Artificial Intelligence-based Multi-objective Optimization of a Solar-driven System for Hydrogen Production with Integrated Oxygen and Power Co-generation Across Different Climates

This study develops and optimizes a solar-powered system for hydrogen generation with oxygen and power coproducts, addressing the need for efficient, scalable carbon-free energy solutions. The system combines a linear parabolic collector, a Steam Rankine cycle, and a Proton Exchange Membrane Electrolyzer (PEME) to produce electricity for electrolysis. Thermodynamic modeling was accomplished in Engineering Equation Solver, while a hybrid Artificial Intelligence (AI) framework combining Artificial Neural Networks and Genetic Algorithms in Statistica, coupled with Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) decision support, optimized technical and economic performance. Optimization considered seven key decision variables, covering collector design, thermodynamic inputs, and component efficiencies. The optimization achieved energy and exergy efficiencies of 30.83 % and 26.32 %, costing 47.02 USD/h and avoiding CO2 emissions equivalent to 190 USD/ton. Economic and exergy analyses showed the solar and hydrogen units had the highest costs (38.17 USD/h and 9.61 USD/h), with 4503 kWh of exergy destruction to generate 575 kWh of electricity. A case study across six cities suggested that Perth, Bunbury, and Adelaide, with higher solar irradiance, delivered the highest annual power and hydrogen outputs, consistent with irradiance–electrolyzer correlation. Unlike conventional single-site studies, this work delivers a climate-responsive, multi-city analysis integrating solar thermal and PEME within an AI-driven framework. By linking techno-economic performance with quantified environmental value and co-production synergies of hydrogen, oxygen, and electricity, the study highlights a novel pathway for scalable clean hydrogen, measurable CO2 reductions, and global decarbonization, with future work focused on digital twins and dynamic, uncertainty-aware optimization.