Model Prediction Control-Based Energy Management Combining Self-Trending Prediction and Subset-Searching Algorithm for Hydrogen Electric Multiple Unit Train
Qi Li, Puren Liu, Xiang Meng, Guorui Zhang, Yuxuan Ai, Weirong Chen
Abstract
To apply the actual fuel cell hybrid system, improve the efficiency of fuel cells, and reduce the operating cost, the model predictive control (MPC)-based energy management strategy (EMS) combining self-trending prediction and the subset-searching algorithm is proposed in this article. Compared with the traditional MPC, the proposed EMS reduces the error of speed prediction and simplifies the process of solving the optimal control trajectory, which increases the potential for practical engineering applications. Moreover, the proposed EMS considers the degradation of the power sources, which would lead to low operating costs from a long-term scale perspective. Finally, this article carries out a hardware-in-the-loop experiment to verify the feasibility and superiority of the proposed EMS. The results show that compared with the optimal benchmark, the operating cost is only increased by 7.70%, and the fuel cells operate in the high-efficiency range. Otherwise, compared with the traditional Markov chain and dynamic programming, the root-mean-square error and single time of each rolling optimization, respectively, reduce 27.79% and 64.9% at least. In addition, this article tests the adaptability of the proposed EMS on other track sections. The results show that in any track section, the proposed EMS can maintain the state of charge (SOC) and reduce the operating cost.