Efficient and Accurate Reservoir Modeling Using Adaptive Gridding with Global Scale Up

Abstract

Abstract An accurate and efficient reservoir modeling process is essential for developing and producing hydrocarbon reserves, especially from unconventional resources. In this paper, we address some of the main challenges associated with modeling complex reservoir geometry and heterogeneous reservoir properties. We present recently developed techniques for adaptively constrained 2.5D Voronoi grid generation and for generic global flow-based scale-up. Our novel gridding approach is based on a new constrained Delaunay triangulation algorithm and a rigorous procedure of constructing a Voronoi grid that conforms to piecewise linear constraints. These gridding approaches allow us to generate 2.5D Voronoi grids that precisely honor small faults, intersections of multiple faults, and intersections of faults at sharp angles, as well as adapt the grid cell sizes to a specified density function. By precisely representing geologic structures in our simulation grid and by accurately scaling up fine-scale geologic properties, we improve the consistency between our geologic descriptions and reservoir simulation models, leading to more accurate simulation results. Numerical examples are provided to demonstrate the techniques and the advantages (both in efficiency and accuracy) of using adaptive gridding with global scale-up. Introduction Reservoir modeling is a crucial step in hydrocarbon resource development and management. It provides a venue for integrating and reconciling geologic concepts and data about a reservoir obtained at different scales. The scale of data ranges from a few inches for core plugs, a few feet for well logs, to many square miles for seismic imagies. Due to monetary and time constraints, directly sampled data of reservoir rock and fluid properties is sparse. Therefore, geologic interpretations based on seismic information and geologic concepts are required to supplement the measured data in order to provide adequate descriptions. A key challenge in reservoir modeling is accurate representation of the reservoir geometry of both the structural framework (i.e., horizons/major depositional surfaces that are nearly horizontal and fault surfaces that can have arbitrary spatial size and orientation) and the detailed stratigraphic layering (Figure 1). The structural framework delineates major compartments of a reservoir and often provides the first order controls on in-place fluid volumes and fluid movement during production. Thus, it is important to model the structural framework accurately. However, despite decades of advances in grid generation across many disciplines, grid generation for practical reservoir modeling and simulation remains a daunting task. For typical reservoir geometries with a high aspect ratio of horizontal to vertical dimensions, 2.5D (prismatic) Voronoi grids, constructed by projection or extrusion of a 2D Voronoi grid in vertical or nearly vertical direction, are a natural choice for reservoir simulations. Prismatic grid cells can easily be constrained to resolve horizons or stratigraphic layer boundaries. Voronoi grids (constructed as a dual to Delaunay triangulations) are much more flexible and adaptive than the corner point grids commonly used in commercial reservoir simulators, generally providing much fewer grid cells for the same accuracy of geometry representation and simulation. They also help reduce the grid orientation effect on numerical solutions of fluid transport problems (Verma 1996). Although less problematic than corner point grids, generating 2.5D Voronoi grids is often very challenging in practice due to the constraints these grids have to honor, which include numerous (intersecting) faults, pinch-outs (cf. Lyons et al. 2006), correlated heterogeneities (e.g., permeability extremes), and adaptive refinement as required for efficient and accurate flow simulations.