A Comprehensive study of Machine Learning Techniques used for estimating State of Charge for Li-ion Battery

Electric Vehicles (EVs) are making more and more financial sense as the operational cost of EVs as compared to Internal Combustion Engine Vehicles (ICEV) is becoming much lower. To further increase the confidence of users in EVs, precise State of Charge (SOC) estimation is need of the hour. The SOC of a battery depends on several factors such as current, voltage, age, temperature, etc. SOC estimation of a Lithium-ion based battery chemistry is a highly complex process. This is due to the fact that Lithium-ion batteries are highly nonlinear, time variant and complex electrochemical systems. A comprehensive study of SOC estimation techniques based on Machine Learning algorithms used in Battery Management Systems (BMS) is performed in this paper. Machine Learning algorithms are highly data driven and can give accurate estimation for nonlinear systems. A critical explanation including pros and cons of all these algorithms is presented. The paper also suggests future developments in BMS.