A novel methodology to parameterise lithium-ion cell models for low temperature applications

Electrification of road transportation is widely recognised as a necessary solution to reduce global warming. However, the mass market adoption of electric vehicles (EVs) has been hindered by reduced battery performance in cold weather conditions leading to warranty limitations coupled with inaccurate range estimation that exacerbates customer range anxiety. Today’s market leading EVs driven on typical UK motorways have a range estimation error up to 27% at an ambient temperature of 10 °C, and due to slower battery kinetics it worsens to 45% at −15 ℃. The range estimation accuracy depends upon the performance of models embedded in the Battery Management System (BMS) which estimates battery states (viz. State-of-Charge (SOC) and Stateof- Energy (SOE)). The performance of the models fundamentally depends upon experimentally obtained parameters at different operating temperatures and currents, and validation exercises against legislative drive cycles. The experiments are usually performed in isothermal conditions by using state-of-the-art climatic chambers that maintain a pre-set temperature by forced air convection. Unfortunately, isothermal conditions are not adhered to as the battery operating temperature deviates significantly from the predefined chamber temperature, especially when battery characterisation is undertaken at low ambient temperatures (≤10 ℃). The aim of this thesis is to propose a novel experimental methodology and alternative modelling approaches to improve the range estimation accuracy of EVs at low ambient temperatures by addressing these shortcomings in existing characterisation methodology. A novel experimental methodology is developed to ensure isothermal conditions using immersed oil baths that provides more accurate usable capacity and energy characteristics of lithium-ion cells, especially at low temperatures, by eliminating the effect of rapid heat generation during battery operation. For the first time, it is shown that model parameterisation using oil-based rather than air-based experiments leads to more accurate estimation of battery states (SOC and SOE). The findings in this thesis suggest that the absolute SOC error is reduced from 13.5% to 5.1% and the absolute SOE error is reduced from 20.6% to 4.3% at −15 °C ambient temperature. A detailed study of heat generation using a battery model utilising the polarisation currents demonstrated improved modelled surface temperature and terminal voltage estimates. These results along with accurate parameterisation data, estimated the battery states and terminal voltage better at low ambient temperatures. A power control approach to battery characterisation ensures that the operating current responds dynamically to the changing cell voltage. A comparison based on energy throughput and peak power demand at low temperatures showed that power control is more representative of real-world applications as compared to current control. Therefore, it is recommended that power control be employed as the primary approach to obtain validation data for cell models. The work demonstrates and provides insight on new aspects for improving the range estimation accuracy of EVs operating under cold weather conditions. Advances from this work enable increased adherence to the rigid environmental conditions necessary for global lithium-ion battery testing standards and battery modelling, leading to the increased uptake of EVs