Scale consistent geostatistical modeling for reservoir characterization

In modern oil industry, it is common that a variety of data are available for reservoir modeling. These data include core and log data, seismic attributes, and conceptual geological models. Data scale, reliability, coverage and availability must be taken into account in integrating these data into numerical reservoir models. Geostatistical reservoir models can be built at different scales for different purposes, such as a large scale model for resource estimation and a fine scale model of heterogeneity for flow simulation. Different modeling techniques and different usage of the input data usually cause inconsistency between the models at different scales. It is desired that all the models are scale consistent. The scale consistent reservoir modeling scenario developed in this thesis aims to build reservoir models that are scale consistent and reproduce the data at different scales.; The basic idea is to construct a large scale model by integrating all available data, and then downscaling or upscaling for different modeling purposes. The downscaling must reproduce exactly the large scale model to ensure consistency. There are three major steps: (1) construct a large scale model over the entire lease by integrating multivariate information; Gaussian-based Bayesian updating technique can be used; local uncertainty assessment is provided, (2) perform petroleum resource estimation with global uncertainty assessment from the large scale model; a spatial/multivariate simulation approach can be used to account for the spatial and multivariate correlations among the local uncertainties, (3) construct fine scale 3-D models of heterogeneity that are consistent with the large scale model and well data using the exact downscaling techniques.; This modeling scenario is developed from the oil sands geostatistical modeling projects for the Surmont lease, Alberta, Canada. An application of the modeling techniques to the Surmont will be presented.