Energy Consumption Prediction of Electric City Buses Using Multiple Linear Regression

The widespread electrification of public transportation is increasing and is a powerful way to reduce greenhouse gas (GHG) emissions. Using real-world driving data is crucial for vehicle design and efficient fleet operation. Although electric powertrains are significantly superior to conventional combustion engines in many aspects, such as efficiency, dynamics, noise or pollution and maintenance, there are several factors that still hinder the widespread penetration of e-mobility. One of the most critical points is the high costs-especially of battery electric buses (BEB) due to expensive energy storage systems. Uncertainty about energy demand in the target scenario leads to conservative design, inefficient operation and high costs. This paper is based on a real case study in the city of Seville and presents a methodology to support the transformation of public transportation systems. We investigate large real-world fleet measurement data and introduce and analyze a second-stage feature space to finally predict the vehicles' energy demand using statistical algorithms. Achieving a prediction accuracy of more than 85%, this simple approach is a proper tool for manufacturers and fleet operators to provide tailored mobility solutions and thus affordable and sustainable public transportation.