Predictive Modelling for State - of - Charge Estimation of Li-Ion Electric Vehicle Batteries: A Comparative Study _* of Regression Models

This paper presents SOC estimation criteria for lithium-ion batteries (LIBs) with varying profiles using regression models. The primary state that is explicitly known to the users and is tracked by the battery management system (BMS) is the state of charge (SOC), which is the foremost importance of a well-monitored BMS. This comparative study demonstrated four regression-based machine learning (ML) methods: support vector regression (SVR), random forest (RF), extra tree regression (ET), and extreme gradient boosting (XGB) regression for the estimation of SOC. The models were evaluated based on the root mean squared error (RMSE), mean absolute error (MAE), and computational time. From the study, we found that the XGB regressor outperforms the other ML algorithms in both error rate and computation time (CT).