Well Location Validation Using CNN Satellite Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Hydrocarbon exploration systems face challenges in efficiently validating and correcting vast amounts of well data, particularly well location errors, which can compromise decision-making due to the unsuitability of rule-based methods for large data sets.
Innovation Solution
A hydrocarbon exploration system utilizing a convolutional neural network to analyze satellite images for well pad identification, coupled with a collaborative correction mechanism allowing users to propose correct well locations, and an administration system to finalize corrections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If rule-based methods are used for data validation, then the validation process is simple to implement, but the methods are unsuitable for large data sets and become inefficient
Solution Approach 1:
The patent replaces rule-based mechanical validation methods with machine learning models that automatically learn validation rules from data patterns. The system uses trained models to validate well data, substituting manual rule-definition with automated pattern recognition that scales efficiently to large datasets while maintaining ease of operation through the interface.
2Reliability
If machine learning models are used for data validation, then the accuracy and reliability of validation improves, but the system complexity increases
Solution Approach 1:
The patent introduces an intermediary layer consisting of pre-trained machine learning models that act as mediators between raw well data and validation rules. These models handle the complexity of pattern recognition internally, while presenting simplified validation results to users through an intuitive interface, thus maintaining reliability without exposing system complexity to end users.
Solution Approach 2:
The system performs preliminary training of machine learning models using historical well data before deployment. This preliminary action creates pre-trained models that encapsulate complex validation logic, allowing the operational system to use these models without needing to re-process the complexity of rule generation during actual validation operations.
3Manufacturing precision
If collaborative correction mechanisms are implemented, then the quality of corrected data improves through user input, but the time required for correction increases
Solution Approach 1:
The patent implements a feedback mechanism where users can review suggested corrections made by the machine learning model and provide input to confirm or modify them. The system learns from this feedback, continuously improving its correction accuracy. This feedback loop enables the system to achieve high correction quality while minimizing the time users need to spend, as the system handles most corrections automatically and only requires user intervention when necessary.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A hydrocarbon exploration system includes a plurality of computer nodes, each comprising a processor and memory coupled to the processor. The computer nodes are configured to implement a well identification system and a well location system. The well identification system is configured to analyze features of an image at a given set of geographic coordinates, and to determine, based on the features, whether a well is present at the geographic coordinates. The well location system is configured to, responsive to the well identification system determining that a well is not present at the geographic coordinates, render, at the geographic coordinates, a first indication of a potential incorrect location of the well in the image, and to render a second indication of a potential correct location of the well in the image.