Research On Health State Estimation And Retirement Sorting Of Electric Vehicle Battery Based On Data-Driven

In the context of the "dual carbon" goal,major automobile companies have been continuously deepening their research and development of new energy vehicles,and the electric vehicle-related industry has been rapidly improving,resulting in a rapid decrease in production and usage costs of electric vehicles.Electric vehicles have become the mainstream direction of the automotive industry.As the "heart" of electric vehicles,lithium-ion batteries have a critical role in energy supply.However,due to the structural and material characteristics of lithium-ion batteries,safety issues during operation and the disposal of a large number of retired batteries are crucial.This article is based on data-driven methods and takes the lithium-ion battery of electric vehicles as the research object.Following the life cycle of lithium-ion batteries,the article studies the health status estimation and retired battery sorting and utilization issues,with the following main innovative work:Firstly,the changes in the voltage curve,current curve,and IC curve of lithium-ion batteries during charging and discharging under different charging protocols were studied,and multiple external features were extracted from the charging and discharging data under different charging protocols.Then,according to the actual scenario requirements,the task of lithium-ion battery health status estimation was divided into two scenarios.In the scenario of estimating the health status of lithium-ion batteries in electric vehicles,a lithium-ion battery health status interval estimation method that is suitable for different charging protocols was proposed.This method first automatically selects appropriate external feature parameters from different charging protocols,and then combines quantile regression algorithm with support vector regression algorithm to propose a support vector quantile regression model.The improved model can consider the uncertainty information during the operation of lithium-ion batteries and provide more referenceable estimation results.In the scenario of retired battery sorting,a deterministic estimation method for the health status of lithium-ion batteries was proposed.This method first uses the Pearson coefficient to perform feature selection from all features to reduce the input data volume of the model.Then,the encoder structure is used to improve the time convolutional network.The improved model can speed up the estimation speed of the health status of lithium-ion batteries and maintain high estimation accuracy.Then,the problems that may exist in the current commonly used methods for retired lithium-ion battery sorting on different lithium-ion battery objects were studied.Based on this,a comprehensive index sorting method using multiple performance indicators to evaluate the comprehensive state of lithium-ion batteries was proposed.In order to verify the practicality of the methods and models proposed in this article,experimental verification was conducted on multiple lithium ion battery public datasets with different charging protocols.The experimental results show that the external characteristics of the selected lithium battery charging and discharging data can be used for different charging protocols;Health state estimation methods can achieve better estimation results for different application scenarios;The comprehensive index sorting method can avoid unreasonable sorting and utilization caused by inconsistent capacity degradation curves of different lithium batteries,and has certain practicality.