Abstract:

The use of electric buses (EBs) has many benefits, including reduced noise pollution, a more relaxing ride due to the lack of motor vibrations, improved recovery of braking energy, increased energy efficiency compared to diesel-powered buses, and a substantial reduction in emissions as a result of fossil fuel savings and greenhouse gas emissions reductions. As prediction of trip-based electricity consumption has become a prerequisite for the deployment of large-scale EB fleets, this study has established random forest-based models to systematically investigate the impacts of environmental conditions, route characteristics, and dynamic traffic conditions. The models have been performed on real-world data collected from 1024 EBs over five consecutive months in Shenzhen, China. The results show that considering all the influencing variables can significantly enhance the prediction performance, but comparatively speaking, the route characteristics contribute the most among the three categories and involving more variables demonstrates greater advantages within the trip length under 20 km. It is also found that the trip length, the number of bus stops and the number of the traffic lights passed rank the top three most influencing factors, while the wet-dry condition is the least one. In addition, the variations under five operation scenarios show similar trend. The trip length and average travel speed are inversely proportional to the specific electricity consumption, while the number of bus stops visited, traffic lights passed, and ambient temperature exhibit a gentle proportional relationship.  The analysis on the influencing variables in this study can serve as a reference to the selection of the influencing factors and the improvement of the prediction accuracy of the electricity consumption of EBs.