Data Driven Energy Economy Prediction for Electric City Buses Using Machine Learning

The electrification of transport systems is gaining momentum and city buses in particular offer huge potential. An in-depth understanding of real-world driving data is essential to vehicle design and fleet operation. The efficient operation of alternative powertrains requires attention to various technological aspects. Uncertainty of energy demand leads to conservative design, which means inefficiency and high costs. The complexity and interrelation of parameters make it impossible for industry and academia to find analytical solutions to solve this problem. Accurate energy demand forecasting leads to great cost reduction by optimised operations. The aim of this paper is to increase the transparency of energy economy of battery electric buses (BEB). We propose new sets of explanatory variables to describe speed profiles, which are used in powerful machine learning methods. We develop and comprehensively evaluate 5 different algorithms in terms of prediction accuracy, robustness and overall applicability. In combination with sophisticated feature selection, our models performed excellent and achieved a prediction accuracy of more than 94%. The approach outlined offers tremendous opportunity for manufacturers, fleet operators and communities to reimagine mobility and thereby lay the groundwork for sustainable, public transport