Study On Driving Mileage Prediction Of Pure Electric Vehicle Based On Machine Learning

In the face of increasingly severe environmental and energy problems,governments around the world have vigorously promoted the development of electric vehicles.The Chinese government in order to get rid of excessive dependence on oil resources,reduce environmental pollution,safeguard national energy and economic security,and realize the leaping development of China’s automobile industry,the central government will support the new energy vehicles as the future strategic emerging industries and pillar industries,the introduction of a series of relevant policies to promote the rapid development of electric vehicles.However,due to the relative lag of the current electric vehicle technology in China,the running mileage of the pure electric vehicle is short;related facilities are not complete,resulting in charging inconvenience,thus it brings "mileage anxiety" and "charging anxiety",which restricts the promotion and development of electric vehicles.So the prediction of driving mileage based on real historical data is of great significance to the development of electric vehicles.The data used in this paper are from the historical data of the operation of the pure electric passenger car produced by the New Energy Company in the working conditions.Unlike previous simulation studies,electric vehicles run in a dynamically changing system,the driving mileage is influenced by many factors.Due to the complexity of the battery itself and the variability of the traffic environment,the real working condition data is more complex,but it has great significance for building a high accuracy driving mileage prediction model.Due to the complexity of the collected data,the data need to be preprocessed before the modeling of the driving mileage prediction.It includes deletion,interpolation and homogenization,which lays a data base for subsequent modeling.In order to accurately establish the prediction model that accords with the actual working conditions,we dig all kinds of factors that influence the mileage of electric vehicles,and make detailed correlation analysis.In order to understand the coupling relationship in data,partial correlation analysis is conducted.It is concluded that SOC(State Of Charge)and minimum monomer temperature have a significant negative linear correlation with mileage.Based on the linear relationship,a multiple linear regression model of SOC and mileage is established in this paper.Then,we incorporate the potential nonlinear relationship between the variables and the undiscovered data statistics into the model in order to improve the prediction accuracy,this paper uses gradient boosting algorithm(GDBT)to enhance the electric vehicle mileage prediction model,through the process of data preparation,feature extraction and parameter adjustment finally established prediction model.The experimental results show that the model has a higher prediction precision and can meet the actual demand.