In order to solve the problems of poor consistency, difficult screening and low efficiency of decommissioned battery separation, an improved GSAA-Kmeans (GSK) algorithm is proposed from the three dimensions of battery internal resistance, open-circuit voltage and capacity. The improved fuzzy GSK algorithm framework is established by introducing the module paste logic system, and the battery performance is sorted by fuzzy logic, and the batteries with high performance, medium performance and low performance are selected preliminarily. The improved GSK algorithm is used to re-sort the batteries with different performance, and the clustering results are verified by experiments. The average consistency parameter δ of this method is 10.53%, which is 32.56% lower than that of the traditional GSK algorithm, the capacity standard deviation coefficient is 42.39%, and the open-circuit voltage standard deviation coefficient is 52.83%. The results show that this method can effectively screen the battery monomers with high consistency, and verify the superiority of the proposed sorting strategy.