Subject:Research On Braided River Reservoir Modeling Using Multiple-point Geostatistics

Traditional variogram-based geostatistics conditional stochastic simulation is a routine tool in reservoir modeling to model reservoir heterogeneities and assess the spatial uncertainties. However, complex patterns are typically poorly reproduced by traditional geostatistics. Multiple-point geostatistics approach based on training images therefore was introduced to solve this problem. A training image is a numerical prior geological model which contains the facies structures and relationship believed existing in realistic reservoirs. Introducing geological prior knowledge or conceptual models into reservoir modeling through training images is a breakthrough contribution of multiple-point geostatistics to reservoir modeling.In this paper, facies models of Orinoco heavy oil belt MPE3Reservoir are built by using multiple-point gestatistcs method which combine different kinds of data such as geological information, well logs and seismic data. After testing, one model is preferred as the final facies model of O-llb-2layer and used in reservoir parameters modeling processes. An international characterized reservoir modeling technique has been given which can be used as the theoretical guidance and the technical support in the future development of the MPE3and other similar reservoirs.