ABSTRACT
Reservoir modeling of the Chaunoy field was performed by combining a sedimentological study, a sequence stratigraphic analysis, geostatistical simulations, and the analysis of production data and fluid-flow simulations. The reservoir corresponds to the distal part of a Middle Triassic alluvial fan system in the Paris basin (France), and is extremely heterogeneous and layered. The reservoir mostly consists of small ribbon channel deposits interbedded with flood-plain and lacustrine mudstones. The channel amalgamation rate varied with cyclic lake-level variations, which directly controlled the reservoir geometry. Within a base-level cycle, during periods of low accommodation, channels were amalgamated, forming highly heterogeneous sand sheets. As the accommodation increased, channels became progressively isolated within flood-plain mudstones. Finally, a lacustrine transgression deposited lacustrine mud-stones and induced thin but widespread vertical permeability barriers across the field. As accommodation started to decrease, considerable pedogenetic alteration occurred, as shown by dolocretes and groundwater dolomites. Five cycles that constituted the reservoir layering framework were identified. Geostatistical simulations of lithotype distribution within these units were computed using the truncated random Gaussian function method. Horizontal and vertical lithotype proportion curves and variograms were calculated from well data. Because of the wide well spacing, it was not possible to determine the range of horizontal experimental variograms. Three lithotype realizations were simulated within a high-resolution grid to compare short, medium, and long correlation lengths. After assigning petrophysical properties to the lithotypes and upscaling, fluid-flow simulations were performed for the three realizations. The three flow simulations were then compared to the 10-yr production history of the field. The simulations showed quite a good match regardless of the variogram range, except in the northern part of the field, indicating a problem in the reservoir layering in this area. This relative insensibility of the flow simulation to the correlation length probably is due to the high net pay within the amalgamated channel reservoir units and to the high number of conditioning wells; however, the flow simulation performed with the longest correlation length showed the best fit with the production history.