Seismic AI in RealTime for Informed Drilling and Fracking

The US Energy Information Administration (EIA) estimates that Unconventional Oil and Gas (UOG) will enable the US energy security and dominance through 2050. An array of technologies are required to support the EIA’s growth projections. Extended-lateral horizontal drilling and high-volume hydraulic fracturing have challenges, and the current recovery efficiency estimates range from 10 – 20%. Advances in real-time diagnostics and analytics are required to increase the efficiency of fracturing and reduce the environmental risks of UOG. Our team of researchers in machine learning, geophysics, and civil engineering propose to develop a “SeisMic AI in RealTime for hydraulic Fracture Remote-sensing And Characterization” (SMART FRAC) tool thatcan identify the micro-seismic events (event type and focal mechanism) to can help manage operations and mitigate the risk of hazardous seismic events during drilling and fracking operations. This tool builds on the current GTC capabilities of micro-seismic analysis by using novel deep learning algorithms for detailed real-time discrimination. We will leverage our technology developed for the US Air Force, which uses deep learning to discriminate nuclear blast from earthquakes through a global sensor network. Phase I objectives to demonstrate the feasibility of the SMART FRAC tool are: Identify and collect sufficient number of field seismic events with well-documented seismic signals and known focal mechanisms in different settings, for instance, during geothermal or hydraulic fracturing operations. Generate acoustic emission data in lab settings of rock fracture at different temperature-pressure conditions to be used as training data set for machine learning algorithms. Combine laboratory acoustic emission data with field seismic events data to develop a deep learning focal-mechanism solver. Demonstrate ability to discriminate event types (shear, hydraulic, thermal) using deep learning SMART FRAC tool. High pressure-temperature conditions and the requirement of large-volume and long-term circulation of exotic fluids, enhanced geothermal system reservoirs will undergo prolonged thermo-chemo-mechanical perturbations resulting in evolutionary changes in its reservoir properties and stress states. The tool proposed in this study will enable geothermal and fracking operators to detect, diagnose, and characterize the nature or cause of the micro-seismic events and help predict the stress state evolution and reservoir-induced seismicity. Such advancement can significantly improve the safety and efficiency of drilling into hot rocks, controlling the flow through fracture networks, and minimizing risk of inducing unwanted possibly damaging earthquakes. Successful UOG energy exploitation can impact the supply and security of U.S. energy. Knowledge learned from this project will also offer insights into earthquake physics, geomechanics, and the deployment of machine learning to quantify the uncertainties associated with signal quality, diverse rock, and fluid properties, and interwoven hydro-thermo-chemo-mechanical processes in subsurface.