Comparative Analysis of Energy Consumption Prediction of Pure Electric Bus Based on Multiple Influencing Factors

The reduction of operational energy consumption and enhancement of economic efficiency have become focal points for public transport enterprises. This study is based on the GPS operation data and total daily energy consumption data of 11 bus lines in the same operating area of Nanjing from April to June 2021. It divides the data according to the operational cycles and explores the influencing factors of energy consumption from the aspects of line characteristics, vehicle types, and driving conditions. Lasso regression and Random Forest (RF) model are used to construct the daily energy consumption prediction model of pure electric buses. Model comparison results reveal that RF model’s prediction performance outperforms Lasso regression model, with a relative decrease of 4.42% in MAPE. Further, the SHAP algorithm is employed to interpret the RF model, revealing that vehicle manufacturer, daily operational cycle number, vehicle age, and acceleration have a significant impact on energy consumption.