Unsupervised Classification for Accurate Facies Definition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for electrofacies classification in reservoir modeling rely on initial assumptions about data distribution and group numbers, leading to subtle distortions that cannot be corrected, affecting the accuracy of facies characterization.
Innovation Solution
An unsupervised classification procedure is employed, which scales well logging data without assumptions, uses a training set to classify samples, and develops classification functions based on lithofacies analysis, allowing for accurate facies definition without prior group assumptions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If initial assumptions about data distribution and group numbers are used in electrofacies classification, then the classification process can be completed, but subtle distortions are introduced that cannot be corrected
Solution Approach 1:
The patent removes the harmful element of initial assumptions about data distribution and group numbers from the classification process. By extracting these restrictive constraints, the method allows data to be classified based on its inherent characteristics without imposed distortions, thereby improving facies characterization accuracy while maintaining classification completion.
Solution Approach 2:
Instead of imposing assumptions on data and then trying to fit it, the patent inverts the approach by allowing the data to define its own groups and distribution characteristics. The classification emerges from the data itself rather than being forced into pre-defined categories, eliminating the subtle distortions caused by incorrect initial assumptions.
2Ease of operation
If current electrofacies methods with initial assumptions are used, then classification can be performed, but the distortions affect volumetric estimates and well performance predictions
Solution Approach 1:
The patent converts the potential harm of making wrong initial assumptions into a benefit by completely eliminating the need for such assumptions. The method uses the data's natural variability and patterns to define groups, transforming what would be a source of error into a source of accuracy, thereby improving both ease of operation and reliability of predictions.
3Measurement precision
If unsupervised classification without initial assumptions is used, then facies definition accuracy improves, but more complex classification procedures are required
Solution Approach 1:
The patent applies self-service by allowing the data to classify itself without external imposition of assumptions or parameters. The unsupervised classification procedure automatically identifies groups and distribution characteristics inherent in the data, eliminating the need for manual assumption-setting while achieving higher accuracy in facies definition.
Data Source
AI summary
The disclosed embodiments include a method, apparatus, and computer program product for generating facies definition. One embodiment is a computer-implemented method that includes the steps of receiving well logging data indicative of one or more properties of geologic formations penetrated by one or more wellbores, wherein no assumptions are being introduced to the well logging data; determining a type well; developing an appropriate scaling of the well logging data based on the type well; creating a training set by drawing samples from the well logging data at random depths; modifying the training set to remove interfering data; performing an unsupervised classification procedure on the training set to group samples in the training set; comparing a suite of values of the well logging data in the groups to classify lithofacies of the type well; develop classification functions; and classifying unknown wells using the classification functions to generate the facies definition.