Facies modeling forms an integral part of geological numerical modeling. Over the last two decades, different facies modeling methods have been developed using geostatistical algorithms. Most of these methods rely on the assumption of discrete or binary modeling during which each model cell is assigned a single facies. In this study, the size of the cells is on average 100 meters by 100 meters laterally by one meter thick. Based on comparisons to outcrops and subsurface data, such cells should, in fact, include a mixture of facies.
The discrete-facies approach assumes a single facies per cell. The distribution of the facies between wells is described using classical categorical geostatistical algorithms. Reservoir properties are then populated by facies within mapped environments of deposition. This process is well-established and straightforward, especially with regard to tying well data, handling property trends, and applying net rock cut-offs.
A mixed-facies approach can be performed using effective property modeling in which multiple small, fine-scale models are built for each environment of deposition. These models are re-sampled to the full-field cell volume using static and flow-based upscaling methods. The resulting statistics are then used with geostatistics, conditioned to the proportion of each facies present, to populate the full-field model. Such models allow the incorporation of core-scale heterogeneity potentially important in improved oil recovery projects, and may reduce modeling cycle times, especially when multiple iterations are required, such as during history-matching or uncertainty analysis.
This paper compares the impact on simulated fluid flow of modeling facies using discrete modeling versus a mix of facies per cell. Shoreface and subordinate fluvial environments of deposition facies, and five reservoir lithofacies, were modeled.
Fluid-flow simulation of the mixed-facies model, under both primary depletion and pressure maintenance conditions, was smooth and uniform, with a highly conformable flood front. The discrete model was more stratified, with faster and less conformable water movement.
The assignment of discrete facies to large model cells (few hundred meters laterally & few meters vertically) takes less time than a mixed-facies approach and does a better job of preserving organized extremes of permeability important at the production timescale. In the early stages of field development, when there is much uncertainty and a rapid, scenario-based modeling approach is desirable, the discrete approach can be used to flag heterogeneity-related risks more quickly and confidently than the mixed-facies technique. Inaccuracies in performance parameters resulting from the assignment of unscaled discrete values can be corrected using fine-scale sector models tailored to the highest risk cases.