Horizontal and highly deviated well data that target pay intervals are generally clustered, both in a vertical sense, as they target specific stratigraphic levels, and spatially, as they target reservoir facies or fracture corridors. This means that the well data misrepresent the volume of interest, and the estimated facies and petrophysical property distributions are most likely biased. Nonetheless, to build reliable facies and petrophysical property models for petroleum exploration and production, geostatistical simulation methods require debiased input property distributions that are representative of the entire reservoir of interest. In this paper, we propose a new 3D kernel density based declustering algorithm to mitigate the inherent sampling bias in the input spatial distribution model and efficiently integrate horizontal and highly deviated well data as conditioning data in facies and petrophysical modeling workflows.
展开▼
