Research On Intelligent Energy Management Strategy For Plug-in Hybrid Electric Bus

Since the combination of the characteristics of pure electric vehicles and traditional hybrid electric vehicles,plug-in hybrid electric vehicle is an important approach to achieve the goal of "energy saving and emission reduction".As one of the key technologies,energy management has important research significance.In this paper,Plug-in Hybrid Electric Bus(PHEB)is taken as the research object,an intelligent energy management strategy(RL-PMP)based on the blend of Reinforcement Learning(RL)and Pontryagin’s Minimum Principle(PMP)is proposed to solve the problem of adaptive control in unknown driving cycles.The specific work can be expressed as:First,based on the index of dynamic performance and the driving demand,the engine,motor,the parameter matching design and component selection of engine,motor,power battery and transmission system are implemented.Then,a mathematical model is established to satisfy the requirements of energy management control.Second,PHEB energy management strategy based on PMP is studied.Since the strong coupling existed in energy management and transmission system,the control method considering throttle and shift schedule is proposed,and the control vector is designed as one-dimensional compact format(regarding throttle and shift schedule as one-dimensional integrated control vector).Moreover,to find the best sampling point,the control vector is sampled in different sampling points based on the optimal Latin hypercube algorithm.Third,an energy management strategy based on RL-PMP algorithm is proposed.The difference between the reference SOC(battery state of charge)and the feedback SOC is innovatively designed as the state,and the state space is designed as a 50-column one-dimensional matrix.In addition,to reduce the sensitivity of the optimal action to the state,the optimal control of the vehicle is solved by PMP,and the co-state is designed as the action in the RL algorithm.The research results demonstrate that the proposed RL-PMP algorithm can realize off-line self-learning,the trained Q table can be directly applied to unknown driving cycles,and adaptive control can be realized.Compared with the CDCS algorithm,the fuel economic can be improved by 22.85% on average.Fourth,based on D2 P rapid control prototype system,the hardware-in-loop platform of RL-PMP algorithm is built.Firstly,the simulation model of PHEB is established,and the RL-PMP algorithm is compiled based on D2 P to generate the product-level code.Then,the code is downloaded to the Hybrid Control Unit(HCU)to realize real-time communication between the PHEB model and HCU by the CAN bus.Finally,RL-PMP algorithm is verified in real time and robustness.The hardware-in-loop simulation shows that the RL-PMP algorithm can realize real-time control and has great application potential.