Geophysical inversion with convolutional neural networks

Patent metadata

Field	Value
Publication number	US-10996372-B2
Application number	US-201816057356-A
Country	US
Kind code	B2
Filing date	Aug 7, 2018
Priority date	Aug 25, 2017
Publication date	May 4, 2021
Grant date	May 4, 2021

Abstract

A method including: storing, in a computer memory, geophysical data obtained from a survey of a subsurface region; and extracting, with a computer, a subsurface physical property model by processing the geophysical data with one or more convolutional neural networks, which are trained to relate the geophysical data to at least one subsurface physical property consistent with geological prior information.

First claim

What is claimed is:

1. A method comprising:
storing, in a computer memory, geophysical data obtained from a survey of a subsurface region;
training one or more convolutional neural networks to relate the geophysical data to at least one subsurface physical property consistent with geological prior information; and
extracting, with a computer, a subsurface physical property model by processing the geophysical data with the one or more convolutional neural networks, wherein one or more of the convolutional neural networks is built with a j-net architecture.

2. The method of claim 1 , wherein the geophysical data includes one or more of seismic, time-lapse seismic, magnetic, electrical, electromagnetic, gravity, gradiometry, well log, well pressure, or well production data.

3. The method of claim 1 , wherein the subsurface physical property includes one or more of acoustic, elastic, anisotropy, attenuation, electrical, magnetic, or flow properties.

4. The method of claim 1 , wherein the method further includes training the convolutional neural network with synthetically generated subsurface physical models consistent with the geological prior information and computer simulated data generated from the synthetically generated subsurface physical models.

5. The method of claim 4 , wherein the method further includes generating the computer simulated data based on an acoustic wave equation, an elastic wave equation, coupled acoustic-elastic wave equations, Maxwell's equations, or potential-field equations, and the appropriate boundary conditions.

6. The method of claim 1 , wherein the method includes training the convolutional neural network with a training set of measured geophysical data and subsurface models associated with the training set of measured geophysical data.

7. The method of claim 1 , wherein the method includes training the convolutional neural network with a blend of synthetic geophysical data and a training set of measured geophysical data and their associated subsurface models.

8. The methods of claim 1 , wherein the method further includes training the convolutional neural network with geophysical training data that represents prior geological knowledge about the subsurface region, the geophysical training data including environment of deposition, well information, stratigraphy, subsurface structural patterns and geophysical property ranges.

9. The method of claim 1 , wherein the convolutional neural network is a convolutional neural network including one or more operations of convolution, filtering, downsampling, upsampling, upconvolution, thresholding, or non-linear activation.

10. The method of claim 1 , wherein the convolutional neural network is built with a ResNet architecture.

11. The method of claim 1 , wherein the method further comprises training the convolutional neural network with a gradient descent algorithm or a stochastic gradient descent algorithm.

12. The method of claim 1 , wherein the method further comprises monitoring a geophysical survey that is obtaining the geophysical data based on the subsurface physical property model.

13. The method of claim 1 , wherein the method further comprises modifying a design of a geophysical survey that is obtaining the geophysical data during the geophysical survey based on the subsurface physical property model.

14. The method of claim 1 , wherein the method further includes inputting the subsurface physical property model into subsurface interpretation, hydrocarbon exploration or hydrocarbon production process.

15. The method of claim 14 , wherein the method further includes inputting the subsurface physical property model into a geophysical imaging process.

16. The method of claim 14 , wherein the method further includes inputting the subsurface physical property model as a starting model of a geophysical inversion process.

17. The method of claim 14 , wherein the method further includes identifying reservoirs and hydrocarbon deposits based on the subsurface physical property model.

18. The method of claim 14 , wherein the method further includes constructing a reservoir model based on the subsurface physical property model.

19. A system, comprising:
a ship including sources and receivers that acquire geophysical data of a subsurface region; and
a non-transitory computer readable storage medium, encoded with instructions, which when executed by the computer causes the computer to:
store, in a memory of the computer, the geophysical data obtained from a survey of the subsurface region;
train, with the computer, one or more convolutional neural networks to relate the geophysical data to at least one subsurface physical property consistent with geological prior information; and
extract, with the computer, a subsurface physical property model by processing the geophysical data with the one or more convolutional neural networks, wherein one or more of the convolutional neural networks is built with a J-net architecture.

20. A non-transitory computer readable storage medium encoded with instructions, which when executed by the computer causes the computer to implement a method comprising:
storing, in a computer memory, geophysical data obtained from a survey of a subsurface region;
training, with a computer, one or more convolutional neural networks to relate the geophysical data to at least one subsurface physical property consistent with geological prior information; and
extracting, with a computer, a subsurface physical property model by processing the geophysical data with the one or more convolutional neural networks, wherein one or more of the convolutional neural networks is built with a J-net architecture.

Assignees

• Denli Huseyin
• Subrahmanya Niranjan A
• Exxonmobil Upstream Res Co

Inventors

• Denli Huseyin
• Subrahmanya Niranjan A

Classifications

• G06N3/045

  Combinations of networks · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N3/09

  Supervised learning · CPC title

• G01V11/00

  Prospecting or detecting by methods combining techniques covered by two or more of main groups G01V1/00 - G01V9/00 · CPC title