Machine Learning Prediction of Systems Tracts from Sea Level Curves
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Interpretation of systems tracts in well data is challenging due to spatial and temporal variations in sedimentary signals, which can lead to incorrect application of sequence stratigraphy, affecting subsurface modeling and decision-making in well operations.
Innovation Solution
A learning machine, such as a neural network, is trained using data from forward stratigraphic models to predict systems tracts based on eustatic and subsidence curves, enabling accurate placement of sequence stratigraphic surfaces and reducing uncertainty in subsurface models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sequence stratigraphy interpretation is applied to predict facies patterns and sediment properties, then the ability to model subsurface systems is improved, but the accuracy deteriorates due to spatial and temporal variations in sedimentary signals that can lead to incorrect interpretation
Solution Approach 1:
The patent applies preliminary action by training a machine learning model in advance using forward stratigraphic models and well data before actual prediction. The model learns the relationship between sea level curves, sediment supply, subsidence, and systems tract boundaries beforehand, so that when new data is input, the prediction can be made accurately without manual interpretation errors
Solution Approach 2:
The patent replaces the manual mechanical interpretation process with an automated machine learning system. Instead of relying on human experts to visually interpret well logs and seismic data to identify systems tract boundaries, a neural network or other ML algorithm automatically analyzes the data, eliminating subjective errors and improving consistency in interpretation
2Ease of manufacture
If manual interpretation of well data is used to identify systems tracts, then the process can be performed with existing tools, but the productivity deteriorates due to the time-consuming and challenging nature of interpreting spatial and temporal variations
Solution Approach 1:
The patent applies self-service by enabling the system to automatically interpret systems tracts without requiring continuous human intervention. The machine learning model, once trained, can independently analyze well logs, seismic data, and geological models to predict systems tract boundaries and facies patterns, freeing experts from repetitive manual interpretation tasks
Solution Approach 2:
The patent replaces the manual mechanical interpretation process with an automated machine learning system. Instead of relying on human experts to visually interpret well logs and seismic data to identify systems tract boundaries, a neural network or other ML algorithm automatically analyzes the data, eliminating subjective errors and improving consistency in interpretation
Data Source
AI summary
In some implementations, a method comprises generating a training dataset including a plurality of sample systems tracts each associated with a respective sample rate of change of subsidence and a respective sediment supply. The method also may comprise training a learning machine to indicate predicted systems tracts for wells based on the plurality of sample system tracts and their respective sample rate of change of subsidence and respective sample sediment supplies.