State of Health and Remaining Useful Life Prediction for Lithium-Ion and Sodium-Ion Batteries Based on Segmental Incremental Capacity Features

Accurate estimation of state of health (SOH) and remaining useful life (RUL) under partial data conditions remains a significant challenge for practical battery management systems, as most existing methods rely on complete charging cycles or task-specific degradation indicators. To address this limitation, this study proposes a segment-based joint prediction framework that utilizes incremental capacity (d Q /d V ) features extracted from partial charging data. The main contribution lies in combining statistical descriptors with voltage-binned d Q /d V features to capture degradation information from incomplete measurements and integrating them into an XGBoost-based multi-output regression model for simultaneous SOH and RUL estimation. The proposed approach is validated on multiple lithium-ion and sodium-ion battery datasets, demonstrating strong generalization capability across different battery chemistries. Experimental results show that the model achieves a mean absolute error of 0.219% and an R 2 value of 0.985 for SOH estimation, while the root mean square error for RUL prediction is as low as 0.888 cycles. The average inference time is only 6 ms, indicating suitability for real-time battery management applications. Ablation studies further verify the contributions of data filtering, voltage binning, and hyperparameter optimization, whereas SHapley Additive exPlanations (SHAP) analysis reveals that voltage kurtosis and mid-range d Q /d V features are the most influential indicators of battery degradation.