Static and Dynamic Evaluation of Stochastic Reservoir Model Workflows with Varying Amounts of Geological Detail – Eunice Monument South Unit (EMSU), New Mexico

W. Scott Meddaugh
ChevronTexaco Exploration and Production Technology Company, Bellaire, TX

Earth models generated for use in fluid flow simulation may incorporate varying amounts of geological detail. Generally, as more geological detail is included the project costs increase, sometimes dramatically when the cost includes “special” data acquisition or interpretation outside the scope of normal Opco activities. The increased costs associated with a detailed earth model are likely worthwhile if the uncertainly of reservoir performance predictions is significantly reduced. Earth models for the EMSU reservoir were generated using three workflows with increasing amounts of geological detail as summarized below:

Simple – Structural and stratigraphic framework based on a gridded surface for the top of Grayburg Fm and the well picks for the Grayburg and four stratigraphic horizons between the Grayburg Fm. and Penrose Fm. The model was populated with porosity via sequential Gaussian simulation (SGS) algorithm using the layer appropriate porosity data and semivariograms. Fifteen porosity realizations were generated. Permeability was added using a transform equation.

Facies – Same framework as the simple workflow model but with a stochastic distribution of shoal and lagoon facies using a sequential indicator simulation (SIS) based algorithm constrained by layer facies maps and core data. Porosity was first distributed using the SGS constrained by facies using well log data from cored wells only and stratigraphic layer. The porosity distribution was modified slightly by collocated cokriging with SGS using all wells with porosity logs (cored well porosity distribution as the secondary data, correlation coefficient of 0.99). Permeability was added using the facies-specific transforms.

Lithology-based – Using the same framework and stochastic facies distribution as the facies-based workflow plus a stochastic distribution of seven lithology-types (mudstone, wackestone, mud-rich packstone, mud-poor packstone, grainstone, rudstone, and sandstone) constrained by cored wells and stratigraphic layer. Porosity was added using the SGS algorithm constrained by lithology using well log data from cored wells only and stratigraphic layer and modified by collocated cokriging with SGS as noted above. Permeability was added using lithology-specific transforms.

The realizations were up-scaled from 400 layers to 18 layers and flow characteristics evaluated using a 3D streamline-based flow simulator. Surprisingly little difference exists between the fluid flow characteristics of earth models generated by the three different modeling workflows. Mean recovery factor (and standard deviation) for the three workflows are as follows: Simple = 0.311 (0.006), Facies-based = 0.294 (0.010), and Lithology-based = 0.315 (0.008). There is, however, a large spread in fluid flow results obtained from individual realizations as shown by the water injection rate which varied by a factor of three across the 15 realizations.

Vertical scale-up was shown to have little or no effect on fluid flow characteristics over the range studied (400 layers to 9 layers). Areal scale-up, however, has a significant effect on fluid flow characteristics. As areal scale-up increases reservoir flow become significantly more optimistic (e.g. less water injected for same oil recovery).

Evaluation of the models using simple statistical measures of static properties (mean permeability, permeability thickness, OOIP etc.) showed inconsistent correlation with fluid flow simulation results. A easy and rapidly calculated static model connectivity measure based on number geobodies for a given porosity cutoff or a geobody derived net to gross number based on porosity cutoff seems to correlate well with the 3D streamline results though more work needs to be done to verify the applicability of the porosity cutoff to other reservoirs.