Estimation and Influencing Factor Analysis of Lithium-Ion Batteries State of Health Based on Features Extraction

State of health (SOH) estimation is the key function of battery management systems and an important prerequisite for ensuring the safe and stable operation of electrochemical energy storage systems. There are several major problems with existing estimation methods, such as low accuracy, and high complexity. In addition, the interpretability of the estimation model is often ignored resulting in poor reliability of SOH estimation. Recently, some methods extract health features from battery charging curves for SOH estimation to mitigate these problems but most of them suffered from uninformative features and complex features correlation. To bridge the gap, this paper customizes a method that combines estimation and influencing factor analysis of lithium-ion batteries' state of health based on feature extraction. Firstly, six health features were extracted from the charging curve of lithium-ion batteries, including two new similarity features. they are proposed to quantify the similarity between the initial cycle charging voltage curve and the current cycle charging voltage curve, namely the dynamic time warping distance and the Wasserstein distance. Secondly, the CatBoost method is applied to estimate battery SOH, it is an efficient ensemble learning framework based on gradient boosting decision tree. Finally, the Shapley additive explanations (SHAP) method is introduced to analyze the impact of health features on the estimation results and the coupling relationship between features. The proposed method can further improve the interpretability of the method on the premise of ensuring the accuracy of SOH estimation. Validation experiments are conducted using multiple batteries from the University of Maryland battery aging dataset. The results show that the proposed SOH estimation method has high accuracy with the average RMSE and MAE are only 2.20% and 1.16%. Compared with the existing methods, the estimation accuracy is improved by more than 12%, and the speed is increased by more than 40%. In addition, the method can still achieve accurate evaluation for batteries with different initial charging SOC and different capacities, which proves that the proposed method has strong adaptability. Finally, the analysis of the evaluation results shows that the similarity feature can effectively characterize the changing trend of the internal resistance in batteries. The CCCT and CVTCT in the health feature have a positive impact on the battery SOH, and DTW and WAS have a negative impact on the battery SOH, and there is a linear coupling relationship between health features. The following conclusions can be drawn from the analysis: (1) The similarity features obtained based on dynamic time warping distance and Wasserstein distance have a strong correlation with the battery internal resistance, and can be used as new health features for battery SOH estimation. (2) Multiple battery data verification experiments show that the battery SOH estimation accuracy of the CatBoost method has been improved by more than 12% compared to existing methods, and its generalization ability is stronger. (3) The contribution rate and influence degree of each health feature were quantified by the SHAP method, which revealed the coupling relationship between features and improved the interpretability of the model. The analysis results showed that CCCT, CVTCT, DTW and WAS characteristics were the key factors affecting SOH assessment. Among them, CCCT and CVTCT have a positive impact, while DTW and WAS have a negative impact on the estimation model. In addition, there is a negative correlation between the constant current charging time and the dynamic time warping distance features and a positive correlation with the stable time of dV/dt. © 2023 Chinese Machine Press. All rights reserved.