Unsupervised Clustering Inversion for Hydrocarbon Reservoir Location
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining the location of hydrocarbon reservoirs in subsurface formations are inefficient due to the inability to directly measure subsurface properties with sufficient resolution, leading to ill-posed inverse problems with non-unique solutions that often violate physical constraints.
Innovation Solution
A method involving iterative clustering regularization inversion, where a parameter vector representing subsurface properties is updated using a composite objective function that includes a data misfit function and a clustering term, guided by a membership matrix and cluster centers, until convergence, to accurately determine the location of hydrocarbon reservoirs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If inverse problems are solved using conventional methods, then solutions can be obtained, but the solutions are non-unique and often violate physical constraints
Solution Approach 1:
The patent transforms the inversion problem from a conventional approach to a clustering-based approach by changing the parameter representation. Instead of directly solving for subsurface properties, the method uses cluster centers and membership matrices to represent the parameter space, which inherently enforces physical constraints and produces unique, reliable solutions.
Solution Approach 2:
The patent introduces cluster centers and membership matrices as intermediary elements between the observed data and the subsurface property distribution. These intermediaries serve as regularizers that guide the inversion process toward physically consistent solutions while maintaining mathematical uniqueness.
2Measurement precision
If subsurface properties are directly measured, then high resolution data can be obtained, but measurements cannot be made throughout the entire subsurface volume
Solution Approach 1:
The patent segments the subsurface volume into discrete clusters, where each cluster represents a region with homogeneous properties. This segmentation allows the inversion method to achieve high resolution by characterizing each cluster center accurately, while the membership matrix extends the solution coverage to the entire subsurface volume by indicating the degree of membership of each location to each cluster.
3Productivity
If conventional inversion methods are used, then computational processing can be performed, but the methods are inefficient and produce inaccurate reservoir location determination
Solution Approach 1:
The patent changes the parameter space from continuous subsurface properties to discrete cluster centers with associated membership values. This transformation simplifies the computational problem by reducing the dimensionality of the search space, thereby improving processing efficiency while simultaneously enhancing the accuracy of reservoir location determination through the regularizing effect of clustering.
Data Source
AI summary
A method for determining a location of a hydrocarbon reservoir. The method may include receiving a cluster number and a parameter vector, wherein the parameter vector represents a spatial distribution of a property over a subsurface. The method may further include receiving observed data and iteratively performing a series of steps until the parameter vector is converged. The steps may include determining a cluster center for each cluster in a plurality of clusters and determining a membership matrix. The steps may further include processing the parameter vector with a forward operator to produce predicted data and determining an update parameter vector guided by a composite objective function composed of a data misfit function and a clustering term. The steps may still further include updating the parameter vector with the update parameter vector. Once converged the parameter vector is used to determine the location of the hydrocarbon reservoir.