Machine Learning (ML) for subsurface geothermal resource analysis and development

This dissertation presents a comprehensive examination of the integration of Machine Learning (ML) methodologies to enhance the analysis and development of subsurface geothermal resources, a pivotal element within the renewable energy spectrum. Despite geothermal energy's substantial potential for sustainable energy production, its exploitation is impeded by numerous geological and technical obstacles. This research endeavors to overcome these impediments by harnessing sophisticated ML algorithms, aiming to augment the predictability and operational efficiency in the exploration and characterization of geothermal resources. Utilizing an extensive dataset from the Northeastern United States, the study conducts a thorough analysis and refinement of prevailing approaches in data preprocessing, feature engineering, and hyperparameter optimization. It assesses the efficacy of various ML models in forecasting subsurface temperatures and geothermal gradients, underscoring the significance of meticulous data examination and model refinement strategies, including outlier detection, data normalization, and the employment of grid search techniques for hyperparameter fine-tuning. The outcomes reveal that ML applications can markedly improve the precision and dependability of predictions concerning geothermal resources, thereby diminishing the financial and technical uncertainties inherent in geothermal project development. The enhanced predictive models formulated through this research facilitate more strategic decision-making and resource allocation within the geothermal energy domain. This dissertation highlights the transformative potential of amalgamating ML technologies with geothermal energy exploration and development practices, paving new avenues for minimizing the environmental impact of energy generation. It proposes future research avenues, such as the integration of diverse data inputs and the implementation of real-time ML monitoring systems, to further propel advancements in the field. By melding conventional geothermal resource assessment methodologies with the latest in ML innovations, this work establishes a foundational framework for significant advancements in the sustainability and efficacy of geothermal energy, reinforcing its role as an indispensable component of the global renewable energy portfolio.