Prediction of Electric Vehicle Energy Consumption by Combining Real Vehicle Data and Machine Learning Methods

Zhengqian Wu¹, Xiaobing Chen²* and Yugang Jiang^3
1Hunan Mechanical Electrical Polytechnic, Hunan Changsha, China
²Hunan Province Motor Vehicla Technician College, Hunan Shaoyang, China
³Hunan Biological and Electromechanical Polytechnic, Hunan Changsha, China
*Corresponding Author: Xiaobing Chen, Hunan Province Motor Vehicla Technician College, Hunan Shaoyang, China.

Published: February 22, 2023

DOI: 10.55162/MCET.04.115

View Pdf

Abstract  

A highly nonlinear relationship exists between complex driving conditions, external influencing factors and vehicle energy consumption. Considering the spatiotemporal characteristics of vehicle operation, the significant feature parameters are extracted to improve the accuracy of vehicle energy consumption prediction. In this paper, an electric vehicle energy consumption prediction method that integrates real vehicle operation data and machine learning methods is proposed. Based on a large amount of real vehicle operation data, data cleaning and data integration methods are used to divide different kinematic segments. The key features of vehicle operation are extracted from the kinematic segment, and the correlation coefficient analysis is used to screen important feature values. Based on the XGBoost algorithm, the vehicle termination SOC prediction model is established to further obtain the results of the vehicle energy consumption. Through real vehicle operation data verification, the energy consumption prediction error within 0.04kWh, the results indicate that the proposed method gives out high accuracy.

Keywords: electric vehicle; energy consumption analysis; XGBoost algorithm; kinematics segment