Performance Evaluation of Different Optimization Techniques for SoC Estimation in EV Battery Management System

Battery management system (BMS) is an embedded system in electric vehicles (EVs) that is responsible for maintaining the performance and efficiency of a battery which allows proper charging and discharging by estimating state-of-charge (SoC), balancing cell voltages, etc. This article compares how well two machine learning (ML) models: specifically, long short-term memory (LSTM) and recurrent neural networks (RNN) estimate the SoC of EV batteries. The performance of each model is assessed by their mean squared error (MSE) values, which are the main evaluation metrics. Preliminary results indicate that the LSTM model, which has more sophisticated handling of long-stage dependencies, classifies the better SoCs than other method like RNN. These results represent an important contribution to the selection of optimal optimization techniques used in battery management, and provide a basis for further research and development. Within the Python programming environment, the models are generated and tested using data from the "Panasonic 18650PF" battery.