A novel differential voltage segment-based rejection and clustering method for enhancing short and long term consistency in reuse of retired batteries

Abstract

Retired battery reuse presents significant economic and environmental advantages, with battery consistency being a critical factor for optimizing performance in second-life applications. Current research methods have the following deficiencies: Feature selection lacks a good cost-efficiency balance, hindering practical scalability; Clustering algorithms are deficient in robustness, adaptability, and initialization sensitivity, restricting the use in multidimensional battery datasets; Verification experiments often ignore comprehensive metrics, failing to consider long-term consistency in real applications. To address these issues, this paper introduces an innovative approach to enhance the uniformity of retired battery systems, incorporating four key advancements: First, coefficient of variation (CV) analysis is applied to differential voltage (DV) curves to identify critical DV segments. Second, a hybrid algorithm integrating the nearest neighbor technique with fuzzy C-means (FCM) clustering is introduced for optimal battery grouping, improving long-term capacity retention by at least 4.8–9.9 % compared to conventional methods. Third, a distance-based outlier detection method is proposed, which enhances long-term system consistency by 15.7 % relative to density-based approaches. Finally, a weighted-average methodology is developed to systematically assess battery consistency. Collectively, these innovations establish a comprehensive framework for improving the reliability and sustainability of retired battery systems in secondary use scenarios.