Curvelet 4D Denoising Seismic Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional 4D seismic data processing methods struggle to effectively differentiate between actual changes in hydrocarbon reservoirs and noise, particularly due to differences in acquisition and processing parameters, leading to contamination of results with coherent and incoherent noise.
Innovation Solution
A computer-implemented method using a curvelet domain approach to denoise seismic datasets by formulating a common noise template from an initial 4D difference, cross-equalizing datasets, and iteratively updating the 4D difference until predetermined criteria are met, utilizing a complex-valued, directional, multi-resolution transform to separate signal from noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional subtraction method is used for 4D seismic difference, then processing simplicity is maintained, but signal-to-noise ratio deteriorates due to coherent and incoherent noise contamination
Solution Approach 1:
The patent segments noise from signal by transforming both baseline and monitor seismic datasets into the curvelet domain, where noise and signal occupy different regions in the transform space. This allows selective processing of noise components while preserving signal components, resolving the contradiction between simple processing and high signal-to-noise ratio.
Solution Approach 2:
The patent introduces a common noise template as an intermediary element that captures coherent noise patterns from the baseline and monitor datasets. This template is then used to attenuate noise in both datasets before computing the 4D difference, enabling effective noise removal while maintaining processing efficiency.
2Measurement precision
If iterative denoising process is applied to improve signal-to-noise ratio, then measurement precision improves, but processing time and complexity increase
Solution Approach 1:
The patent performs preliminary denoising by computing a common noise template from the baseline and monitor datasets before the iterative process. This pre-processing step captures the dominant noise patterns, allowing the subsequent iterative iterations to focus only on residual noise, thereby reducing total processing time while maintaining high signal-to-noise ratio.
Solution Approach 2:
The patent applies partial denoising by focusing computational effort on the most significant noise components identified in the curvelet domain. Rather than attempting to remove all noise components equally, the method targets the dominant coherent and incoherent noise patterns, achieving sufficient denoising with fewer iterations and reduced processing time.
3Reliability
If cross-equalization is applied to match baseline and monitor datasets, then reliability of 4D difference improves, but processing complexity increases
Solution Approach 1:
The patent replaces conventional time-domain cross-equalization methods with curvelet-domain processing. By transforming datasets into the curvelet domain, the method automatically handles phase and timing differences through the inherent multi-resolution and directional properties of the transform, simplifying the cross-equalization process while improving reliability.
Data Source
AI summary
A computer implemented method for denoising a set of seismic datasets, specifically belonging to different 3D subsets of a 4D survey the method including: (a) receiving a baseline and a monitor seismic dataset which were acquired by surveying over the same subsurface formation over different periods of time; (b) cross-equalizing the monitor seismic dataset to match to the baseline seismic dataset in terms of amplitude, frequency, phase and timing of events; (c) computing an initial 4D difference between the monitor and baseline seismic datasets; (d) formulating a common noise template from the initial 4D difference; (e) de-noising the baseline and monitor seismic datasets, independently, using the common noise template in a curvelet domain; (f) updating the initial 4D difference to form an updated 4D difference, which reflects de-noised baseline and monitor datasets from step (e); and iterating the steps (d) through (F) until the updated 4D difference satisfies a predetermined criteria.