Fracture Network Modeling With AI for Real-Time Reservoir Characterization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques for characterizing sub-surface fracture networks in rock formations provide data that requires further interpretation and represent a past state, failing to account for dynamic changes, leading to inaccurate modeling that can be financially and environmentally costly.
Innovation Solution
A bipartite technique using unsupervised learning and artificial intelligence to identify clusters of temporospatial points from sensor data, applying an AI algorithm trained on templates of fracture network topologies for real-time characterization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If traditional sensor techniques are used to characterize fracture networks, then data can be collected from the subsurface, but the data requires further interpretation and represents only a past state rather than real-time conditions
Solution Approach 1:
The system transitions from static, historical fracture network characterization to dynamic, real-time monitoring by continuously collecting sensor data and updating the fracture network model as new data becomes available, allowing the system to adapt to changing subsurface conditions
Solution Approach 2:
The system implements a feedback loop where sensor data is continuously collected, processed through machine learning algorithms, and used to update the fracture network model in real-time, which then informs subsequent drilling or extraction decisions
2Measurement precision
If fracture network models are updated in real-time using sensor data and machine learning, then accuracy and precision of modeling improve, but system complexity and computational requirements increase
Solution Approach 1:
The system divides the complex task of fracture network characterization into distinct modules: sensor data collection, unsupervised learning for pattern recognition, supervised learning for classification, and model updating, allowing each component to be optimized independently
Solution Approach 2:
Machine learning algorithms serve as intermediaries that automatically process sensor data and extract meaningful fracture network characteristics, reducing the need for manual interpretation and simplifying the overall system operation
3Productivity
If inaccurate fracture network models are used for drilling location selection, then drilling operations can proceed with existing data, but financial and environmental costs increase due to incorrect drilling locations
Solution Approach 1:
The system performs preliminary characterization of the fracture network using sensor data and machine learning before drilling operations begin, allowing optimal drilling locations to be identified in advance and avoiding costly mistakes during actual drilling
Data Source
AI summary
Accurately and precisely characterizing a fracture network in a material, such as rocks of a rock formation, in real-time is challenging. However, an accurate and precise characterization can be generated by receiving sensor data from sensors deployed within the material, on the material, or both. The sensor data is converted into processed data from which cluster data is derived. The cluster data includes information corresponding to a geometry and an internal structure of at least a portion of the fracture network. One or more models of the fracture network is generated based on the cluster data. The one or more models are evaluated based on simulations.