3D Seismic Depth Conversion Using Neural Network Velocity Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for converting 3D seismic data from a time domain to a depth domain are not 100% accurate, leading to loss of resolution and detail in reservoir characterization studies, particularly for seismic attributes, which are often interpreted in the time domain and rarely converted to the depth domain effectively.
Innovation Solution
The method involves training a multi-perception back-propagation artificial neural network to predict sonic logs from well log data, allowing for high-resolution conversion of 3D seismic data and attributes from a time domain to a depth domain, incorporating steps like image-ray correction, well log data processing, and interpolation to create a detailed time-depth model, and transferring seismic attributes to a structurally correct depth volume.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional velocity model methods (check-shot surveys, tomography, acoustic inversion) are used for depth conversion, then the conversion process is computationally feasible and can be implemented with existing tools, but the velocity model becomes too simple to maintain high resolution in reservoir intervals, causing smoothing or loss of detail
Solution Approach 1:
The patent segments the velocity model construction into multiple components: well log-derived velocities, seismic stacking velocities, and tomographic velocities. Each component addresses specific spatial scales and geological features, with well log data providing high-resolution interval velocities for reservoir intervals, stacking velocities providing regional framework, and tomographic velocities providing structural context. This segmentation allows the final depth-converted volume to maintain high resolution in reservoir intervals while remaining computationally feasible.
Solution Approach 2:
The patent applies local quality by using different velocity modeling approaches for different spatial locations and depth ranges. In reservoir intervals where high resolution is critical, well log-derived velocities are used to maintain detailed stratigraphic features. In overburden and regional contexts, simpler tomographic and stacking velocity models are sufficient. This localized application of different model complexities optimizes both accuracy and computational efficiency.
2Manufacturing precision
If high-resolution velocity models are used to maintain detail in reservoir intervals, then depth conversion accuracy improves, but the computational complexity and data processing requirements increase significantly
Solution Approach 1:
The system segments the seismic volume into different depth intervals and processing zones. Well log data are used specifically for reservoir interval depth conversion where high resolution is needed, while overburden conversion uses simpler velocity models. This segmentation reduces overall computational complexity by applying complex processing only where necessary.
Solution Approach 2:
The patent performs preliminary depth conversion using a baseline velocity model to create an initial depth framework. Then, high-resolution well log-derived velocities are applied in a second pass to refine reservoir intervals. This preliminary action approach allows the system to benefit from high-resolution processing without requiring all data to be processed at maximum complexity simultaneously.
3Adaptability or versatility
If seismic attributes are converted to depth domain along with seismic data, then the attributes provide value in depth domain for reservoir interpretation, but the attributes suffer from the same resolution losses as the seismic data, and time-based attributes become difficult to interpret
Solution Approach 1:
The patent inverts the conventional approach by computing seismic attributes in the time domain first, then transferring them to the depth domain using the high-resolution velocity model. This allows attributes to be calculated where high temporal resolution is preserved, then mapped to depth coordinates using the detailed velocity information. The result is depth-domain attributes that retain the sharpness and interpretability of time-domain calculations while providing depth context for reservoir interpretation.
Data Source
Figure 1
Figure 2
AI summary
The present invention meets the above needs and overcomes one or more deficiencies in the prior art by providing systems and methods for the conversion of stacked, or preferably, time migrated 3D seismic data and associated seismic attributes from a time domain to a depth domain. In one embodiment, the present invention includes a method for convening threedimensional seismic data from a time domain to a depth domain.