Capillary Pressure and Permeability Prediction in Carbonate Rocks - New Methods for Fractures Detection and Accurate Matrix Properties Prediction

Abstract

Abstract Carbonate rocks are known for their heterogeneity and petrophysical complexity. Core-measured permeability values can span orders of magnitude for a given porosity due to the diversity of pore types, as well as the propensity for carbonate rocks to contain (micro)fractures. Multimodal pore throat size distributions resulting from capillary pressure measurements are also an expression of the carbonate rock complexity. These complexities present challenges for accurate prediction of permeability and saturation, and the resulting uncertainties in reservoir models limit the reliability of fluid storage and flow predictions. Special issues arise when core porosity-permeability values are quality checked for low porosities (typically < 5 pu) as errors could have profound effects, especially in "tight" formations. Given these challenges, the goal for this work is a more accurate methodology for permeability prediction in carbonates using pore throat data to develop a quantitative technique that allows differentiation between matrix and fractures properties in porosity-permeability data. Our previous work showed that use of a normalized pore throat radius parameter yields a straight line relationship with permeability data over 6 orders of magnitude. Accuracy of the technique is validated when comparing predicted and measured permeability values from samples from diverse carbonate reservoirs from various periods of geologic time. An important result is that the method can define a valid permeability range, covering a wide range of pore types in carbonate rocks. Permeability values that exceed this valid range are most likely fracture-influenced (artificial or natural). Screening criteria for identification of fractured samples are independently validated by comparing unstressed and stressed permeability data. These newly developed methods are applied to conventional core data, leading to an improved accuracy of the modeled matrix properties.