Logging Curve Stratigraphic Extraction Using Wavelet BiLSTM
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing stratigraphic division methods in petroleum exploration are subjective and result in low repeatability due to human-made conclusions based on personal experience and scientific evidence, leading to inconsistent results among researchers.
Innovation Solution
A model training method utilizing multi-scale wavelet decomposition and a multi-level bidirectional long short-term memory network is employed to process logging curves, automatically extracting stratigraphic division conclusions by learning temporal features from multiple scales without manual analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If stratigraphic division conclusions are obtained through human-made methods based on personal experience and scientific evidence, then the method can be performed with existing tools, but the repeatability and consistency of results deteriorate due to subjectivity
Solution Approach 1:
The patent replaces the manual mechanical analysis process with an automated computer-based system. The computer automatically extracts stratigraphic division conclusions from logging curves using algorithmic processing, eliminating human subjectivity while maintaining operational simplicity. This substitution of manual mechanical analysis with automated computational analysis resolves the contradiction between ease of operation and repeatability.
2Loss of time
If traditional logging curve analysis methods are used, then the process is simple and quick, but the feature extraction capability is insufficient leading to lower accuracy
Solution Approach 1:
The patent segments the logging curve analysis into multiple hierarchical levels (first level, second level, third level conclusions). Each level extracts different types of features and patterns, with deeper levels building upon previous levels. This multi-level segmentation allows comprehensive feature extraction without requiring excessive processing time, as each level focuses on specific feature types.
Solution Approach 2:
The patent performs preliminary feature extraction and pattern recognition at each level before final conclusion generation. The computer pre-identifies key characteristics, anomalies, and patterns in the logging curves, organizing them into structured conclusions that can be quickly interpreted. This preliminary processing enhances accuracy while maintaining efficient processing time.
3Adaptability or versatility
If manual stratigraphic division is performed by researchers, then the method can accommodate complex geological interpretations, but the subjectivity leads to inconsistent results among different researchers
Solution Approach 1:
The patent replaces manual geological interpretation with automated computer-based analysis. The computer systematically applies consistent criteria and algorithms to extract stratigraphic division conclusions, eliminating researcher subjectivity while maintaining the ability to handle complex geological patterns through sophisticated data processing capabilities.
Solution Approach 2:
The patent implements a feedback mechanism where the computer continuously refines its analysis based on extracted features and patterns. The system processes logging curve data, generates conclusions, and can iteratively improve accuracy by learning from results. This feedback loop ensures consistent application of interpretation criteria while adapting to complex geological scenarios.
Data Source
AI summary
The present application provides a model training method and a stratigraphic division conclusion extraction method. The model training method includes: obtaining multiple first logging curves of a target logging well, where the multiple first logging curves are respectively configured to indicate depth, natural gamma, natural potential, acoustic time difference and bilateral resistivity; pre-processing the multiple first logging curves to obtain a second logging curve corresponding to each of the first logging curves; performing a multi-scale wavelet decomposition on target segment lengths of multiple second logging curves to obtain a multi-scale component corresponding to each of the second logging curves; and inputting sample point data of the target segment length of each of the second logging curves and the corresponding multi-scale component into a pre-constructed multi-level bidirectional long short-term memory network so as to obtain a first stratigraphic division conclusion corresponding to each of the second logging curves.