MAPPING AND MACHINE LEARNING INTERPRETATION OF FRACTURES IN GEOLOGICAL CARBON STORAGE SITES USING 9C SEISMIC DATA (Invited Presentation)

To reduce CO2 concentration in the atmosphere, geological carbon storage (GCS) can play a significant role. In GCS, carbon dioxide is captured at stationary emission sites and transported through high-pressure pipelines in its supercritical state to an injection site to be stored permanently in geological formations at depths. Therefore, mapping subsurface swarms of small-scale discrete fractures is critical in the geological carbon storage site selection to avoid potential leakage of the supercritical CO2 liquid and possible induced seismicity. We show a novel seismic characterization method for subsurface fractures using a 9-C seismic dataset acquired by 3-C sources and 3-C receivers. Our new method, called the double-beam method, forms a directional incident beam made using the point sources and extracts the wavefield interference patterns recorded by the receivers due to scattering among discrete fractures. We use the interference pattern to find the subsurface distribution of discrete fractures whose locations and orientations are interpreted using a neural network machine learning algorithm. The 9-C dataset provides an opportunity to use shear waves to map fractures. The shear waves are more sensitive to fractures than the P wave because fractures could be liquid filled. The P-wave and S-wave double-beams can operate independently but they should yield the same results. Therefore, our method is a self-verifying method useful in GCS applications.