The study of depletion performance of naturally-fractured reservoirs has gained wide interest in the petroleum industry during the last few decades and poses a challenge for the reservoir modeler. The presence of a retrograde gas-condensate fluid incorporates an additional layer of complexity to the problem Upon depletion, reservoir pressure may fall below the dew-point of the hydrocarbon mixture which results in liquid condensation at reservoir conditions. Condensate would first appear in the high-conductivity channels supplied by the fracture network and around the external edges of the matrix blocks which are the zones prone to faster depletion. The presence of condensate around the edges of the matrix block would hinder the flow of hydrocarbons from the inner portions of the matrix blocks and severely obstruct their recovery. Since the bulk of hydrocarbon storage resides inside the matrix, it is critical to answer the question whether this trapped gas has been irreversibly lost or not. It is believed that the interplay of Darcian-type flow and Fickian-type flow (multi-mechanistic flow) is the key to answering the questions about depletion performance and ultimate recovery in these reservoirs, especially for the cases of extremely tight, naturally fractured reservoirs where matrix permeabilities may be less than 0.1 md. In this study, recovery from a single matrix block surrounded by an orthogonal matrix network---the fundamental building block of the full-scale system---is investigated. The dominant flow processes and recovery mechanisms taking place in naturally-fractured gas-condensate reservoirs are shown and the depletion performance of these systems is described in order to provide guidance for the development of this class of reservoirs. Additionally, artificial neural network (ANN) technology or soft-computing was proven instrumental in establishing an expert system capable of learning the existing vaguely understood relationships between the input parameters and output responses of complex hard-computing protocols such as compositional simulations of gas-condensate reservoirs. During the training phase of the artificial neural network, an internal mapping that accurately estimates the corresponding output for a range of input parameters is created. As a result, a powerful screening and optimization tool for the production of gas-condensate reservoirs is presented.
展开▼
