On mapping fracture networks onto continuum

Discrete fracture network (DFN) and stochastic continuum (SC) are two common modeling approaches used for simulating fluid flow and solute transport in fractured media. Fracture continuum approaches combine the merits of each approach; details of the fracture network are preserved, and a computationally efficient grid is utilized for the solution of fluid flow by assigning a conductivity contrast between the grid cells representing the rock matrix and those representing fractures. In this paper, we propose a fracture continuum approach for mapping individual fractures onto a finite‐difference grid as conductivity fields. We focus on several issues that are associated with this approach, such as enhanced connectivity between fractures that would otherwise not be in connection in a DFN simulation and the influence of grid cell size. To addresses these issues, both DFN and the proposed approach are used to solve for fluid flow through two‐dimensional, randomly generated fracture networks in a steady‐state, single‐phase flow system. The DFN flow solution is used as a metric to evaluate the robustness of the method in translating discrete fractures onto grid cell conductivities on four different regularly spaced grids: 1 × 1 m, 2 × 2 m, 5 × 5 m, and 10 × 10 m. Two correction factors are introduced to ensure equivalence between the total flow of the grid and the original fracture network. The first is dependent on the fracture alignment with the grid and is set to account for the difference between the length of the flow path on the grid and that of the fracture. The other correction is applied for areas in the grid with high fracture density and accounts for the artificial degree of connectivity that exists on the grid but not in the DFN. Fifteen different cases are studied to evaluate the effect of fracture statistics on the results of the proposed approach and by taking average results of 100 realizations in each case in a stochastic Monte Carlo framework. The flow equation is solved for the DFN, and total flow is obtained. The flow is also solved separately for the four‐grid resolution levels, and comparisons between the DFN and the grid total flows are made for the different cases and the different grid resolution levels. The approach performed relatively well in all cases for the fine‐grid resolution, but an overestimation of grid flow is observed in the coarse‐grid resolution, especially for cases wherein the network connectivity is controlled by small fractures. This overestimation shows minor variation from one realization to another within the same case. This allowed us to develop an approach that depends on solving limited number of DFN simulations to obtain this overestimation factor. Results indicate that the proposed approach provides improvements over existing approaches and has a potential to provide a link between DFN and SC models.