The Generation of Grid Block Permeabilities from Core Data

Abstract

Abstract The grid blocks used in reservoir simulations typically have dimensions in the order of tens or hundreds of metres, whereas the core data which are used to generate the grid block permeabilities have a scale of only a few centimetres. Thus it is necessary to scale up the data over several orders of magnitude to derive the "effective" permeability distribution at the grid block scale. This is commonly done by taking standard measures of the mean of the data (e.g. arithmetic, geometric or harmonic mean) and making ad hoc assumptions about the rescaled variance. Such an approach can lead to costly errors. However, the obvious alternative — to iteratively scale up from the core to the grid block scale via detailed flow simulations — is often not feasible due to computational limitations. This problem has been resolved in recent years by the development of fast and accurate algorithms for computing the effective permeability of large systems, containing up to millions of grid blocks. This paper describes a computationally efficient software package which has been developed to generate grid block permeabilities for reservoir simulations from core data. Upscaling is performed using algorithms based on random walks and real-space renormalization group methods. Geological correlations and anisotropy at the core scale can be incorporated using geostatistical techniques (e.g., sequential gaussian simulation). The paper emphasises a practical application of the software to field data.