Reservoir Characterization Using ReSampled Seismic Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional petrophysical inversion methods using seismic data often result in inaccurate layer thickness and petrophysical property estimates due to band-limitation issues, which limit the resolution of reservoir characterization in hydrocarbon management.
Innovation Solution
A one-step inversion approach is employed, utilizing an iterative method that alternates between optimization and learning steps, with a prior model updated based on learned information to improve high-frequency representation and resolution, and regularization techniques are applied to enhance layer thickness precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional two-step inversion methods are used, then computational complexity is reduced, but measurement precision of layer thickness and porosity deteriorates due to band-limitation issues
Solution Approach 1:
The patent combines the optimization step and learning step into a single integrated inversion process. The optimization step updates the prior model based on seismic data misfit, while the learning step updates the prior model based on learned sub-seismic information, and these steps are performed iteratively within one unified framework rather than as separate sequential processes.
Solution Approach 2:
The patent performs preliminary resampling of seismic data to a finer grid before inversion to recover sub-seismic information. This pre-processing step prepares the data at higher resolution before the inversion process begins, enabling the subsequent optimization and learning steps to operate on enhanced input data that contains frequency information beyond the original seismic bandwidth.
2Measurement precision
If seismic data is resampled to finer grid, then measurement precision of petrophysical properties improves, but productivity decreases due to increased computational requirements
Solution Approach 1:
The patent applies partial refinement by resampling seismic data to an intermediate finer grid rather than the maximum possible resolution. This intermediate resolution provides sufficient sub-seismic information to improve petrophysical property predictions while avoiding the excessive computational burden that would result from using the finest possible grid, thus achieving an optimal balance between precision and productivity.
Solution Approach 2:
The patent applies different processing resolutions to different aspects of the inversion process. The seismic data is resampled to a finer grid for the optimization step to capture sub-seismic information, while the learning step operates on the learned representations at appropriate resolutions. This local differentiation of quality allows the system to achieve high precision where needed without uniformly increasing computational requirements across all processing stages.
3Manufacturing precision
If prior model is updated iteratively with learned information, then manufacturing precision of reservoir characterization improves, but loss of time increases due to multiple iteration cycles
Solution Approach 1:
The patent maintains continuous useful action by ensuring that each iteration cycle performs both optimization and learning steps that contribute to improving the prior model. Rather than performing redundant or non-contributory processing in each iteration, the method ensures that every cycle advances the reservoir model accuracy through coordinated updates based on both seismic data misfit and learned sub-seismic information, making the extended processing time productive.
Solution Approach 2:
The patent implements feedback mechanisms where the prior model is updated based on feedback from both the optimization step (seismic data misfit) and the learning step (learned sub-seismic information). This dual-feedback approach ensures that each iteration refines the model using information from multiple sources, improving convergence efficiency and reducing the total number of iterations needed to achieve high-precision reservoir characterization compared to methods with single feedback loops.
Data Source
AI summary
A method and apparatus for generating an image of a subsurface region including obtaining geophysical data/properties for the subsurface region; resampling the geophysical data/properties to generate a resampled data set; iteratively (a) inverting the resampled data set with an initial prior model to generate a new model; and (b) updating the new model based on learned information to generate an updated prior model; substituting the initial prior model in each iteration with the updated prior model from an immediately-preceding iteration; and determining an end point for the iteration. A final updated model may thereby be obtained, which may be used in managing hydrocarbons. Inversion may be based upon linear physics for the first one or more iterations, while subsequent iterations may be based upon non-linear physics.