Smarter forecasting of electric vehicle consumption: Integrating physics and machine learning

Details

Supervisors

Pelsser, Cristel ; Rogister, Fabien

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] en sciences informatiques, à finalité spécialisée

Abstract

enThis thesis explores the use of data science and machine learning techniques to understand and predict the energy consumption of electric vehicles. My study begins with the analysis of two real-world test campaigns. Initial assessments reveal the limitations of a simplified physics-based model used to estimate consumption. Then, I progressively improves this model by incorporating the effects of acceleration and temperature, resulting in a significant increase in predictive accuracy. Building upon this foundation, I finally employ machine learning algorithms to predict the residual errors of the improved physical model, further enhancing overall performance. The resulting hybrid approach, which combines physical modeling with artificial intelligence, achieves a prediction error of approximately 5%. This level of accuracy is comparable to, or even better than, state-of-the-art methods, although the validation is performed on a relatively limited dataset. This work was conducted within AISIN Europe’s LBS department. Its results can certainly help plan electric vehicle routes, establish driver profiles, and anticipate overconsumption.