Predicting the remaining useful life of nickel-manganese-cobalt batteries using ensemble gradient boosting with probabilistic estimation

The paper presents a comprehensive analysis of supervised machine-learning models for predicting the Remaining Useful Life (RUL) of Nickel-Manganese-Cobalt (NMC) batteries subjected to ageing tests. The predictive performance of three models – K-Nearest Neighbors (KNN), Multi-Layer Perceptron (MLP) and Light Gradient Boosting Machine (LGBM), each representing the class of instance-based, neural network and gradient boosting frameworks respectively are evaluated. A systematic hyperparameter analysis is firstly conducted to obtain optimal hyperparameters followed by training the model and evaluating the model performance on the test dataset. The R² score for the KNN, MLP and LGBM model on the test dataset were 0.987, 0.977 and 0.990 respectively, suggesting the superior performance of the LGBM model. Following this, the performance of the LGBM model is further enhanced by creating an ensemble framework with the state-of-the-art Natural Gradient (NG) boosting algorithm, with an improved R² score of 0.995 on the test data. In addition to improving the performance of the standalone LGBM model, the ensemble framework with NG boosting also provided probabilistic uncertainty estimates supporting enhanced decision-making and preventive maintenance in battery health management.