The inclusion of dynamic constraints in stochastic conditional simulation

Throughout the life of a petroleum reservoir, information about the reservoir--typically seismic, core and log data, rates and pressures--is collected. These data represent perspectives of the reservoir at widely-differing scales. The challenge therefore, is to integrate these different scales of data into a reservoir description process while honoring the scale level they represent. In addition, algorithmic efficiency may prevent the consideration of certain types of data, for example 'dynamic' data, such as rate and pressure data. These represent valuable information because they represent a heterogeneity scale which is closer to simulation gridblock scales than the typical well core/log data. So far, dynamic data have been used on a limited basis because it then becomes necessary to use flow simulation, making the algorithm inefficient.;An efficient algorithm for using stochastic conditional simulation in which we directly incorporate dynamic information--namely, rate and pressure information is presented. Modeling the spatial description on a fine scale and the flow on an upscaled grid reduces the flow simulation execution time and allows for a faster conditional simulation algorithm. Upscaling approaches are described which result in flow performance matching between the fine or 'true' scale and the upscaled grid. The simulated annealing method is used with a 2-part objective function consisting of a variogram constraint and a flow simulation constraint with each part being appropriately weighted.;Improvements in the reservoir description when the dynamic constraint is included are demonstrated. Also the upscaling techniques used are shown to be effective in matching the flow performance of the reservoir grid between scales. The procedure simultaneously generates fine scale description which can be used for future evaluation, and a coarse scale description which honors the production data. Using this procedure, up to a 10,000-gridblock description has been simulated.