Microseismic Waveform Processing via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional reservoir imaging techniques are cumbersome, require highly skilled engineers, and are limited in their ability to efficiently denoise data, leading to inaccurate results and increased costs in hydrocarbon recovery operations.
Innovation Solution
A method using machine learning algorithms to process microseismic waveform data, involving data training, dictionary creation, denoising, and probability function determination to enhance data accuracy and speed of analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional reservoir imaging techniques are used, then data can be evaluated with existing methods, but the process is cumbersome and requires highly skilled engineers
Solution Approach 1:
The patent replaces manual mechanical analysis methods with automated machine learning algorithms. The system uses trained neural networks to automatically process microseismic waveform data, substituting the need for highly skilled engineers with automated computational systems that perform denoising, event detection, and characterization without human intervention.
Solution Approach 2:
The machine learning model performs self-service by automatically denoising data, detecting events, and characterizing geological features without requiring external expert intervention. The system trains on labeled data and then independently processes new field data, making autonomous decisions about data quality and event significance.
2Measurement precision
If conventional denoising techniques are applied to field data, then some noise can be removed, but the techniques are limited in effectiveness and time-consuming
Solution Approach 1:
The system performs preliminary action by pre-training machine learning models on extensive labeled datasets before deployment. The models are pre-conditioned to recognize patterns and distinguish signal from noise, enabling rapid processing of field data without requiring time-consuming manual analysis or iterative denoising attempts.
Solution Approach 2:
The patent applies parameter changes by transforming the denoising problem from traditional signal processing parameters to machine learning parameter spaces. The system uses learned parameters from training data to dynamically adjust denoising behavior, allowing effective noise removal across varying data conditions without manual parameter tuning.
3Loss of information
If more information is gathered about geological features, then better decisions can be made about hydrocarbon field development, but conventional methods are too slow to provide real-time analysis
Solution Approach 1:
The patent replaces slow conventional analysis methods with fast machine learning-based processing. The system automatically extracts multiple geological features (event locations, magnitudes, frequencies, spatial distributions) simultaneously through automated pipeline processing, delivering comprehensive information at speeds impossible with manual analysis.
Solution Approach 2:
The system enables continuous processing of incoming microseismic data streams without interruption. The machine learning model continuously analyzes new data as it arrives, providing uninterrupted flow of geological information that supports real-time decision-making throughout the monitoring period.
Data Source
AI summary
Embodiments presented provide for a method for performing waveform processing. In one embodiment, a synthetic dictionary is created and then, using a machine learning process, data is processed to produce a result.