Fractured Reservoir Uncertainty Analysis and Connectivity-Based Model Classification

Abstract

AbstractUncertainties associated with fracture properties are usually large, and significantly impact the reservoir model flow behaviour. The analysis of these uncertainties is therefore a necessary task for performing more reliable production forecasts and optimization via fractured reservoir flow models. However this task may quickly become intractable considering the overwhelming uncertain fracture properties ranges to investigate. A workflow is presented that allows one to embrace large multi-scale fracture uncertainties and to analyze their impact on geologically-consistent reservoir models in term of reservoir-related connectivity properties. Fractured reservoir model classification is then performed based on the ranges of connectivity values that result from the fracture properties uncertainties, thus facilitating the analysis of the effect of fracture uncertainties on the reservoir flow models.A stochastic fractal fault model has been used to investigate the effects of fault uncertainties at seismic and sub-seismic scales on a reservoir model. Uncertainty analyses have been performed for the following parameters: (1) the fractal dimension that controls the fault spatial distribution; (2) the fault length distribution defined from a power-law statistical distribution. The sub-seismic fault orientation is constrained by the seismic fault network. The fault network orientation is assumed to have a NW-SE trend. The reservoir model is a synthetic, but geologically-realistic, 12 km by 15 km reservoir with three facies and seven layers. A five-spot well configuration involving four injectors and one producer is considered. The injector-producer connectivity is evaluated via a single-source shortest path graph algorithm that computes the injector-producer distances via the reservoir cell transmissivities and volumes. The effects of the fault uncertainties on the injector-producer connectivity are estimated from a large representative sets of fault network realizations.The different steps of the workflow are rather fast: the fault network generation takes a few seconds, the conversion to an equivalent reservoir flow model takes a few minutes, and the graph connectivity analysis only a few seconds. This low processing time allows one to analyze a large number of fault network realizations, thus better estimating the impact of the large fracture uncertainties on the reservoir model. The distribution of the injector-producer connectivity properties is rather multi-modal and dispersed, however relevant classes of fractured reservoir models could be identified accordingly to the fault fractal dimension. These different classes can be used to estimate fractured reservoir model occurrences based on specific reservoir connectivity properties, thus allowing one to identify the most probable flow reservoir model configurations considering large fracture properties uncertainties.The proposed workflow can be used to analyze the effects of the multi-scale fracture uncertainties on equivalent fractured reservoir flow models. Thus facilitating the classification and identification of the most relevant flow models, used subsequently for production forecasts and optimization.