| Title | Characterizing In-Situ Stresses at Utah FORGE Using a Multi-Component Approach Combining Laboratory Experiments, Field Measurements, Physicsâ€Based Modeling, and Machine Learning Methods |
|---|---|
| Authors | Guanyi LU, Ayyaz MUSTAFA, Andrew BUNGER |
| Year | 2025 |
| Conference | Stanford Geothermal Workshop |
| Keywords | in-situ stress estimation, enhanced geothermal systems, Utah FORGE, velocity-to-stress correlation, stress-induced wave speed anisotropy, core-based laboratory experiments, field measurements, physics-based modeling, machine learning, machine learning |
| Abstract | This study presents a novel multi-component approach that combines core-based laboratory triaxial ultrasonic velocity (TUV) experiments, field measurements, machine learning (ML) models, and physics-based finite element simulations to characterize in-situ principal stresses near a geothermal well. Laboratory TUV experiments were conducted on core samples from well 16A(78)-32 at Utah FORGE under true-triaxial stress conditions to establish velocity-to-stress relationships. These relationships were then used to develop ML models that predict near-field stresses based on sonic logging data. While the ML predictions successfully estimate the three principal stresses, near-wellbore stress concentrations and thermo-poro-elastic disturbances induced by drilling and pre-cooling significantly influence the results. To address this, we coupled the ML derived stresses with a physics-based thermo-poro-mechanical finite element model to translate near-field stresses into far-field stresses. Simulations account for realistic pre-cooling and warmup scenarios, material properties, and boundary conditions to quantify stress evolution near geothermal wells. Results demonstrate the critical impact of thermal effect on stress distributions, with notable stress variations observed near the borehole that diminish radially. Moreover, thermo-poro-mechanical effect amplifies the difference between the two horizontal principal stresses in the near-field. Therefore, the undisturbed far-field stresses are generally more isotropic than near-field predictions. The integration of laboratory and field measurements, ML predictions, and physics-based modeling provides a robust framework for accurately characterizing far-field stresses in geothermal reservoirs. The findings have significant implications for the development and optimization of enhanced geothermal systems and other subsurface energy applications. |