Advanced Estimation of SoC and SoH for Li-Ion EV Batteries Using Soft Computing Techniques

The transition towards battery-powered technologies, driven by environmental sustainability goals and the need to reduce reliance on finite fossil fuels, has accelerated the demand for efficient energy storage solutions. Batteries, pivotal in enabling clean energy adoption and powering electric vehicles (EVs), necessitate robust Battery Management Systems (BMSs) for optimal performance and longevity. A critical function of BMSs is the accurate estimation of State of Charge (SoC) and State of Health (SoH), which measures the remaining energy in a battery relative to its full capacity. This paper explores advanced methodologies, including machine learning (ML) techniques, for enhancing SoC estimation accuracy in lithium-ion batteries. The study evaluates several ML models - Convolutional Neural Networks (CNNs), Feedforward Neural Networks (FNNs), and Long Short-Term Memory (LSTM) - using real-world data from controlled charging and discharging cycles. The results demonstrate that LSTM models exhibit superior performance in SoC and SoH estimation, achieving low Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) values across different operational scenarios. The research underscores the importance of model selection and data preprocessing techniques such as normalization and feature engineering in optimizing SoC and SoH estimation accuracy. Furthermore, the paper discusses the implications of battery aging and operational profiles on estimation methods and battery health monitoring.