Publication

Abstract : The State of Health (SoH) needs to be accurately estimated for Electric Vehicle (EV) batteries to manage performance, safety and longevity. This study aims to propose a data-driven method with the Random Forest Regression (RFR) model to accurately predict the SoH. This approach builds on historical battery performance data to train the RFR model, which is particularly useful for capturing complex nonlinear relationships between input features and the SoH metric. Whereas model-based methods require electrochemical models, and data-driven methods often rely on extensive laboratory testing, our method demonstrates a pathway to a computationally efficient, flexible, and accurate approach that works across a diversity of battery types and use cases. It used key features like voltage, current, temperature, and charge/discharge rates as predictors, which allows a comprehensive examination of the current and former battery behaviours. This model has been evaluated against various benchmark datasets and has shown a high level of accuracy and robustness.