Prediction of battery critical parameters using machine learning algorithms for electric vehicles

To enhance the adaptability of electric vehicles (EVs) and mitigate the intermittent nature of renewable energy sources, energy storage via batteries is imperative. Accurate forecasting of battery performance parameters is vital for optimal utilisation. This study introduces a machine learning algorithm for electric vehicle battery management systems (BMS), focusing on predicting state of charge (SoC) efficiently and precisely. Utilising linear regression and long short-term memory (LSTM) models, the algorithm constructs and deploys predictions. Training data, obtained from Li-ion battery packs during charge-discharge cycles via smart BMS, enables precise modelling. Predicted values are validated against empirical results, and the resultant error guides algorithm refinement for enhanced accuracy. The algorithm, integrated into a web application using Streamlit, achieved a remarkable 99% R2_score, indicating its robust performance. This framework advances EV battery management, facilitating informed decision-making and optimising energy utilisation in conjunction with renewable sources.