Research On SOH Estimation And RUL Prediction Methods For Electric Vehicle Lithium-ion Battery Based On Deep Learning

As the number of electric vehicle owners rises,lithium-ion battery for electric vehicles are being used on a large scale.To ensure that electric vehicle lithium-ion battery is used safely throughout their lifecycle,reliable and accurate methods for the state of health(SOH)estimation and remaining useful life(RUL)prediction of electric vehicles are required.As a result,it is critical to conduct research on the SOH estimation and RUL prediction of electric vehicle lithium-ion battery.In this thesis,in order to further improve the accuracy and effectiveness of the SOH estimation and RUL prediction of electric vehicle lithium-ion battery,the use of deep learning methods,convolutional neural networks(CNN)and recurrent neural networks(RNN)are applied to lithium-ion battery SOH estimation and RUL prediction,and a series of improvements and optimizations are made to the model.The main contents are as follows:1.Analyze the working principle and life decay process of electric vehicle lithium-ion battery,process and study the electric vehicle lithium-ion battery cycle data,further elaborate the process of extracting the health factors of electric vehicle lithium-ion battery,and the lithium-ion battery health factors are extracted from the Center for Advanced Life Cycle Engineering(CALCE)battery dataset at the University of Maryland.Additionally,the health factors are filtered using Pearson and Spearman correlation coefficients,which clarifies the input parameters for SOH estimation of electric vehicle lithium-ion battery.The dataset processing for lithium-ion battery SOH estimation is explained,followed by the dataset construction.2.For the respective characteristics of SOH estimation and RUL prediction of electric vehicle lithium-ion battery,CNN-GRU deep learning models with different structures are created,and experiments on the CALCE battery dataset are performed.The results show that the CNN-GRU deep learning model is able to better achieve SOH estimation and RUL prediction for electric vehicle lithium-ion battery,thus the validity of the method is verified,and further demonstrating the model’s higher accuracy through comparative experiments.3.Based on the characteristics of the predicted correlation relationship between lithium-ion battery SOH estimation and RUL prediction,this thesis proposes a hybrid model-based joint prediction method of lithium-ion battery SOH and RUL for electric vehicles.By constructing the CNN-Bi GRU hybrid model,the joint prediction method is used to achieve lithium-ion battery state prediction,and sets up a comparison experiment.The experimental results show that the joint prediction method can provide more comprehensive and useful information on the status of an electric vehicle lithium-ion battery,improve the accuracy of SOH estimation and RUL prediction for electric vehicle lithium-ion battery,aid in battery life extension,optimize electric vehicle lithium-ion battery management,and ensure the safe operation of an electric vehicle.