Sustainable freshwater/energy supply through geothermal-centered layout tailored with humidification-dehumidification desalination unit; Optimized by regression machine learning techniques

Abstract

The application of regression machine learning techniques is crucial for the analysis and optimization of energy systems based on geothermal energy to produce freshwater, power, and heating. This study applied regression machine learning techniques to investigate and optimize a geothermal tri-generation energy system that combines double-flash geothermal energy, humidification-dehumidification desalination, and transcritical carbon dioxide Rankine cycle. The goal was to produce freshwater, power, and heating. The algorithms performed remarkably well, with R-squared values surpassing 96 %. It is worth mentioning that for specific parameters like freshwater production, heating capacity, and efficiency, the R-squared values exceeded an impressive 99 %. The optimum conditions include maintaining a Rankine pressure ratio of 3, a mass flow rate of 30 kg/s, a geothermal temperature of 220 °C, and a turbine pressure ratio of 2.2. By following these ideal parameters, it is possible to achieve the highest level of system performance. This will lead to the production of 7.88 kg/s of freshwater, the generation of 1283 kW of power, a heating capacity of 30.8 kg/s, and an impressive system efficiency of 24.98 %. The findings reveal that the integration of regression machine learning algorithms in geothermal energy systems for freshwater, power, and heating production holds great promise for a sustainable and efficient future.