Recent high-resolution tools and novel characterization techniques lead to simulate very detailed geological descriptions of heterogeneous reservoirs. To perform full-field displacement simulations, scaling-up techniques must be used because of the required coarser grids. Pseudoization methods are commonly used to get the relative permeability and capillary pressure curves. But they all consider 2D flow to realize the scaling-up. We propose a three-dimensional approach. When capillary forces are important inside the reservoir, the capillary pressure may be considered as nearly constant inside a coarse grid-block. This assumption enables to compute directly the fine grid-block saturations and the average coarse grid-block saturation for a given capillary pressure. For each fluid phase, we then compute the effective phase permeabilities of the coarse grid-block by means of an absolute permeability scaling-up method applied to phase permeabilities of the fine grid-blocks in 3D. Then renormalizing the resulting phase permeabilities by the average absolute permeability yields the average relative permeabilities curves for the three directions of space. Application of this technique, using an algebraic computation of average permeability adapted to heterogeneous media, will be presented for complex fluviatile cases, characterized by a large amount of shale inclusion at small-scale. Comparison of numerical simulations of oil-water flow on high-resolution descriptions, and on the corresponding scaled-up coarse models, shows the efficiency of this method. We recommend to apply this method to reservoir with low gravity segregation, with rocks exhibiting medium to low permeability and high capillary forces, and moderate injection rates.

A good understanding of the internal geometry of a reservoir is necessary for making realistic oil recovery forecasts. In answer to this need, software has been developed to generate 3D lithological images of a reservoir by using a geostatistical approach. Based on previous images, a petrophysical description is then previous images, a petrophysical description is then obtained on a geologic mesh scale. And to perform flow simulation, petrophysical parameters must then be scaled up to reservoir mesh scale. This paper shows an application of the complete procedure of processing from geostatistical imaging to fluid flow modeling. Introduction When dealing with production forecasts in heterogeneous reservoirs, petroleum engineers are rapidly faced with the problem of the description of the internal geometry of reservoirs. Geostatistical methods are of great use in providing images of the reservoir. Therefore a complete procedure has been developed using three complementary software, including the 3D lithological and petrophysical description of the reservoir, the flow simulation, and the visualization of the results. This integrated data processing approach aims at making easier work dealing with several disciplines: geology, geostatistics, and reservoir engineering. And the software tools used for a reservoir study have to be efficiently connected to each other to help the different specialists to work together. Each step of this procedure is explained here, and an application example is described. 2. DESCRIPTION OF THE PROCEDURE FROM GEOSTATISTICAL IMAGING TO FLUID FLOW SIMULATION The main steps of the procedure are the following:*The geostatistical simulator, called HERESIM 3D, for HETEROGENEITIES-RESERVOIR-SIMULATION in 3 dimensions of space, leads to a petrophysical characterization of the reservoir. Its main feature is to establish the link between lithological and petrophysical well data and fluid flow reservoir petrophysical well data and fluid flow reservoir simulators, using conditional geostatistical simulations and petrophysical parameter scaling-up methods.*The second step is flow simulation, performed with a conventional flow simulator (SCORE).*Then, a graphic software is used to visualize the flow simulation results. Data exchange between the 3 software is operated through files having a standard data structure. All the reservoir characterization steps are performed on a work station connected to a network of other work stations and mainframe computers. And flow simulation can be performed either on the work station or on a mainframe computer. 3 CHARACTERISTICS OF THE GEOSTATISTICAL SOFTWARE The geostatistical simulator was developed to establish a link between the geologic model and fluid flow simulations. The main philosophy behind this software is that a correct reservoir characterization can be achieved only through a realistic geologic description. The four main tasks of this modular software are reviewed below. P. 223

An integrated study combining sedimentological and sequence stratigraphy analysis, geostatistical simulations and validation by production history simulation has been performed in the Chaunoy field reser:oir. The reservoi~is a part of a distal alluvial fan system, In the Keuper senes of the Paris Basin (France). RFT surveys suggest a layered behavior. Nevertheless, the low extension of reservoir bodies (small ribbon channels interbedded with flood-plain and lacustrine mudstones) made it difficult to set ut a satisfactory correlation scheme in the past studies, resulting in difficulties in matching the production history.Core and log analysis yielded the conceptual geological model. Geostatistical analysis enabled to its validation and quantification. Inside each genetic unit, repartition of lithotypes and of petrophysical properties were modelled by geostatistical simulations. As the horizontal~ariogram ranges proved to be below the avera~e w~ll spacing: three high resolution models were built, WIth dIfferent honzontal variogram ranges, all smaller than well-spacing (118, 1/2 or 1 times the well spacing). After scaling-up of the resulting petrophysical J:P:od~ls, siIl1lI1ations of the ten-year pro.duction history were performed in each model. By companng the simulated production data with the field production history, the characterization study was validated. A satisfactory match was obtained on the main zone of the reservoir, validating the correlation scheme and explaining the observed layered behavior.

Various geostatistical models have been proposed to generate possible descriptions of the internal structure of heterogeneous resesvoirs. For actual reservoir engineering studies, there are practical problems in dealing with large

No abstract