Machine Learning for Geothermal Energy Systems: Prediction, Optimization, and Physics-Informed Hybrid Methods

Geothermal energy is a clean, stable, and low-carbon renewable resource with increasing strategic importance in sustainable energy transitions. In recent years, machine learning has been progressively applied in geothermal systems, including resource exploration, reservoir characterization, drilling and operational optimization, heat-transfer performance prediction, and ground source heat pump (GSHP) control. Machine learning enables high-dimensional data mining, complex mapping, and surrogate modeling. While it can often deliver higher efficiency and accuracy compared to conventional methods, these advantages are not absolute; rather, they are highly contingent upon the quality and scale of the dataset, the rigorousness of the validation method, and the appropriateness of the selected algorithm. This review systematically summarizes major machine learning applications in geothermal energy, highlighting representative methods, data sources, target tasks, and research features. It identifies critical limitations in current studies, including dataset scale and quality, model generalization, physical consistency, interpretability, and deployment feasibility. The review further emphasizes the evolution of geothermal machine learning from pointwise prediction tools to frameworks supporting intelligent analysis, optimization, and decision-making. Future directions include multi-source heterogeneous data fusion, embedding of physical mechanisms, interpretable modeling, and integration with digital twin systems. These approaches aim to shift machine learning from high-accuracy prediction toward trustworthy and deployable decision support for geothermal energy systems.