Correcting borehole images using machine-learning models

Patent metadata

Field	Value
Publication number	US-11898435-B2
Application number	US-202017032015-A
Country	US
Kind code	B2
Filing date	Sep 25, 2020
Priority date	Sep 25, 2020
Publication date	Feb 13, 2024
Grant date	Feb 13, 2024

Abstract

Borehole images can be corrected using machine-learning models. For example, a system can train a machine-learning model based on a training dataset. The training dataset can include a first set of borehole images correlated to a second set of borehole images, where the second set of borehole images are less precise versions of the first set of borehole images. The system can then execute the trained machine-learning model in relation to an input borehole image to receive a corrected borehole image as output from the trained machine-learning model. The corrected borehole image can be a visually corrected version of the input borehole image. The system may then perform one or more operations based on the corrected borehole image, such as generating a graphical user interface that includes the corrected borehole image for display on a display device.

First claim

The invention claimed is:

1. A system comprising:
a processor; and
a memory including instructions that are executable by the processor for causing the processor to:
generate a training dataset using a three-dimensional model, the three-dimensional model being for simulating a tool in a geological subterranean formation;
train a machine-learning model based on the training dataset to generate a trained machine-learning model, the machine-learning model being different than the three-dimensional model, the training dataset including a first set of borehole images correlated to a second set of borehole images, the second set of borehole images being less precise versions of the first set of borehole images;
execute the trained machine-learning model in relation to an input borehole image to receive a corrected borehole image as output from the trained machine-learning model, the corrected borehole image being a visually corrected version of the input borehole image, wherein the input borehole image is a two-dimensional digital image that includes colored pixels representing a property of a subterranean formation into which a borehole is drilled for extracting hydrocarbons from the subterranean formation, and wherein the corrected borehole image is also a two-dimensional digital image; and
generate a graphical user interface that includes the corrected borehole image for display on a display device.

2. The system of claim 1 , wherein the input borehole image is a resistivity image and the property is a resistivity of the subterranean formation.

3. The system of claim 1 , further comprising an imaging tool positionable in the borehole, wherein the memory further includes instructions that are executable by the processor for causing the processor to generate the input borehole image based on sensor measurements from the imaging tool.

4. The system of claim 1 , wherein the machine-learning model is a convolutional neural network that includes a convolutional layer, a batch normalization layer, a concatenation layer, a summation layer, a max pooling layer, and an up-sampling layer.

5. The system of claim 1 , wherein the tool includes a virtual imaging tool, and wherein the memory further includes instructions that are executable by the processor for causing the processor to:
receive simulated measurements of the property of the geological subterranean formation; and
generate the second set of borehole images based on the simulated measurements.

6. The system of claim 5 , wherein the memory further includes instructions that are executable by the processor for causing the processor to generate the second set of borehole images by performing an inversion process on the simulated measurements.

7. The system of claim 1 , wherein the corrected borehole image is more accurate than the input borehole image.

8. A method comprising:
generating, by a processor, a training dataset using a three-dimensional model, the three-dimensional model being for simulating a tool in a geological subterranean formation;
training, by the processor, a machine-learning model based on the training dataset to generate a trained machine-learning model, the machine-learning model being different than the three-dimensional model, the training dataset including a first set of borehole images correlated to a second set of borehole images, the second set of borehole images being less precise versions of the first set of borehole images;
executing, by the processor, the trained machine-learning model in relation to an input borehole image to receive a corrected borehole image as output from the trained machine-learning model, the corrected borehole image being a visually corrected version of the input borehole image, wherein the input borehole image is a two-dimensional digital image that includes colored pixels representing a property of a subterranean formation into which a borehole is drilled for extracting hydrocarbons from the subterranean formation, and wherein the corrected borehole image is also a two-dimensional digital image; and
generating, by the processor, a graphical user interface that includes the corrected borehole image for display on a display device.

9. The method of claim 8 , wherein the property includes a resistivity of the subterranean formation.

10. The method of claim 8 , further comprising generating the input borehole image based on sensor measurements from an imaging tool in the borehole.

11. The method of claim 8 , wherein the tool includes a virtual imaging tool, and further comprising:
receiving simulated measurements of the property of the geological subterranean formation; and
generating the second set of borehole images based on the simulated measurements.

12. The method of claim 11 , further comprising generating the second set of borehole images by performing an inversion process on the simulated measurements.

13. The method of claim 8 , wherein the corrected borehole image has a higher resolution than the input borehole image, or the corrected borehole image includes a visual feature that is absent from the input borehole image.

14. A non-transitory computer-readable medium comprising program code that is executable by a processor for causing the processor to:
generate a training dataset using a three-dimensional model, the three-dimensional model being for simulating a tool in a geological subterranean formation;
train a machine-learning model based on the training dataset to generate a trained machine-learning model, the machine-learning model being different than the three-dimensional model, the training dataset including a first set of borehole images correlated to a second set of borehole images, the second set of borehole images being less precise versions of the first set of borehole images;
execute the trained machine-learning model to generate a corrected borehole image based on an input borehole image, the corrected borehole image being visually corrected in at least one way relative to the input borehole image, wherein the input borehole image is a two-dimensional digital image that includes colored pixels representing a property of a subterranean formation into which a borehole is drilled for extracting hydrocarbons from the subterranean formation, and wherein the corrected borehole image is also a two-dimensional digital image; and
execute one or more operations based on the corrected borehole image.

15. The non-transitory computer-readable medium of claim 14 , wherein the one or more operations involves generating a graphical user interface including the corrected borehole image for display on a display device.

Assignees

• Halliburton Energy Services Inc

Inventors

• Pan Li
• Fan Yijing
• Wu Hsu-Hsiang
• Ma Jin

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• E21B47/0025 Primary

  generating an image of the borehole wall using down-hole measurements, e.g. acoustic or electric · CPC title

• G01V3/20

  operating with propagation of electric current · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title