Multi-level Reservoir Simulation Solver for Giant Grids
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer simulations of giant subsurface reservoirs are computationally intensive and time-consuming due to the linear solution of large-scale systems of equations, especially when using fine-grid models, and even petascale computing systems struggle with inaccurate initial guesses, leading to wasted iterations.
Innovation Solution
The method involves partitioning the reservoir model into a coarse cell grid, performing preconditioned conjugate-gradient extrapolation, and transforming results back to the original fine grid to achieve an iterative linear solution, utilizing parallel processing and interpolation techniques to refine initial guesses efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fine-grid models are used to represent subsurface reservoirs, then simulation precision is improved, but computational time and complexity increase significantly
Solution Approach 1:
The patent applies segmentation by dividing the fine-grid reservoir model into multiple coarser grid levels. The fine-grid model is systematically coarsened through multiple refinement levels, allowing the simulation to be solved on a hierarchy of grids. This segmentation enables the use of computational resources more efficiently by performing detailed simulations only where necessary while using coarser grids for broader regions, thereby reducing overall computational time while maintaining precision.
Solution Approach 2:
The patent implements preliminary action by using coarser grid models to obtain initial solution estimates before performing detailed simulations on fine-grid models. The coarser grid simulations serve as preliminary computations that provide starting conditions for the fine-grid simulations, reducing the number of iterations needed and thereby decreasing total computational time while maintaining simulation precision.
2Measurement precision
If iterative linear solution methods are used for large-scale systems, then solution accuracy is improved, but the number of iterations and computational cost increase
Solution Approach 1:
The patent applies preliminary action by using coarser grid models to compute initial solution estimates that serve as starting points for fine-grid simulations. These preliminary solutions from coarser grids reduce the number of iterations required for convergence on fine-grid models, thereby improving iteration efficiency while maintaining solution accuracy.
Solution Approach 2:
The patent uses copying by transferring solution estimates from coarser grid models to fine-grid models. The solution fields obtained from coarser grids are copied and interpolated onto the fine-grid model as initial conditions, allowing the fine-grid simulation to start closer to the final solution and reduce the number of iterations needed, thus improving productivity without sacrificing accuracy.
3Power
If petascale computing systems are used to solve large linear systems, then computational power is increased, but inaccurate initial guesses lead to wasted iterations
Solution Approach 1:
The patent applies preliminary action by using coarser grid models to generate accurate initial guesses before starting fine-grid simulations on petascale systems. These preliminary solutions ensure that the iterative linear solvers start from good initial conditions, significantly reducing the number of wasted iterations and improving the efficiency of petascale computational resources.
Solution Approach 2:
The patent uses copying by transferring solution estimates from coarser grids to fine-grids on petascale computing systems. This copying of solution fields as initial conditions ensures that the high-performance computing resources are utilized efficiently by avoiding wasted iterations on poor initial guesses, thereby reducing time loss while maintaining high computational power.
Data Source
AI summary
In solving linear systems of billions of equations resulting from reservoir simulator pressure equations for giant reservoirs organized into fine-grid models, a multi-level methodology acquires a fast and coarser grid solution to be used as a more accurate initial estimate for the fine grid problem. Solver iterations performed on the coarser grid in reduced processing time generate a better initial estimate, resulting in fewer of the more expensive iterations at the fine-grid level.