Machine Learning-Informed Analysis of Remote Sensing Data to Identify Favorable Areas for Enhanced Geothermal Systems
Mines PI: Sebnem Duzgun (Mining Engineering) and NREL PI: Nicole Taverna (Data Scientist)
Virtual via Zoom
Wednesday, April 17, 2024 at 2PM MTN
Enhanced geothermal systems (EGS) are growing closer to wide-scale deployment, which brings a growing and timely need to develop a data-driven methodology for identifying priority locations for EGS deployment. Currently, a lack of research exists concerning the exploration of enhanced geothermal-related resources. On top of that, stress state modeling, which is essential for EGS success, is currently very time consuming and not efficiently scalable to regional scales.
This concept proposes a methodology which will use widely available satellite remote sensing and other geoscientific data to produce labels (positive and negative for EGS), informed by the known positive Utah Frontier Observatory for Research in Geothermal Energy (FORGE), Project Red, & EGS demonstration sites. The algorithms will be trained on the labeled training sites to identify spatial patterns of EGS indicators in exploration datasets, specifically within the Innovative Geothermal Exploration through Novel Investigations of Undiscovered Systems (INGENIOUS) project area which provides a relatively complete geoscientific dataset. Performance will be evaluated using a held-out subset of the labelled sites. These algorithms will be built with scalability and interpretability in mind so that the NREL national-scale EGS resource map could be updated as part of future work, and so that non-experts may understand the process. Streamlining the process for identifying favorable areas for EGS development will ultimately lower the cost and enhance reproducibility of EGS.
