An ORNL research paper explores the use of machine learning for faster and more accurate processing of seismic data to help reduce geothermal risk.
A research project led by geophysicist Chiang Ping Chai of the Department of Energy’s Oak Ridge National Laboratory aims to develop methods to process seismic data more quickly and accurately, reducing the risk of geothermal and carbon capture projects. there is This is done by leveraging machine learning techniques and advanced seismic data processing algorithms.
Chai also works closely with ORNL scientist Monica Maceira. Monica Maceira already managed ORNL’s seismology portfolio when Chai was still a graduate student at Pennsylvania State University.
The goal of this project is to use seismology to better understand the location and motion of subsurface cracks in both geothermal and carbon capture projects. Accurate earthquake information can enhance both the safety and effectiveness of such undertakings.
“I need to use these tools before, during and after these projects.” Chai said. “Choosing a location and drilling a borehole requires an understanding of the underground. While pumping water or carbon dioxide, it is necessary to monitor how the fluid moves and how the destruction progresses. And after that, we have to see if the fracture remains as intended or is progressing as we would like it to. “
Earthquake monitoring with machine learning
research paper “Enhancing 100 km and 10 m scale seismic monitoring with machine learning” The Chai and Maceira work can be accessed at: https://www.osti.gov/biblio/1845768
This project involves developing a machine-learning-enhanced seismic monitoring workflow for rapid and automated processing of seismic event catalogs. This workflow combines state-of-the-art machine learning techniques with advanced seismic data processing algorithms. It is suitable for monitoring both natural and induced earthquakes over a wide range of spatial scales from 10 meters to 100 kilometers.
Data were collected from two sites. A 10-kilometer scale Oklahoma area and his 10-meter scale EGS collaboration site.
About 235,000 ternary body-wave seismometers were downloaded in the Oklahoma scenario. Manual picking to measure signal arrival time is very labor intensive. A deep learning model speeds up this phase selection task, completing it in just 38 minutes versus a human analyst taking her over 100 days.
Deep learning has also improved the accuracy of earthquake occurrence locations. A tighter linear trend was observed for deep learning-derived picks compared to manual picks. The deep learning results also closely match the moment tensor solution from the Saint Louis University earthquake catalog.
Read the paper to learn more about the project and its results.
Get high resolution information faster
According to Maseria, the activity of extracting something from the subsurface or injecting something into the subsurface can cause changes in stress states and trigger induced earthquakes. This was also the case in Oklahoma, where seismic increases coincided with the introduction of hydraulic fracturing.
Systems that can process rapid, high-resolution seismic data help maintain safety by enabling operators to make on-the-fly decisions about operational changes. Geothermal projects can also benefit from accurate seismic models as they help identify cracks that may act as permeable channels for fluids.
Chai is still refining the technique developed from his research, but preliminary results from a deep learning approach show that a 99.9% reduction in processing time is possible.
Source: ORNL
