AI in Geothermal Resource Exploration – A Critical Review

Geothermal exploration carries its heaviest risk exactly where knowledge is thinnest: before drilling confirms the resource. This Geologic®-led 2024 review examines artificial intelligence specifically through the lens of geothermal resource exploration, the processes that discover potential before a commercial resource is confirmed, where several earlier reviews had covered AI across geothermal applications generally. It organises a decade of machine learning research into six application domains: play fairway analysis, integrated subsurface datasets, geochemical data, geophysical data, thermal data and other data sources, and critically evaluates the algorithms, results and limitations of each.

Neural networks are the most widely used technique across the exploration groups; geophysical data analysis is the most developed application; and one reviewed deep learning classifier located geothermal sites with 92 to 95% accuracy on independent test data. Several reviewed AI geothermometers and resistivity-to-temperature models outperformed their classical counterparts in the studies that tested them. The main barriers are data access, data quality and the shortage of tailored data science skills in the geothermal workforce. The review closes with a practical agenda of data engineering, shared repositories and transfer learning to carry AI from research into exploration programs.