Abstract
Many IOR/EOR recovery processes such as chemical, miscible and steam flooding are often associated with complex flow mechanisms that manifest at the displacement front. Viscous fingering, polymer/surfactant dilution and mixing effects are some of these mechanisms. Accurate modeling of these phenomena requires simulations on high resolution grids to properly capture the physics in the vicinity of the displacement front. Unfortunately high grid resolutions incur longer simulation times. Thus, past efforts at running full-field gas or Chemical EOR simulations were frequently deemed impractical.
The advancement in computational power from software, hardware and parallelism has indeed pushed the limits towards higher resolution simulations. However, this may not be practical in workflows that require simulations on many models to manage uncertainties. Dynamic gridding is one approach that attempts to adjust the grid resolution as needed during the run time. No a priori knowledge is assumed regarding the fluid flow pathways. The simulator can track the location of the displacement front, refine the neighborhood cells, and later coarsen them back as the front progresses. The advantage is reducing the number of grid-blocks, and therefore the computational time, compared to the fully refined grid, while preserving the fluid-flow physics. Although this technology is not new in reservoir simulation, there are persisting challenges in the existing methods related to the computational overhead associated with cell re-mapping, transmissibility re-calculation, and grid up-scaling and down-scaling.
A new dynamic gridding functionality has successfully been implemented into our in-house simulator. The key achievements are: 1) eliminate grid re-mapping and transmissibility re-calculation at the run time, 2) capture heterogeneity associated with all levels of grid refinements, 3) model complex geology with non-uniform gridding, and 4) track multiple fronts associated with surfactant/polymer and chase water slugs. We discuss how we overcame the bottlenecks to leverage this technology from prototypes to complex cases. We also demonstrate our method on prototypes and pilot cases under CEOR recovery processes.
