Uncertainty estimation of subsurface resistivity solutions

Patent metadata

Field	Value
Publication number	US-9746577-B2
Application number	US-201414271131-A
Country	US
Kind code	B2
Filing date	May 6, 2014
Priority date	Jun 3, 2013
Publication date	Aug 29, 2017
Grant date	Aug 29, 2017

Abstract

Method for estimating uncertainty in a physical property model generated by inverting measured geophysical data, for example, a resistivity model inferred from electromagnetic field data. The method involves as few as one data inversion coupled with a number of forward simulation. Alternative solutions (models) are generated by probing a perturbation space defined from a reduced model space, resulting from a principal component decomposition of the inverted model, and selecting some of the larger components. Statistical analysis techniques may be applied (to those alternative solutions remaining after thresholding) to generate quantitative uncertainty estimates applicable to the inverted model.

First claim

The invention claimed is:

1. A computer-implemented method for quantitatively estimating uncertainty in a model of a subsurface physical property inferred by inversion of measured geophysical data, comprising:
performing, with a computer, a principal component decomposition of the model by transforming the model to a principal component domain, using a computer;
reducing, with a computer, the principal component domain to an N-dimensional reduced domain by selecting N, less than all, principal components, based on a criterion favoring larger principal components over smaller principal components;
defining, with a computer, an D-dimensional perturbation space within the N-dimensional reduced domain, where D<N;
generating, with a computer, perturbed models by perturbing the model within the perturbation space and interpolation within the perturbation space;
applying, with a computer, a statistical analysis technique to the perturbed models to compute the uncertainty in said model inferred by inversion of measured geophysical data, or compute a statistically updated model and associated uncertainty;
performing a computer simulation with the perturbed models, wherein the simulation includes generating simulated geophysical data from the perturbed models, comparing the simulated geophysical data to the measured geophysical data, and transforming a misfit between the simulated geophysical data and the measured geophysical data to an uncertainty in the subsurface property inferred by the inversion of the measured geophysical data; and
prospecting for hydrocarbon in a subsurface region, wherein a hydrocarbon exploration decision is based on the subsurface property inferred by the inversion of the measured geophysical data and the uncertainty in the subsurface property.

2. The method of claim 1 , further comprising after the generating but before the applying a statistical analysis technique:
forward modeling the perturbed models, using a computer, to generate synthetic geophysical data, and computing misfit between the synthetic geophysical data and the measured geophysical data; and
filtering out at least one of the perturbed models based on a selected data misfit threshold, and using remaining perturbed models for the uncertainty computation by statistical analysis.

3. The method of claim 2 , wherein the misfit threshold is based on misfit associated with the model in said inversion of measured geophysical data.

4. The method of claim 1 , wherein the geophysical data are electromagnetic field data, and the physical property is resistivity or conductivity.

5. The method of claim 4 , wherein the method is performed, separately or jointly, for both a horizontal resistivity model and a vertical resistivity model, wherein for a joint uncertainty estimation, a joint perturbation is performed in a combined perturbation space.

6. The method of claim 1 , wherein the statistically updated model and associated uncertainty is a most probable model and a standard deviation.

7. The method of claim 1 , further comprising performing an additional inversion of the measured geophysical data, resulting in a second inverted model due to inversion non-uniqueness, and performing the method on the second inverted model to generate perturbed models associated with the second inverted model, and then combining perturbed versions of the perturbed models and the perturbed models associated with the second inverted model for the statistical analysis.

8. The method of claim 1 , wherein the principal component decomposition is singular value decomposition.

9. The method of claim 1 , wherein D<N is achieved by omitting one or more of the N principal components in the reduced domain, or by combining one or more of the N principal components to form a single dimension, or by a combination of both.

10. The method of claim 1 , wherein the perturbation space has m+1 principal components and is defined to be 2-dimensional, in which a first dimension is spanned by the principal component having a largest magnitude while a second dimension is given by a subspace of the remaining m principal components.

11. The method of claim 10 , wherein m+1=N.

12. The method of claim 1 , wherein a 3D window in Cartesian space is used for calculating the principal components or perturbing the model or both.

13. The method of claim 1 , wherein D≧2.

Assignees

• Ciucivara Adrian
• Willen Dennis E
• Exxonmobil Upstream Res Co

Inventors

• Ciucivara Adrian
• Willen Dennis E

Classifications

• G01V2210/62

  Physical property of subsurface · CPC title

• G01V3/38 Primary

  Processing data, e.g. for analysis, for interpretation, for correction · CPC title

• G06F30/00

  Computer-aided design [CAD] · CPC title

• G06F17/50

  Physics · mapped topic