Study Of Derived Battery Features for Real Time Estimation of SOH And RUL Of EV Battery Using Data Analysis

Lithium-ion batteries have been widely used in EV’s. It is desired to predict the SOH of batteries to achieve optimal operation and health management. The most significant obstacle to accurately predicting battery health is choosing battery features. In this study, numerous data analysis strategies are introduced to manage feature irrelevancy and help in determining which features can be selected and used in real time. The first step in manually crafting features is to analyses the evolution pattern of numerous essential characteristics of battery. Second, correlation between selected features and degraded capacity analyses. Then, selected features are fed into representative machine learning regression model to effectively predict remaining capacity of battery to find the SOH status. Finally, the remaining capacity of battery is selected as feature to predict the RUL in terms of remaining charge-discharge cycles.