Simulation of Gas Oil Gravity Drainage: Comparison of the Dual Continuum with the Discrete-Fracture and Matrix Approach

Gas-oil gravity drainage (GOGD) is perhaps the most effective recovery mechanism for those naturally fractured reservoirs that are lithologically stratified, the oil is stored in rock matrix, and vertical fractures cut through (multiple) low-permeability layers, providing the necessary permeability in the direction of gravity-driven flow (eg., Gilman 2003, 2011). The conceptual process model for GOGD is that fracture saturation with gas that is much less dens than the oil creates a pressure differential that drains the oil into the fractures, where it flows downward and can be produced from wells placed directly above the oil water contact. Prerequisites are that saturated conditions can be obtained, the fractures are sufficiently tall, and the oil is mobile enough (e.g., Kazemi et al., 1976). Since Gilman and Kazemi (1983) and Saidi (1983) GOGD has predominantly been simulated using the oil industry’s adaptation of the dual continuum (DC) approach (e.g., Warren and Root, 1963),using separate fracture and matrix grids coupled via source terms for fracture-matrix transfer (FMT),see Abushaika and Gosselin (2009). However, it is challenging to derive DC input parameters such as matrix-block size, shape factor or fracture permeability from field data because they tend to obey power-law or multi-modal distributions (e.g., Bonnet et al., 2001). Recently, new numerical fracturematrix upscaling methods have been proposed to overcome this DC model parameterization problem (Sarda et al., 1997, Karimi-Fard et al., 2006), however, they only address aspects of pseudo steadystate FMT and entail other restrictive assumptions. In many cases, history matching of DC models is only possible through over-calibration, raising doubts about the predictive capabilities of such models (e.g., Gilman, 2003). Since fractures have no discrete representations in DC models, these cannot be used to test long-standing hypotheses about the details of FMT (cross flow, capillary continuity, reimbibition of oil, see Fung, 1991; Saidi, 1993), or evaluate the balance between gravitational – viscous and capillary forces and the mutual influence of respective flow processes within the reservoir. Assumptions about how saturation changes in matrix blocks and the pathways that the oil takes on its way to the production wells cannot be tested. Discrete-Fracture and Matrix models (DFM) provide an alternative for such purposes (e.g., Kim and Deo 2000). Using a spatially adaptive poly-element type unstructured grid, complex fracture geometries can be represented in the model explicitly, permitting the modelling of realistic fracture geometries and property statistics (e.g., Matthai et al., 2007). In 2011, Bazrafkan and Matthai,presented a new hybrid FEM node-centered FVM methods where finite elements also serve as control volumes, permitting the correct representation of saturation jump discontinuities at material interfaces and also the realistic simulation of capillary FMT. Here, we apply this novel DFM approach to simulate GOGD using an outcrop analog model that has a fracture geometry that is simple and regular enough to construct an equivalent dual permeability model. The latter is run using a commercial simulator and here we present, for the first time, a comparison of DFM and DCM results, using the same saturation functions and parameters. The DFM results further allow an investigation of the flow patterns in individual fractures and matrix blocks so that the validity of the assumptions that underpin the DC approach can be evaluated.