Multiple-Point Geostatistical Modeling with Training Images

In this paper, some aspects of practical application of multiple-point statistical simulation are considered. Multiple-point simulation is aimed to overcome the certain restrictions, which exist in classical geostatistics. Methods of multiple-point geostatistics include an ensemble of sequential simulation algorithms which use the training image as input data for the description of spatial structure of geological object. The results of modeling with various training images are presented. The next kinds of training images were used: stationary, nonstationary, categorical (for facial modeling), continuous (for modeling of effective thickness). In the DV-Geo module of multiple-point statistics the Direct Sampling algorithm (Mariethoz et al., 2010) is realized. It is possible to say that Direct Sampling algorithm successfully operates on fields with small and large number of wells. The demanded computer memory and CPU time do not depend on the volume of well information