Energy Management Strategy Of Parallel Hybrid Electric Vehicle Based On Deep Reinforcement Learning

With the increasingly serious problems of environmental pollution and energy shortage,Hybrid electric vehicle(HEV),which combines the advantages of traditional fuel vehicle and pure electric vehicle,has become the most promising vehicle at present.HEV refers to a vehicle whose power system is composed of two or more power sources.In this paper,parallel hybrid electric vehicle(PHEV)was selected as the research object.Fuzzy logic control theory and reinforcement learning technology are used to research the energy control strategy.The main work is as follows:(1)The modeling method of HEV is introduced in this paper,and utilize ADVISOR simulation software,the mathematical models of the whole vehicle dynamics,engine,motor,battery,wheel and other modules of PHEV are established.(2)According to the low efficiency of motor and high fuel consumption of electric assist control strategy,the working mode of HEV is introduced,the fuzzy logic control theory is used to design a hierarchy fuzzy logic control strategy which considering the vehicle speed factor,the fuzzy controller taking the difference between the required torque of the whole vehicle and the optimal torque of the engine at the current speed,the SOC of the battery and vehicle peed as the inputs,the engine output torque factor K as the output.Two kinds of energy control strategy are simulated in the software,simulation results show that the fuzzy logic control strategy is superior to electric assist control strategy in the fuel economy and emissions.(3)Q learning strategy based on reinforcement learning was introduced so as to address the lack of adaptive ability of rule-based control strategy.The control strategy could output action sequence from system state absence of prediction information and default rules.In DQN algorithm,demand torque and SOC were regarded as state space.As a consequence,engine motor was selected as control action.The simulation results show that effect of control strategy based on DQN effectively reduced fuel consumption and emissions.(4)Due to the frequency start-stop of hybrid electric vehicle in actual operation,the barking energy management strategy was introduced to attain the aim of energy conservation and emission reduction by improve the efficiency of energy recycle.Since the control object of the EMS of hybrid electric vehicle is a continuous system,the deep deterministic policy gradient(DDPG)was applied in control of HEV.The barking energy management controller based on oriented sampling DDPG is introduced to address stochastic problem of experience pool.The simulation result shows that the effective of control strategy and control strategy could improve the efficiency of energy recycle.