Rapid Estimation of Lithium Battery Health Status Based on Complementary Short-Term Features

Abstract The conventional method for assessing the health status of lithium batteries typically necessitates comprehensive data from complete charging and discharging cycles. The prolonged duration required for the collection of such data may lead to issues including time inefficiency and delays in battery state estimation processes. In response, this paper presents a rapid method for estimating health status based on local information from a short process. Additionally, to address the situation where the correlation of features is low in specific regions of the entire voltage domain, a complementary strategy for features is proposed. This strategy allows for a quick and accurate estimation of battery health status using only data from short process charging and discharging intervals across the entire voltage domain. First, the features of the short process that can represent the health status across the full voltage domain are extracted. Subsequently, a multi-feature fusion approach combined with the LightGBM algorithm is employed to construct a health status estimation model. Finally, the effects of various battery types, different operating conditions, and diverse sampling window sizes on estimation accuracy are analyzed through experiments, thereby demonstrating the feasibility and effectiveness of the proposed approach.