Machine Learning Reservoir Characterization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inaccurate or unreliable subterranean formation models due to limited well data hinder effective hydrocarbon well placement decisions, leading to uncertainties in exploration operations.
Innovation Solution
Application of machine-learning techniques to generate a composite seismic parameter from seismic attributes, using trained models to classify and analyze seismic data, which helps identify optimal well locations by integrating multiple seismic attributes and reducing uncertainty in reservoir characterization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional modeling methods are used with limited well data, then the modeling process is simple and fast, but the model accuracy and reliability are poor
Solution Approach 1:
The patent introduces machine learning algorithms as an intermediary between limited well data and formation model generation. The ML models process seismic data and integrate it with well data to create more accurate formation models, acting as a mediator that extracts additional information from available data without requiring extensive well data
Solution Approach 2:
The patent replaces traditional deterministic modeling methods with probabilistic machine learning approaches. Instead of using conventional geological modeling techniques that rely heavily on well data, the system uses ML algorithms that can handle uncertainty and generate multiple possible formation scenarios, improving model accuracy while managing complexity through computational methods
2Reliability
If more well data is collected to improve model reliability, then the model becomes more accurate, but the exploration cost and time increase
Solution Approach 1:
The patent uses machine learning to create virtual representations of subsurface formations based on patterns learned from existing well data and seismic data. These ML-generated formation models serve as copies or proxies that provide reliable predictions without requiring additional physical wells to be drilled for data collection, thus maintaining reliability while avoiding time loss
Solution Approach 2:
The patent performs preliminary machine learning model training and validation using available well data before actual exploration decisions are made. This preliminary action creates a predictive framework that can be applied immediately to new seismic data, providing reliable model predictions without the need for time-consuming additional well drilling and data collection
3Measurement precision
If machine-learning techniques are applied to generate composite seismic parameters, then the reservoir characterization accuracy improves, but the computational complexity and processing time increase
Solution Approach 1:
The patent segments the complex reservoir characterization process into distinct machine learning stages: seismic data preprocessing, feature extraction, ML model training, and formation model generation. Each segment handles a specific aspect of the problem, allowing the system to achieve high characterization accuracy while managing computational complexity through modular processing
Solution Approach 2:
The patent transforms raw seismic data into composite seismic parameters through machine learning processing. The ML algorithms change the parameter representation from conventional seismic attributes to enhanced composite parameters that better characterize reservoir properties, improving accuracy while the computational complexity is managed through efficient parameter transformation techniques
Data Source
AI summary
A system can determine a location for future wells using machine-learning techniques. The system can receive seismic data about a subterranean formation and may determine a set of seismic attributes from the seismic data. The system can block the set of seismic attributes into a set of blocked seismic attributes by distributing the set of seismic attributes onto a geo-cellular grid representative of the subterranean formation. The system can apply a trained machine-learning model to the set of blocked seismic attributes to generate a composite seismic parameter. The system can distribute the composite seismic parameter in the subterranean formation to characterize formation locations based on a predicted presence of hydrocarbons.