Optimal Gridding Selection for Field-Scale Reservoir Simulation of a Channelized Deepwater System

Abstract

Sub-seismic scale heterogeneity in channelized deepwater reservoirs can lead to significant uncertainty in reservoir connectivity and predicted performance. Though bed-scale heterogeneity can influence reservoir performance, reservoir simulation typically requires cell sizes much greater than the scale of internal channel architecture. Fine-scale sector models consisting of two stacked channel segments were used previously to quantify the influence of bed-scale heterogeneity and channel stacking on reservoir performance. Systematic upscaling and local grid-refinement are utilized here to optimize model construction for field-scale reservoir performance prediction while preserving key flow behaviors observed in fine-scale sector modeling.

Outcrop characterization of deepwater channels from the Tres Pasos Formation of the Magallanes Basin in Chile provided the framework (internal architecture and cross-sectional geometry) for each dynamic simulation. Twelve fine-scale sector models (cells 2 m x 2 m x 0.25 m) representing a range of potential offset angles and distances between two straight channel segments served as the “base-case” for upscaled model performance evaluation. Flow was induced between a single injector-producer well-pair completed 500 meters apart in opposing channel margins. Two different styles of gridding were tested for each of the 12 scenarios: (1) cartesian; and (2) conformable gridding. A range of vertical (0.25 to 2 m) and lateral (2 to 100 m) cell sizes were tested for each gridding style. Effective reservoir properties were calculated initially using flow-based averaging. History matching of upscaled and fine-scale results evaluated cumulative production, well flow rates, pressure, recovery efficiency, water breakthrough timing, and simulation runtime. Conformable gridding with cells 30 m x 30 m x 1 m most efficiently replicated high resolution simulation results for all 12 stacking scenarios. To obtain greater flexibility in defining channel fill architectures (a deterministic fill pattern was used previously), observed trends and distributions were used to geostatistically define net sand on the 30 m X 30 m x 1 m grids. Porosity and directional permeability were then assigned using the distributions of these properties and relationship to net sand taken from earlier upscaled models. Simulated results from the non-deterministic upscaled models match earlier fine-scale simulation results within 5% error. The methodology and results presented here provide the framework for future efforts to evaluate the influence of inter- and intra-channel architecture on reservoir performance prediction at the field-scale.

AAPG Datapages/Search and Discovery Article #90291 ©2017 AAPG Annual Convention and Exhibition, Houston, Texas, April 2-5, 2017