Synthesis of sequential, spectral, and time-series data

Patent metadata

Field	Value
Publication number	US-11640525-B2
Application number	US-201916592531-A
Country	US
Kind code	B2
Filing date	Oct 3, 2019
Priority date	Jul 23, 2018
Publication date	May 2, 2023
Grant date	May 2, 2023

Abstract

A method comprises performing data pre-processing of initial signals to obtain pre-processed initial signals; building a first machine learning model based on the pre-processed initial signals; generating output signals using the first machine learning model; computing ranks of the output signals; computing classifications of the output signals; and building a set of stacked machine learning models based on the ranks and the classifications. The set of stacked machine learning models may be used to generate subsurface well log data, NMR data, or other data.

First claim

What is claimed is:

1. A method comprising:
performing processing of a training dataset to obtain a processed training dataset;
building a first machine learning model based on the processed training dataset;
generating output signals using the first machine learning model, wherein the output signals are dielectric-dispersion (DD) logs or nuclear magnetic resonance (NMR) logs;
computing ranks of the output signals based on an accuracy of prediction of the first machine learning model;
computing classifications of the output signals, wherein the classifications are based on a composition, an abundance of pores, a deviation of pore sizes, a number of peaks in a pore size distribution, or a dominant pore size;
building a set of stacked machine learning models based on the ranks and the classifications;
using the set to predict conductivity or a permittivity of a soil, generate subsurface well log data, or generate NMR data; and
implementing an oil and gas operation based on the conductivity, the permittivity, the subsurface well log data, or the NMR data.

2. The method of claim 1 , wherein the training dataset comprise at least one of a training input signal, a training output signal, or a deployment input signal.

3. The method of claim 1 , wherein the first machine learning model is a neural network model, a support vector machine model, or a random forest model.

4. The method of claim 1 , wherein the classifications are further based on a pore size distribution.

5. The method of claim 1 , further comprising further computing the classifications using a classification algorithm.

6. The method of claim 5 , wherein the classification algorithm is a k-nearest neighbor (kNN) algorithm or a random forest classifier algorithm.

7. The method of claim 1 , further comprising further using the set to predict the conductivity or the permittivity of a soil.

8. The method of claim 1 , further comprising further using the set to generate the subsurface well log data.

9. The method of claim 1 , further comprising further using the set to generate the NMR data.

10. The method of claim 1 , wherein the conductivity and the permittivity are frequency-dependent.

11. An apparatus comprising:
a memory configured to store instructions; and
a processor coupled to the memory and configured to execute the instructions to cause the apparatus to:
perform processing of a training dataset to obtain a processed training dataset;
build a first machine learning model based on the processed training dataset;
generate output signals using the first machine learning model, wherein the output signals are dielectric-dispersion (DD) logs or nuclear magnetic resonance (NMR) logs;
compute ranks of the output signals based on an accuracy of prediction of the first machine learning model;
compute classifications of the output signals, wherein the classifications are based on a composition, an abundance of pores, a deviation of pore sizes, a number of peaks in a pore size distribution, or a dominant pore size;
build a set of stacked machine learning models based on the ranks and the classifications;
use the set to predict a conductivity or a permittivity of a soil, generate subsurface well log data, or generate NMR data; and

implement an oil and gas operation based on the conductivity, the permittivity, the subsurface well log data, or the NMR data.

12. The apparatus of claim 11 , wherein the training dataset comprises at least one of a training input signal, a training output signal, or a deployment input signal.

13. The apparatus of claim 11 , wherein the first machine learning model is a neural network model.

14. The apparatus of claim 11 , wherein the classifications are further based on a pore size distribution.

15. The apparatus of claim 11 , wherein the processor is further configured to further compute the classifications using a classification algorithm.

16. The apparatus of claim 15 , wherein the classification algorithm is a k-nearest neighbor (kNN) algorithm.

17. The apparatus of claim 11 , wherein the processor is further configured to execute the instructions to cause the apparatus to further use the set to predict the conductivity or the permittivity of the soil.

18. The apparatus of claim 11 , wherein the processor is further configured to execute the instructions to cause the apparatus to further use the set to generate the subsurface well log data.

19. The apparatus of claim 11 , wherein the processor is further configured to execute the instructions to cause the apparatus to further use the set to generate the NMR data.

20. The apparatus of claim 11 , wherein the conductivity and the permittivity are frequency-dependent.

21. A computer program product comprising instructions that are stored on a non-transitory computer-readable medium and that, when executed by a processor, cause an apparatus to:
perform processing of a training dataset to obtain a processed training dataset;
build a first machine learning model based on the processed training dataset;
generate output signals using the first machine learning model, wherein the output signals are dielectric-dispersion (DD) logs or nuclear magnetic resonance (NMR) logs;
compute ranks of the output signals based on an accuracy of prediction of the first machine learning model;
compute classifications of the output signals, wherein the classifications are based on a composition, an abundance of pores, a deviation of pore sizes, a number of peaks in a pore size distribution, or a dominant pore size;
build a set of stacked machine learning models based on the ranks and the classifications;
use the set to predict a conductivity or a permittivity of a soil, generate subsurface well log data, or generate NMR data; and
implement an oil and gas operation based on the conductivity, the permittivity, the subsurface well log data, or the NMR data.

22. The computer program product of claim 21 , wherein the ranks are based on accuracy of prediction.

Assignees

• Univ Oklahoma

Inventors

• Misra Siddharth
• He Jiabo

Classifications

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/08 Primary

  Learning methods · CPC title

• G06F16/2477 Primary

  Temporal data queries · CPC title

• G06N20/00

  Machine learning · CPC title