period. The RLS method’s accuracy is alsoreduced by selecting a large sampling time. In practice, it is therefore important thatthe optimum sampling time is used. The optimal sampling time varies from one cellchemistry to another, but typically a sampling time of 1 - 2 seconds are used.
It is clear that the characteristics of the battery influence the accuracy that the practi-cal system requires in order to successfully implement the RLS algorithm. The simulationwas adjusted to investigate the accuracy required by a practical system, to accuratelyestimate the battery parameters using the RLS algorithm. This was achieved by addinga quantisation error to the input (current) and output (voltage) data. A quantisationerror typically occurs when the measurements are measured with an ADC. The simulateddata is rounded off to the third decimal in order to achieve a quantization error largeenough to affect the output of the RLS algorithm. This effect can be seen in Figure 5.8as an example of the RLS algorithm’s behaviour during such an input. It is clear thatthe RLS algorithm is extremely sensitive to low resolution data, since a 0.1mV accuracyis required on a voltage signal with an amplitude of up to 50V. This sensitivity is furthersupported by the fact that the covariance matrix of the current system used in the RLSalgorithm is ill-conditioned. Thus, a small change in input data has a great effect on theoutput data of the algorithm.
0 1,000 2,000 3,000 4,000 5,0000
2
4
6
8
10
Time [s]
τ t2[s]
T2 estimation
RLS est. τt2τt2RLS est. τt2 with quantization
Figure 5.8: τt2 quantization error result
5.5 ConclusionThis chapter discussed the mathematical derivation of an ECM to the RLS algorithm’sformat. The battery model was simulated in Simulink. The generated data was used incombination with the RLS algorithm in order to estimate the parameters of the simulatedbattery parameters in an on-line manner. The derivation was proven to be accurate andthe RLS algorithm was implemented successfully.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 5. ON-LINE PARAMETER ESTIMATION 65
The designed BMS’s accuracy unfortunately does not comply with the required spec-ifications for the RLS algorithm. This resulted in the on-line estimation being inaccurateand can thus not be implemented in the BMS. Further work will need to be conducted inorder to make the algorithm function properly with less accurate data. Other work thatneed to be investigated includes the minimisation of the noise on the measured data. Itmight be necessary to extend the current RLS algorithm into the recursive extended leastsquares (RELS) algorithm and or the Kalman filter in order to limit the impact of noise.
Stellenbosch University https://scholar.sun.ac.za
Chapter 6
Results
6.1 IntroductionThis section discusses the results of all the different components designed within thisthesis. This includes the SSC as well as the different components for the BMS.
6.2 Battery Management SystemThis section investigates the performance of all the different components within the BMS.This includes the voltage and current measurements and how these measurements areaffected.
6.2.1 BQ76940 Analogue to Digital Converter Accuracy
The voltage measurements obtained using the BQ76940 battery monitoring chip arediscussed within this section. After initial testing, the discovery was made that theBQ76940’s ADC measurement output at certain voltage references remains constant eventhough the voltage changes. An example of this behaviour is shown in Figure 6.1. Themeasurements were obtained after a discharge pulse during which the cells were giventime to reach equilibrium. All the cell voltages rise as can be expected. The measuredcell voltages also rise but at certain reference voltages are clamped to a constant valueas is shown in Figure 6.1. The reference value varies from cell to cell. This behaviour atone cell also does not influence the measurement of the other cell voltages. The sourceof this behaviour was investigated but was not found.
It is important to note that this non-linear behaviour’s impact on the accuracy of themeasurements are within the accuracy of the BQ76940 chip. A curve was fitted to themeasured data in order to calculate the voltage measurement error more accurately. Themaximum error was found to be 4mV, which is within the accuracy specification of theBQ76940 battery monitoring chip.
66
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 67
50800 51000 51200 51400 51600 51800 52000Time [s]
3.185
3.190
3.195
3.200
3.205
3.210
3.215
3.220
Volta
ge[V
]
cell 2cell 3cell 4cell 5
Figure 6.1: ADC error
6.2.2 Measurement Noise
This section investigates the measurement noise of the proof of concept BMS and thefull scale BMS. A lot of effort was put into minimising both the voltage and currentmeasurement noise on the full scale BMS. This included upgrading to the TPS54060buck regulator. The buck regulator has a small voltage ripple which in turn reduces thenoise induced on to the cell voltage and battery current measurements. The full scaleBMS design also makes use of a 4 layer PCB to minimise noise, where one layer was usedfor grounding and another layer was used for supply power to the PCB.
6.2.2.1 Voltage Measurements
The cell voltage measurement noise of the proof of concept BMS is compared to thatof the full scale BMS in this subsection. Unfortunately, the BQ76940 cannot measure avoltage of 0V. Therefore, the voltage measurement noise are isolated off-line by means ofa filter. The Python code is shown in Appendix B.3. Both signals are filtered with exactlythe same filter: A Butterworth filter with a sample rate of 1 Hz and a cut-off frequencyof 0.01 Hz. The filtered data is subtracted from to the measured data to estimate thenoise. The estimated noise of both systems are shown in Figure 6.2. It is clear that thefull scale BMS’s measurement noise is lower than that of the prototype BMS.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 68
0 200 400 600 800 1000Time [s]
−0.005
−0.004
−0.003
−0.002
−0.001
0.000
0.001
0.002
0.003
0.004Vo
ltage
[V]
Proof of concept BMS noiseFull scale BMS noise
Figure 6.2: Voltage measurement noise comparison
The power within the signals are calculated using
P =1
N
N−1∑n=0
|x[n]|2. (6.2.1)
The above equation quantifies the difference in power of the two noise signals. Theprototype BMS was calculated to have a noise power level of 21.9 µW/Ω compared to10.7 µW/Ω of the full scale BMS which confirms that the noise is attenuated in the fullscale design.
6.2.2.2 Current Measurements
One of the main motivations for minimising the voltage ripple of the supply voltage is tominimise the current measurement noise. The sensor used converts the analogue currentmeasurement into a reference voltage that is then measured using the MCU’s ADC. Acurrent value of up to ±150A is scaled down to a corresponding voltage of between zeroand the supply voltage of 3.3V. This makes the sensor extremely sensitive to noise. Thesensor output at a current of 0A is shown in Figure 6.3. The magnitude of the noise is1.6A. This results in a 0.53% error which is well within the desired error range of 1%.
In order to further decrease the noise, an infinite impulse response (IIR) filter isapplied in software. A simple low pass filter is used with the difference equation given inthe time domain by
y(k) = y(k − 1) + a[x(k)− y(k − 1)].
The variables y(k) and x(k) refer to the output and input data, respectively, while a =1− e−ωcTs . The cut-off frequency is selected as 4Hz and the corresponding filter outputis shown in Figure 6.4. The cut-off frequency is chosen according to the cell voltage
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 69
0 100 200 300 400 500Samples
−1.0
−0.5
0.0
0.5
1.0
1.5
Cur
rent
[A]
Noise
Figure 6.3: Current sensor noise
sampling frequency. The measurement noise is reduced to 150mA, which is less than10% of the original magnitude.
100 200 300 400 500Samples
−0.05
0.00
0.05
0.10
Cur
rent
[A]
Noise
Figure 6.4: Filtered current sensor noise
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 70
6.2.3 Current Sensor Thermal Test Result
A thermal test was conducted at a constant current of 116A, at room temperature,for a time of 5 minutes. A Fluke thermal camera was used to measure the maximumtemperature of the board and the result is shown in Figure 6.5. A temperature riseof 21C was measured. The temperature is seen to rise higher than anticipated due tothe power dissipated within the internal resistance of the current sensor, which was notincluded into the temperature calculation. The temperature rise is however still withinthe safe range of operation.
Figure 6.5: Current sensor thermal performance
6.2.4 Terminal Connections
The high current rating of the battery proposes various problems. One such problem isthe amount of contact required between the battery terminal and connector. A test wasundertaken to investigate the influence of a weak terminal connection. The battery packunderwent a 66A pulse discharge test. The cell measurements of the surrounding cellsare shown in Figure 6.6.
Cell 1 was purposefully connected with a weak terminal connection while cell 2 andcell 3 where connected normally. From Figure 6.6 it is clear that the cell voltage ofcell 1 is greatly affected compared to cell 2 and cell 3. The weak terminal connectionintroduces a high contact resistance. When large amount of current flows into or fromthe battery terminal, the high contact resistance dissipates a large amount of power. Thisgenerates heat at the terminal of the battery, which in turn reduces the available capacityof the cell. The voltage across the contact resistance also forces the two parallel cells todischarge unevenly. As soon as the connection is tightened, the two cells are shown toreach equilibrium again. This test also shows that basic algorithms can be used to ensurethat all the terminal connections are connected properly, by measuring the difference incell voltage.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 71
13000 14000 15000 16000 17000 18000 19000Time [s]
2.6
2.7
2.8
2.9
3.0
3.1
3.2
3.3
Volta
ge[V
]
cell 1cell 2cell 3
Figure 6.6: Weak terminal connections
6.2.5 Balancing
This section discusses the performance of the balancing circuits and the balancing al-gorithm. Cell balancing is implemented to maximise the usable capacity of the batterypack. The balancing was implemented according to the balancing algorithm discussed inChapter 4. The balancing reference voltage was chosen as 3.45V. This resulted in thecells being balanced until their voltage falls below the chosen reference voltage, as shownin Figure 6.7. Additional cell measurements can be seen in Appendix B.4.
The cell voltage reference value can be further reduced if the cells require more bal-ancing. Unfortunately, this will also result in a longer balancing time. The cell balancing,for this test, was achieved within 36 minutes. Further field testing is however required inorder to find the optimal balancing current and time, since these parameters are typicallydependant upon the application that the battery is used for.
During balancing cell 6 and cell 11’s voltage measurement are influenced as shown inFigure 6.8. This can be attributed to the operation of the BQ79640 battery monitoringchip. It is internally divided into three sections, each working independently. Each ofthese sections monitors five cells and are stacked on top of each other to achieve an overallcell count of 15. While cell balancing is active, these sections typically turn off balancingfor a short duration of time to measure the cell voltages. Unfortunately, in some instancesthese dead-times are not synchronised between the different sections. It is thus possiblefor one cell to be measured while another cell is being balanced. Take for instance cell10 and cell 11 in Figure 6.8, where a clear influence is notable during balancing.
All the terminal connections connected to the main control PCB are fused except theground connection. Cell voltages are measured from the one positive terminal connectionto the positive connection of the next cell’s positive terminal to reduce the amount ofconductors connected to the main control PCB.
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 72
0 500 1000 1500 2000
Time [s]
3.35
3.40
3.45
3.50
3.55
3.60V
olta
ge[V
]
cell 3cell 4cell 5
Figure 6.7: Cell balancing
500 1000 1500 2000
Time [s]
3.40
3.45
3.50
3.55
3.60
3.65
3.70
Vol
tage
[V]
cell 10cell 11
Figure 6.8: Impact of protection fuse during cell balancing
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 73
In this instance the balancing current of cell 10 flows through a 2A fuse, which affectsthe voltage measurement of cell 11. The fuse has an internal resistance of 50mΩ. Abalancing current of 1A thus results in a 50mV voltage offset, as seen in Figure 6.8. For-tunately, this behaviour can be ignored by implementing software changes. The analogueovervoltage threshold of the BQ76940 chip is also adjusted to account for this possiblevoltage offset. The fuse can be moved within the circuitry or can be replaced by a fusewith a smaller resistance, in future iterations of the BMS, to solve this problem.
6.3 Battery Off-line Parameter Estimation
6.3.1 Introduction
In this section the off-line parameter estimation of the DP Thévenin ECM is estimated forthe LFP battery pack. Firstly, the total battery pack capacity was validated accordingto the standard discharge test of the battery manufacturer. The capacity was determinedas only 150Ah, which is 75% of the rated 200Ah capacity specified. A later discussionwith the Chinese manufacturer revealed that they sent the wrong capacity cells. The85Ah version of the cells where shipped, not the 100Ah cells. This is very peculiarsince the manufacturer claimed that both versions have the same physical size and thatcurrently the 100Ah cells are out of stock. Unfortunately, this discussion only occurredafter the off-line tests were already conducted and the battery was returned to the microelectric vehicle manufacturer. For these reasons all the tests were performed at the 150Ahcapacity.
The DP Thévenin ECM is again chosen for off-line parameter estimation due to itshigh accuracy. For convenience the model is repeated here and shown in Figure 6.9. Themodel has six different parameters that require to be estimated. All these parameters areestimated according to [16]. The test consists of pulse discharging the battery through aconstant load. Unfortunately, no HPPC tests where concluded since during the testingphase, there where no high power chargers available. The hysteresis effect of the LFPcells were also ignored, as in [16].
The voltage of the battery during the pulse discharge test is shown in Figure 6.10. Theduration for the battery to reach equilibrium was determined by discharging the batteryat a constant rate before giving it time to settle at a constant voltage. The duration forthe dynamic behaviour of the battery to reach 98% of it’s final value was determined as
VOC(SOC)
Rs
Rt1 Rt2
Ct1
Vt1+ −
Ct2
Vt2+ −+
−
V(t)
I(t)
Figure 6.9: Dual polarization Thévenin equivalent circuit model
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 74
0 10000 20000 30000 40000 50000 60000Time [s]
42
44
46
48
50
52Vo
ltage
[V]
Voltage
Figure 6.10: Pulse discharge test
90 minutes. The datasheet’s standard discharge current (66A) was chosen as the averagedischarge rate of the test. The battery was discharged 10% at a time. The coulombcounter of the BMS was used during discharge to monitor the SOC of the battery.
6.3.2 Parameter Estimation
The estimation of the model parameters are discussed in the following section. The SOCrelationship with the OC voltage is shown in Figure 6.11. A polynomial curve fit wasimplemented in order to estimate a continuous relationship between the two variables.
Secondly, resistor Rs was calculated by using the immediate voltage rise during onesample period after the contactor opened, while the current was measured just before thecontactor opened. The assumption was made that for a step input, the the dynamic effectsover one sampling period is negligibly small compared to that of the static componentsin the model. The resistance Rs is shown in Figure 6.12 with respect to SOC. FromFigure 6.12 it can be seen that the battery’s ohmic resistance increases at low- and highSOC. The magnitude of the resistor Rs rises 13% from its minimum to maximum value.
Lastly, the remaining parameters were estimated using Matlab’s curve fit toolbox.The dynamic behaviour of the battery is approximated by
Vdynamic(t) = Ibat[Rt1e
−tRt1Ct1 +Rt2e
−tRt2Ct2 ]. (6.3.1)
A curve fit of this form is applied onto the battery data. As an example, the curvefit of the dynamic behaviour at 90% SOC is shown in Figure 6.13. It is clear that thecurve fit approximates the battery’s behaviour accurately. This confirms the model’sability to accurately model the battery dynamics. The required parameters were then
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 75
0 20 40 60 80 100State of charge [%]
44
46
48
50
52O
pen
circ
uitv
olta
ge[V
]
Open circuit voltage datasetPolynomial curve fit
Figure 6.11: SOC vs OC voltage
extracted from the curve fit result. The approximated resistance Rt1 and Rt2 is shown inFigure 6.14.
The total battery resistance, Rtotal = Rs + Rt1 + Rt2, is shown in Figure 6.15. Thebattery’s maximum internal resistance, as expected, is found at a low SOC. The inter-nal resistance decreases as the SOC increases up to 60%, where the minimum internal
10 20 30 40 50 60 70 80 90State of charge [%]
0.0100
0.0105
0.0110
0.0115
0.0120
0.0125
Res
ista
nce
[Ω]
Ohmic battery resistance Rs
Figure 6.12: Ohmic resistance characteristic curve of the battery
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 76
0 1000 2000 3000 4000 5000 6000State of charge [%]
0.0
0.2
0.4
0.6
0.8
Tim
eco
nsta
nt[s
]
Battery dynamic behaviourCurve fit
Figure 6.13: Curve fit of dynamic behaviour
10 20 30 40 50 60 70 80 90State of charge [%]
0.004
0.006
0.008
0.010
0.012
Res
ista
nce
[Ω]
Concentration polarization resistance Rt2
Electrochemical polarization resistance Rt1
Figure 6.14: Dynamic resistance characteristic curve of the battery
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 77
resistance is found. This is typically where Li-ion batteries are most stable. The internalresistance increases gradually as the SOC increases to 100%. The magnitude of the totalinternal resistance rises 37.5% from its minimum to maximum value.
10 20 30 40 50 60 70 80 90State of charge [%]
0.018
0.020
0.022
0.024
0.026
0.028
Res
ista
nce
[Ω]
Total battery resistance
Figure 6.15: Total battery resistance characteristic curve
The two estimated time constants, τt1 and τt2, is shown in Figure 6.16 and 6.17respectively. τt1 and τt2 represents the time constants RT1CT1 and RT2CT2. Time constantτt1 stays relatively constant with a rise of 9% from its minimum to maximum value. Thetime constant τt2 varies significantly over the discharge cycle, up to 41% from its minimumto maximum value. It is not uncommon for the concentration polarisation to vary to thismagnitude as can also be seen in [16].
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 78
10 20 30 40 50 60 70 80 90State of charge [%]
1250
1300
1350
1400
1450
1500Ti
me
cons
tant
[s]
Time constant t1
Figure 6.16: Time constant τt1 characteristic curve
10 20 30 40 50 60 70 80 90State of charge [%]
50
55
60
65
70
75
80
85
90
Tim
eco
nsta
nt[s
]
Time constant t2
Figure 6.17: Time constant τt2 characteristic curve
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 79
6.4 Solid State ContactorIn this section the performance of the SSC is discussed. This includes the normal opera-tion, fault conditions and thermal analysis of the SSC.
6.4.1 Test Setup
The performance of the designed SSC is investigated within the following subsection. Thetest set-up is shown in Figure 6.18. A 48V lead-acid battery is used to deliver the requiredcurrent through the SSC. Fuse F1 is used to limit the battery current in case of a faultcondition. Switch S1 is an emergency switch that can be used to disconnect the batteryfrom the load in the case of an emergency. The current through the SSC is measuredwith a current probe. The SSC is powered by a small 12V supply. A signal generatoris used to generate an input to the SSC. Diode D1 is used as a freewheeling diode whenthe SSC is opened in order to provide a path for the current to flow. This decreases theamount of power that is to be dissipated by the TVS within the SSC. Resistor R1 is avariable resistor that is used to control the magnitude of the current flowing through theSSC. The high current carrying components are indicated with thicker lines.
48 V
400 A
F1
S1
Current probe
SSC
D1 R1
Supply
Input
Figure 6.18: SSC test set-up
6.4.2 Normal Operation
The SSC’s normal operation is tested first. The SSC is required to operate at a steadystate current of up to 120A without the overcurrent protection tripping. The resistiveload is setup to allow a maximum continuous current of 120A. The SSC is closed by thesignal generator. The SSC current and the input voltage to the MOSFET gate driver areshown in Figure 6.19. It is clear that the overcurrent was not activated.
The current sense pin voltage and the current reference pin voltage of the MOSFETdriver are shown in Figure 6.20. The trip condition is activated when the current sensevalue exceeds the reference value minus 25mV as stated by the datasheet of the UCD7100MOSFET driver. The current sense value remains below this reference value as shown
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 80
−0.0001 0.0000 0.0001 0.0002 0.0003 0.0004 0.0005 0.0006Time [s]
0
20
40
60
80
100
120C
urre
nt[A
]
SSC currentSSC gate driver input
0
2
4
6
8
10
12
Volta
ge[V
]
Figure 6.19: SSC current at maximum load
−0.00005 0.00000 0.00005 0.00010 0.00015 0.00020Time [s]
−0.1
0.0
0.1
0.2
0.3
0.4
0.5
Volta
ge[V
]
Current referenceCurrent sense
Figure 6.20: Current sense and reference pin at maximum load
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 81
in Figure 6.20. It can be seen that the reference value is affected by the large change incurrent of the SSC. As soon as the SSC current settles at a constant value, the referencevalue also settles. The change in current induces a voltage on the reference pin. Thiscoupling effect will have to be eliminated in the next iteration of the SSC.
The thermal test of the SSC is performed at a constant current of 120A. The testis conducted at room temperature. The temperature of the the SSC is measured witha Fluke thermal camera. The SSC is closed for a period of five minutes before thetemperature is measured. The test results of the thermal camera is shown in Figure 6.21.The SSC reaches a maximum temperature of up to 51C around the MOSFETs. Thisresults in a temperature rise of about 26C which confirms the thermal design. Thethermal design was implemented for a maximum temperature rise of 30C. The openPCB areas were not included in the thermal design, even though it does contain thermalconduction. This explains why the temperature rise is lower than anticipated.
Figure 6.21: SSC temperature at maximum load
6.4.3 Overcurrent Trip
This section investigates the SSC’s ability to open in the case of an overcurrent condition.This feature is of great importance in order to make the SSC more robust. A mechanicalcontactor is very robust in the sense that its maximum failing current is much higherthan its rated current. For example, the contactor used for this project has a maximumbreaking capacity of eight times its rated capacity. The BMS thus has a relatively longtime period to react in the case of a fault condition.
Unfortunately, the SSC is much more sensitive. When the current that is required tobe disconnected is too large, the risk of the MOSFETs not closing increases. This thusmakes the SSC less robust. The SSC has over an overcurrent trip with a fast reactiontime for this reason. This feature protects the SSC from a short circuit fault as well asfrom thermal runaway.
The resistive load, for this test, is decreased until the SSC’s overcurrent trip is acti-vated. Decreasing the resistive load increases the SSC current. An example of the SSCovercurrent trip is shown in Figure 6.22. It opens at a current of 133A even though the
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 82
−0.00005 0.00000 0.00005 0.00010 0.00015 0.00020Time [s]
0
20
40
60
80
100
120C
urre
nt[A
]
SSC currentSSC gate driver input
0
5
10
15
Volta
ge[V
]
Figure 6.22: SSC trip
input to the system is still active. The input is required to be reset in order to resetthe SSC. This proves that the resistance of the MOSFETs can be used, to relative accu-racy, to measure the current flowing through them. The feedback loop also performed asdesigned. The SSC’s overcurrent trip protects it from a fault condition.
The current sense pin voltage and the current reference pin voltage of the MOSFETdriver is shown in Figure 6.23. The trip condition is shown to be activated when thecurrent sense value is larger than the reference value minus 25mV. This leads to theMOSFET driver opening the MOSFETs. The change in current that induces a voltageon the reference pin can once again be seen. Fortunately, it is not large enough to impactthe operation of the SSC.
Both the reference and sense pins of the MOSFET driver, experience a transientresponse during the initial start of switching, as shown in Figure 6.24. The source ofthese oscillations were not found. It is present at no load conditions as well as at a lowerMOSFET switching time. This will need to be investigated further in future work.
The dead-time of the current sense pin voltage is also shown in Figure 6.24. The deadtime is 800 ns, which confirms the design choice of a dead-time below 1µs.
One of the concerns with using MOSFETs to break a current, is how to protect theMOSFET’s terminal voltage. The MOSFET typically switch very fast, which could leadto a voltage spike on the terminal of the MOSFET (VDS) because of the inductance withinthe conductors. Transient voltage suppressors (TVSs) are therefore used to ensure theMOSFETs are protected against overvoltage conditions. The current and voltage of theSSC are shown in Figure 6.25, during turn-off. There is a voltage spike at the beginning assoon as the MOSFETs close. The voltage spike activates the TVSs. The TVSs conducts
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 83
−0.00005 0.00000 0.00005 0.00010 0.00015 0.00020Time [s]
−0.1
0.0
0.1
0.2
0.3
0.4
0.5Vo
ltage
[V]
Current referenceCurrent sense
Figure 6.23: Current sense and reference pin
0.0000005 0.0000010 0.0000015 0.0000020 0.0000025Time [s]
−0.2
0.0
0.2
0.4
0.6
Volta
ge[V
]
Current senseCurrent reference
Figure 6.24: Dead time
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 6. RESULTS 84
−0.00001 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005Time [s]
0
20
40
60
80
100
120C
urre
nt[A
]SSC currentSSC terminal voltage
0
20
40
60
80
100
120
Volta
ge[V
]
Figure 6.25: SSC during turn-off
the current until the current reaches zero, thus effectively clamping the voltage below themaximum voltage of 100V. When the current reaches zero, the TVSs deactivate until thenext turn-off cycle.
6.5 ConclusionThis section discussed the various results of the BMS. The BMS’s voltage and currentmeasurement noise were investigated, both of which were reduced due to design changesimplemented from the proof of concept design BMS. The balancing algorithm was imple-mented within the BMS and the BMS was successfully used to balance the LFP batterypack. The BMS was also successfully used to estimate the DP Thévenin ECM parametersof the LFP battery pack in an off-line manner.
This section also discussed the various results of the SSC. The SSC’s overcurrenttrip function was implemented and it successfully protected the SSC from an overcurrentcondition. The thermal design of the SSC was verified. The overvoltage protection whilebreaking the rated current of the SSC was also validated.
Stellenbosch University https://scholar.sun.ac.za
Chapter 7
Conclusion
7.1 IntroductionThe conclusion to the different components and algorithms investigated within this thesisare discussed in the following section. Possible future work is also discussed.
7.2 ConclusionAn overview is presented to show how the thesis addressed the objectives as set out inChapter 1. These were:
• Development of a proof of concept Li-ion BMS to ensure the basic working principleof the BMS.
• Development of a full scale Li-ion BMS for the purpose of a micro EV.
• The successful implementation of a balancing algorithm within the BMS.
• Off-line parameter estimation using the BMS.
• The development of an on-line parameter estimation algorithm using an appropriatebattery model.
• Development of a proof of concept SSC.
7.2.1 Development of a proof of concept Li-ion BMS to ensurethe basic working of the BMS
The proof of concept BMS was designed, manufactured and tested. The system func-tioned with sufficient performance and proved the operation of the various subsystems.The system was shown to be ideal to monitor small battery packs. It also provideda good basic understanding of BMSs and good insight into what was required for thedevelopment of the full scale BMS.
85
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 7. CONCLUSION 86
7.2.2 Development of a full scale Li-ion BMS for the purpose ofa micro EV
The full scale BMS was also designed, manufactured and tested. The system deliveredimprovements in performance compared to that of the proof of concept version. Thiswas mainly due to the full scale BMS’s ability to minimise the supply voltage ripple.All of the different subsystems performed according to the design specifications. Thesesubsystems include the current sensor circuit, balancing circuit and main control circuitof the full scale BMS.
7.2.3 The successful implementation of a balancing algorithmwithin the BMS
Passive battery cell balancing was successfully implemented within the BMS. Balancingwas achieved by the integration of the hardware design as well as the software design.
7.2.4 Off-line parameter estimation using the BMS
The BMS was used to monitor the states of the battery during a pulse discharge cycle.This data was used off-line to estimate the parameters of a DP Thévenin ECM. Theseparameters can be used in future work to monitor for example how the battery ages overtime.
7.2.5 The development of an on-line parameter estimationalgorithm using an appropriate battery model
A derivation of the DP Thévenin ECM mathematics to the RLS algorithm format waspresented in Chapter 5. A Simulink simulation was used to simulate a DP Thévenin ECMthrough a discharge cycle. The derivation was confirmed with the RLS algorithm in anon-line manner. The RLS algorithm was found to be accurate, but extremely sensitiveto measurement accuracy and noise.
7.2.6 Development of proof of concept SSC
A proof of concept SSC was designed, manufactured and tested. The performance of theSSC, in terms of efficiency, was shown to be competitive compared to that of the me-chanical contactor. The SSC was shown to be more energy efficient within the operatingconditions that the SSC was designed for. Unfortunately, it is very expensive comparedto the mechanical contactor.
7.3 Future work
7.3.1 Battery management system
The addition of an optional real time calibration system can be added to increase themeasurement accuracy of the BQ76940. This typically entails externally calibrating the
Stellenbosch University https://scholar.sun.ac.za
CHAPTER 7. CONCLUSION 87
measurements using the host MCU. Further software development can also decrease thepower consumption of the BMS for example using the MCU’s sleep mode.
The current sensor’s noise sensitivity can also be reduced by powering it with 5V andnot 3.3V. At 3.3V it typically produces a measurement offset. This will unfortunatelyresult in having an extra supply rail.
7.3.1.1 Battery cell balancing
The optimal magnitude of the balancing current will need to be investigated for thisspecific BMS. This will possibly lead to increasing the battery balancing current to shortenthe battery balancing time. Other balancing algorithms can also be investigated to findan algorithm that can balance the cells while the battery is actively being used.
The 2A protection fuse used in the balancing circuit’s design will have to be recon-sidered. This could possibly result in moving the fuse or to chose a fuse with a smallerinternal resistance.
7.3.2 On-line parameter estimation
Methodologies for making the RLS algorithm less susceptible to measurement accuracywill have to be investigated in future work. The impact of noise will also have to beinvestigated. This could include upgrading the RLS algorithm to a Kalman filter.
7.3.3 Solid state contactor
The cost of the SSC will need to be reduced in order to make it a financially viableoption. This will have to be investigated in future work. It could typically be achievedby decreasing the size of the PCB and reducing the amount of MOSFETs used. Thedesign needs to be optimised in terms of cost and efficiency with the specific load inmind. The transient response of the current feedback also requires to be investigated andminimised. This can lead to the SSC being more robust.
Stellenbosch University https://scholar.sun.ac.za
Bibliography
[1] M. Taylor, K. Daniel, A. Ilas, and E. Young, “Renewable Power Generation Costsin 2014,” International Renewable Energy Agency (IRENA), Tech. Rep., 2015.[Online]. Available: www.irena.org/publications
[2] R. Xiong, F. Sun, X. Gong, and C. Gao, “A data-driven based adaptive state ofcharge estimator of lithium-ion polymer battery used in electric vehicles,” AppliedEnergy, vol. 113, pp. 1421–1433, 2014.
[3] Energy Storage Association, “Redox Flow Batteries.” [Online]. Available:http://energystorage.org/energy-storage/technologies/redox-flow-batteries
[4] Cornell University, “Battery Anodes.” [Online]. Available: http://www.emc2.cornell.edu/content/view/battery-anodes.html
[5] F. Badin, Hybrid Vehicles From Components to System. Editions Tech-nip, 2013. [Online]. Available: http://gen.lib.rus.ec/book/index.php?md5=AAD1D2CA3838E8CAD1D1416B8EB23F7A
[6] J. Molenda and M. Mole, “Composite Cathode Material for Li-Ion Batteries Based onLiFePO4 System.” in Metal, Ceramic and Polymeric Composites for Various Uses.InTech, jul 2011, ch. 30.
[7] B. Scrosati, J. Garche, and W. Tillmetz, Advances in Battery Technologies for Elec-tric Vehicles, 1st ed. Woodhead Publishing, 2015.
[8] M. Bingeman and B. Jeppesen, “Improving Battery Management System Perfor-mance and Cost with Altera FPGAs,” Altera, Tech. Rep. January, 2015.
[9] A. Hausmann and C. Depcik, “Expanding the Peukert equation for battery capacitymodeling through inclusion of a temperature dependency,” Journal of Power Sources,vol. 235, pp. 148–158, 2013.
[10] L. Qian, Y. Si, and L. Qiu, “SOC estimation of LiFePO4 Li-ion battery using BPNeural Network,” in 28th International Electric Vehicle Symposium and Exhibition,2015, pp. 1–7. [Online]. Available: http://www.a3ps.at/site/sites/default/files/downloads/evs28/papers/A4-03.pdf
[11] D. Jiani, L. Zhitao, W. Youyi, and W. Changyun, “A fuzzy logic-basedmodel for Li-ion battery with SOC and temperature effect,” in 11th IEEEInternational Conference on Control & Automation (ICCA). IEEE, jun 2014, pp.
88
Stellenbosch University https://scholar.sun.ac.za
BIBLIOGRAPHY 89
1333–1338. [Online]. Available: http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=6871117
[12] J. Jiuchun and Z. Caiping, Fundamentals and Application of Lithium-ion Batteriesin Electric Drive Vehicles, 1st ed. Wiley, 2015. [Online]. Available: http://gen.lib.rus.ec/book/index.php?md5=11AA3536910E17170FF614920CB688D4
[13] “FreedomCAR Battery Test Manual For Power-Assist Hybrid Electric Vehicles,”Idaho National Engineering and Environmental Laboratory, Tech. Rep., 2003.
[14] H. He, R. Xiong, and J. Fan, “Evaluation of Lithium-Ion Battery EquivalentCircuit Models for State of Charge Estimation by an Experimental Approach,”Energies, vol. 4, no. 12, pp. 582–598, mar 2011. [Online]. Available:http://www.mdpi.com/1996-1073/4/4/582/
[15] H. Hongwen He, R. Rui Xiong, X. Xiaowei Zhang, F. Fengchun Sun, andJ. JinXin Fan, “State-of-Charge Estimation of the Lithium-Ion Battery Usingan Adaptive Extended Kalman Filter Based on an Improved Thevenin Model,”IEEE Transactions on Vehicular Technology, vol. 60, no. 4, pp. 1461–1469, may2011. [Online]. Available: http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=5739545
[16] D. Gandolfo, A. Brandão, D. Patiño, and M. Molina, “Dynamic model of lithiumpolymer battery: Load resistor method for electric parameters identification,” Jour-nal of the Energy Institute, vol. 88, no. 4, pp. 470–479, 2015.
[17] A. Emadi, Advanced Electric Drive Vehicles. CRC Press, 2014.
[18] “EIS Measurement of a Very Low Impedance LithiumIon Battery,” Gamry instruments, Tech. Rep., 2011.[Online]. Available: http://www.gamry.com/application-notes/EIS/eis-measurement-of-a-very-low-impedance-lithium-ion-battery/
[19] L. Lu, X. Han, J. Li, J. Hua, and M. Ouyang, “A review on the key issues forlithium-ion battery management in electric vehicles,” Journal of Power Sources, vol.226, pp. 272–288, 2013.
[20] D. Andre, C. Appel, T. Soczka-Guth, and D. U. Sauer, “Advanced mathematicalmethods of SOC and SOH estimation for lithium-ion batteries,” Journal of PowerSources, vol. 224, pp. 20–27, 2013.
[21] M. Corno, N. Bhatt, S. M. Savaresi, and M. Verhaegen, “Electrochemical Model-Based State of Charge Estimation for Li-Ion Cells,” IEEE Transactions on ControlSystems Technology, vol. 23, no. 1, pp. 117–127, jan 2015. [Online]. Available:http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=6786991
[22] J. Xu, B. Cao, J. Cao, Z. Zou, C. C. Mi, and Z. Chen, “A Comparison Studyof the Model Based SOC Estimation Methods for Lithium-Ion Batteries,” in 2013IEEE Vehicle Power and Propulsion Conference (VPPC). IEEE, oct 2013, pp.1–5. [Online]. Available: http://ieeexplore.ieee.org/document/6671653/
Stellenbosch University https://scholar.sun.ac.za
BIBLIOGRAPHY 90
[23] L. W. Juang, P. J. Kollmeyer, T. M. Jahns, and R. D. Lorenz, “Implementationof online battery state-of-power and state-of-function estimation in electric vehicleapplications,” in 2012 IEEE Energy Conversion Congress and Exposition (ECCE).IEEE, sep 2012, pp. 1819–1826. [Online]. Available: http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=6342591
[24] Analog Devices, “Lithium Ion Battery Monitoring System AD7280A,” 2011.[Online]. Available: http://www.analog.com/media/en/technical-documentation/data-sheets/AD7280A.pdf
[25] Linear Technology, “Multicell Battery Stack Monitor LTC6802-1.” [Online].Available: http://cds.linear.com/docs/en/datasheet/68021fa.pdf
[26] Texas Instruments, “bq769x0 3-Series to 15-Series Cell Battery MonitorFamily for Li-Ion and Phosphate Applications,” 2013. [Online]. Available:http://www.ti.com/lit/ds/slusbk2g/slusbk2g.pdf
[27] I. Aizpuru, U. Iraola, J. M. Canales, M. Echeverria, and I. Gil, “Passivebalancing design for Li-ion battery packs based on single cell experimentaltests for a CCCV charging mode,” in 2013 International Conference on CleanElectrical Power (ICCEP). IEEE, jun 2013, pp. 93–98. [Online]. Available:http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=6586973
[28] Texas Instruments, “TMS320F2803x Piccolo Microcontrollers,” 2009. [Online].Available: http://www.ti.com/lit/ds/symlink/tms320f28035.pdf
[29] ——, “Constant On-Time Synchronous Buck Regulator LM5017,” 2012. [Online].Available: http://www.ti.com/lit/ds/symlink/lm5017.pdf
[30] ——, “LMX9838 Bluetooth Serial Port Module,” 2007. [Online]. Available:http://www.ti.com/lit/ds/snosaz9e/snosaz9e.pdf
[31] ——, “SN65HVD23x 3.3-V CAN Bus Transceivers,” 2002. [Online]. Available:http://www.ti.com/lit/ds/symlink/sn65hvd235.pdf
[32] Future Technology Devices International Ltd, “UMFT201XB, UMFT220XB andUMFT230XB Datasheet.” [Online]. Available: http://www.ftdichip.com/Support/Documents/DataSheets/Modules/DS_UMFT201_220_230XB.pdf
[33] Texas Instruments, “TPS54060 Step Down DC-DC Converter,” 2009. [Online].Available: http://www.ti.com/lit/ds/symlink/tps54060.pdf
[34] D. G. Brooks and J. Adam, “Trace Currents and Temperatures Revisited,” Tech.Rep., 2015.
[35] Allegro MicroSystems, “ACS758 Hall Effect-Based Linear Current Sensor.” [On-line]. Available: http://www.allegromicro.com/en/Products/Current-Sensor-ICs/Fifty-To-Two-Hundred-Amp-Integrated-Conductor-Sensor-ICs/ACS758.aspx
[36] Albright International, “SW80 D.C. contactor.” [Online]. Available: http://www.albrightinternational.com/products/sw80/
Stellenbosch University https://scholar.sun.ac.za
BIBLIOGRAPHY 91
[37] Murata Power Solutions, “Isolated 1W Single Output DC/DC ConvertersMEU1 Series.” [Online]. Available: http://power.murata.com/data/power/ncl/kdc_meu1.pdf
[38] Fairchild, “General purpose phototransistor optocouplers 4N35.” [Online]. Available:https://www.fairchildsemi.com/datasheets/4N/4N35M.pdf
[39] Micro Commercial Components, “Transient Voltage SuppressionDiodes.” [Online]. Available: http://pdf.datasheet.company/datasheets-1/micro_commercial_components/5KP54CA-AP.pdf
[40] Texas Instruments, “Digital Power Control Driver UCD7100.” [Online]. Available:http://www.ti.com/product/UCD7100/technicaldocuments
[41] H. He, X. Zhang, R. Xiong, Y. Xu, and H. Guo, “Online model-basedestimation of state-of-charge and open-circuit voltage of lithium-ion batteries inelectric vehicles,” Energy, vol. 39, no. 1, pp. 310–318, 2012. [Online]. Available:http://dx.doi.org/10.1016/j.energy.2012.01.009
[42] Y.-H. Chiang, W.-Y. Sean, and J.-C. Ke, “Online estimation of internal resistanceand open-circuit voltage of lithium-ion batteries in electric vehicles,” Journal ofPower Sources, vol. 196, no. 8, pp. 3921–3932, 2011.
[43] T. Feng, L. Yang, X. Zhao, H. Zhang, and J. Qiang, “Online identificationof lithium-ion battery parameters based on an improved equivalent-circuitmodel and its implementation on battery state-of-power prediction,” Journalof Power Sources, vol. 281, pp. 192–203, 2015. [Online]. Available: http://dx.doi.org/10.1016/j.jpowsour.2015.01.154
[44] Xidong Tang, Xiaofeng Mao, Jian Lin, and B. Koch, “Li-ion batteryparameter estimation for state of charge,” in Proceedings of the 2011American Control Conference. IEEE, jun 2011, pp. 941–946. [Online]. Available:http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=5990963
[45] A. Vahidi, A. Stefanopoulou, and H. Peng, “Recursive least squares with forgettingfor online estimation of vehicle mass and road grade: theory and experiments,”Vehicle System Dynamics, vol. 43, no. 1, pp. 31–55, jan 2005. [Online]. Available:http://www.tandfonline.com/doi/abs/10.1080/00423110412331290446
Stellenbosch University https://scholar.sun.ac.za
Appendices
92
Stellenbosch University https://scholar.sun.ac.za
Appendix A
Calculations
A.1 LM5017 regulator designDesign specifications: Maximum input voltage of 54.6 V with a minimum input voltageof 41.6 V. The regulated output voltage is 5 V. Design is done for a maximum outputcurrent of 500 mA. The design can be seen below:
A.1.1 Feedback resistor selection
VOUT = VFB x (RFB2
RFB1
+ 1)
where VFB = 1.225 V. Standard values of 3 kΩ and 1 kΩ are chosen for RFB2 and RFB1
to result in an output voltage of 4.9 V.
A.1.2 Frequency selection
fSW (MAX) =1−DMAX
TOFF (MIN)
=1− 5/41.6
200 ns= 4.4 MHz
fSW (MAX) =DMIN
TON(MIN)
=5/54.6
100 ns= 0.916 MHz
A conservative switching frequency of 200 kHz is selected.
fSW =VOUTK.RON
= 200 kHz
RON =VOUTK.fSW
=5
1x10−10.200x103= 250 kΩ
RON is chosen smaller at 220 KΩ to ensure the switching frequency is always largerthan the reference 200 KHz.
93
Stellenbosch University https://scholar.sun.ac.za
APPENDIX A. CALCULATIONS 94
A.1.3 Inductor selection
The minimum inductance is selected to limit the output ripple to 15 % of the maximumload current.
∆IL =VIN − VOUTL1.fSW
xVOUTVIN
The maximum ripple is observed at maximum input voltage. Substituting VIN = 54.6V and ∆IL = 15 % x IOUT (MAX) results in L1 = 3.028 µH. The next higher standardvalue of 330 µH is chosen.
A.1.4 Output capacitor selection
The output capacitor is selected to minimise the output voltage ripple. The ripple canbe calculated by:
COUT =∆IL
8.fSW .∆Vripple
A 22 µF ceramic capacitor is chosen. This results into a voltage ripple of 2.13 mV.
A.1.5 Input Capacitor selection
The input capacitor is selected to minimise the input voltage ripple. The ripple can becalculated by:
CIN ≥IOUT (MAX)
8.fSW .∆VIN
Two 2.2 µF ceramic capacitors are placed in parallel to result in a total of 4.4 µF.This results into a voltage ripple of 82 mV.
A.1.6 Undervoltage lockout resistors selection
These resistors set the undervoltage lockout (UVLO) and hysteresis according to:
VIN(HY S) = IHY S x RUV 2
andVIN(UV LO, rising) = 1.225 x (
RUV 2
RUV 1+ 1)
where IHY S = 20 µA. RUV 1 is chosen as 6.8 kΩ and RUV 2 as 120 kΩ. This result in aUVLO threshold and hysteresis of 22.8 V and 2.4 V respectively.
A.1.7 Ripple configuration selection
The minimum ripple configuration is chosen to minimise the output voltage ripple. Thevalues are chosen as follows:
Cr = 3.3 nF
Stellenbosch University https://scholar.sun.ac.za
APPENDIX A. CALCULATIONS 95
Cac = 100 nF
Rr are chosen according to
RcCr ≤(VIN(MIN) − VOUT ) x TON
25 mV
which result in 100 kΩ.
Stellenbosch University https://scholar.sun.ac.za
Appendix B
Code
B.1 Python USB listener
"""Created on Wed May 04 16:38:09 2016@author: B. Horn"""import serialimport structimport csv
try:ser = serial.Serial(port=’COM8’,baudrate=9600,parity=serial.PARITY_NONE,stopbits=serial.STOPBITS_ONE,bytesize=serial.EIGHTBITS,xonxoff=False,timeout=5# set input buffer size = read_size # doesn’t seem necessary in pyserial
)except serial.SerialException:
print ’Could not open port’, ’port’ + ’.’print ’Use \’dmesg -T | grep -i usb\’ to find appropriate ports.’
myfile = open(’document.csv’, ’ab’)
try:while 1:
outputList = []while 1:
data_in = bytearray()while not data_in:
data_in = ser.read(size=4)
96
Stellenbosch University https://scholar.sun.ac.za
APPENDIX B. CODE 97
if len(data_in) == 4:(float_out,) = struct.unpack(’>f’, data_in) #>foutputList.append(float_out);data_in = 0#print ’Output received and stored:’#print float_out
if len(outputList) == 18:#append write to csvwr = csv.writer(myfile, dialect=’excel’)wr.writerow(outputList)myFormattedList = [ ’%.3f’ % elem for elem in outputList ]#print ’Output received and stored:’print myFormattedListoutputList[:] = []break
except KeyboardInterrupt:myfile.close()ser.close()
B.2 MATLAB Symbolic solver
syms Voc Rin Rp Rq Cp Cq Vt dVt dV2t I1 dI d2I Vp Vq dVp dVq dV2p dV2q
sol = solve(’-Vt + Voc - I1*(Rin) - Vp - Vq’,’dVt +dI*Rin + dVp + dVq’,’dV2t +d2I*Rin + dV2p + dV2q’,’Vp/Rp + dVp*Cp - I1’,’ dVp/Rp + dV2p*Cp -dI’,’-I1 +Vq/Rq + dVq*Cq’,’-dI + dVq/Rq + dV2q*Cq’,’Vp’ ,’Vq’ , ’dVp’ , ’dVq’,’dV2p’, ’dV2q’,’Vt’);
sol.Vt
%Voc - I1*Rin - I1*Rp - Cp*Rp*dVt - Cp*Rin*Rp*dI%Voc%- I1*Rin - I1*Rp - I1*Rq%- Cp*Rin*Rp*dI - Cq*Rin*Rq*dI - Cp*Rp*Rq*dI - Cq*Rp*Rq*dI%- Cp*Cq*Rin*Rp*Rq*d2I%- Cp*Rp*dVt - Cq*Rq*dVt%- Cp*Cq*Rp*Rq*dV2t
syms Theta2 Theta3 Theta4 Theta5 Theta6 Tp Tq
sol2 = solve(’Theta2+Rin+Rp+Rq’,’Theta3+Rin*(Tp+Tq)+Rq*Tp+Rp*Tq’,’Theta4 +Tp*Tq*Rin’,’Theta5 + Tp+Tq’,’Theta6 +Tp*Tq’, ’Rin’ , ’Rq’, ’Rp’, ’Tp’,’Tq’);
Stellenbosch University https://scholar.sun.ac.za
APPENDIX B. CODE 98
% sol2 = solve(’Theta2+Rin+Rp+Rq’,’Theta3 + Cp*Rin*Rp + Cq*Rin*Rq + Cp*Rp*Rq +Cq*Rp*Rq’,’Theta4 + Cp*Cq*Rin*Rp*Rq’,’Theta5 + Cp*Rp + Cq*Rq’,’Theta6+Cp*Cq*Rp*Rq’, ’Rin’ , ’Rq’, ’Rp’, ’Cp’,’Cq’);
sol2.Rinsol2.Tpsol2.Tqsol2.Rpsol2.Rq
B.3 Noise filter
B.3.1 Noise filter Python code
# -*- coding: utf-8 -*-"""Created on Fri May 27 13:31:24 2016
@author: B. Horn"""import numpy as npfrom scipy.signal import butter, lfilter, freqz, filtfiltimport matplotlib.pyplot as plt
def butter_lowpass(cutoff, fs, order=5):nyq = 0.5 * fsnormal_cutoff = cutoff / nyqb, a = butter(order, normal_cutoff, btype=’low’, analog=False)return b, a
def butter_lowpass_filter(data, cutoff, fs, order=5):b, a = butter_lowpass(cutoff, fs, order=order)z = filtfilt(b, a, data)return z
Voltage,current =loadtxt(’Data_lipo.CSV’,delimiter=’,’,skiprows=1,usecols=(0,1), unpack = 1)
current2, Voltage2 = loadtxt(’test6_long.CSV’,delimiter=’,’,skiprows=0,usecols=(15,17), unpack = 1)
order = 1fs = 1 # sample rate, Hzcutoff = 0.01 # desired cutoff frequency of the filter, Hz
y = butter_lowpass_filter(Voltage, cutoff, fs, order)
Stellenbosch University https://scholar.sun.ac.za
APPENDIX B. CODE 99
x = butter_lowpass_filter(Voltage2, cutoff, fs, order)
Lipo = Voltage[11800:12800]Lipo_filt = y[11800:12800]
LFP = Voltage2[28000:29000]LFP_filt = x[28000:29000]
noise1 = Lipo_filt-Liponoise2 = LFP_filt-LFP
noise11 = (sum((abs(noise1))**2))/sys.getsizeof(noise1)noise22 = (sum((abs(noise2))**2))/sys.getsizeof(noise2)
B.3.2 Noise filter result
0 20 40 60 80 100Time [s]
−0.005
−0.004
−0.003
−0.002
−0.001
0.000
0.001
0.002
0.003
0.004
Volta
ge[V
]
Prototype BMS noiseUp-scaled BMS noise
Figure B.1: Noise of prototype BMS compared to that of the up-scaled BMS
Stellenbosch University https://scholar.sun.ac.za
APPENDIX B. CODE 100
B.4 Battery balancing
500 1000 1500 2000Time [s]
3.35
3.40
3.45
3.50
3.55
3.60
Volta
ge[V
]
cell 1cell 2cell 6cell 7
Figure B.2: Battery balancing
500 1000 1500 2000Time [s]
3.35
3.40
3.45
3.50
3.55
3.60
3.65
3.70
Volta
ge[V
]
cell 8cell 9cell 10cell 11
Figure B.3: Battery balancing
Stellenbosch University https://scholar.sun.ac.za
APPENDIX B. CODE 101
500 1000 1500 2000Time [s]
3.35
3.40
3.45
3.50
3.55
3.60
Volta
ge[V
]cell 12cell 13cell 14cell 15
Figure B.4: Battery balancing
Stellenbosch University https://scholar.sun.ac.za
Appendix C
PCB Schematics
C.1 Full Scale BMS Schematic
C.1.1 BMS Control Schematic
11
22
33
44
55
66
DD
CC
BB
AA
3
Mel
lowC
abs
Fran
shoe
kSo
uth
Afric
a
7
BMS
31.
120
16/1
1/22
09:3
9:54
PM
C:\U
sers
\166
8781
7\D
ropb
ox\M
eest
ers\
PCB
des
igns
\PC
B 1
.1\P
WR
524A
_AFE
1.Sc
hDoc
Title
Size
:N
umbe
r:
Dat
e:Fi
le:
Rev
isio
n:
Shee
tof
Tim
e:Ta
bloi
dB
arth
o H
orn
Engi
neer
:
DSG
1C
HG
2
VSS
3
SDA
4SC
L5
TS1
6C
AP1
7R
EGO
UT
8R
EGSR
C9
VC
5X10
NC
11
NC
12
TS2
13C
AP2
14
VC
10X
15
NC
16
NC
17
TS3
18C
AP3
19B
AT
20
NC
21N
C22
NC
23
VC
1524
VC
1425
VC
1326
VC
1227
VC
1128
VC
10B
29
VC
1030
VC
931
VC
832
VC
733
VC
634
VC
5B35
VC
536
VC
437
VC
338
VC
239
VC
140
VC
041
SRP
42
SRN
43
ALE
RT
44
U1
PartNum
ber
GN
D
GN
D
GN
D
GN
D
C5
C4
C3
C1
C0
C2
C6
C7
C8
C9
GN
D
C10
C8
C7
C6
C5
C4
C3
C2
C1
C0
C0
ALE
RT
SRN
SRP
VC
10B
VC
11V
C12
VC
13V
C14
VC
15
BA
T
REG
SRC
CH
GD
SG
VC
10X
CA
P1
REG
OU
T
SCL
SDA
C1
C2
C3
C4
C5
C7
C6
C9
VC
5X
VC5X
VC
5X
REG
SRC
VC
5X
GN
D
GN
D
BA
TT+
C9
C8
123
P3 3 w
ay 2
.54
mol
ex C5F
C10
FC
15F(
48V
)C
15F(
48V
)
C5F
C10
F
BQ
_Ena
ble
2.2u
FC
19
1KR2
4.7K
R3
Z2 Zene
r
1 2
T3
1 2
T2
1 2
T1
3.3n
FC
16
3.3n
FC
17
3.3n
FC
18
1µF
C13
1µF
C14
1µF
C15
1KR1
C5F
SM4T
6V7A
YD
1
4.7u
FC
20
SM4T
6V7A
YD
2
SM4T
6V7A
YD
3
SM4T
6V7A
YD
4
SM4T
6V7A
YD
5
SM4T
6V7A
YD
6
SM4T
6V7A
YD
7
SM4T
6V7A
YD
8
1234
POW
ER
4 W
ay 2
.54m
m m
olex
hea
der
123
D12
BA
V99
SM4T
6V7A
Y
D11
SM4T
6V7A
Y
D9
SM4T
6V7A
Y
D10
123
P2 3 w
ay 2
.54
mol
ex
123
P1 3 w
ay 2
.54
mol
ex
C10
Z1 Zene
r
470nF
C2
470nF
C12
470nF
C11
470nF
C10
470nF
C9
470nF
C8
470nF
C1
470nF
C7
470nF
C6
470nF
C5
470nF
C4
470nF
C3
G1
S2
D3
N ty
pe
U2B
PIC101 PIC102COC
1
PIC201 PIC202COC
2
PIC301 PIC302COC
3
PIC401 PIC402COC
4
PIC501 PIC502COC
5
PIC601 PIC602COC
6
PIC701 PIC702COC
7
PIC801 PIC802CO
C8
PIC901 PIC902COC
9
PIC1001 PIC1002CO
C10
PIC1101 PIC1102CO
C11
PIC1201 PIC1202CO
C12
PIC1301 PIC1302COC13
PIC1401 PIC1402CO
C14
PIC1501 PIC1502CO
C15
PIC1601 PIC1602COC16
PIC1701 PIC1702COC17 PIC1801 PIC1802
COC18
PIC1901 PIC1902COC19
PIC2001PIC2002COC20
PID101PID102COD
1
PID201PID202CO
D2
PID301PID302COD
3
PID401PID402COD
4
PID501PID502CO
D5
PID601PID602COD
6
PID701PID702CO
D7
PID801PID802COD
8
PID901
PID902
COD9
PID1001
PID1
002
COD10
PID1101
PID1
102
COD11
PID1
201
PID1
202
PID1
203CO
D12
PIP101
PIP102
PIP103
COP1
PIP201
PIP202
PIP203
COP2
PIP301
PIP302
PIP303
COP3
PIPO
WER0
1
PIPOWER02
PIPO
WER0
3
PIPO
WER0
4
COPOWER
PIR101
PIR1
02COR1
PIR2
01PIR202COR2
PIR301PIR302 COR3
PIT101
PIT102 COT
1
PIT201
PIT202 COT
2
PIT301
PIT302 COT
3
PIU101
PIU102
PIU103
PIU104
PIU105
PIU106
PIU107
PIU108
PIU109
PIU1
010
PIU1
011
PIU1
012
PIU1
013
PIU1
014
PIU1
015
PIU1
016
PIU1
017
PIU1
018
PIU1
019
PIU1020
PIU1
021
PIU1
022
PIU1
023
PIU1024
PIU1
025
PIU1
026
PIU1
027
PIU1
028
PIU1
029
PIU1
030
PIU1
031
PIU1
032
PIU1
033
PIU1034
PIU1
035
PIU1
036
PIU1037
PIU1
038
PIU1
039
PIU1
040
PIU1041
PIU1042
PIU1
043
PIU1
044CO
U1
PIU201
PIU202
PIU203
COU2
B
PIZ101PIZ103COZ
1
PIZ201PIZ203COZ
2PI
U104
4
PID1
102
PIU1020
PIU203
PID1
201
PID801PI
POWE
R01
PIR1
02NLC
0
PIC1102 PIC1201
PID701 PID802
PIP301
PIU1
040
NLC1
PIC1002 PIC1101
PID601 PID702PIP302
PIU1
039
NLC2
PIC902 PIC1001
PID501 PID602
PIP303
PIU1
038
NLC3
PIC802 PIC901
PID502
PIP201
PIU1037
NLC4
PIC801
PID901
PIP202
PIU1
036
NLC5
PID401
PIPOWER02
PIR202
NLC5F
PIC602 PIC701
PID301 PID402
PIP203
PIU1034
NLC6
PIC502 PIC601
PID201 PID302
PIP101
PIU1
033
NLC7
PIC402 PIC501
PID101 PID202
PIP102
PIU1
032
NLC8
PIC302 PIC401PID102
PIP103
PIU1
031
NLC9
PIC301
PID1001
PIU1
030
NLC10
PIPO
WER0
3NL
C10F
PIPO
WER0
4NL
C15F
(48V
)
PIC1301
PIU107
PIU102
PIU101
PIC202
PIC1202
PIC1302
PIC1802
PIC1902
PIC2001PIT102
PIU103
PIZ201
PIC101PI
R201
PIU1
035
PIZ103
PIC201PIR101
PIU1041
PIC1401
PIU1
016
PIU1
017
PIU1
019
PIC1501
PIU1
011
PIU1
012
PIU1
014
PIC1601PIT301
PIU1
018
PIC1701PIT201
PIU1
013
PIC1801
PIR302
PIT101
PIU106
PIZ203
PID1
202
PIR301PI
D120
3
PIU1
021
PIU1
022
PIU1
023
PIC2002
PIU108
PIC1901
PIU109
PIU202
PIU105
PIU104
PIU1
043
PIU1042
PIC102
PIC702PIC1502
PIC1702
PID902
PIT202
PIU1
010
PIU201
PIZ101NLVC5X
PIU1
029
PIC1402PIC1602
PID1
002
PIT302
PIU1
015
PIU1
028
PIU1
027
PIU1
026
PIU1
025
PID1101
PIU1024
102
Stellenbosch University https://scholar.sun.ac.za
APPENDIX C. PCB SCHEMATICS 103
11
22
33
44
55
66
DD
CC
BB
AA
4
Mel
lowc
abs
Fran
shoe
kSo
uth
Afric
a
7
BMS
41.
120
16/1
1/22
09:5
0:18
PM
C:\U
sers
\166
8781
7\D
ropb
ox\M
eest
ers\
PCB
des
igns
\PC
B 1
.1\P
WR
524A
_AFE
2.Sc
hDoc
Title
Size
:N
umbe
r:
Dat
e:Fi
le:
Rev
isio
n:
Shee
tof
Tim
e:Ta
bloi
dB
arth
o H
orn
Engi
neer
:
10µF
C29
C12
C13
C14
C15
GN
D
C10
F
BA
TT+
C5F
VC
15V
C14
VC
13V
C12
VC
11V
C10
B
VC
10X
VC
5X
BA
T
C15
F(48
V)
1KR4
1KR5
1KR6
1KR7
10µF
C27 10
µFC28
470nF
C22
C4
C9
SM4T
6V7A
YD
15
SM4T
6V7A
YD
16
SM4T
6V7A
YD
17
SM4T
6V7A
YD
18
SM4T
6V7A
YD
19
SM4T
6V7A
YD
13
SM4T
6V7A
YD
14
123
P4 3 w
ay 2
.54
mol
ex
C10
C10
C11
123
P5 3 w
ay 2
.54
mol
ex
Z3 Zene
r
470nF
C23
470nF
C24
470nF
C25
470nF
C26
470nF
C21
PIC2101 PIC2102CO
C21
PIC2201 PIC2202CO
C22
PIC2301 PIC2302CO
C23
PIC2401 PIC2402CO
C24
PIC2501 PIC2502CO
C25
PIC2601 PIC2602CO
C26
PIC2701 PIC2702CO
C27
PIC2801 PIC2802CO
C28
PIC2901 PIC2902CO
C29
PID1301PID1302CO
D13
PID1401PID1402CO
D14
PID1501PID1502CO
D15
PID1601PID1602CO
D16
PID1701PID1702CO
D17
PID1801PID1802CO
D18
PID1901PID1902CO
D19
PIP401
PIP402
PIP403
COP4
PIP501
PIP502
PIP503
COP5
PIR4
01PI
R402COR
4
PIR5
01PI
R502COR
5
PIR601
PIR602COR
6
PIR7
01PI
R702COR
7
PIZ301PIZ303COZ
3
PIC2901
PIR7
01PI
R702
PID1401PID1402PI
R402
PID1301
PIP501
NLC1
0
PID1302
PID1901
PIR5
02
PIR602
PIC2502 PIC2601
PID1801 PID1902
PIP502
NLC1
1
PIC2402 PIC2501
PID1701 PID1802
PIP503
NLC1
2
PIC2302 PIC2401
PID1601 PID1702
PIP401
NLC1
3
PIC2202 PIC2301
PID1501 PID1602
PIP402
NLC1
4
PIC2201PID1502
PIP403
NLC1
5
PIC2802
PIC2702
PIC2801PI
R401
PIC2101PIR601
PIZ303PIC2102
PIC2602
PIC2701
PIC2902
PIR5
01
PIZ301
Stellenbosch University https://scholar.sun.ac.za
APPENDIX C. PCB SCHEMATICS 104
11
22
33
44
55
66
DD
CC
BB
AA
5
Mel
lowC
abs
Fran
shoe
kSo
uth
Afric
a
7
BMS
51.
120
16/1
1/22
09:5
2:38
PM
C:\U
sers
\166
8781
7\D
ropb
ox\M
eest
ers\
PCB
des
igns
\PC
B 1
.1\P
WR
524A
_BQ
7835
0DB
T.Sc
hDoc
Title
Size
:N
umbe
r:
Dat
e:Fi
le:
Rev
isio
n:
Shee
tof
Tim
e:Ta
bloi
dB
arth
o H
orn
Engi
neer
:
SCL
SDA
ALE
RT
12K
R16
300K
R11
30.9
K
R22
3.3V
22uFC
42
10nF
C35
BA
TT+
GPI
O22
1
GPI
O32
2
GPI
O33
3
GPI
O23
4
GPI
O42
5
GPI
O43
6
VD
D7
VSS
8
nXR
S9
nTR
ST10
AD
CIN
A7
11
AD
CIN
A6
12
AD
CIN
A5
13
AD
CIN
A4
14
AD
CIN
A3
15
AD
CIN
A2
16
AD
CIN
A1
17
AD
CIN
A0
18
Vre
fhi
19
VD
DA
20
VSS
A21
Vre
flo22
AD
CIN
B0
23
AD
CIN
B1
24
AD
CIN
B2
25
AD
CIN
B3
26
AD
CIN
B4
27
AD
CIN
B5
28
AD
CIN
B6
29
AD
CIN
B7
30
GPI
O27
31
CA
NTX
32
CA
NR
X33
SCL
34
VSS
35
VD
DIO
36
GPI
O26
37
Test
238
GPI
O9
39
SDA
40G
PIO
1841
GPI
O17
42G
PIO
843
GPI
O25
44G
PIO
4445
GPI
O16
46G
PIO
1247
GPI
O41
48G
PIO
749
GPI
O6
50X
251
X1
52V
SS53
VD
D54
GPI
O19
55G
PIO
3956
GPI
O38
57G
PIO
3758
GPI
O35
59G
PIO
3660
GPI
O11
61G
PIO
562
GPI
O4
63G
PIO
4064
GPI
O10
65G
PIO
366
GPI
O2
67G
PIO
168
GPI
O0
69V
DD
IO70
VSS
71V
DD
72V
rege
nz73
GPI
O34
74G
PIO
1575
GPI
O13
76G
PIO
1477
GPI
O20
78G
PIO
2179
GPI
O24
80U
3
F280
35
GN
D
GN
D
GN
D
GN
D
3.3V
GN
D
GN
D
3.3V
12
34
56
78
910
1112
1314
P6 Hea
der 7
X2
3.3V
TCK
TMS
TDI
TDO
3.3V
TMS
TDI
TDO
TCK
2.2K
R13
GN
D GN
D
GN
D
3.3n
F
C37
TRST
TRST
3.3V
GN
D2.2u
FC
31
3.3V
3.3V
3.3V
GN
DG
ND
Driv
e_se
nse
Key
GN
D
LED
1LE
D2
LED
3
LED
1LE
D2
LED
3
SCI-
rece
ive
SCI-
trans
mit
BT1
Can
TxC
anR
x
D1
GN
D2
Vcc
3
R4
EN5
CL
6C
H7
Rs
8U
6
CA
N-S
N65
HV
D23
4
GN
D3.
3V
CA
N_E
N
CA
N_E
N
CA
NTx
CA
NR
xG
ND
GN
D 10K
R43
GN
D
62R46
BQ
_Ena
ble
GPI
O_0
1
GPI
O_1
2
GPI
O_2
3
GPI
O_3
4
RES
ET_L
5
SER
IAL_
46
SER
IAL_
37
SER
IAL_
28
SER
IAL_
19
SER
IAL_
010
GN
D11
VD
D12
NC
13
N14
N15
L16
L17
GN
D18
GN
D19
GN
D20
GN
D21
GN
D22
GN
D23
GN
D24
GN
D25
U7
Gre
enPH
Y m
ini s
tam
p
GN
D
3.3V
GN
D
GN
D
SPI-
MO
SI
SPI-
MIS
OSP
I-C
LK
SPI-
CLK
SPI-
MIS
OSP
I-M
OSI
SPI-
CS
SPI-
CS
Gre
enP-
R
Gre
enP-
R
Gre
enP-
I
Gre
enP-
I
Gre
enP-
GPI
O0
Gre
enP-
GPI
O1
Gre
enP-
GPI
O2
Gre
enP-
GPI
O0
Gre
enP-
GPI
O1
Gre
enP-
GPI
O2
100n
F
C39
100n
FC
30
1KR37
1KR36
4.7K
R14
4.7K
R10
2.2K
R12
10K
R8
10K
R9
10K
R31
10K
R32
10K
R35
10K
R33
1 2
CA
N1
2.2u
FC
32
47nF
C49
LED
1LE
D2
LED
3
REG
OU
T10
K
R17
62R47
NC
1
RES
ET2
GN
D3
GN
D4
NC
5
MV
cc6
PG6
7
XO
SCEN
8
Vcc
_CO
RE
9
Vcc
10
Vcc
_IO
11
RX
D12
TXD
13
RTS
14
CTS
15
OP3
16
GN
D17
GN
D18
PG7
19
Aud
io20
Aud
io21
Aud
io22
Aud
io23
GN
D24
OP5
25O
P4/P
G4
2632
K+
2732
K-
28G
ND
29G
ND
30G
ND
31G
ND
32N
C33
NC
34N
C35
NC
36N
C37
NC
38N
C39
NC
40U
5
LMX
9838
Blu
etoo
th m
odul
e
GN
D
GN
D
GN
D
GN
D
3.3V
3.3V
GN
D
BT-
LED
1
BT-
LED
2
OP3
OP5
OP4
BT-
LED
2
D23
3.3V
330K
R25
BT-
LED
1
D22
3.3V
330K
R23
1KR26
1KR27
1KR28
3.3V
3.3V
3.3V
OP3
OP4
OP5
BT1
SCI-
rece
ive
SCI-
trans
mit
GN
D
Gre
en-P
HY
Drive
GN
D10K
R44
Contactor_out
GN
D
12345678910
P7
3.3V
con
trolle
d
GN
D Gre
en-P
HY
GN
D
3.3V
GN
D
3.3V
GN
D
3.3V
GN
D
3.3V
Pre-charge_out
I_measure
I_m
easu
reC
HG
GN
D
DSG
10K
R42
Key
Sw
itch
Blue
toot
h
Buck
reg
ulat
or 3
.3V
Cur
ent P
CB
conn
ecto
r
123
D30
BA
V99
123
D29
BA
V99
123
D20
BA
V99
123
D24
BA
V99
123
D21
BA
V99
1 2
OP3
1 2
OP4
1 2
OP5
100
R18
100
R24
100
R20
510
R34
510
R45
RTS
CTS
RTS
CTS
3.3V
A1
GN
D2
B3
Y4
Vcc
5C
6U
8
SN74
LVC
1G11
3.3V
GN
D
100K
R39
330K
R38
100K
R41
330K
R40
GN
DG
ND
CH
G_o
utD
SG_o
ut
Con
tact
or
Con
tact
or
Contactor_out
Cha
rge_
sens
e
Key
_sen
seK
ey_s
ense
Driv
e_se
nse
Charge
GN
D10K
R29
GN
D
3.3V
123
D25
BA
V99
510
R30
Cha
rge_
sens
e
AN
D G
ate
3.3n
FC
36
GN
D
Pre-
char
ge_o
utK
ey
Driv
e
Cha
rge
CA
N C
omm
unic
atio
n
Gre
en-P
HY
Com
mun
icat
ion
Stat
us L
EDS
BT S
tatu
s LED
S an
d In
puts
4.7u
FC
43
60 O
hm
L2
60 O
hm
L1
60 O
hmL4
DSG_out CHG_out
G1
S2
D3
U9B
3.3V
10K
R48
3.3V
con
trol
3.3V control
3.3V
con
trolle
d
Cur
rent
mea
sure
on/
off c
ontr
ol
GN
D10K
R49
GN
D
3.3V
123
D26
BA
V99
510
R50
12V
_sen
se
12V
_sen
se
12V
_sen
se_m
easu
re
12V_sense_measure
2.2u
FC
52
22uFC
53
GN
D
GN
D
R68
KR
2110
pFC
34
GN
D
330K
R15
10K
R19
100n
F
C38
GN
D
100V
D27
Dio
de150u
H
L3 Indu
ctor
10uH
L5 Indu
ctor
22uFC
45
GN
D
Boo
t1
VIN
2
EN3
SS/T
R4
CLK
5PW
RG
D6
Vse
nse
7C
OM
P8
GN
D9
PH10
U10
TPS5
4060
2.2u
FC
33
2.2u
FC
40
2.2u
FC
41
2.2u
FC
44
2.2u
FC
48
2.2u
FC
47
PIC3001PIC3002CO
C30
PIC3101PIC3102CO
C31
PIC3201PIC3202CO
C32
PIC3301PIC3302CO
C33
PIC3401 PIC3402CO
C34
PIC3501PIC3502CO
C35
PIC3601 PIC3602CO
C36
PIC3
701
PIC3
702COC3
7
PIC3
801
PIC3
802
COC3
8
PIC3
901
PIC3
902
COC3
9
PIC4001PIC4002CO
C40
PIC4101PIC4102CO
C41
PIC4201PIC4202CO
C42
PIC4301PIC4302CO
C43
PIC4401PIC4402CO
C44
PIC4501PIC4502CO
C45
PIC4701PIC4702CO
C47
PIC4801PIC4802CO
C48
PIC4901 PIC4902CO
C49
PIC5201PIC5202CO
C52
PIC5301PIC5302CO
C53
PICAN101
PICAN102
COCAN1
PID2001
PID2002
PID2
003CO
D20
PID2101
PID2102
PID2
103CO
D21
PID2201PID2202CO
D22
PID2301PID2302CO
D23
PID2
401
PID2
402
PID2
403COD2
4
PID2
501
PID2
502
PID2
503CO
D25
PID2
601
PID2
602
PID2
603CO
D26
PID2701PID2702CO
D27
PID2901
PID2902
PID2
903CO
D29
PID3001
PID3002
PID3
003CO
D30
PIL101
PIL1
02
COL1
PIL2
01PIL202
COL2
PIL301
PIL3
02
COL3
PIL401PIL402 COL4
PIL5
01PIL502
COL5
PILED101PILED102COLED1
PILED201PILED202COLED2
PILED301PILED302COLED3
PIOP
301
PIOP
302 CO
OP3
PIOP
401
PIOP
402 CO
OP4
PIOP501
PIOP502 CO
OP5
PIP601
PIP602
PIP603
PIP604
PIP605
PIP606
PIP607
PIP608
PIP609
PIP6010
PIP6011
PIP6
012
PIP6013
PIP6
014
COP6
PIP701
PIP702
PIP703
PIP704
PIP705
PIP706
PIP707
PIP708
PIP709
PIP7
010
COP7
PIR801PIR802 COR8
PIR901PIR902 COR9
PIR1001PIR1002 COR1
0
PIR1101PIR1102CO
R11
PIR1201PIR1202 COR1
2
PIR1301PIR1302 COR1
3
PIR1401PIR1402 COR1
4PIR1501PIR1502 COR1
5
PIR1601PIR1602 COR1
6
PIR1
701
PIR1
702COR1
7
PIR1
801
PIR1
802COR1
8
PIR1901PIR1902 COR1
9
PIR2
001
PIR2
002COR2
0
PIR2101PIR2102CO
R21
PIR2201PIR2202 COR2
2
PIR2301PIR2302 COR2
3
PIR2
401
PIR2
402COR2
4
PIR2501PIR2502 COR2
5
PIR2601PIR2602CO
R26
PIR2701PIR2702CO
R27
PIR2801PIR2802 COR2
8
PIR2901PIR2902CO
R29
PIR3001PIR3002CO
R30
PIR3101PIR3102CO
R31
PIR3201PIR3202CO
R32
PIR3301PIR3302 COR3
3
PIR3401PIR3402 COR3
4
PIR3501PIR3502CO
R35
PIR3601PIR3602 COR3
6
PIR3701PIR3702 COR3
7
PIR3801PIR3802 COR3
8
PIR3901PIR3902 COR3
9
PIR4001PIR4002CO
R40
PIR4101PIR4102CO
R41PIR4201PIR4202 CO
R42
PIR4301PIR4302 COR4
3
PIR4401PIR4402 COR4
4
PIR4501PIR4502 COR4
5
PIR4601PIR4602CO
R46
PIR4701PIR4702CO
R47
PIR4
801
PIR4
802COR4
8
PIR4901PIR4902CO
R49
PIR5001PIR5002CO
R50
PIU301
PIU302
PIU303
PIU304
PIU305
PIU306
PIU307
PIU308
PIU309
PIU3010
PIU3011
PIU3012
PIU3013
PIU3014
PIU3015
PIU3016
PIU3017
PIU3018
PIU3019
PIU3020
PIU3021
PIU3022
PIU3023
PIU3024
PIU3025
PIU3026
PIU3027
PIU3028
PIU3029
PIU3030
PIU3031
PIU3032
PIU3033
PIU3034
PIU3035
PIU3036
PIU3037
PIU3038
PIU3039
PIU3040
PIU3041
PIU3042
PIU3043
PIU3044
PIU3045
PIU3046
PIU3047
PIU3048
PIU3049
PIU3050
PIU3051
PIU3052
PIU3053
PIU3054
PIU3055
PIU3056
PIU3057
PIU3058
PIU3059
PIU3060
PIU3061
PIU3062
PIU3063
PIU3064
PIU3065
PIU3066
PIU3067
PIU3068
PIU3069
PIU3070
PIU3071
PIU3072
PIU3073
PIU3074
PIU3075
PIU3076
PIU3077
PIU3078
PIU3079
PIU3080
COU3
PIU501
PIU502
PIU503
PIU504
PIU505
PIU506
PIU507
PIU508
PIU509
PIU5010
PIU5011
PIU5012
PIU5013
PIU5014
PIU5015
PIU5016
PIU5017
PIU5018
PIU5019
PIU5020
PIU5021
PIU5022
PIU5023
PIU5024
PIU5025
PIU5026
PIU5027
PIU5028
PIU5029
PIU5030
PIU5031
PIU5032
PIU5033
PIU5034
PIU5035
PIU5036
PIU5037
PIU5038
PIU5039
PIU5040
COU5
PIU601
PIU602
PIU603
PIU604
PIU605
PIU606
PIU607
PIU608
COU6
PIU701
PIU702
PIU703
PIU704
PIU705
PIU706
PIU707
PIU708
PIU709
PIU7010
PIU7011
PIU7012
PIU7013
PIU7014
PIU7015
PIU7016
PIU7017
PIU7018
PIU7019
PIU7020
PIU7021
PIU7022
PIU7023
PIU7024
PIU7025
COU7
PIU801
PIU802
PIU803
PIU804
PIU805
PIU806
COU8
PIU901
PIU902
PIU903
COU9
B
PIU1001
PIU1002
PIU1003
PIU1004
PIU1005
PIU1006
PIU1007
PIU1008
PIU1009
PIU10010
COU1
0
PIC4302
PIC4401
PIC4502
PIC4702
PIC4801
PID2002
PID2102
PID2
402
PID2
502
PID2
602
PID2902
PID3002
PIL101
PIL2
01
PIL401
PIL502
PIP605
PIR802 PIR901 PIR1001PIR1401
PIR2302PIR2502
PIR2602PIR2702
PIR2802
PIU506
PIU5010
PIU5011
PIU603
PIU7012
PIU805
PIU902
PIR4
801
PIU3045
NL303V controlPIP704
PIU903
PIP703
PIR4902NL12
V0se
nse PI
D260
3PIR4901 PIR5002
PIU3027
NL12V0sense0measure
PIU3039
PIC3102PIC5202
PIR1502PIU1002
PIU3066
PIU3050
PIU502
NLBT
1
PID2201
PIU507
NLBT
0LED
1
PID2301
PIU5019
NLBT
0LED
2
PIR4301
PIU3031
NLCAN0
EN
PIR3701
PIU3033
NLCANRx
PIR3602
PIU3032
NLCANTx
PID2
503
PIR2901 PIR3002
PIU3075
NLCharge
PIP707
PIR2902NLCh
arge0sen
se
PIR3802 PIR3801 PIR3902
PIU803
NLCHG0out
PIU3078
PIU806NL
Cont
acto
r
PID2
003
PIR2
001
PIU804
NLContactor0out
PIU3077
PIU5015
NLCT
S
PID3
003
PIR4401 PIR4502
PIU3055
NLDrive
PIP708
PIR4402NLDrive0sense
PIR4002 PIR4001 PIR4102
PIU801
NLDSG0out
PIC3002
PIC3101
PIC3201 PIC3301
PIC3402
PIC3501
PIC3602
PIC3
701
PIC3
902
PIC4001
PIC4101
PIC4201
PIC4301
PIC4402
PIC4501
PIC4701
PIC4802
PIC4902
PIC5201
PIC5301
PID2001
PID2101
PID2
401
PID2
501
PID2
601
PID2701
PID2901
PID3001
PILED101PILED201
PILED301
PIP604
PIP606
PIP608
PIP6010
PIP6
012
PIP701
PIR1101
PIR1302
PIR1601
PIR1901
PIR3001PIR3401
PIR3901PIR4101
PIR4202
PIR4501PIR5001
PIU308
PIU3021
PIU3022
PIU3035
PIU3052
PIU3053
PIU3071
PIU3073
PIU503
PIU504
PIU5017
PIU5018
PIU5024
PIU5027
PIU5029
PIU5030
PIU5031
PIU5032
PIU602
PIU704
PIU7011
PIU7014
PIU7015
PIU7018
PIU7019
PIU7020
PIU7021
PIU7022
PIU7023
PIU7024
PIU7025
PIU802
PIU1009
PIP7
010
PIU7016
PIU7017
NLGreen0PHY
PIU301
PIU701
NLGr
eenP
0GPI
O0
PIU302
PIU702
NLGr
eenP
0GPI
O1
PIU303
PIU703
NLGr
eenP
0GPI
O2
PIU304
PIU7010
NLGreenP0I
PIU305
PIU705
NLGreenP0R
PIC3601PI
D210
3PI
R180
1
PIU3026
NLI0measurePI
D290
3PIR3301 PIR3402
PIU3080
NLKey
PIP709
PIR3302NLKey0sense
PIR3102
PIU3062
NLLED1
PIR3202
PIU3063
NLLED2
PIR3502
PIU3064
NLLED3
PIC3001PI
L102
PIU3036
PIC3202PIU307
PIC3302PIL202
PIR1201
PIU3020
PIC3401PIR2102
PIU1008
PIC3502
PIU1004
PIC3
702
PIR2101
PIC3
801
PIU1001
PIC3
802
PID2702PIL301
PIU10010
PIC3
901
PIL402PIU3070
PIC4002PIU3054
PIC4102PIU3072
PIC4202PIC5302
PIL3
02PI
L501
PIR2202
PIC4901PIR4601 PIR4702
PICAN101
PIR4602
PIU607
PICAN102
PIR4701
PIU606
PID2202PIR2301
PID2302PIR2501PILED102PIR3101
PILED202PIR3201
PILED302PIR3501
PIOP
301
PIR2601
PIOP
401
PIR2701
PIOP501
PIR2801
PIP6013
PIR1002PI
P601
4
PIR1402
PIP702
PIR1
802
PIP705
PIR2
402
PIP706
PIR2
002
PIR1102PIU1005
PIR1202PIU309
PIR1501 PIR1602
PIU1003
PIR1
701PIU3037
PIR1902PIR2201
PIU1007
PIR3601PIU601
PIR3702PIU604
PIR4201PIU608 PIR4302
PIU605
PIR4
802
PIU901
PIU306
PIU3011
PIU3012
PIU3013
PIU3014
PIU3015
PIU3016
PIU3017
PIU3018
PIU3019
PIU3023
PIU3024
PIU3025
PIU3028
PIU3029
PIU3030
PIU3038
PIU3044
PIU3048
PIU3051
PIU3056
PIU3061
PIU3065
PIU3067
PIU3068
PIU3069
PIU3074
PIU501
PIU505
PIU508
PIU509
PIU5020
PIU5021
PIU5022
PIU5023
PIU5028
PIU5033
PIU5034
PIU5035
PIU5036
PIU5037
PIU5038
PIU5039
PIU5040
PIU7013
PIU1006
PIOP
302
PIU5016
NLOP
3
PIOP
402
PIU5026NL
OP4
PIOP502
PIU5025NLO
P5
PID2
403
PIR2
401
PIU3079
NLPre0charge0out
PIR1
702
PIU3076
PIU5014
NLRT
S
PIU3049
PIU5013
NLSC
I0re
ceiv
e
PIU3047
PIU5012
NLSC
I0tr
ansm
it
PIR801PIU3034
PIR902PIU3040
PIU3041
PIU709
NLSPI0CLK
PIU3043
PIU708
NLSPI0CS
PIU3042
PIU707
NLSP
I0MI
SO
PIU3046
PIU706
NLSP
I0MO
SI
PIP609
PIP6011
PIU3057NL
TCK
PIP603
PIU3059NL
TDI
PIP607
PIU3058NL
TDO
PIP601
PIU3060NL
TMS
PIP602
PIR1301
PIU3010
NLTRST
Stellenbosch University https://scholar.sun.ac.za
APPENDIX C. PCB SCHEMATICS 105
11
22
33
44
55
66
DD
CC
BB
AA
2
Mel
lowC
abs
Fran
shoe
kSo
uth
Afric
a
7
BMS
21.
120
16/1
1/22
09:5
7:09
PM
C:\U
sers
\166
8781
7\D
ropb
ox\M
eest
ers\
PCB
des
igns
\PC
B 1
.1\P
WR
524A
_FET
s.Sch
Doc
Title
Size
:N
umbe
r:
Dat
e:Fi
le:
Rev
isio
n:
Shee
tof
Tim
e:Ta
bloi
dB
arth
o H
orn
Engi
neer
:
BA
TT-
BA
TT-
GN
D
C0
BA
TT+
SRP
SRN
-
0 - 6
0V
Inpu
t
+
10µF
C50
1SM
A54
AT3
G
D31
10µF
C51
PIC5001 PIC5002CO
C50
PIC5101 PIC5102CO
C51
PID3101PID3102CO
D31
PIC5001PID3102
PIC5102PID3101
NLBATT0
PIC5002 PIC5101
Stellenbosch University https://scholar.sun.ac.za
APPENDIX C. PCB SCHEMATICS 106
C.1.2 Balance Schematic
11
22
33
44
DD
CC
BB
AA
*
* * * * **
Bala
nce
circ
uit *
*20
16/1
1/22
09:1
8:05
PM
C:\U
sers
\166
8781
7\D
ropb
ox\M
eest
ers\
PCB
des
igns
\Bal
ance
PCB
\Bal
ance
PCB
.Sch
Doc
Title
Size
:N
umbe
r:
Dat
e:Fi
le:
Rev
isio
n:
Shee
tof
Tim
e:A
4En
gine
er:
B. H
orn
F1 SSQ
123
D1
BA
V99
dio
de
0.8K
R1
Res
3
10K
R2
Res
3V
sens
e/nB
alan
ce
CEL
L-G
ND
CEL
L++ -
t°
RT1
NTC
10K
1%
Bla
nce_
Res
isto
rGen
eric
boa
rd
P1 Pow
er
1 2
P6 2 w
ay 2
.54m
m m
olex
SM4T
6V7A
YD
2G
1
S2
D3
U3B
P5 GN
D
P3 Mea
sure
PID101
PID102
PID103COD
1
PID201PID202COD2
PIF101
PIF102COF
1
PIP101COP
1
PIP301COP
3
PIP501COP
5
PIP601
PIP602COP
6
PIR101
PIR102
COR1
PIR201
PIR202
COR2
PIRT101
PIRT102CO
RT1
PIU301
PIU302
PIU303
COU3B
PIF102
NL0
PID101
PID201
PIP501
PIU303
NL0
NLBlance0Res
istor
PID102
PID202
PIF101
PIP101
PIR101
PIU302
PID103
PIR201
PIU301
PIP601
PIRT102
PIP602
PIRT101
PIP301
PIR102
PIR202
NLVs
ense
0nBa
lanc
e
Figure C.1: Balance schematic
Stellenbosch University https://scholar.sun.ac.za
APPENDIX C. PCB SCHEMATICS 107
C.1.3 Current Sense Schematic
11
22
33
44
DD
CC
BB
AA
1
* * * * *1
Cur
rent
Sen
sor *
3.1
2016
/11/
2109
:24:
34 P
MC
:\Use
rs\1
6687
817\
Dro
pbox
\Mee
ster
s\PC
B d
esig
ns\C
urre
nt3.
1 pc
b\C
urre
ntPc
b.Sc
hDoc
Title
Size
:N
umbe
r:
Dat
e:Fi
le:
Rev
isio
n:
Shee
tof
Tim
e:A
4En
gine
er:
B H
orn
Gre
enPH
Y
GN
D
Batte
ry P
lus
Hig
h cu
rren
t
Vcc
1
GN
D2
VIO
UT
3
IP+ 4IP-5
U2
AC
S758
Cur
rent
sens
e
12345678910
P2
3.3V
I_m
easu
re
GN
D
3.3V
I_m
easu
re
GN
D
3.3V
123
D4
BA
V99
11
H2
M6
hole
1 1
H1
M6
hole
SMA
J54C
AD
3+
1
-2
Load
3
Load
4U
1
CPC
1008
N R
elay
NO
GN
D
GN
D
123
D6
BA
V99
3.3V
Pre_Charge
3.3V
470
R3
G1
S2
D3
Q1B
SMA
J54C
AD
1
G1
S2
D3
Q2B
GN
D
10nF
C2
Main_Contact
GN
D
123
D5
BA
V99
3.3V
Pre_ChargeMain_Contact
Key
_sen
seD
rive_
sens
eC
harg
e_se
nse
Key
_sen
seD
rive_
sens
eC
harg
e_se
nse
Mai
n_C
onta
ct_o
ut
Pre_
Cha
rge_
out
Pre_
Cha
rge_
out
Mai
n_C
onta
ct_o
ut
48V
(5A
fuse
d)
48V
(5A
fuse
d)
GreenPHY
Batte
ry O
utpu
t
In connectorOut connector
Mai
n C
onta
ctor
Dri
ver
Pre-
char
ge R
esist
or D
rive
r
Con
tact
or w
ord
gedr
yf d
eur d
ie 4
8V(5
A fu
se).
Mai
n_C
onta
ct_o
ut w
ord
gebr
uik
as d
ie sk
akel
aar.
F1 5 A
D2
SB51
00-T
100
R8
100
R6
100
R7
1MR4
1MR2
10K
R1
3KR5
Bat
t_ou
t
Bat
t_ou
t
10m
H
L1 Indu
ctor
1234567
P1 7 w
ay 2
.54m
m m
olex
12V
_sen
se
12V
_sen
se
2.2u
FC
1
2.2u
FC
3
10K
R9
PIC101PIC102COC
1
PIC201 PIC202COC
2
PIC301PIC302COC
3
PID101PID102COD
1
PID201
PID202CO
D2
PID301PID302COD
3
PID401
PID402
PID403
COD4
PID501
PID502
PID503COD
5
PID601
PID602
PID603COD
6
PIF101
PIF102
COF1
PIH101
COH1 PIH201
COH2
PIL101
PIL102
COL1
PIP101
PIP102
PIP103
PIP104
PIP105
PIP106
PIP107
COP1
PIP201
PIP202
PIP203
PIP204
PIP205
PIP206
PIP207
PIP208
PIP209
PIP2
010
COP2
PIQ101
PIQ102
PIQ103
COQ1B
PIQ201
PIQ202
PIQ203
COQ2B
PIR101
PIR102COR
1
PIR201PIR202 COR2
PIR301
PIR302
COR3
PIR401PIR402 COR4
PIR501 PIR502COR5
PIR601 PIR602
COR6
PIR701
PIR702
COR7
PIR801
PIR802COR
8
PIR901
PIR902CO
R9
PIU101
PIU102
PIU103
PIU104
COU1
PIU201
PIU202
PIU203
PIU204PIU205COU2
PIC102
PIC302
PID402
PID502
PID602
PIL102PIU101
PIU201
PIP101
PIP203NL
12V0sense
PID202
PIP107
NL48
V(5A
fus
ed)
PID301
PIF102
PIH101PIU103
PIU205NLBatt0out
PIP106
PIP207NL
Char
ge0s
ense
PIP105
PIP208NL
Driv
e0se
nse
PIC101
PIC201
PIC301
PID101
PID401
PID501
PID601
PIP201
PIQ102
PIQ202
PIR201
PIR401
PIU202
PID201
PIF101
PIP2
010
NLGreenPHY
PIP202
PIR601NLI0measure
PIP104
PIP209NL
Key0sense
PID503
PIR102
PIR801
NLMain0Contact
PID102
PIP103
PIQ103
NLMain0Contact0out
PIC202PID403PIR602
PIR701
PID302PIR301
PIU104
PIH201PIU204
PIL101
PIP204
PIP205
PIR902
PIP206
PIR802
PIQ101
PIR101 PIR202
PIQ203 PIR501 PIR502
PIU102
PIR702PIU203
PID603
PIQ201
PIR402
PIR901
NLPre0Charge
PIP102
PIR302
NLPre0Charge0out
Figure C.2: Current sense schematic
Stellenbosch University https://scholar.sun.ac.za