Robustness Evaluation of Energy Management Strategies for Hydrogen-Based Railway Vehicles

This paper presents a robustness analysis of different energy management strategies designed for hydrogen-based rail vehicles. From the analyzed 4 strategies, one is based on dynamic programming, two are rule-based strategies optimized by genetic algorithms, and the last one is a learning-based approach. The strategies are optimized for (or learn from) a nominal scenario. Then, the number of passengers and the consumption of the auxiliaries are varied to generate new scenarios. The results obtained in the new scenarios show that the trends obtained in the nominal scenario are generally respected. Additionally, some other conclusions are also obtained: 1) GA-based strategies get closer to the global optimization of DP as the average demand of the route is higher, and 2) ANFIS should improve the degradation that the FC suffers, as it is increased overmuch compared to the other strategies.