# US 12,182,226 B2 · Neural-network-based mapping of potential leakage pathways of subsurface carbon dioxide storage > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,182,226 B2 - Document ID: US12182226B2 - Status: Granted - Kind: B2 - Application number: 17/471,601 - Filed: 2021-09-10 - Published: 2023-03-16 - Granted: 2024-12-31 - Earliest claimed: 2021-09-10 - Assignees: Microsoft Technology Licensing, LLC - Applicants: Microsoft Technology Licensing, LLC - Inventors: Qie Zhang; Harpreet Kaur - Web page: https://patentgrep.com/patent/US12182226B2 ## CPC - G06F 18/2148: {characterised by the process organisation or structure, e.g. boosting cascade} - G06N 3/082: modifying the architecture, e.g. adding, deleting or silencing nodes or connections - G01V 1/30: Analysis (G01V1/50 takes precedence) - G01V 1/301: {for determining seismic cross-sections or geostructures} - G01V 20/00: Geomodelling in general - G06F 18/2193: {based on specific statistical tests} - G06N 3/045: Combinations of networks - G06N 3/08: Learning methods - G06V 30/2504: {Coarse or fine approaches, e.g. resolution of ambiguities or multiscale approaches} - G01V 2210/642: Faults - G01V 2210/644: Connectivity, e.g. for fluid movement - G06N 3/09: Supervised learning - G06N 3/0985: Hyperparameter optimisation; Meta-learning; Learning-to-learn - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G01V: GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS (means for indicating the location of accidentally buried, e.g. snow-buried, persons A63B29/02) - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract The disclosed technology is generally directed to carbon capture and storage. In one example of the technology, a first neural network is trained with synthetic data that is associated with seismic images of synthetic simulated subsurfaces. The first neural network extracts features from multiple resolutions of the seismic images of the synthetic simulated subsurfaces. The ground truth includes synthetic labels that indicate probabilities of potential carbon dioxide leakage pathways of the synthetic simulated subsurfaces. A seismic image of a first subsurface is received. At least the trained first neutral network is used to generate output labels that indicate probabilities of potential leakage pathways of carbon dioxide storage of the first subsurface. ## Claims ### Claim 1 An apparatus, comprising: a device including at least one memory having processor-executable code stored therein, and at least one processor that is adapted to execute the processor-executable code, wherein the processor-executable code includes processor-executable instructions that, in response to execution, enable the device to perform actions, including: training a first neural network with synthetic data that is associated with seismic images of synthetic simulated subsurfaces, wherein the first neural network extracts features from multiple resolutions of the seismic images of the synthetic simulated subsurfaces, with ground truth including synthetic labels that indicate probabilities of potential carbon dioxide leakage pathways of the synthetic simulated subsurfaces, and wherein the synthetic data includes simulated elastic deformations of the potential carbon dioxide leakage pathways of the synthetic simulated subsurfaces; receiving a seismic image of a first subsurface; and using at least the first neural network to generate output labels that indicate probabilities of potential leakage pathways of carbon dioxide storage of the first subsurface. ### Claim 2 (depends on 1) The apparatus of claim 1, wherein the synthetic data incorporates linear transformations. ### Claim 3 (depends on 1) The apparatus of claim 1, where the training is performed on a first group of neural networks including the first neural network and at least one additional neural network. ### Claim 4 (depends on 3) The apparatus of claim 3, wherein the training of the first group of neural networks also provides uncertainty estimations for the output labels. ### Claim 5 (depends on 4) The apparatus of claim 4, wherein at least one neural network in the first group of neural networks is trained using a dropout layer. ### Claim 6 (depends on 3) The apparatus of claim 3, wherein the training incorporates ensemble learning. ### Claim 7 (depends on 6) The apparatus of claim 6, wherein the ensemble learning includes combining results of each of the neural networks in the first group of neural networks. ### Claim 8 (depends on 3) The apparatus of claim 3, wherein the first neural network includes a first model, and wherein the at least one additional neural network includes a second neural network that includes a second model that is different from the first model. ### Claim 9 (depends on 8) The apparatus of claim 8, wherein the at least one additional neural network further includes a third neural network and a fourth neural network, wherein the third neural network is trained based on the first model and using dropout relative to the first neural network, and wherein the fourth neural network is trained based on the second model and using dropout relative to the second neural network. ### Claim 10 A method, comprising: via at least one processor, training at least one neural network with synthetic data, wherein the synthetic data includes seismic images of synthetic simulated subsurfaces, at least a first neural network of the at least one neural network extracts features from multiple resolutions of the seismic images of the synthetic simulated subsurfaces, and wherein the training is also accomplished with a ground truth that includes synthetic labels that are associated with probabilities of potential carbon dioxide leakage pathways of the synthetic simulated subsurfaces, and wherein the synthetic data includes simulated elastic deformations of the potential carbon dioxide leakage pathways of the synthetic simulated subsurfaces; receiving a seismic image of a first subsurface; and using the at least one neural network to provide output labels that are associated with probabilities of potential carbon dioxide leakage pathways of the first subsurface. ### Claim 11 (depends on 10) The method of claim 10, wherein the synthetic data incorporates linear transformations. ### Claim 12 (depends on 10) The method of claim 10, wherein the at least one neural network further includes at least one additional neural network, and wherein the training incorporates ensemble learning. ### Claim 13 (depends on 12) The method of claim 12, wherein the first neural network includes a first model, and wherein the at least one additional neural network includes a second neural network that includes a second model that is different from the first model. ### Claim 14 (depends on 12) The method of claim 12, wherein at least one neural network among the at least one neural network is trained using a dropout layer. ### Claim 15 (depends on 12) The method of claim 12, further comprising training the at least one neural network to provide uncertainty estimations of the output labels. ### Claim 16 A processor-readable storage medium, having stored thereon processor-executable code that, upon execution by at least one processor, enables actions, comprising: performing training of a first neural network with synthetic data that is associated with images of synthetic simulated subsurfaces, wherein the first neural network extracts features from multiple resolutions of the images of the synthetic simulated subsurfaces, with ground truth including synthetic labels that indicate probabilities of potential carbon dioxide leakage pathways of the synthetic simulated subsurfaces, and wherein the synthetic data includes simulated elastic deformations of the potential carbon dioxide leakage pathways of the synthetic simulated subsurfaces; and providing output labels from an image of a first subsurface via at least the first neural network, such that the output labels indicate probabilities of potential leakage pathways of carbon dioxide storage of the first subsurface. ### Claim 17 (depends on 16) The processor-readable storage medium of claim 16, wherein the synthetic data incorporates linear transformations. ### Claim 18 (depends on 16) The processor-readable storage medium of claim 16, wherein the training is performed on a first group of neural networks including the first neural network and at least one additional neural network, and wherein the training incorporates ensemble learning. ### Claim 19 (depends on 18) The processor-readable storage medium of claim 18, further comprising training the first group of neural networks to provide uncertainty estimations of the output labels.