Abstract:
A well-developed hydraulic fracture network with sufficient extension into the target zone and minimal interaction with the adjacent layers can enhance the effective permeability of the rock mass by orders of magnitude, and thereby, increase the recovery factor of the reservoir significantly. However, the simulation of such a hydraulically driven fracture remains a computationally expensive and difficult task. This paper presents a hydraulic fracture simulator based on the eXtended Finite Element Method (XFEM), which is a flexible means for simulating evolving discontinuities without having to assume a fracture path a priori. The simulator is a coupled dynamic model capable of simulating the propagation of two-dimensional non-planar hydraulic fractures in heterogeneous reservoirs. It is demonstrated that the presence heterogeneities with different stiffness leads to deviated fracture paths.
Introduction
Simulation of Hydraulic Fractures in its simplest form contains coupling of at least three physical processes: deformation of the solid matrix, fluid flow within the fracture network/porous media, and propagation of fractures (including both reopening of existing fractures and initiation of new fractures) (Shlyapobersky, 1985 and Adachi et al., 2007). At small scales, rock blocks in unconventional reservoirs are considered to have negligible permeability, and therefore, fracture network is the only pathway for fluid flow (Lange et al., 2013). Linear elasticity is the most prevalent constitutive model for the rock mass in small scale studies (Carter et al., 2000); however, application of linear elasticity results in an unrealistic solution as infinite stresses arise at the fracture tip. A very common solution for such a problem is application of cohesive models. Using a cohesive model ensures that stress at the fracture tip does not exceed the material strength, and that the fracture opens smoothly behind the fracture tip.
Cohesive elements are very common in Finite Element (FE) models of hydraulic fractures (Chen et al., 2009); nevertheless, such an approach requires the location of cohesive elements to be pre-assumed. As a result the model is unable to predict fracture trajectories under complex conditions.
