Research on Battery Inconsistency Detection Based on Improved Weighted Outlier Test

Cadmium-nickel battery pack as the power source of the auxiliary power supply system for CRH trains, its inconsistency will more affect the overall performance of the battery pack and the safety of train operation. This article proposes a method that integrates the objective weighting algorithm (CRITIC) with the subjective weighting algorithm optimized by Improved Particle Swarm Algorithm (IPSO-AHP). It combines the Local Outlier Factor Algorithm (MDS-LOF) based on the multidimensional scaling analysis for the battery group inconsistency test. The MDS-LOF algorithm is applied to reduce the dimensionality of the discharge data and map it onto a onedimensional time scale, and calculate the outlier detection score of each node in time units to obtain a battery inconsistency criterion evaluation model. Experimental verification was conducted using real discharge data of CRH cadmium-nickel batteries. The diagnostic results of the model indicate that the fusion weight assignment method can improve the consistency ratio of parameter weight values. The fusion score criterion algorithm can provide an intuitive evaluation of the differences in cell inconsistency between battery packs. This research holds significant practical value.