Estimation in Electric Vehicle Battery Systems

As the electric vehicle (EV) industry continues to grow, accurate State-of-Charge (SoC) estimation of the Battery Management System (BMS) has become increasingly crucial for the efficient and safe operation of EVs. Traditional methods often need help with battery behaviour’s inherent complexity and nonlinearity. This paper investigates the potential of Long Short-Term Memory (LSTM) networks, a form of recurrent neural network well-suited to time-series data, to improve the accuracy and reliability of SoC estimation. Through extensive simulations and real-world tests, the proposed LSTM-based approach demonstrates superior performance, capable of accounting for intricate battery dynamics and improving prediction accuracy. Furthermore, the robustness of LSTM networks to various driving and environmental conditions is analysed. The findings suggest that LSTM-based models represent a promising approach to advance SoC estimation methods in EV BMS, potentially contributing to enhanced EV performance and lifespan. This research paves the way for further investigations into deep learning applications within the EV industry, leading to innovative and sustainable transportation solutions.