Comparative Machine Learning Approach to Predict Electric Vehicle Battery Life

The growth of electric vehicles (EVs) has been significantly influenced by advances in battery energy storage. To improve the reliability and efficiency of EVs, accurately predicting the health of the battery and its remaining useful life (RUL) stands acute. In a battery management system (BMS), monitoring state temperature is critical. While several algorithms have been explored for battery health management, there remains substantial scope for further research. The BMS estimates the battery’s health state and RUL. Machine learning techniques improves the prediction accuracy. The research reports the challenges in battery health with the increasing electric vehicles demand. It leverages machine learning techniques using NASA battery dataset in predicting the RUL of batteries. The prediction models implemented include Long short term memory model, Support Vector Regression model (SVR), Random Forest model and Gradient Boosting model. Features of the dataset includes voltage, load current, current load and temperature are analyzed to predict battery life. The models are compared by measuring the performance metrics, with the Gradient Boosting model outperforming Mean Square error and Mean Absolute Error.