Stochastic Reservoir Parameter Estimation via Joint Inversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional deterministic inversion methods for estimating reservoir parameters are local and provide limited uncertainty information, relying on initial values and lacking global robustness.
Innovation Solution
A stochastic method using a Bayesian framework with Markov random fields to jointly invert multiple types of geophysical data, incorporating categorical variables and hierarchical models to estimate reservoir parameters and their uncertainties, independent of initial values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If deterministic inversion methods are used, then computational efficiency is improved, but global robustness and uncertainty information deteriorate
Solution Approach 1:
The patent replaces deterministic mechanical optimization methods with a stochastic probabilistic framework. Instead of using deterministic algorithms like Gauss-Newton or conjugate gradient that converge to local minima, the invention employs probabilistic sampling methods to explore the entire parameter space, thereby achieving global robustness while maintaining computational feasibility through efficient sampling strategies.
Solution Approach 2:
The invention transforms the inversion problem from a deterministic parameter optimization task into a probabilistic parameter estimation problem. By changing from fixed initial values to probability distributions and from deterministic updates to stochastic sampling, the method achieves global robustness while managing computational complexity through efficient probabilistic algorithms.
2Device complexity
If deterministic inversion methods are used, then computational simplicity is improved, but uncertainty information deteriorates
Solution Approach 1:
The patent segments the inversion problem into distinct probabilistic components: prior distributions for model parameters, likelihood functions for data fitting, and posterior distributions for results. This segmentation allows each component to be handled separately with appropriate computational methods, managing complexity while comprehensively capturing uncertainty information throughout the inversion process.
Solution Approach 2:
The invention introduces probability distributions as intermediary representations between the input data and the final model parameters. These distributions serve as mediators that carry and propagate uncertainty information through the inversion process, allowing comprehensive uncertainty characterization without excessive computational complexity.
3Reliability
If stochastic inversion methods are used, then global robustness and uncertainty information are improved, but computational complexity increases
Solution Approach 1:
The patent applies partial sampling strategies where instead of exhaustively sampling the entire parameter space, it uses carefully selected samples that capture the essential features of the posterior distribution. This partial action approach achieves global robustness through targeted stochastic sampling while keeping computational complexity manageable through efficient sampling algorithms.
4Loss of information
If stochastic inversion methods are used, then uncertainty information is improved, but computational time increases
Solution Approach 1:
The invention uses partial sampling approaches where a limited number of strategically selected samples are used to characterize the posterior distribution. This allows comprehensive uncertainty information to be obtained through efficient sampling methods that converge faster than exhaustive approaches, reducing computational time while maintaining thorough uncertainty characterization.
Data Source
AI summary
A system and a method for estimating a reservoir parameter are provided. The method includes calculating a plurality of priors using a Markov random field, the plurality of priors comprising probability distributions of a plurality of litho-types; calculating posterior distributions based on the priors, the posterior distribution depending upon measured geophysical data, geophysical attributes and reservoir parameters; and determining at least a portion of litho-types in the plurality of litho-types that correlate most with the measured geophysical data.