Estimation of Battery State of Charge Using Machine Learning Techniques

Battery management systems plays an important role in electric vehicle to ensure the safety of the battery. It is important to ensure the accuracy of the battery state of charge (SOC) for the safe, reliable and efficient operation of the electric vehicle battery. This paper presents an investigation into the estimation of battery SOC using machine learning models, specifically LSTM, Bidirectional LSTM and Vector Autoregression (VAR), as well as a combined model of VAR and LSTM. In this study, historical battery SOC data is utilized to train and evaluate the performance of the individual LSTM, Bidirectional-LSTM, and VAR models. Additionally, a combined model of VAR and LSTM is proposed to leverage the strengths of both approaches. Root Mean Square error (RMSE) has been considered as a performance metrices for the comparison of the models. The results indicate the effectiveness of the LSTM, Bi-LSTM, and VAR models in accurately estimating battery SOC. The combined VAR and LSTM model shows further improvement in prediction accuracy. The findings highlight the potential of machine learning algorithms, particularly LSTM, Bi-LSTM, and VAR, in battery SOC estimation.