A method for generating one or more reservoir models using machine learning is provided. Generating reservoir models is typically a time-intensive idiosyncratic process. However, machine learning may be used to generate one or more reservoir models that characterize the subsurface. The machine learning may use geological data, geological concepts, reservoir stratigraphic configurations, and one or more input geological models in order to generate the one or more reservoir models. As one example, a generative adversarial network (GAN) may be used as the machine learning methodology. The GAN includes two neural networks, including a generative network (which generates candidate reservoir models) and a discriminative network (which evaluates the candidate reservoir models), contest with each other in order to generate the reservoir models.

A method and apparatus for automated seismic interpretation (ASI), including: obtaining trained models comprising a geologic scenario from a model repository, wherein the trained models comprise executable code; obtaining test data comprising geophysical data for a subsurface region; and performing an inference on the test data with the trained models to generate a feature probability map representative of subsurface features. A method and apparatus for machine learning, including: an ASI model; a training dataset comprising seismic images and a plurality of data portions; a plurality of memory locations, each comprising a replication of the ASI model and a different data portion of the training dataset; a plurality of data augmentation modules, each identified with one of the plurality of memory locations; a training module configured to receive output from the plurality of data augmentation modules; and a model repository configured to receive updated models from the training module.

Provided are a device, method, and computer-readable recording medium for detecting an earthquake in a microelectromechanical system (MEMS)-based auxiliary seismic observation network. The method includes performing detrending of removing a moving average from original acceleration data received from single sensors of an MEMS-based auxiliary seismic observation network to preprocess the acceleration data, calculating a short-term average/long-term average (STA/LTA) value using a filter parameter value specified on the basis of the preprocessed acceleration data, generating an event occurrence message or event end message on the basis of the calculated STA/LTA value and transmitting the event occurrence message or event end message, when the event occurrence message is generated, calculating an earthquake probability through an earthquake detection deep learning model using the preprocessed acceleration data as an input, and analyzing noise by calculating a power spectral density (PSD) from the original acceleration data which is merged at certain intervals.

A method for weathered layer correction of seismic data includes identifying arrival times in the seismic data corresponding to a weathered layer velocity gradient. A velocity model of the weathered layer is generated using the arrival times. The seismic data are time adjusted using the velocity model.

Methods of and systems for forming an image of a subterranean region of interest are disclosed. The method includes obtaining an observed seismic dataset and a seismic velocity model for the subterranean region of interest and generating a simulated seismic dataset based on the seismic velocity model and the source and receiver geometry of the observed seismic dataset. The method also includes forming a plurality of time-windowed trace pairs from the simulated and the observed seismic datasets, and forming an objective function based on a penalty function and a cross-correlation between the members of each pair. The method further includes determining a seismic velocity increment based on the extremum of the objective function and forming an updated seismic velocity model by combining the seismic velocity increment and the seismic velocity model, and forming the image of the subterranean region of interest based on the updated seismic velocity model.

Geologic modeling methods and systems disclosed herein employ an improved simulation meshing technique. One or more illustrative geologic modeling methods may comprise: obtaining a geologic model representing a faulted subsurface region in physical space; providing a set of background cells that encompass one or more partial faults within the subsurface region; defining a pseudo-extension from each unterminated edge of said one or more partial faults to a boundary of a corresponding background cell in said set; using the pseudo-extensions and the background cell boundaries to partition the subsurface region into sub-regions; deriving a simulation mesh in each sub-region based on the horizons in each sub-region; and outputting the simulation mesh.

In a first step, a defined scope value is selected for each of a plurality of hydrodynamic input parameters. A simulated topographical result is generated using the selected scope values and a forward model. A detailed seismic interpretation is generated to represent specific seismic features or observed topography. A calculated a misfit value representing a distance between the simulated topographical result and a detailed seismic interpretation is minimized. An estimated optimized sand ratio and optimized hydrodynamic input parameters are generated. In a second step, a genetic algorithm is used to determine a proportion of each grain size in the estimated optimized sand ratio. A misfit value is used that is calculated from thickness and porosity data extracted from well data and a simulation result generated by the forward model to generate optimized components of different grain sizes. Optimized hydrodynamic input parameters and optimized components of different grain sizes are generated.

Systems and methods for simulating subterranean regions having multi-scale fracture geometries. Non-intrusive embedded discrete fracture modeling formulations are applied in conjunction with commercial simulators to efficiently and accurately model subsurface transient flow characteristics in regions having complex hydraulic fractures, complex natural fractures, or a combination of both, and geometries including corner point grids.

A seismic source is positioned at the surface of a geologic formation and a plurality of seismic receivers is positioned in a wellbore of the geologic formation. Seismic wavefield data is obtained based on the seismic source outputting seismic energy into the wellbore and the plurality of seismic receivers receiving the seismic energy. A velocity profile is determined along the wellbore based on the seismic wavefield data. P and S wave data in a downgoing direction is separated from the seismic wavefield data based on an inversion and the velocity profile. The P and S wave data in the downgoing direction is adaptively subtracted from the seismic wavefield data to form residual wavefield data. The P and S wave data in a upgoing direction is separated from the residual wavefield data based on the inversion and an updated velocity profile. The P and S wave data in the upgoing and downgoing direction is output.

Method for characterizing subterranean formation is described. One method involves simulating a poroelastic pressure response of known fracture geometry utilizing a geomechanical model to generate a simulated poroelastic pressure response. Compiling a database of simulated poroelastic pressure responses. Measuring a poroelastic pressure response of the subterranean formation during a hydraulic fracturing operation to generate a measured poroelastic pressure response. Identifying a closest simulated poroelastic pressure response in the library of simulated poroelastic pressure response. Estimating a geometrical parameter of a fracture or fractures in the subterranean formation based on the closest simulated poroelastic pressure response.