Using machine learning approaches to model and optimize a combined solar/natural gas-based power and freshwater cogeneration system

Abstract

Problems of sustainable development and environmental protection pose a challenge to humanity unprecedented in scope and complexity. Whether and how the problems are resolved have significant implications for human and ecological well-being. Accordingly, this work proposes an electricity and freshwater cogeneration system using solar energy and natural gas dual sources. The proposed system is a combination of a heliostat field with a gas turbine cycle as the top systems and thermal vapor compression-multi effect desalination, steam Rankine cycle, organic Rankine cycle, and thermoelectric generator as subsystems. For performance analysis, energy, exergy, exergoeconomic, economic, and environmental analyses were performed. To optimize the system, a coupled model of the support vector regression, multi-objective grey wolf optimization algorithm, multi-objective Grasshopper optimization algorithm, and two different decision-making methods were suggested. According to obtained results, the best swirling number and pressure ratio were 0.95 and 7, respectively. Moreover, the exergy efficiency, total product cost rate, and CO2 emission were chosen as the best optimization scenario, which led to 45.6 % of exergy efficiency, 2.716 $/GJ of total product cost rate, and 30.26 kg/s of freshwater. Moreover, the total exergy destruction rate decreased from 15153 kW to 14820 kW after the optimization of the system.