Speed-prediction-based hierarchical energy management and operating cost analysis for fuel cell hybrid logistic vehicles

This paper devises a generalized two-layer predictive energy management strategy with a comprehensive operating cost analysis for fuel cell logistic vehicles under different application scenarios. In the upper layer, an improved speed predictor based on long-and-short-term memory neural network and fuzzy C-means clustering is proposed, which can recognize driving states in real time and select corresponding sub-models for speed forecasting. In the lower layer, a multi-objective cost function including hydrogen consumption cost and power-source degradation cost is established and the optimal control action is derived within each receding horizon using sequential quadratic programming. Moreover, the performance discrepancies caused by various factors such as optimization weighting coefficients, prediction horizon length, velocity prediction methods and solution method are analyzed. Compared with benchmark strategies, the proposed strategy could reduce vehicular total operating cost by 0.76 %–32.83 % and fuel cell aging cost by 0.75 %–16.04 % across all the cycles. In addition, the operating cost distribution law with respect to different logistic vehicle types and different component sizes are analyzed via a comparative study, which could be used as a guideline for prospective designers in control strategy development.