Explainable XGBoost-SHAP-Based Classification of Battery Degradation Environments Using Multi Features

With the rapid growth of the electric vehicle (EV) industry driven by carbon neutrality policies, the number of retired lithium-ion batteries (LIBs) is increasing. Although batteries reach end of life in high-power applications, they still retain residual capacity for reuse in low-power systems, emphasizing the need for non-destructive and reliable state assessment. However, conventional indicators such as capacity or internal resistance are insufficient to distinguish complex degradation mechanisms that vary with environmental and operational history. This study integrates incremental capacity analysis (ICA), differential voltage analysis (DVA), and distribution of relaxation times (DRT) to extract physically meaningful features that reflect static and dynamic electrochemical behavior. Using these multi-domain features, an XGBoost-based classification model was developed to identify degradation environments. The model achieved classification accuracy with limited experimental data. To enhance interpretability, SHAP (Shapley Additive Explanations) analysis was employed to quantify feature contribution, linking the model’s decision to electrochemical phenomena including ICA and DRT peak variations. The proposed ICA-DVA-DRT fusion with XGBoost-SHAP framework enables explainable and reliable inference of degradation environments, contributing to efficient repurposing of second-life LIBs.