Feature selection and SOH prediction of lithium-ion batteries based on XGBoost

This study focuses on identifying the most relevant features or characteristics of lithium-ion batteries to improve predictions of their State of Health (SOH), which measures their remaining lifespan and overall condition. It involves selecting the most important data attributes and using them to monitor and assess the battery’s performance over time. Specifically, the XGBoost algorithm is used for feature selection and pruning with a lithium-ion battery dataset. Features unrelated to SOH prediction is dropped to reduce computational costs during training and prediction. The CALCE battery dataset is divided into training and validation sets to predict the State of Health (SOH). The results show that there is a strong correlation between the predicted values and the actual values.