Research On Energy Dynamic Management Strategy For Fuel Cell Vehicle In Full Life Cycle

Severe environmental challenges and energy crises make the development of fuel cell vehicle technology rise to the level of national energy strategy.The power source of fuel cell vehicle is generally composed of fuel cells and batteries.Energy Management Strategy(EMS)is one of the key technologies of fuel cell vehicles,which plays an important role in slowing down the life attenuation of dual power sources and reducing the hydrogen consumption of vehicles.Fuel cells aging can lead to changes in system efficiency,while battery aging can lead to batteries capacity reduction and internal resistance increment.The untimely identification of the aging state of the dual power sources and adjustment of the corresponding EMS systems will accelerate dual power sources life decay and increase vehicle hydrogen consumption.This paper takes a Fuel Cell Electric Vehicle(FCEV)as the research object to carry out specific research on the long-term performance of EMS.Firstly,according to the technical requirements of FCEV power performance,the parameter matching design of vehicle powertrain was completed,which provides a basis for the modeling and simulation of long-term energy management strategy.Secondly,the influencing factors of fuel cell lifetime were analyzed and the aging model of dual power sources was established.The aging model can accurately reflect the aging state of dual power sources.Then,a condition recognition method based on time and spatial dimension memory was proposed,and an energy management strategy for optimal parameters based on working condition segment eigenvalue was established to reduce hydrogen consumption.Finally,on the basis of the above optimization parameter strategy,the long-term energy management strategy based on deep reinforcement learning was explored by considering the total cost of fuel cell life attenuation and vehicle hydrogen consumption.This strategy can varies with the decline of dual power sources,to realize the long-term adaptation with the characteristics of dual power sources,the reduction of hydrogen consumption,and mitigation of life attenuation of dual power sources.The specific research contents include:(1)The vehicle powertrain was designed according to the technical requirements of FCEV.The powertrain configuration was determined,and the parameters of driving motor,fuel cell and battery were completed;Then,a novel DC/DC converter with low input current ripple was developed.The multi-phase switching mode of DC/DC converter was adopted to improve its working efficiency,and the stability of topology structure was analyzed to ensure its stable operation.Based on the functional requirements of FCEV,the hardware design of Vehicle Control Unit(VCU)was completed,which laid the hardware foundation for the subsequent EMS development.(2)Based on the various operating conditions of FCEV,the influencing factors of fuel cell life are analyzed,and the aging model of dual power sources was developed.Since the singlephase switching control strategy of DC/DC converter will produce a low frequency narrow pulse current,whose influence on the fuel cell life was analyzed from the perspectives of gas metering ratio,relative humidity,water content and pressure difference between the two sides of the membrane,based on the results of analysis,the limiting conditions for maximizing the fuel cell life is provided.The influence of on-board operating conditions on fuel cell life was analyzed,and the influence degree was quantified;The fuel cell aging model built in this paper has the function of identifying the fuel cell aging state and fitting the fuel cell system efficiency curve.Besides,the battery aging model can simulate performance under different health states.(3)In order to reduce the hydrogen consumption of vehicle,an energy management strategy for optimal parameters based on working condition segment features was proposed.The features of working condition segment can reflect the driving cycle variation of vehicle,so a method of identifying the working condition segment features based on time and space dimension memory was proposed.According to the features of different working conditions and the aging state of dual power sources,the optimized parameter energy management strategy was developed to reduce the hydrogen consumption.Simulation results showed that the proposed strategy can effectively reduce the vehicle hydrogen consumption.(4)A long-term energy management strategy based on deep reinforcement learning was proposed considering fuel cell loss and hydrogen consumption cost.The instant reward function of deep reinforcement learning is defined by combing the loss cost of fuel cells and the energy management strategy based on the features of working segments.The Actor network and Critic network were designed simultaneously.Based on the optimization results of the previous energy management strategy,the optimization process of long-term energy management strategy combined with deep reinforcement learning was discussed.(5)The proposed long-term energy management strategy based on deep reinforcement learning was verified by bench test.Firstly,a fuel cell simulator was developed according to the output characteristics of a fuel cell,a dynamometer to simulate the load of vehicle,the power level of battery and the driving motor were reduced equivalently to realize the design of a bench test platform.Then,utilizing the combined control methods of computer and VCU,the training of deep reinforcement learning network was completed based on the actual collected vehicle operating conditions.The test results show that the proposed energy management strategy has advantages in slowing down the life attenuation of dual power sources and reducing the hydrogen consumption of vehicle.The long-term energy management strategy based on deep reinforcement learning proposed in this paper can realize strategy updating adaptively in the whole life cycle of dual power sources.Compared to conventional thermostat strategies,it reduces the total cost of fuel cell loss and hydrogen consumption by 8.31%,which promotes the industrialization process of fuel cell vehicles.Meanwhile,it is of great significance to apply the deep reinforcement learning algorithm to the control field of FCEV.