W. Scott Meddaugh1
(1) Chevron Petroleum Technology Company, Houston, TX
ABSTRACT: Geostatistical Modeling of the Elk Hills 26R, 31S, and Northwest Stevens Reservoirs, California
Detailed stochastic models for three Elk Hills, California reservoirs were built to support flow simulation projects. The general workflow included data quality assessment and basic statistical analysis, detailed geostatistical analysis including semivariogram modeling, and stochastic property distribution. The effective porosity models were generated using sequential indicator simulation. A cloud transform-based technique calibrated with field specific core data was used to generate the corresponding permeability distribution.
Reservoir model sizes ranged from 370,000 cells for the 26R model to 3,200,000 cells for the 31S model. Cell dimensions were approximately 200' x 200' x 2'. Exponential semivariogram models with areal (xy) ranges between 1200' and 2600' were used for all geological layers in all models. Analysis showed only minor semivariogram anisotropy. Shale-rich intervals have generally higher semivariogram ranges than sand-rich intervals.
Multiple realizations (6-10) were generated for each reservoir. Visual inspection was used to select one model for scale-up and 3D-flow simulation. Scale-up was done using fluid-flow based software. After scale-up, the flow simulation models contained 30,000 to 50,000 cells.
Although the stochastic reservoir models were generated using all available geological markers (10-14 "sub-markers") to control the model framework, subsequent 2D flow simulation studies have shown that there is very little difference in flow characteristics between detailed models and models generated using only the top and bottom markers.
AAPG Search and Discovery Article #90906©2001 AAPG Annual Convention, Denver, Colorado