Li-ion battery state-of-health estimation based on the combination of statistical and geometric features of the constant-voltage charging stage

State-of-health (SOH) estimation is critical in ensuring safe and reliable operation of Li-ion batteries. The first step in the estimation process is extracting features that reflect the SOH. This study proposes a novel method that utilizes both statistical and geometric features of Li-ion batteries to improve the accuracy of SOH estimation. Moreover, feature extraction is performed from the constant-voltage (CV) charging stage as it is unaffected by the randomness of charging onset point and does not require long resting after a full charge. Firstly, features are extracted from both statistical and geometric perspectives. Subsequently, these features are combined with the mean CV charging current to create a feature combination. Finally, the XGBoost algorithm is used to construct the SOH estimation model. The effectiveness of the proposed model is validated using three types of battery datasets. In all the experiments, the root mean square error and the mean absolute error of the proposed model are less than 1.3 % in the overall test set. Moreover, the proposed model achieves high accuracy for all three battery types and demonstrates good adaptability to different discharge current rates. Furthermore, the model achieves high accuracy, even with only the first 50 % of the CV charging data.