Power Battery Grouping Technology Research Based On U-shapelets

With the continuous expansion of electric vehicle production capacity,power battery as the core component of electric vehicle is considered to be an important part of the development of electric vehicle market.Its performance is directly related to the cruising range and safety performance of electric vehicle.The output voltage and output energy of a single power battery are small.Therefore,in applications such as electric vehicles,power batteries often need to be connected in series and parallel to form a higher voltage and a larger capacity.Like a Tesla electric car with a cruising range of 500 kilometers,it consists of more than 7,000 18650 lithium batteries.The consistency of the single cells in the group not only determines the output energy of the power battery pack and the performance of the battery pack,but also seriously affects the service life of the power battery pack.In this paper,several methods for improving the consistency of the battery pack are proposed for the reasons of the inconsistency of the power battery pack.In the battery grouping process,the battery grouping technology is to classify the performance parameters of the single battery,and classify the single eells with similar performance parameters to ensure the consistency within the battery.Compared with the existing battery grouping method,this paper proposes an effective battery grouping strategy to convert the battery grouping problem into a time series clustering problem,so the battery curve clustering based on feature sub-shapes(u-shapelets)is proposed.This paper introduces the implementation process of battery curve clustering based on feature sub-shapes(u-shapelets).The extraction algorithm of the battery curve feature sub-shape(u-shapelets)is essentially a violent search algorithm,which searches for the local features of the battery curve with unrecognized conditions without supervision.The algorithm first takes all subsequences of the battery curve as candidate sub-sequences of u-shapelets,and performs quality assessment on all sub-sequences in the u-shapelets candidate set to find the best u-shapelets.Battery curve clustering based on feature sub-shapes(u?shapelets),the self-organizing feature mapping network(SOM)in the neural network is selected on the clustering algorithm to perform battery curve clustering.The clustering results were analyzed by using the unsupervised algorithm's evaluation index contour coefficients.Finally,the paper compares the multi-parameter grouping method and the grouping strategy proposed in this paper.The battery data is from Zhejiang Nandu Power Power Co.,Ltd.and the laboratory.The battery storage performance test and battery cycle performance experiment were carried out on the results of battery grouping,and the capacity and self-discharge rate of the battery were selected as the comparison parameters for the result analysis.The results show that the results of the power battery grouping technology based on the feature sub-shape are better.