Conditional Simulation of Complex Geological Structures Using Multiple-Point Statistics

Conditional Simulation of Complex Geological Structures Using Multiple-Point Statistics

January 2002 | Sebastien Strebelle
The paper "Conditional Simulation of Complex Geological Structures Using Multiple-Point Statistics" by Sebastien Strebelle addresses the challenges of modeling curvilinear geological structures, such as sand channels in clastic reservoirs, using traditional two-point statistics. The proposed approach leverages multiple-point statistics (MPS) from training images to infer joint variability at three or more points simultaneously. This method is more effective than traditional pixel-based algorithms, which struggle with random geometries and cannot accurately reproduce complex structures. The algorithm is tested on a fluvial hydrocarbon reservoir with meandering channels, demonstrating its simplicity, generality, and efficiency in handling large 3D simulation grids. The paper also reviews existing methods, including parametrization of specific shapes, Boolean object-based algorithms, and pixel-based approaches like simulated annealing and Markov Chain Monte Carlo (MCMC) simulation, highlighting their limitations and the advantages of the proposed MPS approach.The paper "Conditional Simulation of Complex Geological Structures Using Multiple-Point Statistics" by Sebastien Strebelle addresses the challenges of modeling curvilinear geological structures, such as sand channels in clastic reservoirs, using traditional two-point statistics. The proposed approach leverages multiple-point statistics (MPS) from training images to infer joint variability at three or more points simultaneously. This method is more effective than traditional pixel-based algorithms, which struggle with random geometries and cannot accurately reproduce complex structures. The algorithm is tested on a fluvial hydrocarbon reservoir with meandering channels, demonstrating its simplicity, generality, and efficiency in handling large 3D simulation grids. The paper also reviews existing methods, including parametrization of specific shapes, Boolean object-based algorithms, and pixel-based approaches like simulated annealing and Markov Chain Monte Carlo (MCMC) simulation, highlighting their limitations and the advantages of the proposed MPS approach.
Reach us at info@study.space