State of health estimation of lithium-ion battery aging process based on time-frequency fusion characteristics

Lithium-ion battery decay is closely linked to various side reactions within the battery. Limiting attention to the time-domain aging performance only cannot reflect the internal degradation state of the battery comprehensively and profoundly, which may not accurately estimate the state of health (SOH). Thus, a SOH estimation method for the lithium-ion battery aging process based on time-frequency fusion features is proposed. Firstly, a series of information such as incremental capacity (IC) curve, relaxation voltage curve, internal resistance, etc. are extracted and constructed as time-domain aging features. Here, dynamic time warping (DTW) with a sliding window is designed to construct a stable alternative feature, to overcome the challenge of disappearing IC peaks. Secondly, electrochemical impedance spectroscopy (EIS), which includes frequency-domain aging mechanism information, is decoupled by the distribution of relaxation times (DRT), combined with Pearson analysis, to obtain health indicators that reflect both the time and frequency domain aging behavior. Finally, sparse spectrum Gaussian process regression (SSGPR) is utilized as a more flexible SOH prediction model. On the NCA battery dataset at 25 °C, 35 °C, and 45 °C, the model's mean absolute error (MAE) and root mean square error (RMSE) are as low as 0.59 % and 0.71 %, respectively.