Abstract: Aiming at the problem of low screening accuracy and efficiency in the reused retired batteries, a retired battery screening method based on a MLP-SVR-LR hybrid model is proposed. Based on a new voltage segment feature extraction, the capacity-related features are obtained from the constant current charging voltage segments. Combined with the MLP-SVR-LR hybrid model, the highly-precised and highly-efficient screening of the retired batteries is achieved. Meanwhile, a particle swarm optimization-based K-means algorithm and an improved equal number strategy based on the Silhouette Coefficient (SC) are utilized to evenly recombine the screened batteries according to their capacities and internal resistances, achieving the direct group utilization of the retired batteries. A simulation and experimental platform for the lithium iron phosphate batteries is built to analyze the proposed battery screening and regrouping strategies. The results show that the proposed method improves the screening accuracy by more than 2% compared with the similar methods, verifying the correctness and effectiveness of the proposed strategies from the perspectives of retired battery screening accuracy, screening efficiency, and consistency of the recombined batteries.