ML Formation Property Prediction Automation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining formation properties in subterranean formations, such as water saturation and total porosity, are manual, subjective, and non-repeatable, relying on crossplots of drilling data and mud gas logs, which are prone to errors and lack automation.
Innovation Solution
A computing system utilizing Machine Learning algorithms, specifically Random Forest and Support Vector Machines, processes Mud Gas Logs and Drilling Data to predict formation properties like water saturation and total porosity, providing automated, accurate, and repeatable results, even in the absence of direct measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual crossplot methods are used to determine formation properties, then flexibility in interpretation is maintained, but accuracy and repeatability deteriorate due to subjectivity and human error
Solution Approach 1:
The patent replaces the manual mechanical interpretation process with an automated computer-based system that uses machine learning algorithms and statistical methods to determine formation properties, eliminating human subjectivity while maintaining interpretative flexibility through configurable parameters and multiple analysis methods
Solution Approach 2:
The system transforms the interpretation process from manual parameter selection to automated parameter optimization by using statistical methods to identify optimal baseline parameters and formation property values based on drilling data patterns, improving accuracy through data-driven parameter selection
2Reliability
If automated Machine Learning methods are used to predict formation properties, then accuracy and repeatability improve, but computational complexity increases
Solution Approach 1:
The patent segments the complex prediction task into distinct computational modules: data preprocessing, feature extraction, model training, and prediction generation. Each module handles a specific aspect of the analysis, making the overall system more manageable and maintainable while achieving high reliability through consistent automated processing
Solution Approach 2:
The system incorporates self-training capabilities where the machine learning model automatically improves by learning from new drilling data without requiring manual reconfiguration. The system performs self-validation and adjusts parameters autonomously, reducing operational complexity while maintaining high prediction reliability
3Productivity
If manual interpretation processes are used, then operational simplicity is maintained, but productivity deteriorates due to time-consuming analysis
Solution Approach 1:
The patent implements continuous automated processing of drilling data as it becomes available, eliminating the intermittent manual analysis process. The system continuously ingests data, updates models, and generates predictions in real-time, significantly improving productivity through uninterrupted automated operation while managing complexity through efficient data streaming and incremental learning
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
A method includes receiving well log data for a plurality of wells. A flag is generated based at least partially on the well log data. The wells are sorted into groups based at least partially on the well log data, the flag, or both. A model is built for each of the wells based at least partially on the well log data, the flag, and the groups.