Multi-model Ensemble Framework with SHAP Analysis for the Battery State of Health Estimation and Remaining Useful Life Prediction

Lithium-ion batteries are central to modern energy storage systems, and accurate health management is critical for ensuring battery safety and prolonging cycle life. However, achieving an accurate state of health (SOH) estimation and remaining useful life (RUL) prediction remains challenging, particularly under varying operating conditions, across diverse chemistries, and when generalizing to unseen tasks. To address these challenges, this study proposes an ensemble framework that integrates five base models. Model weights are determined through a hybrid approach that combines the entropy weight method with the Technique for Order Preference by Similarity to an Ideal Solution, improving stability and predictive reliability. Furthermore, a weighted Shapley Additive Explanations (SHAP) approach is introduced to produce unified and consistent interpretability outputs. Evaluations on public datasets show that the proposed framework is robust across diverse battery chemistries. For SOH estimation, it achieves a root-mean-square error below 2%, with a best case of 0.11%. The framework also generalizes well to RUL prediction. By applying weighted SHAP to physics-based health factors, the study reveals distinct feature dependencies for SOH estimation and RUL prediction. This scalable framework delivers accurate, reliable prognostics and decision support for electric vehicles and large-scale energy storage.