Object-Based modeling of fluvial/deepwater reservoirs with fast data conditioning: methodology and case studies

摘要

This paper presents a new method for fast fluvial and deepwater stochastic simulation. Sinusoidal channels and attached levees and crevasses splays are generated to honor global net-to-gross ratios, vertical and areal proportions and well data. Fast conditioning to well data is achieved without resorting to time-consuming iterative procedures such as simulated annealing. The undulating channel centerline is created by 1-D conditional sequential Gaussian simulation. Direct conditioning is also used for the reproduction of 2-D areal proportions. The method has been applied to a fluvial reservoir in the North Sea with 16 wells and to a braided channel reservoir with 110 wells in Mexico. The dependence of CPU time to a number of different factors was explored. The CPU time for the proposed method and annealing based algorithm are compared. The method with direct well conditioning is an order of magnitude faster than annealing based method.