DCNN Permeability Ratio Determination from Core Photographs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining reservoir permeability, especially in oil sands and heavy oil reservoirs, face challenges due to scarce vertical permeability data, biased sampling, and computational inefficiency, leading to overestimation and limited accuracy in permeability modeling.
Innovation Solution
A deep convolution neural network system is employed to process core photographs and Vshale logs, applying thresholds to generate binary images and iteratively determine optimal thresholds for calculating permeability ratios, significantly reducing computation time and resource requirements compared to traditional micro-modelling techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional micro-modelling techniques are used to determine permeability ratios, then measurement precision may be improved, but productivity deteriorates due to computational inefficiency and long processing times
Solution Approach 1:
The patent replaces traditional mechanical micro-modelling techniques with a deep convolutional neural network (DCNN) system. The DCNN processes core photographs directly through computational algorithms, substituting the physical mechanical modelling process with an optimized computational approach that achieves comparable accuracy while dramatically reducing processing time and computational resource requirements.
Solution Approach 2:
The patent transforms the permeability determination problem by changing the input parameters from complex 3D numerical models to simplified 2D core photographs. By converting the problem space from requiring extensive physical modelling to analyzing image-based parameters through neural networks, the system achieves faster processing while maintaining measurement precision.
2Measurement precision
If complex 3D numerical models are generated with multiple inputs including porosity, saturation, and permeability, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent extracts the essential information needed for permeability ratio determination from complex 3D numerical models by focusing specifically on core photographs and Vshale logs. This extraction approach isolates the critical input data required for accurate permeability ratio calculation, eliminating the need to process all the complex parameters (porosity, saturation, etc.) required in traditional 3D modelling.
Solution Approach 2:
The patent segments the reservoir characterization process into distinct components: core photograph acquisition, binary image generation through thresholding, and permeability ratio calculation via DCNN. This segmentation allows each component to be optimized independently, reducing overall system complexity while maintaining measurement precision.
3Manufacturing precision
If iterative threshold optimization is applied to core photographs, then manufacturing precision is improved, but loss of time increases due to multiple comparison cycles
Solution Approach 1:
The patent applies preliminary action by pre-processing core photographs into binary images using thresholding algorithms before feeding them to the DCNN. The iterative optimization of thresholds is performed in advance to generate optimized binary images, which then can be rapidly processed by the trained neural network, separating the time-consuming optimization step from the rapid prediction step.
Data Source
AI summary
Systems and methods for field of reservoir characterization, and more specifically to more accurate and processor efficient methods of permeability modeling. The systems efficiently determine a permeability of a reservoir using a deep convolution neural network and core photographs and Vshale logs. In some aspects, the core photographs are windowed to determine a continuous permeability ratio for the reservoir.