准确快速的分选仍是实现退役锂离子电池安全再利用的关键挑战。本研究提出一种改进的两步聚类法以解决该问题。首先构建包含容量特征、直流内阻和温升速率的三维特征向量来表征电池老化状态。该容量特征可从局部充电曲线高效提取，从而缩短测试时长。其次提出基于排序点识别聚类结构算法，以剔除异常电池并获取可能的聚类数量随后，定义了K-means算法中质心初始化的参考范围，并综合放电电压曲线的数值特征与形态学特征进行二次分选。开发了融合静态与动态指标的双维度评估框架。采用200个实际退役电池进行的实验验证表明，该方法使静态一致性参数和动态一致性参数均获得显著提升，最大改善率分别达到64%和80%。再利用实验结果表明：通过两步聚类法组装的退役电池组的潜在循环次数，较随机筛选退役电池组装的电池组提升了68%。该方法为退役电池回收利用提供了高效解决方案。K defines the initialization reference range for centroids in the K-means algorithm, and numerical and morphological features of discharge voltage curve are combined to perform secondary sorting. A dual-dimensional evaluation framework integrating both static and dynamic indicators has been developed. Experimental validation using 200 real-world retired batteries demonstrates significant improvements in both static and dynamic consistency parameters, with maximum enhancement rates reaching 64% and 80%, respectively. The result of the reuse experiment reveals the potential cycle numbers of reuse pack manufactured via two-step clustering is 68% higher than that of reuse packs assembled from randomly selected retired batteries. This method provides an efficient solution for the recycling of retired batteries.
