US 2026/0211136 A1 · SEISMIC VELOCITY PREDICTION USING DISPERSION IMAGES AND AUTOREGRESSIVE MACHINE LEARNING

Abstract

Methods and systems are disclosed. The methods may include generating training pairs and training a machine learning (ML) model using the training pairs to produce a predicted seismic velocity profile from a dispersion image. The methods may further include obtaining seismic data from a subterranean region of interest, where the seismic data is sorted into gathers and, for each of the gathers, determining, using a transform, a dispersion image from each gather, inputting the dispersion image into the ML model, and producing a predicted seismic velocity profile from the ML model based on the dispersion image. The methods may still further include determining a predicted seismic velocity model using the predicted seismic velocity profile for each gather.

Claims20

Dependent claims

A method of training a machine learning (ML) model comprising:

generating a plurality of training pairs,
wherein each of the plurality of training pairs comprises an mth training seismic velocity profile and an associated mth training synthetic dispersion image,
wherein m is a count of the plurality of training pairs, and
wherein generating the plurality of training pairs comprises:
collecting the mth training seismic velocity profile;
generating an mth training seismic velocity model by extending the mth training seismic velocity profile along a dimension;
determining, using forward modeling, mth training synthetic seismic data using the mth training seismic velocity model;
and
determining, using a transform, the associated mth training synthetic dispersion image using, at least in part, the mth training synthetic seismic data;
and
training the ML model using the plurality of training pairs,
wherein the ML model is trained to produce a predicted seismic velocity profile from a dispersion image.

9 more elements

A method of determining a predicted seismic velocity model comprising:

obtaining seismic data from a subterranean region of interest,
wherein the seismic data is sorted into a plurality of gathers;
for each gather among the plurality of gathers:
determining, using a transform, a dispersion image from each gather,
inputting the dispersion image, at least in part, into a machine learning (ML) model,
wherein the ML model is trained to produce a predicted seismic velocity profile from the dispersion image, and
producing the predicted seismic velocity profile from the ML model based, at least in part, on the dispersion image;
and
determining the predicted seismic velocity model for the seismic data associated with the subterranean region of interest using the predicted seismic velocity profile for each gather.

6 more elements

Dependencies are inferred from the claim text.

Seismic velocity prediction using dispersion images and autoregressive machine learning

Life of this patent

Seismic velocity prediction using dispersion images and autoregressive machine learning

Life of this patent

Abstract

Claims20