Given the longitudinal average velocity and energy consumption of a Full Electric Vehicle (FEV) for any given part of a targeted road trip, this work solves the problem of online remaining range estimation, i.e., predicting, at any given travelled distance from the beginning of the trip, the actual distance the vehicle can still be driven before recharging is required. Modelling the remaining range is closely related with modelling the energy consumption of an electric vehicle. The latter remains an open problem due to unknown context data that may apply such as driving trip speed, vehicle load and topographical characteristics. In this work, a regression model is formulated in order to learn, from time/location-variant real driving data, a relationship between the future energy consumption on one side, and the following related factors, which are considered known, on the other side: i) the difference in average velocity between the future and the past ii) the difference in elevation rate between the future and the past and iii) the recent past energy consumption. Experimental results onaround 2000km of discharge trips, demonstrate the effectiveness of the method over a conventional method that is based solely on historic energy usage evidence. An average Mean Absolute Error (MAE) of 1.64 km and of 1.95 km is obtained when the regression model is evaluated on a model trained without and with elevation respectively.
Online prediction of an electric vehicle remaining range based on regression analysis
01.12.2014
1159578 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Enhancing Electric Vehicle Remaining Range Prediction Through Machine Learning
Springer Verlag | 2024
|Remaining driving range estimation of electric vehicle
IEEE | 2012
|Electric vehicle remaining driving range prediction method and device considering road environment
Europäisches Patentamt | 2024
|Digital Twin-Based Remaining Driving Range Prediction for Connected Electric Vehicles
SAE Technical Papers | 2023
|