The invention relates to a vehicle-mounted fuel cell system control method and system based on deep reinforcement learning, in particular to the technical field of hydrogen fuel cells. The method comprises the following steps: constructing a fuel cell model, wherein the fuel cell model comprises a thermal management system model and an output power system model; the method comprises the following steps of: constructing a reward function by taking required power of a vehicle-mounted fuel cell system, hydrogen and oxygen pressure difference on two sides of a membrane and temperature difference in a battery at front and back moments as states, taking hydrogen and air inlet pressure and cooling water temperature as actions and using power difference, and training a Q network of each action by adopting DDQN (Double Data Quality Network); and controlling the fuel cell system according to the Q network trained by each action. The durability of the fuel cell can be improved.
本发明涉及一种基于深度强化学习的车载燃料电池系统控制方法及系统,具体涉及氢燃料电池技术领域。所述方法包括构建燃料电池模型,所述燃料电池模型包括:热管理系统模型和输出功率系统模型;以车载燃料电池系统的需求功率、膜两侧氢氧压差和前后时刻电池内部温差为状态,以氢气、空气入口压力和冷却水温度为动作,以功率之差构建奖励函数,采用DDQN对各动作的Q网络进行训练;根据各动作训练后的Q网络对所述燃料电池系统进行控制。本发明可以提高燃料电池的耐久性。
Vehicle-mounted fuel cell system control method and system based on deep reinforcement learning
基于深度强化学习的车载燃料电池系统控制方法及系统
2022-12-30
Patent
Electronic Resource
Chinese