Automated Characterization of Fracture Conductivities from Well Tests Inversion (SPE-121172)

摘要

Methodologies and numerical tools are available to construct geologically-realistic models of fracture networks, and to turn these models into simplified conceptual models usable for field-scale simulations of multi-phase production methods. However a critical intermediate step remains that of validating the geometry of the geological fracture network and characterizing it in terms of flow properties. This characterization step involves the calibration of available field data, e.g. well tests, with simulated data. It is usually an ill-posed problem: available data are insufficient to fully characterize the fracture properties. Therefore prior estimates of these properties are necessary, but these are usually uncertain, resulting in a wide range for the properties values to be investigated. This paper presents a complete methodology to address this inverse problem: a covariance matrix adaptation-evolution strategy is used to explore efficiently the properties space; this optimizer is coupled with a surface response methodology in order to detect several solutions simultaneously, and to estimate the properties global sensitivities to field data during optimization.An application involving a horizontal well and a fracture network with two sets of systematic joints and one fault set is presented. For a given realization of the fracture network geometry, it is shown that the optimization of fracture conductivities via the present methodology results in the characterization of two distinct, although both data-consistent below a 0.1% error in pressure, sets of solutions for the joints conductivities. The fault set conductivity is shown to have a low sensitivity to well test data.As sensitivities are estimated during optimization, by using a surface response methodology and the Sobol decomposition technique, the optimization cost may be reduced by eliminating low sensitive parameters. The effectiveness of the methodology to characterize physically meaningful and data- consistent fracture properties is discussed further.