Lithium-ion Battery Life And Macro Mileage Prediction For Real Vehicle Conditions

Lithium-ion batteries have become the main power source of electric vehicles due to their environmental characteristic,high charging efficiency and high output power.The scientific and healthy management of the battery system has also become one of the key technical issues in the development of electric vehicles.In the data-based and intelligent era with cloud computing,research on data-driven health management of lithium batteries has become a hot issue of widespread concern and research.The relevant theoretical research based on the experimental working condition data is quite abundant,but there is still a big gap with the actual application working condition of the battery,and it is difficult to directly migrate the application.Aiming at the management and application problems faced by lithium-ion batteries under actual application conditions,this paper conducts research from the following aspects:First,starting from the working principle of lithium batteries,the aging mode and aging mechanism of lithium battery cells are deeply studied.In addition to the impact of temperature,depth of discharge,charge and discharge rate and other important parameters on aging,the actual factors that affect the aging of on-board battery packs are analyzed.It lays a theoretical foundation for the life prediction of lithium batteries and the cycle mileage prediction of electric vehicles in the subsequent chapters.Second,based on the multi-segment fast charging data,the final life of the battery is predicted.Considering the charging strategy of real vehicle working conditions,this paper extracts a variety of processing features and measurement features based on the multisegment fast charging data of the MIT-Stanford public data set.The correlation coefficients between each feature and battery life are compared.Among the features,four features are substituted into various machine learning model for training.When changing the proportion of training data and model type,the optimal model prediction error is only9.34% and the average life prediction error is 92.56 cycles,which can be used to predict battery life accurately.Then,the cycle mileage prediction of electric vehicles under the macro-time scale is realized.Based on the actual data of car A and car B,the multi-dimensional features that characterize the aging state of the battery are extracted and substituted into the support vector regression model for training,with the final prediction error of within 2%.The future features are forecasted to predict the future cycle mileage of car A and car B.This algorithm helps provide personalized and refined maintenance suggestions,which can effectively extend the service life of the battery and improve the user experience.Finally,in order to meet the research needs of battery practitioners,the battery data analysis software AILi On is developed and improved based on MATLAB APP Designer.The software covers the entire process of data analysis,including modules such as data import,basic data display,data preprocessing,feature extraction,model selection,and result visualization.Based on the original software of laboratory,the above two prediction algorithms are integrated and applied to simplify complex codes and facilitate users to use.