SMART ENERGY FORECASTING FOR ELECTRIC BUSES THROUGH MACHINE LEARNING TECHNIQUES

Abstract

Transportation systems are becoming more and more electrified; city buses in particular offer tremendous possibilities. Designing vehicles and managing a fleet requires a thorough grasp of real-world driving data. To operate alternative powertrains effectively, a number of technical factors need to be taken into account. When energy consumption is uncertain, cautious design is used, which suggests inefficiency and high prices. Because of the intricacy and interdependence of the factors, both industry and academics fail to find analytical answers to this challenge. By optimising processes, accurate energy demand forecast allows for considerable cost savings. The purpose of this study is to make the energy economy of battery electric buses (BEBs) more transparent. To describe speed profiles, we provide new sets of explanatory variables that we use in effective machine learning techniques. We create and thoroughly evaluate five distinct algorithms in terms of their general applicability, robustness, and prediction accuracy. When combined with the careful feature selection, our models demonstrated exceptional performance, achieving a prediction accuracy of over 94%. The suggested approach has the potential to revolutionise mobility and open the door for sustainable public transportation for municipalities, fleet operators, and manufacturers.