An interpretable state of health estimation method for lithium-ion batteries based on multi-category and multi-stage features

Accurate state of health (SOH) estimation of lithium-ion batteries is essential for quality control and improving safety and efficiency of electric vehicles and energy storage systems. Due to inadequate consideration of different categories of measurement data and stages of battery working in constructing degradation features, current SOH estimation methods are limited in accuracy and fail to facilitate development of specific battery management strategies. To address these issues, this paper proposes a novel SOH estimation method for lithium-ion batteries based on multi-category and multi-stage (MC-MS) features and an input optimization interpretable multi-variable long short-term memory (IO-IMV-LSTM) model. First, a MC-MS degradation feature construction scheme is presented based on working principles and degradation mechanism of lithium-ion batteries. Then, an improved interpretable estimation model called IO-IMV-LSTM is introduced, which uses feature importance to identify the most effective features and optimize model input. Finally, influence of battery measurement data from different categories and working stages on SOH estimation results is analyzed, and the findings can be used to optimize battery utilization and maintenance strategies. A public dataset together with degradation dataset of our testbed is used for validation experiments. Results show that the MC-MS features are superior to existing features and the performance of the IO-IMV-LSTM model is significantly improved through input optimization. The proposed method showcases remarkable improvements and advantages compared with state-of-art methods. Specifically, the proposed SOH estimation method achieves satisfactory performance on four evaluation indicators: mean absolute error, mean absolute percentage error, root mean square error, and R-square, with average values of 0.20%, 0.22%, 0.24%, and 99.34%, respectively.