The production of hydrocarbons trapped in unconventional plays depends on the ability to drain efficiently and safely the wellbore area within tight formations. To date, hydraulic fracturing is the main operational technique to turn a shale formation into an economically exploitable resource. The injection of a high pressure fluid can indeed induce new channels in the rock and enhances the ultimate recovery. However the result of this technique remains difficult to assess due to the complexity of the involved mechanisms and the heterogeneities governing flows in such environment. Considering the amount of data available, like microseismic event records, it becomes clear that pre-existing weaknesses such natural fractures and bed boundaries are reactivated during fracturing jobs.
Therefore in order to study the efficiency of fracturing and to better understand recovery mechanisms, we present a workflow based on the characterization of the natural fracture network and the simulation of injection in this a fracture network submitted to in-situ stress. This workflow integrates the available geological data in a 3D Discrete and Deformable Fracture Network (DDFN) on which fluid flows are simulated. We first emphasize how to build a consistent DDFN based on the natural fracture observation in the wells. Then we use an innovating meshing algorithm with corresponding un-structured numerical scheme, allowing to simulate flows on 3D DDFN made of more than 100.000 deterministic and stochastic fractures (natural or not). A simplified model simulating fracture reactivation during fluid injection is introduced. The results of these simulations are finally compared with the microseismicity and monitored injection pressure.
This new simulator is applied on an injection phase of a representative real field case in Bakken formation. We demonstrate how the integration of a geological and mechanical interpretation provides a better understanding of the processes involved during fracturing. A nonlinear mechanical behavior for each fracture in the DFN is mandatory to reproduce the measured trends.