Integrating Well Logs, Drilling Parameters, and Machine Learning for Subsurface Characterization of Volcanic Geothermal Rocks

Abstract Characterizing subsurface rock properties in high-temperature geothermal systems is challenging because extreme conditions often limit the availability and reliability of wireline, cuttings, and core data. This study applies a data-driven workflow to well THG13 in northeastern Iceland, combining objective-driven preprocessing of wireline and drilling data with unsupervised machine learning to derive electrofacies associated with rock composition and relative porosity. These electrofacies are then integrated with rate of penetration and weight on bit to investigate the mechanical response of volcanic units. After minimizing operational effects, variations in drilling parameters reveal porosity-related contrasts in rock strength and drillability. The results demonstrate that integrating carefully processed wireline and drilling datasets can improve geological and mechanical characterization of volcanic geothermal reservoirs under data-constrained conditions.