ABSTRACT: Stochastic Simulation of Facies/Lithology for Reservoir Characterization

Vinicio Suro-Perez, Andre G. Journel

Reservoir performance prediction calls for a prior modeling of the porosity/permeability field as input to the flow simulator. The reservoir architecture, a geometric problem, should be defined prior to interpolation/simulation of tensorial and continuous variables such as permeability and porosity. Three-dimensional geometric interpolation of several intermingled facies is a difficult problem that cannot be fully automated. In the best case, such interpolation would result in one unique, overly homogeneous, geometric model when the reality of uncertain interpolation would call for several alternative reservoir images with their differences reflecting that uncertainty. The geological data at each sampled location x is coded as an indicator vector [EQUATION] of K mutually xclusive facies, where all indicator elements are zero except for the one I_k(x) = 1 corresponding to the facies k prevailing at x. These indicator data can stem from multiple information sources. The spatial autocorrelations and cross-correlations between the various facies indicators I_k(x), I_k^prime(x) characterize the relative geometry of these facies and are modeled by a limited number of principal component spatial covariances. Stochastic simulations from these principal component covariances allow generation of equiprobable reservoir geometries yet honor all facies indicator data. One or more of these reservoir architectures can be selected for further simulations of permeability/porosity within homogeneous facies. The geological integrity of the fin l numerical model(s) of permeability/porosity is thus preserved. A case study illustrates the proposed algorithm.

AAPG Search and Discovery Article #91003©1990 AAPG Annual Convention, San Francisco, California, June 3-6, 1990