As one of the most important supervisory control functionalities, the energy management strategy (EMS) of a hybrid electrified vehicle (HEV) optimizes the use of onboard energy resources for energy conservation and emission mitigation. Engine Start-up proved to have great contribution to fuel consumption and emission. A deep reinforcement learning based EMS is proposed for a power-split HEV to reduce the energy consumption and emission by recognizing start-up conditions and decreasing the start-up frequency. The EMSs based on Proximal Policy Optimization (PPO) and Twin-delayed Deep Deterministic Policy Gradient (TD3) are also compared in transient working condition frequency. Simulation study is conducted to demonstrate the advantage of the proposed energy management method. The EMS considering fuel consumption minimization and irrational actions avoidance is optimized by running the vehicle model under the WLTC condition repeatedly. PPO can get 9.02% lower fuel consumption, 25.6% lower start-up times and 8.2% transient working condition percentage than TD3. PPO is more suitable in the EMS domain.
Keywords Energy Conservation and Emission Reduction, Energy Management Strategy, Deep Reinforcement Learning, Engine Start-up Conditions, Hybrid Electric Vehicle