Research On Energy Management System Optimization Of Extended Range Plug-in Hybrid Vehicle Based On Reinforcement Learning

Under the guidance of the development strategy of "carbon compliance" and "carbon neutrality",traditional fuel vehicles are accelerating their transformation to electrification.Range-extended plug-in hybrid electric vehicles have dual energy systems of oil and electricity,which is one of the important technical ways for new energy vehicles.The energy management is responsible for controlling the reasonable and efficient work of each power component,which is the key technology of range-extended plug-in vehicles and is of great significance to energy saving and carbon reduction.This thesis focuses on the energy management problem of range-extended plug-in hybrid vehicles.Firstly,the influence of power components on energy management is considered,and each power component is reasonably selected and matched.Secondly,to address the problems of long development period,poor portability and non-global optimal solution in traditional rule-based energy management methods,a reinforcement learningbased energy management method has been proposed to solve the above problems through the good self-learning optimization capability of reinforcement learning itself.Further,the energy management method of hierarchical reinforcement learning is investigated to solve the problem of long training time for a single time that occurs during reinforcement learning.Finally,due to the possible deviation in the effect of the energy management method between software simulation and practical application,a physical test of the range extender bench is carried out to verify the engineering practicality of the energy management method.The main research work and conclusions of this thesis are as follows:(1)Rational matching of each power component in the energy management system.When matching the parameters of the power components,the power performance index requirements should be met first,and in addition,the power components should work in their high efficiency range as much as possible when the range-extended plug-in vehicle is driven daily.The basis of power component matching is the longitudinal dynamics modeling and performance evaluation design of the whole vehicle,and the selection,matching and modeling of power components according to the influence of performance evaluation index and energy management.The selection,matching and modeling of range extender,drive motor and power battery pack also provide the simulation basis for the subsequent verification of the effect of energy management methods.(2)Different driving modes of range-extended plug-in vehicles are classified based on thresholds,and a rule-based energy management method is developed.Due to the problems in rule-based energy management,an energy optimization management method based on reinforcement Q-learning is proposed.Further,the hierarchical reinforcement learning method is introduced to extend the idea of reinforcement learning method to solve the energy management optimization problem because of the shortage of reinforcement learning method and the speed stratification feature of the test cycle working condition.Simulations are conducted to validate the rule-based,reinforcement Q-learning and hierarchical reinforcement learning-based energy management methods,and the simulation results show that the hierarchical reinforcement learning-based energy management optimization method has better fuel economy.(3)The range extender test bench is built to verify the engineering practicality of the energy management method.The physical verification of the range extender bench with the energy management optimization method is completed under the premise of stable and reliable bench.The results of software simulation and bench test show that the proposed energy management optimization method has certain practical engineering value.Therefore,the research on energy management based on reinforcement learning not only helps to complement and improve the intelligent energy management method of range-extended plug-in hybrid vehicles,but also has great significance for improving vehicle performance and fuel economy.