Modeling Hierarchical Reservoir Architecture with Improved Pattern-Based Multiple-Point Geostatistics Algorithm

Enhanced oil recovery requires the characterization of multiple-scale reservoir architecture. Owing to the deficiency of variogram- and object-based traditional geostatistics, multiple-point geostatistics has been proposed to delineate complex architecture. Pattern-based mul-tiple-point geostatistics is becoming increasingly popular. However, replacement with whole pattern is difficult to be conditioned with dense hard data, which misleads to sequent simulation. In this work, we attempt to increase the suitability of pattern-based multiple-point geostatis-tics for modeling hierarchical reservoir architecture with dense hard data. Pattern-based multiple-point geostatistics includes pattern extraction and pattern reproduction. In the process of pattern extraction, templates on multiple grid are utilized to capture hierarchical reservoir heteroge-neities. Pattern reproduction is accomplished by sequential simulation, in which only the central node instead of data event is replaced with the searched pattern, which confers ease of hard-data conditioning to the algorithm. At last two synthetic models demonstrate that the algorithms perform better than SIMPAT and SNESIM.