A Stochastic Approach to Characterize Discrete Fracture Network Patterns of Heterogeneous Media

Abstract:

Fractured rocks host those aquifers with the highest permeability contrasts. While the matrix permeability of carbonate or crystalline rocks often is negligible for groundwater flow, fractures represent the main pathways for flow and transport. However, despite their relevance, fracture networks are typically complex and can hardly be described in all detail in models. Size and shape of fractures depend on in-situ stress conditions, existing discontinuities, fluid pressure, temperature as well as fluid and rock physical properties.

In this work, an innovative approach is introduced for (discrete fracture network) DFN characterization of heterogeneous fractured media via active thermal tracer testing. By recording multiple temperature breakthrough curves at different observation points, preferential flow paths are recognized and this information is used to construct equally probable DFN patterns. The sequential steps of this procedure are illustrated in Figure 1. It starts with initialization a DFN pattern based on the geological properties of the medium such as fracture orientation, length, spacing and persistency retrieved from outcrop, borehole image logs, etc. Pressure and temperature fields are estimated inside the fracture network by means of an implicit upwind finite difference method, which is used to compute the heat tracer travel times between injection and observation points. A trans-dimensional inversion is adopted to update the DFN model by comparison of proposed and observe travel time. The resulting ensemble of models can be used as an input geometry for deterministic simulations of fractured medium for further applications such as heat exploitation from geothermal resources, enhanced oil recovery from naturally fractured reservoirs, and gas production from fractured shale reservoirs. Aside from deterministic simulation, the obtained ensemble facilitates stochastic flow and transport simulation to account for the uncertainty in DFN characterization.

Figure 1. Schematic representation of the proposed stochastic approach for DFN characterization of heterogeneous aquifers