Generation of Saturation Functions from Historic SCAL Data for Simulation of a Super-Giant Carbonate Reservoir

Abstract

Abstract Modeling of water saturation during simulation of this middle-eastern super giant carbonate reservoir is quite challenging due to its geological complexities. This reservoir is multilayered, thick and very heterogeneous with permeabilities ranging from 0.01 mD to several Darcies. Consequently, the log-derived water saturations as functions of height above the free water level span a wide range in values. For poorer quality reservoir, the range is larger even when the Sw(z) values are subdivided into their respective rock type groups. A database of experimental SCAL data acquired over the last 50 years was utilized to prepare saturation functions to initialize the water saturation in the simulation model. A workflow for generating internally consistent sets of saturation functions for simulation input from capillary pressure and relative permeability data (Kralik, J.G., et. al., 2010) was applied to the SCAL database. The data were reviewed to confirm that they were of acceptable quality for use in saturation function development and organized by rock types. For each rock type, saturation endpoints were first established by curve fitting the appropriate capillary pressure data to a consistent constitutive equation that contains the saturation endpoint as a regression parameter. The resulting primary drainage water-oil Pc model was checked against the available log Sw(z) data, especially those at the top of the formation. Once the saturation endpoints had been established, relative permeability saturation functions were derived, starting with the bounding water-oil primary drainage curves, followed by the primary imbibition krow-krw curves. Three-phase gas-oil primary drainage krog-krg bounding curves were derived in an analogous manner. Because the imbibition krow-krw tests correspond to unique imbibition scanning curves, they were scaled to the bounding primary drainage kro curve to ensure consistency. This process was applied to each group of capillary pressure and relative permeability data. Data gaps were filled with suitable analog SCAL data taken from the rock type groups that were the most similar. The process used above provided for self-consistent primary drainage and imbibition functions. The prepared saturation functions (J-function and relative permeability) were input into the simulation model. The quality of the water saturation match as predicted by the simulation model was checked against log data for 120 wells. A good initialization of the model was achieved with some exceptions. The relatively poor quality unit at the bottom of the reservoir showed a mismatch, especially in the stylolitic intervals. The main mismatch was due to the larger range in permeability values (i.e., two cycles on logarithmic scale) for the same rock type. One of the rock types was further subdivided into distinct rock types for the porous and the stylolitic units. This modification greatly improved the water saturation match of the model. Model Sw was further improved in the northern part of the reservoir by mapping the impermeable dense layers deterministically.