Subsurface Grid Cell Subdivision for Uncertainty Bias
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Subsurface modeling techniques often assume that facies boundaries coincide with grid boundaries, leading to biased representation of uncertainties, which can result in incorrect estimation of uncertainties in simulation outputs and history match actions.
Innovation Solution
The Conditional Property Interface Filter (CoPIF) method, which divides cells into sub-cells and attributes low order property values based on high order properties of adjacent cells, correcting for the bias in uncertainty representation by blurring the image statistically and improving the representation of low order properties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If facies boundaries are assumed to coincide with grid boundaries, then the modeling process is simplified and computation is easier, but the representation of uncertainties becomes biased and inaccurate
Solution Approach 1:
The method segments each grid cell into multiple sub-cells (e.g., 9 sub-cells in 2D, 27 sub-cells in 3D). By dividing cells into sub-cells, the model can represent facies boundaries that do not align with grid boundaries, allowing for more accurate uncertainty representation while maintaining computational feasibility through the structured subdivision approach.
Solution Approach 2:
Different sub-cells within the same grid cell are assigned different facies types based on their position and the probability distribution of facies boundaries. This local differentiation allows the model to capture the heterogeneity and uncertainty of facies boundaries at the sub-cell level, improving accuracy without requiring a complete remeshing of the entire domain.
2Reliability
If multiple realizations are used to capture reservoir uncertainty, then forecast uncertainties are computed, but the bias in facies boundary representation propagates through all realizations
Solution Approach 1:
The method performs preliminary action by establishing the correct sub-cell level facies assignment before generating multiple realizations. By correctly representing facies boundaries at the sub-cell level in the base model, all subsequent realizations inherit this improved structural representation, ensuring that uncertainty propagation occurs on an accurate geometric foundation rather than propagating grid-aligned biases through all simulations.
3Ease of manufacture
If grid geometry is certain while petrophysical properties are uncertain, then modeling is straightforward, but the method cannot handle cases where grid geometry itself is uncertain
Solution Approach 1:
The method adds a new dimension of analysis by introducing sub-cell level representation within the existing grid structure. This dimensional refinement allows the model to handle geometric uncertainty without abandoning the structured grid framework, enabling the representation of uncertain facies boundaries while maintaining the computational advantages of grid-based modeling.
Data Source
AI summary
Disclosed is a method of modelling a subsurface volume. The method comprises: discretizing said subsurface volume to define a grid comprising an array of cells (1, 2, 3), each cell being defined by cell faces, each cell having attributed thereto one or more high order properties (A,B); dividing each cell into a plurality of sub-cells, said plurality of sub-cells comprising: a plurality of central sub-cells adjacent a center of the cell, and a plurality of peripheral sub-cells, each adjacent a corresponding cell face; for each of said central sub-cells, and for each of said peripheral sub-cells for which the cell adjacent to that peripheral sub-cell has attributed thereto the same one or more high order properties as that of the cell in which said peripheral sub-cell is comprised, attributing one or more low order property values (c, d, e, x, y, z) in accordance with the high order property attributed to the cell in which said peripheral sub-cell is comprised; for each of said peripheral sub-cells for which the cell adjacent to that peripheral sub-cell has attributed thereto a different one or more high order properties than that of the cell in which said peripheral sub-cell is comprised, attributing one or more low order property values (c, d, e, x, y, z) in accordance with the high order property attributed to the adjacent cell; and determining one or more low order property values for each cell based on the one or more low order property values attributed to each cell's component sub-cells.