Research On Fast Screening And State Of Health Estimation Of Retired Power Batteries Based On Data Driven Method

Against the background of increasing environmental pollution and non-renewable energy shortage,the electric vehicle industry is booming.Lithium-ion power battery has become the main power source of electric vehicles because of its high energy density,low self-discharge rate and long life.Lithium-ion power batteries always deteriorate when they are used,and they usher in the first peak of retirement in 2020.It is estimated that the total amount of retired power batteries in China will reach 116GWh in 2025.The echelon utilization of retired batteries has become the focus of the sustainable development of electric vehicle industry.However,the echelon utilization technology of retired batteries is still not mature.On the one hand,the screening method of retired batteries before echelon utilization has low efficiency and poor universality.On the other hand,the health state estimation of retired batteries after echelon utilization has poor generalization ability and low accuracy,which restricts the promotion and development of battery echelon utilization.In order to solve the problems of slow screening and difficult estimation,this paper proposes a data-driven method for fast screening and accurate health state estimation of retired lithium batteries.The main contents are listed as follows:Firstly,the inconsistency of the retired lithium-ion battery is evaluated.Taking the retired LiFePO4 battery as the research object,the experimental platform Is built and the life cycle experiment is designed to analyze the inconsistency of the lithium-ion battery in the aging process,so as to provide theoretical support for the necessity of retired battery screening.Secondly,the retired batteries are screened quickly based on fuzzy c-means algorithm.At first,the capacity and internal resistance of 176 retired 1 LiFePO4 batteries are tested,and the distribution of aging parameters is fitted and analyzed by mathematical method.The incremental capacity curve is denoised by Kalman filter,and partial charging curves is intercepted based on incremental capacity analysis.The features related to aging state are extracted as the input of screening model which is built with the fuzzy c-means clustering algorithm.The screening result on 176 LiFePO4 batteries proves the high accuracy and high efficiency of the approach.The screening accuracy can reach 90.9%.With permitted error of 1%,it can be as high as 95.5%.The screening efficiency is about 10.5 times of the traditional separation method.Then,aiming at the problems of low precision of unsupervised clustering screening method and randomness of capacity demarcation,screening method based on BP neural network is proposed.However,because the initial value of the algorithm is uncertain,the improvement of screening precision is not obvious and the result is extremely unstable.On this basis,genetic algorithm is used to optimize the model,and the screening precision of the optimized model reaches 94.4%,and the stability of separation results is improved obviously.Finally,based on the neural network algorithm optimized by genetic algorithm,the health state of the retired battery is estimated accurately.The capacity of 103 retired LiNCM batteries are tested and analyzed under different temperatures and charging and discharging rates.Taking the aging characteristics of partial charging curve as input and the corresponding capacity as output,the SoH estimation model is trained based on GA-BP neural network.The accuracy of SoH estimation under different working conditions is verified,and the effectiveness of the estimation method in LiFePO4 battery is also verified.The maximum error of the proposed estimation method is less than 2%,and the average absolute error and root mean square error are both less than 1.5%.