Seismic image data acquired for a subsurface formation from a data acquisition system is input into a deep neural network to generate fault detection data for the subsurface formation comprising probability values at a grid of locations in the subsurface formation. The fault detection data is preprocessed via downsampling and distributed weighted factors and inputted into a generative adversarial network (GAN) upscaling generator to create high resolution fault detection data with minimized distortion and artifacts. The GAN upscaling generator is pre trained on synthetic fault data in a GAN training system using adversarial training against a GAN upscaling discriminator, and both the GAN upscaling generator and the GAN upscaling discriminator learn to approximate the distribution of the synthetic fault data.

The present disclosure provides a method and a system for high-resolution seismic fault detection by means of an adversarial neural network, including following steps of: training a target adversarial neural network based on a preset training sample set, so as to obtain a trained target adversarial neural network, wherein the preset training sample set includes seismic data and fault labels, the target adversarial neural network includes: a segmentation module, a feature fusion module, and a discriminator module, the segmentation module is a module configured for obtaining a fault feature based on the preset training sample set, and the feature fusion module is a module configured for fusing the fault feature and the seismic data into a global feature map; and performing seismic fault detection on a target seismic image based on the trained target adversarial neural network.

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.

A method for generating a geophysical image of a subsurface region includes defining a computational sub-volume for the geophysical image including a predetermined number of seismic traces of a plurality of seismic traces and a predetermined number of samples per each one of the plurality of seismic traces, generating a data matrix corresponding to a first sub-volume of the subsurface region based on the defined computational sub-volume, the data matrix comprising the predetermined number of samples for the predetermined number of traces of a portion of a seismic dataset corresponding to the first sub-volume. The method also includes estimating a coherence between the predetermined number of traces of the data matrix by performing a sum of a variance of the predetermined number of samples of the data matrix, and assigning the estimated coherence to a location in the geophysical image.

Techniques for generating a geological model from 3D seismic data and rock property data are disclosed. Rock property data and 3D seismic data are received. Based on the rock property data and the 3D seismic data, an adaptive geological model is generated. The adaptive geological model includes a characteristic geological property. Synthetic seismic data is generated from a first region of interest of the adaptive geological model. The synthetic seismic data is adapted to facilitate a comparison between the first region of interest and a corresponding region of interest of the received 3D seismic data. The characteristic geological property is adjusted until the comparison indicates a result that is within a predetermined threshold region of the corresponding value from the rock properties. A validated geologic model is then generated.

A sonic logging method is provided that transmits acoustic signals using a high order acoustic source and processes waveform data to identify a set of arrival events and time picks by automatic and/or manual methods. Ray tracing inversion is carried out for each arrival event over a number of possible raypath types that include at least one polarized shear raypath type to determine two-dimensional reflector positions and predicted inclination angles of the arrival event for the possible raypath types. One or more three-dimensional slowness-time coherence representations are generated for the arrival event and raypath type(s) and evaluated to determine azimuth, orientation and raypath type of a corresponding reflector. The method outputs a three-dimensional position and orientation for at least one reflector. The information derived from the method can be conveyed in various displays and plots and structured formats for reservoir understanding and also output for use in reservoir analysis and other applications.

Hydrocarbon exploration and extraction can be facilitated by determining fault surfaces from fault attribute volumes. For example, a system described herein can receive a fault attribute volume for faults in a subterranean formation determined using seismic data. The fault attribute volume may include multiple traces with trace locations. The system can determine a set of fault samples for each trace location. Each fault sample can include fault attributes such as a depth value, an amplitude value, and a vertical thickness value. The system can determine additional fault attributes such as a dip value and an azimuth value for each fault sample of each trace location. The system can determine fault surfaces for the faults using the fault samples and fault attributes. The system can then output the fault surfaces for use in a hydrocarbon extraction operation.

Geologic modeling methods and systems disclosed herein employ fault face parameterization to constrain and improve the transformation of a faulted physical space geologic model into an unfaulted depositional space geologic model. An illustrative embodiment includes: associating a seismic image with each face of at least one fault in a subsurface region; determining a correspondence map between the seismic images for said at least one fault; parameterizing the faces using the correspondence map to match parameter value assignments for corresponding portions of the faces; creating a displacement map that draws together matching parameter values to align the corresponding portions of the faces; applying the displacement map to the geologic model to create a design space model; modifying the design space model; applying the displacement map in reverse to the modified design space model to obtain a modified geologic model; and outputting the modified geologic model.

A method for detecting an unknown fault in a target seismic volume. The method includes generating a number of patches from a training seismic volume that is separate from the target seismic volume, where a patch includes a set of training areas, generating a label for assigning to the patch, where the label represents a subset, of the set of training areas, intersected by an known fault specified by a user in the training seismic volume, training, during a training phase and based at least on the label and the training seismic volume, a machine learning model, and generating, by applying the machine learning model to the target seismic volume during a prediction phase subsequent to the training phase, a result to identify the unknown fault in the target seismic volume.

A method for subsurface fault extraction using undirected graphs is provided. Extracting faults in the subsurface may assist in various stages of geophysical prospecting. To that end, an undirected graph may be used in order to identify distinctive fault branches in the subsurface. Fault probability data, from seismic data, may be used to establish connections in the undirected graph. Thereafter, some of the connections in the undirected graph may be removed based on analyzing one or more attributes, such as dip, azimuth, or context, associated with the connections or nodes associated with the connections. After which, the undirected graph may be analyzed in order to extract the faults in the subsurface.