MODELING TEXTURAL PARAMETERS OF A FORMATION WITH DOWNHOLE MEASUREMENTS

• US 20200132875A1
• Filed: 10/24/2018
• Published: 04/30/2020
• Est. Priority Date: 10/24/2018
• Status: Active Grant

• Current Assignee: Baker Hughes Holdings LLC (Baker Hughes Company)
• Original Assignee: Baker Hughes, a GE Company, LLC (Baker Hughes Company)
• Inventors: Zhang, Yinxi, Dutta, Sushant
• Granted Patent: US 11,719,849 B2
• Time in Patent Office: Days
• CPC Class Codes: E21B 47/003 Determining well or borehole volumes
• CPC Class Codes: E21B 49/00 Testing the nature of borehole walls; Formation testing; Methods or apparatus for obtaining samples of soil or well fluids, specially adapted to earth drilling or wells
• CPC Class Codes: E21B 49/087 Well testing, e.g. testing for reservoir productivity or formation parameters
• CPC Class Codes: G01V 20/00 Geomodelling in general
• CPC Class Codes: G01V 3/30 operating with electromagnetic waves
• CPC Class Codes: G01V 3/32 operating with electron or nuclear magnetic resonance
• CPC Class Codes: G01V 3/38 Processing data, e.g. for analysis, for interpretation, for correction
• CPC Class Codes: G01V 99/00 Subject matter not provided for in other groups of this subclass
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/048 Activation functions
• CPC Class Codes: G06N 3/08 Learning methods

Abstract

Embodiments include a method that includes obtaining a dielectric measurement of a portion of a downhole formation. The method also includes processing the dielectric measurement via a trained machine learning system. The method further includes determining at least one of a classification or a textural parameter simultaneously with formation water saturation of the downhole formation, via the machine learning system, based at least in part on the dielectric measurements. The machine learning system is based on pre-determined dataset from previous measurements or simulated results of synthetic cases. The method detangled the correlation between water saturation and formation texture through a frequency cascading training process based on sensitivity of complex dielectric spectrum with respect to desired parameters including water saturation and texture parameter. The method also includes assigning at least one of the classification or the textural parameter to the downhole formation.

Claims

1. A method, comprising:

obtaining a dielectric measurement of a downhole formation;

processing the dielectric measurement via a trained machine learning system;

determining water saturation and at least one classification or textural parameter of the downhole formation, via the machine learning system, based at least in part on the dielectric measurement; and

assigning water saturation and at least one of the classification or the textural parameter to the downhole formation.

2. The method of claim 1, further comprising:

obtaining predetermined formation data including at least a formation dielectric spectrum response, a porosity, and a rock composition;

obtaining predetermined formation data including formation textural properties from generated three-dimensional rock geometries; and

training a machine learning system with the predetermined formation data to form the trained machine learning system.

3. The method of claim 1, further comprising:

comparing the at least one classification or textural parameter to a known value in a data store;

determining the at least one classification or textural parameter is within a threshold amount of the known value in the data store; and

verifying the at least one classification or textural parameter.

4. The method of claim 3, further comprising:

storing the water saturation, and the verified at least one classification or textural parameter.

5. The method of claim 1, further comprising:

determining a physical operation, to be performed at a well site, based at least in part on the at least one classification or textural parameter.

6. The method of claim 1, wherein the at least one textural parameter comprises pore size, pore radius, pore locations, pore connectivity, porosity, wettability, or a combination thereof.

7. The method of claim 1, wherein the dielectric measurement of the portion of the downhole formation is not accompanied by a core sample of the portion of the downhole formation.

8. A computing device, comprising:

a microprocessor; and

memory including instructions that, when executed by the microprocessor, cause the computing device to;

generate a synthetic rock geometry, the synthetic rock geometry including at least one estimated parameter of a downhole formation;

compute a textural parameter of the synthetic rock geometry;

simulate a dielectric spectrum response of the synthetic rock geometry, based at least in part on the at least one estimated parameter; and

determine a correspondence between at least one of the textural parameter or the dielectric response and the synthetic rock geometry.

9. The computing device of claim 8, wherein the memory includes instructions that, when executed by the microprocessor, further cause the computing device to:

compare at least one of the textural parameter or the dielectric response to a known value;

determine the at least one of the textural parameter or the dielectric spectrum response is within a threshold value of the known value; and

storing the at least one of the textural parameter or the dielectric spectrum response to a database.

10. The computing device of claim 8, wherein the memory includes instructions that, when executed by the microprocessor, further cause the computing device to:

compare at least one of the formation water saturation of the synthetic rock geometry, the textural parameter of the synthetic rock geometry, or the dielectric spectrum response of the synthetic rock geometry to a known value;

determine at least one of an estimated formation water saturation, an estimated textural parameter, or an estimated dielectric spectrum response based on the comparison to the known value; and

storing at least one of the estimated formation water saturation, the textural parameter, or the dielectric spectrum response to a database.

11. The computing device of claim 8, wherein the memory includes instructions that, when executed by the microprocessor, further cause the computing device to:

determine a classification for the synthetic rock geometry, based at least in part on the correspondence between the at least one of the textural parameter or the dielectric spectrum response and the synthetic rock geometries.

12. The computing device of claim 8, wherein the memory includes instructions that, when executed by the microprocessor, further cause the computing device to:

determine a physical operation, to be conducted at a well site, based at least in part on the determined correspondence between the at least one of the textural parameter or the dielectric spectrum response and the synthetic rock geometry.

13. The computing device of claim 8, wherein the memory includes instructions that, when executed by the microprocessor, further cause the computing device to:

obtain a plurality of rock geometries and corresponding dielectric spectrum responses;

train a machine learning system, based at least in part on the plurality of rock geometries and corresponding high frequency sections of the dielectric spectrum responses; and

determine a water saturation estimation, based at least in part on one or more properties of the plurality of rock geometries and the corresponding high frequency dielectric spectrum responses.

14. The computing device of claim 13, wherein the memory includes instructions that, when executed by the microprocessor, further cause the computing device to:

combine the estimated water saturation with plurality of rock geometries and corresponding dielectric spectrum responses;

determine a combined correspondence between the water saturation and the plurality of rock geometries and corresponding dielectric spectrum responses;

compare the combined correspondence with one or more verified correspondences in a data store;

determine at least one of the textural parameters, based at least in part on the combined correspondence; and

update at least one of the computed textural parameters.

15. The computing device of claim 8, wherein the synthetic rock geometry is a plurality of synthetic rock geometries, and wherein the memory includes instructions that, when executed by the microprocessor, further cause the computing device to:

determine the correspondence for each synthetic rock geometries of the plurality of rock geometries; and

aggregate the respective correspondences to generate a relationship between the at least one of the textural parameter or the dielectric response and the plurality of synthetic rock geometries.

16. A system for conducting measurement operations, the system comprising:

a dielectric measurement device forming at least a portion of a downhole tool string, the dielectric measurement device operable to generate measurement data for detecting a dielectric property of a formation;

a microprocessor; and

memory including instructions that, when executed by the microprocessor, cause the system to;

receive the measurement data;

process the measurement data, via a trained machine learning system;

determine a classification of the formation, via the trained machine learning system, the classification being related to a likelihood of recoverability based on one or more textural properties of the formation; and

assign the classification to the downhole formation.

17. The system of claim 16, wherein the memory includes instructions that, when executed by the microprocessor, further cause the system to:

receive a plurality of rock geometries;

receive a plurality of known measurement data corresponding to dielectric properties of respective rock geometries of the plurality of rock geometries; and

train the machine learning system using the plurality of rock geometries and the plurality of known measurement data.

18. The system of claim 16, wherein the memory includes instructions that, when executed by the microprocessor, further cause the system to:

generate a synthetic rock geometry;

compute a textural parameter of the synthetic rock geometry;

simulate a response of complex dielectric spectrum of the synthetic rock geometry;

train the machine learning system using the computed textural parameter and the simulated dielectric response.

19. The system of claim 16, wherein the memory includes instructions that, when executed by the microprocessor, further cause the system to:

determine a physical operation, to be performed at a well site, based at least in part on the classification.

20. The system of claim 16, wherein the measurement data does not include a core sample.