A Multimode Inversion Methodology for the Characterization of Fractured Reservoirs from Well Test Data

Optimizing the exploitation of fractured reservoirs requires a reliable characterization of the most influential heterogeneities on flow properties. To this end, reservoir simulation models should remain interpretable in geological terms. Workflows are available (i) to construct geologically-realistic models of fracture networks, (ii) to turn these models into simplified conceptual models usable for field-scale simulations of multi-phase production methods. However a critical step remains that of characterizing the flow properties of the geological fracture network. The multiscale nature of fracture networks and the associated modeling cost impose a scale-dependent characterization: (i) multi-scale fractures are characterized in local dynamic tests area, e.g. well test drainage area, through the calibration of geologically-realistic Discrete Fracture Network (DFN) models and accurate local flow tests simulations; (ii) large-scale faults are characterized through reservoir-scale production history simulations. This paper presents an efficient "multi-mode" inversion methodology to facilitate the characterization of fracture properties from well test data. The fracture model is optimized sequentially according to different modes of deformation. First, the parameterization is defined such that the fracture model is deformed as a whole (mode 1), then S fracture sub-sets are defined accordingly to their sensitivities to well test data (mode S), where S is increased sequentially. The optimization is performed via an original evolutionary algorithm that has been coupled with a 3D DFN flow simulator. An application is presented on a fractured reservoir model having two facies, four fracture sets and involving a horizontal well test. The characterized fracture properties are the mean size properties, mean conductivity and faciesdependent fracture densities. This "multi-mode" inversion methodology is shown to be much faster to characterize physically meaningful and dataconsistent fracture properties, compared with a direct optimization application on all model parameters. This methodology facilitates the characterization of fractured reservoirs flow properties.