This paper proposes a hierarchical predictive energy management strategy (EMS) for hybrid electric bus (HEB) with an intelligent state of charge (SOC) reference planning method. In the cloud layer, future driving cycle is acquired through the intelligent transportation system (ITS) and well-trained neutral networks of deep deterministic policy gradient (DDPG) are extracted to plan the SOC reference trajectory quickly. In the vehicle layer, back propagation neutral network (BP-NN) is used to predict the velocity in a short term and an optimal controller is designed to distribute power flows optimally. Simulation results show that the fuel economy is improved by 2.12% compared with DDPG and reaches 97.43% of dynamic programming (DP).
Keywords Hybrid electric bus, energy management, intelligent SOC reference planning, model predictive control, deep deterministic policy gradient