Proxy Surfaces for Reservoir History Matching Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for history matching in hydrocarbon reservoirs face challenges in computational efficiency due to high dimensionality and non-linear relationships, especially when dealing with produced fields where exhaustive search methods are prohibitively costly.
Innovation Solution
A method utilizing Genetic Algorithms combined with proxy surfaces, such as kriging, splines, and artificial neural networks, for global optimization in iterative steps to select and filter reservoir models, reducing the need for costly simulations and enhancing convergence by discarding unacceptably high misfit combinations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If exhaustive search methods are used to find models that reproduce historical data, then the accuracy of history matching is improved, but the computing cost becomes prohibitively high
Solution Approach 1:
The patent applies preliminary action by creating proxy surfaces (using kriging, splines, or neural networks) before performing the exhaustive search. These proxies are built from an initial set of simulation results and are used to guide the search, eliminating the need to run full simulations for every candidate model while maintaining history matching accuracy.
Solution Approach 2:
The patent introduces proxy surfaces as intermediary objects that mediate between the objective function evaluation and the optimization process. Instead of directly evaluating the expensive reservoir simulation for every model, the proxy surfaces provide approximate objective function values, dramatically reducing computing cost while preserving the ability to identify accurate history-matched models.
2Productivity
If traditional sampling strategies are used to build proxies, then the computational efficiency is improved, but the ability to identify multiple solutions degrades
Solution Approach 1:
The patent applies dynamics by making the sampling strategy adaptive rather than static. The optimization algorithm dynamically adjusts the sampling points based on the proxy surface quality and the search progress, allowing it to efficiently explore the parameter space and identify multiple distinct solutions while maintaining computational efficiency.
Solution Approach 2:
The patent implements feedback by using the results from proxy evaluations to guide subsequent sampling decisions. The optimization algorithm learns from previous evaluations and adjusts its sampling strategy accordingly, ensuring that computational resources are focused on regions that contribute to finding diverse solutions rather than redundant evaluations.
3Speed
If gradient based algorithms are used for history matching, then the convergence speed is improved, but the performance degrades with problem size and depends strongly on initial guess
Solution Approach 1:
The patent applies copying by replacing the expensive reservoir simulation with a simplified proxy model that replicates the essential behavior. This proxy model can be evaluated much faster and allows gradient-based algorithms to converge quickly even for high-dimensional problems, while the proxy accurately reflects the relationship between parameters and production responses.
Data Source
AI summary
A method for forecasting production from a hydrocarbon producing reservoir, the method includes defining an objective function and characteristics of a history-matched model of a reservoir and acceptable error E. At least one geological realization of the reservoir is created representing a probable geological setting. For each geological realization, a global optimization technique is used to perform history matching in a series of iterative steps to obtain acceptable models. Production of the reservoir is forecasted based upon simulation runs of the respective models.