Fracture Flow Property Identification: An Optimized Implementation of Discrete Fracture Network Models

Abstract

Proposal The hydraulic characterization of naturally fractured reservoirs is a key point to provide valid flow properties to full-field simulation models. This calibration and validation step is based on the simulation of dynamic data such as flowmeter, transient or interference well tests. Recently, a methodology has been proposed to simulate such tests on Discrete Fracture Network (DFN) models derived from field fracture data integration. However, the drainage area of a well test in a fractured reservoir is generally large, thus leading to numerous computation nodes, especially in the presence of dense fracture networks. In such situations, the CPU time required for a single well test simulation for instance can reveal itself prohibitive. This paper proposes an original approach to simplify the DFN model in order to reduce drastically the number of fracture nodes, while keeping the same hydraulic properties. This approach keeps the actual fracture network geometry close to the well and replaces the fracture network far from the well by a simplified discrete fracture network. This DFN simplification is performed through an homogenization procedure keeping the same matrix-fracture storativity ratio and interporosity flow parameter. A sensitivity study was carried out to define the area around the wellbore where the fracture network can be simplified. We also defined a connectivity criterium to ensure a representative flow coupling between the actual near-well fracture network and the simplified distant network. Well tests simulations on this innovative DFN model are compared with simulations on the actual dense fracture network. Then we considered a 3D problem of well testing in a realistic field context. For both cases, the well test signature is not affected by the DFN simplification procedure, but the computation time is reduced drastically, by nearly two orders of magnitude (from 1 hour to 5 minutes CPU for instance). These tests illustrate the major contribution of this new approach. Actually, numerical flow simulation on DFN models can now be carried out at the well drainage area scale and thus be used to tune the field-measured well tests responses: a determinant step to validate the fracture network geometry derived from geological data and to identify fracture conductivities. Introduction The reservoir engineer needs to validate the geometry of the "static" fault/fracture model provided by the geologist as regards fluid flow behaviour, and to calibrate missing flow properties such as conductivities. To that end, a patented simulator was developed within FRACATM software platform. It simulates steady-state and transient well tests on the Discrete Fracture Network (DFN) model built by the geologist, for comparison with the actual field measurements of those tests (Sarda et al.[1], 2002). Its specificity consists in an optimal discretisation of the DFN using cells defined by fracture intersections, and also in an accurate simulation of matrix-fracture transfers. Indeed, the actual distribution of matrix block sizes and shapes is taken into account as each fracture node is associated with the proper matrix volume it can drain at this precise location, such matrix volume being determined through a specific processing of geological images. Thus, the dynamic behaviour of any type of reservoir crossed by fractures or faults can be characterised from well tests, especially in the case of complex multi-scale fractured reservoirs for which no analytical well test solutions are available. However, in case of a large drainage area or when the fracture network is dense, flow simulation may be time consuming due to numerous computation nodes. To overcome this difficulty, we propose to simplify the fracture network far away from the well to lower the number of discretisation nodes and thus decrease the computationcost.