Loading...
Surface and subsurface explainable artificial intelligence models for geothermal exploration
Demir, Ebubekir
Demir, Ebubekir
Citations
Altmetric:
Advisor
Editor
Date
Date Issued
2024
Date Submitted
Collections
Research Projects
Organizational Units
Journal Issue
Embargo Expires
2026-04-04
Abstract
Current artificial intelligence (AI) applications in geothermal exploration are tailored to specific geothermal sites, limiting their transferability and broader applicability. This study aims to investigate the viability of a globally applicable and transferable geothermal AI model to empower the exploration of geothermal resources. We modified an existing two-dimensional (2D) Geothermal AI model that utilizes known indicators of geothermal areas, including mineral markers, land surface temperature (LST), and faults. We tested its performance for three distinct geothermal sites—Brady, Desert Peak, and Coso. Our methodology includes self-testing to understand the unique characteristics of each site, followed by dependent and independent tests to assess cross-compatibility and model transferability.
Additionally, this study introduces a three-dimensional (3D) Geothermal AI utilizing a temperature block model, fault density block model, and lithology block model as inputs to predict subsurface geothermal resources. We have applied 3D Geothermal AI at the Coso site. The results show that 3D Geothermal AI can predict geothermal resources with an overall accuracy of 84%.
Moreover, in this study, an approach is developed to identify the transferability of 2D Geothermal AI models by looking at the impact of inputs such as mineral markers, surface temperature, and faults. The suggested methodology uses explainable AI (XAI) methods, i.e., DeConvNet, deep Taylor decomposition, gradient, guided backpropagation, input*gradient, and LRP-z, to compare three geothermal sites with distinct characteristics. Initially, we conducted self-testing to comprehend the specific characteristics of each geothermal site. Subsequently, we conducted cross-testing to reveal the transferability of 2D Geothermal AI models. This case study showed that Coso and Desert Peak sites are transferable. However, Brady is not transferable to Coso and/or Desert Peak sites. Finally, we have evaluated the XAI methods used in this study and found that the deep Taylor decomposition method is the best-fitting method for 2D Geothermal AI transferability analysis.
Lastly, we used XAI methods to reveal the impact of input features of 3D Geothermal AI. Temperature, fault density, and mc granite were the most influential features in predicting geothermal resources.
Associated Publications
Rights
Copyright of the original work is retained by the author.