State Estimation And Consistency Study Of Lithium Ion Batteries

As the penetration rate of the new energy vehicle industry continues to rise,the sales of new energy vehicles usher in explosive growth.As the core component of new energy vehicles,the installed power battery shows a continuous rising trend.However,there are still some key technical problems unsolved behind its rapid development.Among them,What to do with a large number of decommissioned batteries is an unavoidable problem in the process of the popularization of new energy vehicles.If the huge number of batteries cannot be recycled in time,it will cause great environmental pressure and waste of resources,which goes against the original intention of the promotion of new energy vehicles.Therefore,in order to realize the rapid recovery and step utilization of batteries,it is necessary to further study the key technologies such as the functional state,health state and consistent sorting mechanism of batteries.In view of the above problems,this paper studies the multi-state estimation and consistent sorting methods of lithium-ion batteries based on the battery independent experiment platform.The main research contents and innovations are as follows:(1)Research on State of Charge(SOC)estimation of battery.Aiming at the problem of poor applicability of the battery SOC estimation model in different operating conditions,this paper proposes a Transformer model based on Convolutional Neural Network(CNN)architecture to realize SOC estimation of the battery.In order to improve the accuracy of SOC boundary model estimation,a symmetric filling method is used to improve the convolutional layer of CNN network.In order to verify the accuracy of the improved model,the battery data sets under three working conditions were used for training and verification,and the existing SOC estimation methods were compared.Finally,the experimental data of the designed operating conditions were collected through the autonomous battery experimental platform to test the model generalization.(2)Estimation of battery State of Health(SOH).To solve the problem of low accuracy of SOH estimation,this paper proposes a method of SOH estimation based on optimized Long Short-term Memory(LSTM).Firstly,the characteristics of isobaric drop discharge time and relaxation voltage in the battery charging and discharging process were extracted as sample data,and then used grey correlation analysis and Spearman coefficient for correlation analysis.Sparrow Search Algorithm(SSA)serves as a optimizer to quickly search important parameters of the LSTM model,and the improved softsign activation function is used to solve the problem of vanishing gradient of the model.Finally,data samples are input into the built model for training and testing,and compared with some existing SOH estimation methods.(3)Research on consistent sorting method of battery.Fast sorting strategy is the key to step utilization of battery.In this paper,a multi-classification model of Support Vector Machines in kernlab(KSVM)based on Particle Swarm Optimization(PSO)is proposed to realize fast battery sorting.The battery test is designed to extract the capacity,internal resistance and voltage of 96 lithium iron phosphate battery samples.The weight of three parameters was analyzed by using Critic weight method,and the classification of batteries was realized according to the comprehensive score of each cell.Based on the classified data samples,Gaussian kernel support vector machine was used to train multiple classification models to realize fast sorting of batteries.Finally,the battery module testing equipment was built to evaluate the consistency of batteries before and after classification.