Regrouping and echelon utilization of retired lithium-ion batteries based on a novel support vector clustering approach

Abstract

To avoid severe resource waste and environmental pollution problems, research on the retirement of power lithium-ion batteries (LIBs) for electric vehicles (EVs) has attracted significant attention. Echelon utilization is one of the most prevailing strategies to solve the problems of reusing retired LIBs. In this article, we present a clustering and regrouping framework for retired LIBs based on a novel equal-number support vector clustering (SVC) approach, which provides a new perspective to address above problems. First, we extract the feature parameters used in clustering [including capacity, internal resistance, and remaining useful life (RUL)] and quickly screen out batteries without echelon utilization value. Then, based on the results of SVC, an equal-number clustering strategy is proposed. The consistency within the battery pack after using equal-number SVC approach has been significantly improved, and the battery pack can be directly applied to the different echelon utilization scenarios. Finally, based on the public dataset, 60 batteries equally divided into four clusters were used to verify the proposed approach. In addition, the results show that compared with the initial random grouping, the average standard deviation of capacity, internal resistance, and RUL used to evaluate consistency within the group are reduced by about 1.55 times, 1.53 times, and 3.27 times, respectively. The variance and maximum difference within each group are also reduced. We also compare$k$-means and Gaussian mixture models (GMMs) clustering algorithms, and the results also suggest that the equal-number SVC approach is quite promising. The presented approach is of great significance for applications involving the screening and recycling of retired LIBs for EVs.