Well Log Feature Extraction for Sweet Spot Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods lack an efficient and principled approach to extract meaningful features from high-dimensional well log curves for identifying sweet spots in shale plays, relying on simple summary statistics that lead to variability in modeling results.
Innovation Solution
The method involves extracting one-dimensional features from vertical well logs using functional Principal Components Analysis (fPCA) and interpolating these features onto horizontal well production data coordinates, building a predictive model by regressing production data onto the interpolated features, and visualizing the predicted resource production.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If simple summary statistics are used to extract features from well log curves, then the process is simple and fast, but the modeling results show high variability and lack reliability
Solution Approach 1:
The patent transforms the feature extraction approach by changing from simple summary statistics to functional Principal Components Analysis (fPCA). This parameter change in the mathematical methodology extracts more meaningful features from well log curves while maintaining computational efficiency, thereby improving modeling result consistency without sacrificing productivity.
Solution Approach 2:
The patent replaces the mechanical/statistical approach of summary statistics with a functional analysis approach (fPCA). This substitution introduces a more sophisticated mathematical framework that captures the functional relationships in well log data, leading to more reliable and consistent modeling results while preserving computational efficiency.
2Measurement precision
If functional Principal Components_analysis is used to extract features from well log curves, then the identification of sweet spots becomes more accurate and reliable, but the computational complexity increases
Solution Approach 1:
The patent applies segmentation by decomposing the complex well log curves into principal component functions through fPCA. This breaks down the high-dimensional functional data into a smaller set of dominant components that capture the essential variability, thereby improving sweet spot identification accuracy while reducing the computational burden of processing the full data complexity.
Solution Approach 2:
The patent extracts the most significant features from well log curves by taking out the dominant principal component functions. This extraction process isolates the key information needed for accurate sweet spot identification while discarding redundant or less important variations, thus improving measurement precision without proportionally increasing computational complexity.
3Area of stationary object
If spatial interpolation is performed on extracted features onto horizontal well production data coordinates, then the predictive model coverage is improved, but the processing time increases
Solution Approach 1:
The patent performs preliminary action by extracting features and performing spatial interpolation on the training data before building the final predictive model. This pre-processing step creates a ready-to-use interpolated feature set that accelerates subsequent predictions, thereby improving model coverage without proportionally increasing processing time for actual deployment.
Solution Approach 2:
The patent applies partial action by performing spatial interpolation selectively on the most critical regions or using a subset of interpolated points for model building. This approach achieves sufficient predictive coverage across the field while avoiding the computational overhead of interpolating every possible location, thus balancing coverage improvement with processing time constraints.
Data Source
AI summary
A method for identifying a resource in a field using historic well data including vertical well logs for the resource and historic horizontal well production data for the resource, the method including extracting a plurality of features from the vertical well logs, performing a spatial interpolation of the plurality of features extracted from the vertical well logs onto coordinates of the horizontal well production data to determine a plurality of interpolated features, and building a model predicting production of the resource in the field by regressing the horizontal well production data onto the interpolated features, wherein the model is displayed as a visualization of the resource production predicted in the field.