A robust physics-constrained neural operator framework for efficient geothermal resource development.

Area of Science:

• Geothermal Energy Engineering
• Computational Science
• Artificial Intelligence in Energy

Background:

• Efficient evaluation of geothermal systems is crucial for scalable development.
• Conventional methods face challenges due to high computational costs and limited surrogate model generalizability.

Purpose of the Study:

• To present a physics-constrained neural operator framework for rapid, high-resolution, and physically consistent geothermal system evaluation.
• To enable accurate prediction of subsurface dynamics and surface energy production across diverse conditions.

Main Methods:

• Developed a physics-constrained neural operator framework learning the solution operator of governing partial differential equations.
• Integrated modules for power output estimation and techno-economic assessment.

Main Results:

• Achieved an average relative error of 1.76% for reservoir predictions and 1.70% for target variables.
• Demonstrated approximately 1,400-fold acceleration compared to conventional numerical methods.
• Enabled consistent techno-economic assessment across multiple geothermal applications.

Conclusions:

• The framework offers a scalable pathway for geothermal development by supporting rapid analyses.
• Facilitates resource assessment, uncertainty quantification, and multi-objective optimization for geothermal energy.