Deep Learning Stripping of Strong Reflection Layers in Seismic Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for stripping strong reflection layers in geophysical exploration for petroleum reservoirs, such as spectral decomposition, matching pursuit, and multi-wavelet decomposition, face limitations like poor lateral continuity and lack of quantitative frequency selection, leading to inaccurate reservoir predictions.
Innovation Solution
A deep learning method is employed to establish mapping relationships between seismic wavelets and data, using alternate iterative deep neural networks and U-Net to accurately predict and remove strong reflection layers, thereby enhancing the extraction of weak reflection signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If spectral decomposition is used to reduce strong reflection energy, then the strong reflection energy is reduced, but there is no quantitative standard for frequency selection which may cause pitfalls in reservoir prediction
Solution Approach 1:
The patent transforms the frequency selection from a manual parameter-setting process to an automated learning process. The neural network dynamically learns and adjusts the frequency parameters needed for decomposition based on training data, eliminating the need for empirical frequency selection while maintaining prediction accuracy.
Solution Approach 2:
The system performs self-learning through the neural network that automatically determines the appropriate frequency parameters for spectral decomposition. The model learns from training data what frequency ranges correspond to strong reflections and weak reflections, enabling autonomous parameter selection without human intervention.
2Object-affected harmful factors
If matching pursuit algorithm is used to strip strong reflection, then the strong reflection is removed, but the algorithm has strong sparsity causing poor lateral continuity and reduced lateral accuracy
Solution Approach 1:
The patent replaces the traditional matching pursuit algorithm with a deep learning-based spectral decomposition approach. Instead of using iterative sparse decomposition with fixed dictionaries, the neural network directly learns the spectral characteristics and performs decomposition in a more stable manner that preserves lateral continuity while removing strong reflections.
3Object-affected harmful factors
If multi-wavelet decomposition is used to remove strong reflection signals, then the strong reflection is stripped, but the traditional seismic trace model based on single fixed wavelet struggles with limitations
Solution Approach 1:
The patent transitions from static fixed wavelet models to dynamic adaptive wavelet modeling. The neural network learns wavelet parameters dynamically from data, allowing the wavelet characteristics to adapt to different geological conditions and signal types, thereby improving both strong reflection removal and preserving weak reflection signals.
Data Source
AI summary
Disclosed herein is a method of stripping a strong reflection layer based on deep learning. The method establishes a direct mapping relationship between a strong reflection signal and seismic data of a target work area through a nonlinear mapping function of the deep neural network, and strips a strong reflection layer after the strong layer is accurately predicted. A mapping relationship between the seismic data containing the strong reflection layer and an event of the strong reflection layer is directedly found through training parameters. In addition, this method does not require an empirical parameter adjustment, and only needs to prepare a training sample that meets the actual conditions of the target work area according to the described rules.