Recognition model for eco-driving behavior of electric-buses entering and leaving stops

The speed fluctuation is a typical operating condition of buses, which increases the energy consumption during the process of entering and leaving stops (ELSs). This study analyzes the driving behavior characteristics and identifies eco-driving behavior of inbound and outbound conditions. It first collects natural driving data of electric buses (E-Buses) on BRT lines, and analyzes the driving behavior characteristics of inbound and outbound conditions. A discriminative model based on prior rules is then developed to label the entering and leaving stops driving behavior (ELS-DB) as eco-driving and non-eco-driving. Finally, according to the calibration results, a supervised learning clustering analysis model and a recognition model for eco-driving behavior of ELSs are developed based on machine learning. Afterwards, many algorithms are compared, and evaluated. In addition, the characteristics of eco-driving and non-eco-driving behaviors are analyzed and compared from both macro and micro perspectives based on the recognition results. The obtained results show that the CatBoost model has the highest recognition performance for driving behavior, reaching recognition accuracies of 92.8 % and 96.5 % during the process of entering and leaving stops, respectively.