ABSTRACT Electric vehicles and energy storage technologies use lithium‐ion batteries, which are a significant contributor to the renewable energy market. Battery technology is susceptible to gradual deterioration in battery performance and a shorter lifespan. It is essential to monitor the battery's status and discover any damage as soon as possible to avoid unpredicted breakdown in operation and management and also for ensuring the safety of people. Using data‐driven techniques to forecast a battery's residual useful life is quicker and less complicated than using experimental methods. Therefore, using battery datasets, machine learning models are used in this paper to predict the residual useful life of Li‐Ion batteries. Different regression models such as random forest, adaboost, gradient boost, bagging regression, support vector regression, decision tree, extra tree regression, linear regression, stochastic gradient descent regression, K‐nearest neighbor regressor, XGBoost and long short term memory model are implemented for prediction of residual useful life. Apart from conventional, advanced machine learning model such as Bayesian optimization‐based LightGBM (Light Gradient Boosting Machine) is proposed in this research work. The Bayesian optimization aids in finding the optimal hyperparameters of the LightGBM model to enhance the prediction accuracy of battery RUL and alleviate the computational complexity. To determine which ML model is optimal for residual useful life prediction, a comparative study of the models' performance metrics is also provided.
Patra et al. (Sun,) studied this question.