Reservoir Modeling and Simulation Backed Up with Geological Knowledge

Sun, Ruidong¹, Yuhong Liu² (1) Schlumberger Information Solutions, Houston, TX (2) ExxonMobil Upstream Research Company, Houston, TX

The final goal of reservoir modeling is to forecast hydrocarbon flow response given all available information. Therefore it is important not only to capture realistic geological varia­tion, but also to reproduce those features that are most critical to flow simulations. Various geostatistical approaches are available for reservoir modeling, they all aim at reproducing realistic geological phenomena important to flow. For example, object-based modeling builds facies models by throwing / perturbing / rejecting various geological bodies into the simulation field; traditional two-point geostatistics uses statistics such as variogram to cap­ture spatial variation; the newly developed multiple-point geostatistics reproduces various types of spatial patterns depicted by a training image.

In this paper, we use different technologies to build different static geological models, with the same input data as constraints. Three facies modeling technologies, object-based modeling, sequential indicator simulation, multiple-point statistical simulation, associated with sequential Gaussian simulation, are used to build these models. We then flow simulate them to study their flow performance. It is observed that, factors such as faulting and spa­tial distribution of facies, have more impact on the final flow response than the spatial vari­ation of reservoir properties within the net pay zones (e.g., porosity and permeability). Simply varying the variogram ranges or azimuths, as generally done in the two-point geo­statistical modeling, is not enough to capture the full uncertainty space about the reservoir. Instead, we need to use a full spectrum of existing technologies to capture all those features critical to the flow response.