Bayesian Seismic Inversion for Reservoir Property Uncertainty
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current seismic inversion techniques struggle with providing reliable, quantitative estimates of reservoir properties changes due to the inherent noise and uncertainties in time-lapse seismic data, lacking methods to assess uncertainties effectively.
Innovation Solution
A method that operates directly on seismic difference data, using a Bayesian setting to estimate changes in elastic material properties with uncertainty bounds, by combining prior knowledge with difference data, and employing a forward modeling operator to represent seismic data as a convolutional model, allowing for robust and quantitative statistical treatment of uncertainties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional seismic inversion techniques are used on time-lapse seismic data, then reservoir property changes can be estimated, but the results lack reliable uncertainty assessment due to inherent noise in the data
Solution Approach 1:
The patent transforms the inversion problem by changing the parameter representation from direct elastic property estimation to probability distribution estimation. By representing reservoir properties as probability distributions rather than deterministic values, the method simultaneously provides both point estimates and uncertainty quantification, resolving the contradiction between measurement precision and reliability assessment.
Solution Approach 2:
The patent introduces probability distributions as an intermediary between the noisy seismic difference data and the final reservoir property estimates. This intermediary layer allows the system to propagate uncertainty through the inversion process, providing both accurate estimates and reliable uncertainty assessment without requiring separate analysis steps.
2Loss of information
If seismic difference data is used to estimate reservoir changes, then quantitative estimates can be obtained, but the noise and uncertainties in the difference data reduce the reliability of the results
Solution Approach 1:
The patent converts the harmful effect of noise and uncertainty in seismic difference data into a beneficial feature by using Bayesian probability distributions. Instead of treating noise as an obstacle to be eliminated, the method incorporates uncertainty explicitly into the model, allowing the inversion to produce reliable quantitative estimates with associated confidence levels that reflect the actual data quality.
Solution Approach 2:
The patent applies prior probability distributions as a cushioning mechanism before the inversion process. These priors represent existing knowledge about reservoir properties and serve to stabilize the inversion against noisy difference data, ensuring that estimates remain physically reasonable even when the data quality is poor.
3Productivity
If separate inversions of baseline and repeat surveys are performed, then reservoir changes can be calculated by subtraction, but the uncertainty propagation between the two surveys is difficult to assess
Solution Approach 1:
The patent merges the baseline and repeat survey inversions into a single unified Bayesian inversion framework that processes seismic difference data directly. This combined approach maintains computational efficiency while automatically propagating uncertainty from both surveys through the inversion process, eliminating the need for separate uncertainty assessment steps required by the subtractive method.
Data Source
AI summary
A method of processing seismic data representing a physical system, the method comprising a difference between first and second seismic data representing the system in first and second states, respectively, and inverting the difference in accordance with a parameterized model of the physical system to obtain changes in the parameters of the model.