2019 AAPG Annual Convention and Exhibition:

Datapages, Inc.Print this page

Modeling and monitoring reservoirs over their lifetimes


Large-scale simulation is a tool for understanding how underground reservoirs change during their production lifetimes and how these changes can be detected and quantified by geophysical remote sensing. This paper describes a recent collaborative project between government, industry, and academic research groups to design, build, and simulate an integrated geologic, reservoir, and geophysical model during more than two years of fluid production. The geologic model consisted of 2 billion grid cells, representing a region 12.5 km by 12.5 km in horizontal extent and 5 km in depth and including a 420-m thick reservoir with upper and lower turbidite-fan units separated by an impermeable shale layer and offset by faults. The model was based on a typical shallow Gulf of Mexico reservoir which can also serve as an analog of turbidite fields around the world. The reservoir simulation computed fully coupled three-phase fluid flow and linear geomechanical responses in a production scenario involving 11 production wells and 6 water-injection wells penetrating three reservoir compartments. Seismic surveys were simulated with isotropic elastic wave modeling before the start of production and after 27.5 months of production at a simulated rate of about 67 500 barrels per day, with about 32 500 barrels per day of water injection. Rock properties were updated by petrophysical models calibrated to turbidite systems in the Gulf of Mexico. Analysis of the model and simulations improves understanding of the complex interaction of fluid effects, pressure changes, and rock deformation. For example, compaction in the reservoir may cancel time-lapse fluid effects, and compaction or dilation in the surrounding shales may override the observed reservoir signal in the seismic bandwidth. Strain-induced velocity changes in the shales have a much larger effect on estimated time-lapse time shifts than do strain-induced changes in path lengths. The models and data created by the project are now publicly available.