Summary
A technique for the sequential generation of perpendicular-bisectional
(PEBI) grids adapted to flow information is presented and applied. The
procedure includes a fine-scale flow solution, the generation of an initial
streamline–isopotential grid, grid optimization, and upscaling. The grid
optimization is accomplished through application of a hybrid procedure with
gradient and Laplacian smoothing steps, while the upscaling is based on a
global-local procedure that makes use of the global solution used in the
grid-determination step. The overall procedure is successfully applied to a
complex channelized reservoir model involving changing well conditions. The
gridding and upscaling procedures presented here may also be suitable for use
with other types of structured or unstructured grid systems.
Introduction
Modern geological and geostatistical tools provide highly detailed
descriptions of the spatial variation of reservoir properties, resulting in
fine-grid models consisting of 107 to 108 gridblocks. As
a consequence of this high level of detail, these models cannot be used
directly in numerical reservoir simulators, but need to be coarsened
significantly. Coarsening requires the averaging of rock parameters from the
fine scale to the coarse scale. This process is referred to as upscaling. For
simulation of flow in porous media, the upscaling of permeability is of
particular interest. A large body of literature exists on this topic; for a
comprehensive review of existing techniques, see Durlofsky (2005).
To preserve as much of the geological information of the fine grid as
possible, the grid coarsening should not be performed uniformly, but with more
refinement in areas that are expected to have large impact on the flow,
including structural features, such as faults. Although grid-generation
techniques based on purely static, nonflow-based considerations have been shown
to produce reasonable results(Garcia et al. 1992), the application of
flow-based grids is often preferable. Flow-based grids require the solution of
some type of fine-scale problem. They are then constructed by exploiting the
information obtained from streamlines (and possibly isopotentials) either
directly or indirectly. Depending on the type of grid used, points will be
defined as cell vertices or nodes, resulting in either a corner-point geometry
or point-distributed grid. Several gridding techniques for reservoir simulation
have been introduced along these lines, as we now discuss.
© 2006. Society of Petroleum Engineers
View full textPDF
(
2,086 KB
)
History
- Original manuscript received:
23 March 2005
- Revised manuscript received:
7 March 2006
- Manuscript approved:
31 March 2006
- Version of record:
20 September 2006
View Cart
