Faults and complex wells are two important types of internal boundaries to resolve in reservoir simulation. Faults are physical boundaries which may form local barriers or conduits to fluid flow. In structured-grid simulation, fault surfaces are typically represented as zig-zag cell edges where the depths may be shifted across the fault face. The better representation of fault traces using unstructured gridding has been the subject of research in the petroleum literature for over two decades. The use of long horizontal and multi-branch complex wells for production from tight and heterogeneous reservoirs is also common practice nowadays. These wells can be densely populated which make classical local grid refinement (LGR) methods difficult to apply. It is highly desirable to represent the perforation inflow and the near-wellbore flow more accurately in full-field simulation.
The paper extends the Voronoi gridding method (Fung et al. 2014) for densely-spaced complex wells in full-field simulation to the modeling of faulted reservoirs containing these wells. Well branches and faults may intersect one another. Frequently, in order to honor multiple conflicting internal boundaries, grid congestion may occur which leads to small cells and/or poorly shaped grid cells. The method uses multi-level quad-tree method to achieve the desired resolution in areas of interest and a hierarchical point prioritization/selection procedure to resolve congestion. Grid quality at the desired resolution in congested areas is an important consideration for solution efficiency and robustness in simulation practice.
Following an introduction of unstructured-grid methods in reservoir simulation, the gridding algorithm is discussed in details. This is followed by simulation examples, which includes a full-field compositional simulation example of a faulted gas-condensate reservoir completed with many deviated and horizontal wells. An in-house parallel reservoir simulator is used to run the models. Simulation results using both the structured corner-pointed-geometry (CPG) grid and the unstructured-grid method are compared. The advantages of unstructured approach in complex field-scale simulation are demonstrated.