Prediction Model of Driving Range of Electric Vehicle Based on Driver’s Driving Behavior

Aiming at the problem of "mileage anxiety" caused by the inaccurate estimation of vehicle owners due to the complicated and changeable actual driving conditions, the current study of electric driving range prediction focuses on exploring the battery parameters or fitting the linear relationship between electric vehicle SOC (state of charge) and driving range under ideal conditions of experiments or software simulation, which ignores the actual vehicle driving conditions and the difference in driving range due to differences in driver's driving behaviors. This paper puts forward an innovative prediction model of the driving mileage of pure electric vehicles based on driver's driving behavior. Firstly, taking the driver's real vehicle operation data as the object, the model extracts 2,412 kinematics driving fragments according to the short trip segmentation method, then carries on the PCA(Principal Component Analysis) to the 18 characteristic parameters of the kinematics driving fragments which reflects the driver's driving behaviors, and performs the FCM (Fuzzy C Means) clustering analysis to obtain the clustering center. Secondly, based on the comparison between the drivers’ historical driving fragments and the clustering center, a prediction model for the driving range of pure electric vehicles is established. Finally, the model is verified by the actual vehicle driving data which comes from the actual vehicle driving data for a month. The maximum relative error of the model is 4.5%, which indicates that the model is accurate and feasible.