Research On Data-driven-based Remaining Useful Life Prediction Of Li-ion Battery

The development of new energy vehicles based on electric vehicles has become an important choice for my country’s transportation sector in energy conservation,low carbon emission reduction and energy security strategies.Lithium-ion battery is the core component of electric vehicles.Effective health assessment and life prediction are the key to ensuring the endurance and safe and reliable operation of electric vehicles.It is also an important technical means for the operation and maintenance of electric vehicles.Based on the actual vehicle operating data of electric vehicles,the paper used data-driven methods to analyze and optimize research from multiple aspects of lithium-ion battery data preprocessing,health estimation,and remaining useful life prediction.The safety and reliability of the lithium-ion battery system can be greatly improved in practice,and it has practical value and practical significance.The research work of this article is as follows:(1)The research background and current research status of the remaining life of lithium-ion batteries for electric vehicles were summarized,and the significance and value of the research were clarified.The research ideas were established for estimating and predicting the health of lithium-ion batteries based on actual vehicle monitoring data,and then predicting the remaining battery life.It was analyzed that the working principle and aging characteristics of the battery,and four remaining life prediction models commonly used in time series prediction were introduced.(2)The life prediction process was developed by analyzing the operating data of electric vehicles and the aging mechanism of lithium-ion batteries.By comparing the common methods of outlier detection and missing value filling,the best method was chosen to perform data cleaning and data preprocessing on the actual vehicle data work,then carry out dimensional correction and smooth filtering of the current value to prepare data for subsequent battery state of health estimation and remaining life prediction.(3)The capacity as the evaluation index of the battery state of health was selected.A method was given to calculate the maximum available capacity of the deep charge-discharge segment through the inverse ampere-hour integration method,and obtain the corresponding SOH value.A supervised machine learning model and the SOH tag value of the deep charge and discharge segment were used to perform SOH estimation on the remaining segments.The full-time SOH estimation was realized.The SOH decay curve of the lithium-ion battery was drew,and tag data was provided for the subsequent timing prediction of the remaining life of the lithium-ion battery.(4)Four types of deep learning methods commonly used in time series prediction problems were used to build four remaining battery life prediction models.And the historical data and future data in a relative sense were divided into state of health predictions to evaluate the effectiveness of the models.Combining the two algorithms with better results,a GRU-CNN hybrid prediction model based on orthogonal parameter optimization was proposed.The optimized hybrid model was used to predict the remaining battery life.the actual battery health status and remaining life prediction results of the target vehicle in the future were obtained.This method can integrate the advantages of CNN in the early stage of prediction and GRU in the middle and late stages of prediction,and provide an effective prediction method for the RUL prediction of lithium-ion batteries based on actual vehicle operating data.