Comparative Analysis of Machine Learning Models for Li-Ion Battery SoC Estimation

The SoC estimation problem is a pivotal part of the smart Battery Management System (BMS). Although several research articles discuss useful SoC estimation techniques, the accuracy of SoC estimation module needs improvement considering its utility to other vehicular applications. Citing the same, the article analyses data-driven, supervised Machine Learning (ML) regression models used for the SoC estimation problem. Further, the requirement, working, and limitations of the models are discussed in the article in depth. The ML model selection problem is hence focused on while considering key aspects of the SoC estimation and performance indices of the ML model. The article attempts to provide much-needed clarity on the ML models’ selection, utility, and implementation feasibility for the SoC estimation to advance the same for the possible real-time application. Panasonic 18650PF Li-ion battery cell real-life data was considered for the training and testing of the ML models considered. The script for the ML application was written in Python.