Energy consumption prediction for electric vehicles: methodologies and performance analysis

Details

Supervisors

Pelsser, Cristel

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil en informatique, à finalité spécialisée

Abstract

This master thesis presents a reader-friendly analysis of energy consumption prediction for electric vehicles (EVs), focusing on micro-based prediction models. The study explores various methodologies, including analytical, statistical, and machine learning approaches, to estimate electric energy consumption. The research focuses on data handling and the importance of selecting appropriate segment sizes—where a segment refers to a discrete portion of a trip, typically defined by a specific length or time interval— and features to improve the accuracy of predictions. Through examination of different segment aggregation methods, this work puts on view the challenges associated with predicting energy consumption for segments of varying lengths and the impact on overall trip energy estimation. The findings suggest that while micro-based predictions can offer valuable insights, the choice of evaluation criteria and the scale of aggregation significantly influence the results. The study also discusses the limitations of this current work, particularly in predicting speed profile variables like acceleration and deceleration, which are crucial as energy consumption depends heavily on these factors.