A Skeleton-based Multiple Point Statistics for Reservoir Stochastic Modeling

Yin, Yan-Shu¹ (1) Yangtze University, Jingzhou, China

Traditional stochastic reservoir modeling, including object-based and pixel-based methods, can't solve the problem of reproducing continuous and curvilinear reservoir objects. The recently developed multiple point statistics is hopeful to approach to this goal and represents the future of reservoir stochastic modeling. So a detail study on multiple point statistics is urgent.

This paper first dives into the newly published multiple point statistics, that is, the Snesim and Simpat . The study shows that both the Snesim and the Simpat face the problem of reproducing continuous shape due to the random selection of data patterns. An intelligent choice of data patterns may solve this problem. Based on this consideration, the paper designs a new multiple point statistics algorithm, skeleton-based multiple point statistics. The core idea is using the skeletons of reservoir objects to restrict the selection of data patterns. So the algorithm of skeleton-based multiple point statistics consists of two parts, firstly, constructing the skeleton of the reservoir objects; secondly, forecasting the distributions of reservoir objects using multiple point statistics. The paper proves the skeleton-based multiple point statistics can reproduce the continuous and curvilinear reservoir objects through the modeling of several conceptual fluvial models. During the tests of the skeleton-based multiple point statistics, the paper points out the new method has the ability of solving stationary problem by reservoir skeleton, which has been puzzling geostatistical scientists for years.