Lithofacies Simulation Conditioned to Diverse Data Based on MPG

摘要

Traditional two-point geostatistics (TPG) simulation algorithms, limited to the reproduction of two-point statistics, cannot reproduce complex geological structures. The alternative to these traditional techniques is the use of multiple-point geostatistics (MPG) simulation. The basic idea behind MPG is to go beyond the two-point modeling and to model the reservoir using multiple-point relations. A new algorithm to MPG stochastic simulation is proposed based on the existing algorithms. Subpatterns are extracted using a predefined template. These subpatterns are clustered using fuzzy c-means clustering and a certain number of class centers, which are defined by user, are acquired. A data event is extracted using the same template and compared with all of the class centers using a similarity criterion. The class center most similar to the data event is acquired. The central value of the class center is taken as the conditional probability density function (cpdf) and is used to carry out stochastic simulation. During the stochastic simulation, conditionings to diverse data are considered. Practical applications in Dongying delta Jiyang depression are realized based on this algorithm by using training images, well data, and seismic data. The model data and practical application results show that the proposed method can achieve better reproduction of geological characteristics of the study area and is beneficial for further application and popularization.