Abstract:

The State of Health （SOH）， a critical metric for evaluating battery aging and performance degradation， requires accurate estimation to ensure the safe operation and lifespan management of battery systems. Compared to the well-established lithium-ion battery systems， the aging mechanism and capacity degradation behavior of sodium-ion batteries remains insufficiently understood. In this study， a SOH estimation method for sodium-ion batteries is proposed by fusing incremental capacity （IC） and relaxation voltage （RV） features. The IC curves are employed to analyze phase transition dynamics during charge/discharge processes， while RV features are utilized to characterize electrode polarization recovery patterns during resting periods， thereby comprehensively revealing multi-dimensional aging mechanism. A feature fusion model is developed to enhance the sensitivity and noise immunity of health indicators. By leveraging machine learning algorithms， the mapping relationship between IC/RV-derived features and SOH is established， constructing an LSTM-Attention （Long Short-Term Memory network integrated with an attention mechanism） based estimation model. The experimental results show that the proposed method achieves superior SOH estimation accuracy （RMSE<0.51%，MAE<0.40%） compared to single-feature approaches， providing a robust solution for real-time health monitoring and industrial deployment of sodium-ion batteries.

Key words: Sodium-ion batteries, SOH estimation, incremental capacity, relaxation voltage, LSTM-Attention