Prediction of State of Charge (SOC) of Electric Vehicle (EV) Batteries using Machine Learning Technique

Accurate prediction of the state of charge (SOC) of electric vehicle (EV) batteries is essential for improving the operational efficiency, reliability and lifetime. The SOC of EV batteries is predicted comprehensively in this work utilizing machine learning (ML) techniques and a prototype. Using an on-board diagnostics system (OBD) during real-time drives, information on battery characteristics and SOC was gathered. The best models for predicting SOC were determined to be random forest regression and recurrent neural networks. A Grafana dashboard was created to show battery SOC estimates in real-time after a prototype was produced, tested, and confirmed. This methodology can help EV users to optimize their driving range and reduce battery damage due to over-discharge.