Predicting Effective Permeability to Oil in Sandstone and Carbonate Reservoirs From Well-Logging Data

Abstract

Summary This paper discusses a new method for predicting continuous logs of effective permeability to oil in sandstone and carbonate formations from well-logging data. This important problem in formation evaluation has remained previously unsolved because the conventional approach used to predict reservoir properties from well-log data relies on simple empirical equations and idealized models of reservoir rocks. The use of such equations and models to predict a complex reservoir property (e.g., effective permeability) can result in an inaccurate representation of the formation. This paper shows that accurate values of effective permeability to oil can be predicted using model-independent mapping functions constructed from Gaussian radial basis functions (RBFs), which can be derived from a laboratory database of measurements on partially saturated core samples. The mapping functions replace the empirical equations used in the conventional approach and are model-independent representations of the database measurements. Once the mapping functions are constructed from the database, there are no adjustable parameters. In this study, mapping functions for sandstones and carbonates were derived from a worldwide database of laboratory measurements made on 79 sandstone and 25 carbonate core samples. The laboratory measurements available on each sample included irreducible water saturation, effective permeability to oil, porosity, and nuclear-magnetic-resonance (NMR) T2 distributions. The mapping functions can be used to predict effective permeability to oil of reservoirs at irreducible water saturation from input measurements of porosity, T2 distribution, irreducible water saturation, and knowledge of rock lithology (e.g., sandstone or carbonate). The methodology for deriving the mapping functions is explained in detail. The mapping functions are applied to the database itself to show the accuracy of the effective-permeability predictions on core samples. The method is also applied to log data from both sandstone and carbonate formations in three wells from different parts of the world. The predicted effective permeabilities to oil are shown to be consistent with oil mobilities measured in the formations by fluid-sampling tools.