Graph neural predictive control of a dam–reservoir floating photovoltaic–hydrogen system: modeling and Raspberry Pi validation

Abstract

Hydrogen production from floating photovoltaic (FPV) systems is attracting increasing attention as a promising option for decentralized low-carbon energy generation, particularly in regions with high solar potential and limited land availability. In this study, a predictive coordination strategy is investigated for a solar-hydrogen system associated with a reservoir and conceptually situated at the Martil Dam in Tetouan, Morocco. The proposed control framework is based on Graph Neural Predictive Control (GNPC), which captures the interactions among the main components of the system within a finite-horizon optimization framework. This approach enables coordinated regulation of photovoltaic power extraction, DC-bus voltage, and electrolyzer operation, and its performance is compared with that of two classical maximum power point tracking (MPPT) methods, namely Incremental Conductance (INC) and fuzzy logic control (FLC). The simulation results indicate a tracking efficiency of 98.9% and a settling time of approximately 0.075 s, allowing the system to deliver nearly 79.1 kW from an 80 kW photovoltaic array. For the evaluated temperature range, the cumulative power reduction remains limited to 2.99%, indicating that the controller maintains stable power-tracking performance under the tested thermal conditions. Using irradiance data from Tetouan, the controller exhibits smoother transient behavior, improved DC-bus voltage regulation, and stable hydrogen production, reaching about 5.9 × 10 −4 mol s −1 nder nominal conditions. To assess its practical feasibility, the control algorithm was implemented in real time on a Raspberry Pi-based embedded platform connected to a laboratory PV-DC/DC experimental setup. The experimental results show rapid convergence and very low steady-state oscillations during operation. These findings indicate that predictive coordination can improve the stability and control quality of hydrogen production systems powered by floating solar energy.