Reproduction of channel stacking patterns in geomodeling: Metrics and impact of the modeling strategy on reservoir flow behavior

Abstract

Channelized turbidite systems are often grouped into complexes and exhibit various stacking patterns, which play a crucial role in controlling the connectivity between high-permeability and low-permeability sedimentary bodies. While some studies have analyzed the static connectivity of different stacking patterns, few have quantitatively evaluated the dynamic implications of these patterns on fluid flow circulation at large scale, and no method have started to compare the outcome of different modeling methods. This study addresses this gap by quantitatively investigating the impact of different classes of geostatistical modeling methods on both static and dynamic connectivity using multiple metrics. Using an original object-based simulation method, we have stochastically generated 3 different stacking pattern sets of 100 realizations each with similar facies proportions, that represent: (i) disorganized channels, (ii) independent channels conditioned to a vertical sand proportion map, and (iii) organized stacking that reproduce vertical and lateral migration of channels. Channel internal heterogeneities are neglected and homogeneous properties are attached to the three main facies: channels, inner levees and outer levees. To analyze the hydrodynamic responses, we set up a two-phase system of oil and water. Reservoir simulations are performed for all 300 realizations, providing production curves, saturation field and pressure field. Classical dynamic metrics (breakthrough time and recovery efficiency) are completed by original ones, designed to understand what governs the differences between the sets: at 30% of the Pore Volume Injected, saturation front shapes are extracted and we compute their surface area and sphericity to get objective comparison criteria. Results reveal significant differences in flow behavior across the three studied sets of stacking models: Disorganized stacking patterns, and, to a smaller extent, the conditioned disorganized stacking patterns, exhibit delayed water breakthrough times and more optimistic recovery than organized stacking patterns, which have, however, a higher static connectivity. This unexpected result seems to be related to gravity-driven fluid segregation in the considered reservoir production settings. Overall, these results quantitatively confirm that the ability of the channel simulation method to generate realistic stacking pattern is essential. In addition, the lack of a clear relationship between static and dynamic connectivity metrics suggests the need for further research to develop more effective forecasting techniques for reservoir behavior in turbiditic channel settings.