Classification and reorganization of retired power batteries based on bp and k-means clustering

Abstract

Abstract The article proposes an efficient method for classifying and reorganizing retired power batteries. Firstly, the K-means clustering algorithm is used to enhance electrochemical response consistency in reorganized battery packs by considering the polarization time constant. Secondly, pulse power test parameters and battery health feature parameters estimated by a BP network are utilized for battery classification and reorganization. This approach addresses issues like lengthy grouping time and the inability to determine specific application scenarios. Simulation results show over 97% classification accuracy for individual batteries and good consistency after grouping, indicating high practical value.