Hierarchical predictive energy management of hybrid electric buses based on driver information

Abstract

To improve the energy efficiency of hybrid electric city buses, a hierarchical predictive energy management strategy (HP-EMS) based on driver behavior and type is proposed in this paper. Within the model predictive control (MPC) framework, the k-Nearest Neighbor (kNN) method is applied to identify the driver type, and the deep neural network (DNN) is adopted to predict future speed based on the historical speed, driver type, and driver behavior. Combined with the city bus driving characteristics, the hierarchical strategy aims to reduce the frequent starts of the engine. The upper-level controller implements a rule-based strategy to limit the engine start-stop frequency. The lower-level controller uses dynamic programming (DP) to search for the best control strategy in the prediction horizon. Simulation results show that, compared with speed prediction without driver information, the new method can effectively improve the accuracy of future speed prediction, and RMSE between the prediction and measurement drops from 1.58 m/s to 1.45 m/s. The HP-EMS without driver information can reduce the number of engine starts by 30% while increase only 2% energy consumption compared with predictive energy management without hierarchical control. The paper also studies the benefits of considering driver behavior and type. The same HP-EMS controller is implemented with and without driver behavior and type. The one with the additional information reduces the energy consumption by 3.34% compared to the one without the information.