A comprehensive system simulation from PEMFC stack to fuel cell vehicle

Abstract

The commercialization of fuel cell vehicles (FCVs) is a key method for achieving deep decarbonization in the transportation sector. Boosting powertrain energy conversion and utilization efficiency, especially for fuel cells, is crucial for advancing FCV technology. In the present study, a multi-level FCV system model is developed, and optimization has been carried out at various scales. The results reveal that the Gaussian process regression (GPR) model outperforms other machine learning models in performance prediction accuracy and speed. Then, based on the GPR model, different optimization algorithms are adopted to obtain the optimal operating conditions. Under the hydrogen recirculation architecture of this study, the system efficiency reaches its peak (47.4 %) at a load current of 110 A, which corresponds to the lowest point of hydrogen consumption. By coupling machine learning stack performance prediction models, the dynamic performance and fuel economy of FCVs under the New European Driving Cycle are studied. A novel fuzzy control-based energy management strategy (EMS) is proposed, which can significantly improve energy utilization efficiency while reducing the fuel cell power fluctuations. The multi-level optimization research conducted in this article, from the cell itself to the system and then to FCVs, can be widely applied to the design or control of FCVs' powertrain.