ANN-Based Solar Irradiance Prediction for Photovoltaic-Driven Green Hydrogen Production

The reliable assessment of solar resources represents a fundamental step in the development of photovoltaic-based green hydrogen production systems, since variations in solar irradiance strongly influence photovoltaic energy conversion and, consequently, hydrogen production performance. This study evaluates the feasibility of solar-powered hydrogen generation in Oujda, located in eastern Morocco, where favorable solar conditions are available throughout the year. Experimental measurements of Global Horizontal Irradiance (GHI), collected using a thermopile pyranometer, were considered as reference data for the development and validation of an Artificial Neural Network (ANN) model for solar irradiance predicting. The generated irradiance predictions were then integrated into a photovoltaic system model to estimate electricity production and coupled with a Proton Exchange Membrane (PEM) electrolyzer model to determine the corresponding hydrogen yield. The developed ANN model demonstrated high predictive capability, with strong agreement between predicted and measured irradiance values (R 2 = 0.9875). Moreover, techno-economic evaluation revealed a Levelized Cost of Hydrogen (LCOH 2 ) of about 4.0069 $/kg, indicating that ANN-based solar resource prediction constitutes a reliable approach for evaluating photovoltaic-powered hydrogen production potential in high-solar-resource regions.