Researches On Customer Segmentation And Range Estimation Of Battery Electric Vehicle Based On Data Driven

Battery Electric Vehicle uses electric batteries as the only source of power,it has the characteristics of low pollution,low noise,and high energy conversion efficiency,which can help to alleviate the increasingly severe energy and environmental problems.With the strong advocacy and policy support of governments around the world,the market of battery electric vehicle has great potential for development.However,with the strong growth of battery electric vehicle,the consumption and needs of its customers groups gradually show diversified and hierarchical characteristics,and customers generally have the emotions of "user anxiety".In order to meet the needs of BEV market development and enhance driver confidence,this thesis relies on the real driver data of BEVs to perform cluster analysis on driver groups to portray the different driver groups and provide theoretical basis for the market positioning and precision marketing of enterprise.Meanwhile,a study on range prediction based on the actual operating state data of BEVs is carried out to improve driver satisfaction.In summary,this research topic has important theoretical significance and practical value.The thesis takes the driver data of BEVs as the research object,and uses machine learning algorithm to carry out related research work around customer group segmentation and range prediction.The main research contents are as follows:(1)A study on customer segmentation based on FP-Kmeans algorithm.First combed the study theories and the key technologies of cluster analysis to lay the theoretical foundation and technical support for the follow-up study content;Secondly,based on the customer data of a certain BEV on the market,data cleaning technology and feature correlation analysis technology are used to obtain a data set that includes14 optimal feature variables;Thirdly,introduced the main principles of the traditional K-means algorithm and its advantages and disadvantages,and proposed a FP-Kmeans algorithm based on manifold distance for the disadvantages of the K-means algorithm to cluster drivers of BEVs;Then,conduct a detailed comparative analysis of consumption ability and usage behavior of the five types of clustered driver groups,and consider the actual needs of the company’s business to give advice on choosing customers;Finally,the focus of the drivers is analyzed.The study found that driven distance is the most concerned performance index for derivers,accounting for 46.3%,which is one of the important reasons that hinder customers from purchasing BEVs.Accurate range prediction can help to alleviate the user’s "mileage anxiety" lays the foundation for the next part of the research.(2)A study on range prediction of BEVs based on Light GBM algorithm.First,summarize the relevant theories and analysis techniques of regression prediction analysis;Secondly,in order to improve the reliability and accuracy of subsequent model predictions,after preprocessing the operating state data of BEVs,16 feature variables are obtained as model inputs after segmentation of driving segments and feature expansion operations;Then,establish a Light GBM prediction model and compare it with other mainstream regression prediction models to verify the effectiveness and superiority of the Light GBM algorithm;Finally,use Grid Search CV technology to optimize the Light GBM model,and use the SHAP to explain and analyze the model globally and locally.