AI-assisted multi-objective optimization and performance analysis of a solar-driven hybrid system for power generation, desalination and fuel synthesis

Abstract

This study evaluates the performance of a multipurpose, self-sufficient system comprising a heliostat solar farm, a central thermal receiver, supercritical CO₂ Brayton cycles, an organic Rankine cycle (ORC), a reverse osmosis (RO) unit, a PEM electrolyzer, a CO₂ absorption unit, and a methanol synthesis unit. The aim was to achieve sustainable production through the integration of bioproducts while maximizing the technical and economic potential of the design. The system was assessed to be attractive for investment in terms of both performance and cost-effectiveness. Simulation results indicate that the proposed configuration can generate 4.69 MW of power from the supercritical CO₂ Brayton cycle and 383 kW from the ORC, while producing 0.2 m³/h of freshwater, 88 kg/h of CO₂, 14 kg/h of hydrogen, and 63 kg/h of methanol, with an overall exergy efficiency of 52.6 %. In the final stage, an artificial neural network (ANN) was developed using machine learning techniques, and the system was optimized via the Gray Wolf algorithm considering five decision variables across two-, three-, four-, five-, and six-objective modes. The six-objective scenario achieved the best performance, yielding an exergy efficiency of 54.18 %, a total cost rate of 6376 $/h, hydrogen and methanol production rates of 16.87 kg/h and 65.3 kg/h, respectively, and LCOE and LEIOE values of 0.0744 $/kWh and 20.45 Pts/MWh.