Researchers at Penn State have used machine learning to connect low-magnitude microearthquakes with the permeability of subsurface rocks below the Earth’s surface. This finding could potentially impact the improvement of geothermal energy transfer. Generating geothermal energy relies on having a permeable subsurface to effectively release heat as cold fluids are injected into the rock.This study uncovers the best times for efficient energy transfer using microearthquakes, which are tracked on the surface with seismometers. The team published their results in Nature Communications.
With funding from the U.S. Department of Energy (DOE) and data from the EGS Collab and Utah FORGE demonstration projects, researchers used machine learning to identify the “noise” in the data that was hiding the connection. They then used machine learning to develop a model from one site and successfully applied it to the other site, a process known as transfer learning, indicating that the li…nk was established on the fundamental physics of subsurface rocks, indicating that it is likely universally applicable to all geothermal energy sites, according to the researchers. “The success of transfer learning confirms the generalizability of the models,” explained Pengliang Yu, the lead author of the study and a postdoctoral scholar at Penn State. “This suggests that seismic monitoring could be widely utilized to enhance geothermal energy transfer efficiencies across various sites.” Yu also emphasized the critical importance of increasing rock permeability in a variety of energy extraction methods, affecting both traditional fossil fuel recovery and renewable energy.Hydraulic fracturing, also known as hydrofracturing, is a process that involves injecting cold fluids into the subsurface through porous rock to create high pressures that can break the rock. This results in the generation of microearthquakes, which are similar to natural earthquakes but on a smaller scale. The increased permeability of the rock allows for easier access to energies such as heat and hydrocarbons. According to Yu, their algorithm demonstrated a direct correlation between the strength of the seismic activity and the rock’s permeability. Understanding this link can lead to improvements in various processes, including hydrogen production.The ability to harvest energy while keeping small earthquakes below the level that could cause harm or be noticed by the public is crucial. “Machine learning was essential in revealing the connection between seismic activity and rock permeability,” explained co-author Parisa Shokouhi, a professor of engineering science and mechanics in the College of Engineering. “It assisted in pinpointing the crucial elements of the seismic data for forecasting the evolution of rock permeability. We restricted the machine learning algorithm to guarantee a model that makes sense in the physical world. As a result, the model’s prediction unveiled a previously unknown aspect.
There is still an unknown physical link between seismic data and rock permeability.
Researchers have stated that increasing the availability of geothermal energy could reduce dependence on fossil fuels. They also mentioned that connecting rock permeability to microquakes could help monitor gas movement for carbon sequestration, as well as the production and storage of subsurface hydrogen.
This research is a component of a larger DOE-funded project aimed at reducing the cost and increasing the production of geothermal energy. It also involves using machine learning to better understand and predict earthquakes, including microquakes.
“Yu’s work is a part of our effort to advance…”According to Chris Marone, professor of geosciences at Penn State, machine learning methods are being used for geothermal exploration and energy production. The lab studies have shown connections between the evolution of elastic properties before lab earthquakes, and similar relationships are observed in nature. Ankur Mali, from the University of South Florida and a Penn State graduate, along with Thejasvi Velaga, a research assistant, and Alex Bi, an undergraduate student, are also involved in this research.The research was conducted at Penn State by Pengliang Yu, Ankur Mali, Thejasvi Velaga, Alex Bi, and Chris Marone. The study was also contributed to by Jiayi Yu, a graduate student in the Department of Geosciences, and Derek Elsworth, G. Albert Shoemaker Chair and professor of energy and mineral engineering and geosciences at Penn State.