Health Feature Extraction and Efficient Sorting of Second-Life Lithium-Ion Batteries

Efficient sorting has always been a difficult problem for a large number of second-life batteries. In this article, a sorting method is developed for second-life batteries based on ordering points to identify the clustering structure (OPTICS) algorithm. A three-dimensional vector is created as health feature to map the aging degree of second-life batteries, including capacity feature, direct current resistance and polarization resistance. The vector can be accurately extracted from a partial charging curve to reduce testing time. Then, the OPTICS algorithm is presented to cluster the second-life batteries with a good consistency of aging degree and screen out some outliers. Specifically, a distance graph is used to determine the input parameters of the algorithm and the number of clusters in advance, which essentially improves the accuracy of clustering. 110 second-life batteries are divided into two clusters and outliers, and the clustering results show that the proposed method can significantly improve the aging consistency of second-life batteries. The standard deviations of capacity and polarization resistance are reduced by up to about two and five times compared with other algorithms.