Accurate Prediction of Energy Consumption of Electric Buses Based on Traffic Condition, Vehicle Status, Driving Behavior, and Environmental Condition

Accurate prediction of trip-level energy consumption of electric buses (EBs) is of great significance to alleviate the driver’s mileage anxiety and maintain the health of the battery system. In order to achieve accurate prediction of energy consumption on EBs, actual operation data of 95 204 trips from winter to summer covering long, medium, and short-distance bus routes were collected, and a total of 17 energy consumption-related features were extracted from four aspects, including traffic condition, vehicle status, driving behavior, and environmental condition. A total of 11 features were selected using the variance inflation factor (VIF) method and stepwise regression algorithm (SRA), then the relationships between the selected features and the relationships between features and the energy consumption were analyzed in depth. By combining the prediction of critical features, six machine-learning models were built and compared after hyperparameter optimization. LightGBM, which showed the best performance in cross validation was selected as the final machine-learning model for energy consumption prediction. The proposed method can achieve a mean absolute percentage error (MAPE) of 6.65% on real-world testing data, with an improvement of 23% over the existing methods.