燃料电池汽车最优氢耗马尔科夫决策控制

本文基于马尔科夫决策过程提出一种燃料电池汽车最优等效氢燃料消耗控制策略.控制策略以部分观测量为基础,以马尔科夫转移概率矩阵为条件,采用基于蒙特卡洛马尔科夫(MCMC)算法的Metropolis-Hastings采样方法,获得平均奖励输出,进而通过最优氢燃料消耗代价函数的优化以控制在氢燃料电池系统和动力电池系统间进行能量分配.该策略避免了目前燃料电池汽车控制策略过度依赖未来需求功率的预测以及预测模型的准确性.在建立燃料电池汽车动力模型,燃料电池系统和动力电池系统模型的基础上,进行了包含自学习系统、基于MH采样的平均奖励过滤系统以及控制选择输出系统的控制策略设计.通过仿真和实验结果表明基于马尔科夫决策控制策略的有效性.

Best equivalent hydrogen consumption control for fuel cell vehicle based on Markov decision process-based

In order to realize the best hydrogen consumption for a fuel-cell vehicle, an Markov decision process based energy management strategy is proposed in this paper. The proposed EMS takes the part observation variables as inputs, according to the Markov transition probability matrix, gets the control sequences of the average reward based on the Metropolis-Hastings (MH) method of MCMC, and further by optimizing the hydrogen consumption cost function to distribute the power requirement between the fuel cell system and the power battery. The proposed EMS avoids the strong dependence of the future power requirement on the inaccuracy of the predictive model. On the basis of establishing the fuel cell vehicle model, the fuel cell system model and the battery model, the EMS which includes modules of self training, average reward filter and selector of control action output is designed based on the Markov decision process. Finally the energy management strategy is verified through simulation and experiment by comparing with three other different control strategies, the results show the effectiveness of the MDP-based energy management strategy.