Model predictive control of an on-site green hydrogen production and refuelling station

Abstract

The expected increase of hydrogen fuel cell vehicles has motivated the emergence of a significant number of studies on Hydrogen Refuelling Stations (HRS). Some of the main HRS topics are sizing, location, design optimization, and optimal operation. On-site green HRS, where hydrogen is produced locally from green renewable energy sources, have received special attention due to their contribution to decarbonization. This kind of HRS are complex systems whose hydraulic and electric linked topologies include renewable energy sources, electrolyzers, buffer hydrogen tanks, compressors and batteries, among other components. This paper develops a linear model of a real on-site green HRS that is set to be built in Zaragoza, Spain. This plant can produce hydrogen either from solar energy or from the utility grid and is designed for three different types of services: light-duty and heavy-duty fuel cell vehicles and gas containers. In the literature, there is a lack of online control solutions developed for HRS, even more in the form of optimal online control. Hence, for the HRS operation, a Model Predictive Controller (MPC) is designed to solve a weighted multi-objective online optimization problem taking into account the plant dynamics and constraints as well as the disturbances prediction. Performance is analysed throughout 210 individual month-long simulations and the effect of the multi-objective weighting, prediction horizon, and hydrogen selling price is discussed. With the simulation results, this work shows the suitability of MPC for HRS control and its significant economic advantage compared to the rule-based control solution. In all simulations, the MPC operation fulfils all required services. Moreover, results show that a seven-day prediction horizon can improve profits by 57% relative to a one-day prediction horizon; that the battery is under-sized; or that the MPC operation strategy is more resolutive for low hydrogen selling prices.