A Regional Uplayering Technique for Preserving Reservoir Heterogeneities

Abstract

An automated sequential statistical algorithm is proposed for nonuniform vertical upscaling (also known as uplayering) that uses the averaged property information (defined as trend) to capture reservoir heterogeneity globally. The fine scale geological model is first divided into multiple reservoir zones based on the geological boundary information and/or the vertical trend information. In each reservoir zone, the horizontal trend information is used to partition the zone into a set of regional blocks, with the intrablock property variance being minimized and the inter-block variance being maximized. Each regional block, regardless of its neighboring blocks, is further divided into a certain number of layers. Thus, this new upscaling method generates nonneighboring connections between the adjacent blocks. This technique can be used on continuous and categorical properties or on any property that is a good discriminator of reservoir flow. The other reservoir properties or any of their combinations can be used as the weighting functions, globally or by zone, to modify the uplayering process, such that the detailed reservoir features, such as the shale barriers, thief zones can be preserved. The proposed algorithm is size independent, scenario independent, and CPU efficient. The advantage of using vertical nonuniform upscaling over uniform upscaling is demonstrated with several case studies for different depositional systems.