Rock Image Porosity Correction for Hydrocarbon Saturation Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional digital rock physics modeling underpredicts hydrocarbon saturation due to the failure to account for sub-resolution pore volumes in micron-scale images, leading to inaccurate estimation of effective electrical conductivity and hydrocarbon saturation.
Innovation Solution
A method for estimating hydrocarbon saturation using a backpropagation-enabled model to segment 3D rock images, correcting for sub-resolution porosity by determining a corrected saturation exponent, and applying a correction factor to image-derived porosity to account for missing pore volume.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional digital rock physics modeling is used to estimate hydrocarbon saturation from micron-scale images, then the estimation process is rapid and less expensive, but the accuracy deteriorates due to underprediction of saturation caused by missing sub-resolution pore volumes
Solution Approach 1:
The patent replaces physical laboratory measurements with a computational approach that uses machine learning trained on limited high-quality data. The system substitutes mechanical/electrical measurement systems with an information-processing system that uses 3D images and neural networks to predict saturation exponents, achieving both speed and accuracy.
Solution Approach 2:
The patent changes the approach from directly measuring electrical properties to using image-based porosity measurements combined with machine learning predictions of saturation exponents. This parameter transformation allows the system to bypass the resolution limitations of direct electrical measurements while maintaining accuracy through training on laboratory data.
2Measurement precision
If laboratory measurements are conducted to obtain accurate saturation exponents, then measurement accuracy is improved, but time consumption and cost increase substantially
Solution Approach 1:
The patent performs preliminary actions by training the machine learning model on a limited set of laboratory measurements before deployment. Once trained, the model can rapidly predict saturation exponents for numerous samples without requiring repeated laboratory measurements, thus capturing accuracy benefits upfront while enabling rapid subsequent analysis.
Solution Approach 2:
The patent creates a computational model that copies the relationship between porosity and saturation exponents learned from laboratory data. This digital copy allows the system to replicate accurate measurements without repeating the time-consuming physical experiments, using instead fast image-based porosity measurements combined with the trained model.
3Reliability
If standard sample sizes are required for laboratory measurements, then measurement reliability is improved, but the number of processable samples is limited due to sampling constraints
Solution Approach 1:
The patent transitions from requiring physical sample dimension constraints to using digital image data. By moving to a 3D image-based approach, the system eliminates the need for standardized physical sample sizes, allowing processing of samples in their original dimensions and thereby increasing the number of processable samples while maintaining measurement reliability through the trained model.
Data Source
AI summary
The present invention provides a method for estimating hydrocarbon saturation of a hydrocarbon-bearing rock from a measurement for an electrical property a resistivity log and a rock image. The image is segmented to represent either a pore space or solid material in the rock. An image porosity is estimated from the segmented image, and a corrected porosity is determined to account for the sub-resolution porosity missing in the image of the rock. A corrected saturation exponent of the rock is determined from the image porosity and the corrected porosity and is used to estimate the hydrocarbon saturation. A backpropagation-enabled trained model can be used to segment the image. A backpropagation-enabled method can be used to estimate the hydrocarbon saturation using an image selected from a series of 2D projection images, 3D reconstructed images and combinations thereof.