DAS Microseismic Event Picking for Low-SNR Fracture Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional DAS systems for monitoring hydraulic fracture stimulations in oil and gas wells face challenges with low signal-to-noise ratio (SNR) and require significant processing time, limiting real-time data analysis and control capabilities.
Innovation Solution
An improved method that reduces data volume through preprocessing and utilizes separately trained CNNs with machine learning to identify microseismic events in real-time, employing a cascading workflow of ML models to enhance signal processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional CNN processing is used to extract microseismic signals from DAS data, then signal extraction capability is achieved, but processing speed is slow and analysis time is excessive
Solution Approach 1:
The patent segments the microseismic signal processing into distinct components: correlation template matching for event detection, pick picking for arrival time identification, and phase identification for classification. This segmentation allows each component to be optimized independently, significantly reducing overall processing time while maintaining accuracy.
Solution Approach 2:
The patent implements preliminary action by pre-computing correlation templates from noise-free signal segments and storing them for rapid matching during actual processing. This pre-processing step eliminates the need for complex real-time computations, enabling fast signal extraction without sacrificing detection accuracy.
2Measurement precision
If conventional CNN processing is used to analyze DAS data, then microseismic events can be detected, but the first arrival picking accuracy is insufficient
Solution Approach 1:
The patent replaces conventional mechanical CNN processing with a correlation-based approach that uses template matching against pre-identified noise-free signal segments. This substitution enables precise first arrival picking by comparing incoming signals against known event templates, achieving superior accuracy without the computational overhead of iterative CNN processing.
3Measurement precision
If DAS systems are used for hydraulic fracture monitoring, then spatial resolution of a few meters is achieved, but signal-to-noise ratio remains low
Solution Approach 1:
The patent merges multiple signal processing techniques including correlation matching, pick picking, and phase identification into a unified workflow. This combination allows the system to leverage the strengths of each method: correlation for robust event detection, pick picking for precise timing, and phase identification for accurate classification, thereby improving overall signal reliability despite low input SNR.
Solution Approach 2:
The patent creates idealized copies of microseismic signals by extracting noise-free segments from the DAS data and using them as correlation templates. These template copies serve as reference patterns for detecting and characterizing actual microseismic events, enabling reliable signal extraction even when the original signals are buried in noise.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables real-time processing of microseismic data, reducing analysis time by 1000-fold and improving the accuracy of fracture monitoring, allowing for immediate operational adjustments during hydraulic fracturing.
Implementation Method 1
DAS systems use various techniques, e.g., detecting Rayleigh scattering, and often are interferometric in nature
Data Source
AI summary
Systems and methods for operating a distributed acoustic sensing (DAS) system are disclosed that process the DAS signal by downsampling the received signal and stacking the channels, generating a plurality of sliding windows of the processed signal, analyzing the windows to either identify a microseismic event or determine that the window contains only noise, then discarding all noise windows. A convolutional neural network is used to determine an onset time and a peak channel of each microseismic event and to reduce the dimensionality of the data in time and space around the onset time. A convolutional neural network is used to identify a first arrival pick within the truncated window of all recorded phases, which are then used to determine the physical location of the source fracture.