PNNL

Abstract

Mesh generation lies at the interface of geological modeling and reservoir simulation. Highly skewed or very small grid cells may be necessary to accurately capture the geometry of geological features, but the resulting poorly scaled or small grid cells can have a substantial negative impact on simulator accuracy and speed. One way to minimize numerical errors caused by gridding complex structures is to simulate on high-quality Voronoi meshes, which will reduce grid orientation affects in fluid flow. This work presents a complete methodology to create Voronoi simulation grids, model fluid flow in complex geological systems and visualize the results. A recently developed Voronoi meshing method that can automatically generate provably good unstructured random meshes that conforms to virtually any interior and exterior input surfaces is used. Initially an analytical benchmark simulation is presented to validate the quality of the meshes and simulation results. Next meshes are created for test structures representing four of the most common geological features in the subsurface: layering, pinch-out, an interior lens that tapers to zero thickness on all sides and a fault with offset. Eight benchmark flow simulations are run for each test structure. Finally, a realistic geological example for CO2 injection into an anticline is simulated. Three realizations of the random meshes at the same resolution are generated for the simulations. Each mesh is highly refined near the injection wells and coarse in areas of less interest. These three meshes are used to model the CO2 as it migrates to the top of the structure and then fills downward. Simulations on the randomly generated meshes each give slightly different results for the viscous-unstable buoyant gas flow. The results presented in this work show a promising first step towards accurate meshing and subsurface flow simulations in complex geology using open-source software.