Consistency evaluation and cluster analysis for lithium-ion battery pack in electric vehicles

Abstract Consistency is an essential factor affecting the operation of lithium-ion battery packs. Pack consistency evaluation is of considerable significance to the usage of batteries. Many existing methods are limited for they are based on a single feature or can only be implemented offline. This paper develops a comprehensive method to evaluate the pack consistency based on multi-feature weighting. Firstly, the features which reflect the static or dynamic characteristics of batteries are excavated. Secondly, a weighted method of multi-feature inconsistency is proposed to evaluate pack consistency. In which case, the entropy weight method is employed to determine the weight. Thirdly, an improved Greenwald-Khanna algorithm based on genetic algorithm and kernel function is developed to cluster batteries. Finally, nine months of electric vehicle data are collated to validate the proposed algorithms. Meanwhile, the main factor affecting consistency change is analyzed. The results show that with the usage of batteries, the difference between the cells becomes more serious, which weakens the pack consistency. Besides, the relationship between the consistency attenuation rate and the driving mileage can be approximated by a first-order function. The higher mileages will aggravate the pack inconsistency. Moreover, it has been proven that the improved clustering algorithm has stronger robustness and classification performance.