Dynamic Optimization of Geothermal Production and Heating Systems Based on Machine Learning

ABSTRACT: Geothermal energy is a renewable resource distinguished by its extensive reserves and broad geographic distribution. Significant variations in outdoor temperature lead to pronounced fluctuations in residential heating demand during winter, requiring that geothermal production be dynamically adjusted. Traditional optimization methods for geothermal production and heating systems often rely on human expertise, which can lead to unnecessary waste of geothermal energy, thereby reducing geothermal utilization efficiency. This study combines numerical simulation and machine learning techniques to develop a thermo-hydraulic coupling intelligent surrogate model. This model is employed to investigate heat extraction mechanisms and key factors of geothermal well systems, with a focus on the prediction of the injection-production pressure difference (IPPD) in the well system. Additionally, an index of power difference, determining the geothermal utilization efficiency, is introduced, and the NSGA-II algorithm is applied for real-time optimization of power difference (an index quantifying energy waste due to supply-demand mismatch). In the optimization process, the minimization of power difference and IPPD are defined as the primary objectives. After optimization, the power difference decreased by 98.3%, while the IPPD was reduced by 43.7%. These findings highlight the potential for further research and the application of intelligent technologies to optimize geothermal energy production systems.