|
|
| |
Geothermal Fracture Network Analysis Accelerated by Machine Learning: A Use Case
Guoxiang LIU, Abhash KUMAR, Scott BEAUTZ, Saif Al-dean QAWASMEH, Jeffrey NGUYEN, Dustin CRANDALLl, MacKenzie MARK-MOSER, Luciane CUNHA, Jacqueline Alexandra HAKALA, Kelly ROSE, John ROGER, Huihui YANG, Jay CHEN
[NETL, USA]
The success of geothermal energy production, particularly in Enhanced Geothermal Systems (EGS), relies heavily on a thorough understanding of subsurface fracture networks. However, leveraging multiple datasets from drilling, completion, monitoring, testing, and production to accurately characterize fractures remains challenging due to data scarcity, resolution limitations, and inherent geological complexity. This study presents an advanced machine learning-based framework to integrate and analyze various datasets from a geothermal field, addressing these challenges and enhancing fracture network characterization to support unleashing American energy. The proposed methodology applies multiple machine learning techniques to datasets obtained from the U.S. Department of Energy’s (DOE) Frontier Observatory for Research in Geothermal Energy (FORGE) initiative in Utah. Initially, up to five unsupervised machine learning algorithms are employed to analyze passive seismic data from wells 16A and 16B, quantifying b-values and categorizing diffusivity to delineate the fracture geometry, which consists of both hydraulic and natural fractures. Subsequently, supervised machine learning techniques are applied to well logs, core testing results, stimulation and pumping data, tracer records, and circulation tests, serving as validation measures for the fracture networks derived from seismic analysis. Further validation is conducted by integrating simulated completion and stimulation processes, incorporating treatment, injection, and testing pressure measurements. A key aspect of this approach is the implementation of data fusion techniques to systematically integrate fracture network interpretations across multiple datasets. This multi-level analytical framework enhances fracture characterization, providing a holistic understanding of subsurface fracture behavior. To facilitate visualization and user interaction, a web-based interactive platform has been developed. This platform enables users to explore input datasets, analytical processes, machine learning algorithms, and final fracture network models in an interactive manner, improving accessibility and interpretability. The insights derived from this study offer direct benefits to geothermal operations, including optimization of completion and stimulation designs, fracture interference analysis, and production performance forecasting. Additionally, the developed technology has been successfully tested in hydraulic fracture network characterization for unconventional oil and gas applications, demonstrating innovation in energy AI technologies and broader applicability in subsurface energy extraction.
Topic: Field Studies