Reservoir Simulation with Hydraulic Fractures: Does it Really Matter How We Model Fractures?

Abstract

Abstract Hydraulic fracturing is a key technology to developing unconventional resources. At the planning stage, the prediction of production profiles in the presence of designed hydraulic fractures in the reservoir model is very important. Various analytical methods available are inadequate to produce reliable production profiles by accurately modeling irregular reservoir shape, permeability heterogeneity, multiphase flow and the presence of natural fractures. Any standard reservoir simulator facilitates these modeling features; the key challenge is, however, to model the propped hydraulic fractures accurately in the reservoir. Techniques such as negative skin, equivalent well radius and a uniform conductivity rectangular fracture are common built-in features in almost all available reservoir simulators. The imperative questions are (1) how good are these techniques to reliably model the flow through a non-rectangular fracture having non-uniform conductivity, and (2) how to model the combination of propped hydraulic fractures and stimulated reservoir volume (SRV) in the presence of stimulated natural fractures? This paper presents the results from two case studies that reveal the shortcomings and simulation difficulties with negative skin, equivalent wellbore radius and uniform conductivity rectangular fractures. The paper also presents improved techniques to model propped hydraulic fractures with nonuniform conductivity in nonrectangular geometry (as usually designed) and SRV in the reservoir model for reservoir simulations. The comparison of results and discussion of various techniques are expected to enable engineers to understand why in certain conditions built-in features may not be possible to use because of numerical instabilities during simulation, and the range of variations in prediction from various techniques. The improved technique presented in this paper, which is based on a local grid refinement technique, can provide more reliable production prediction and can overcome numerical instabilities (though it is generally more computation intensive).