Permeability Prediction in Tight Carbonate Rocks using Capillary Pressure Measurements

The prediction of permeability in tight carbonate reservoirs presents ever more of a challenge in the hydrocarbon industry today. It is the aim of this paper to ascertain which models have the capacity to predict permeability reliably in tight carbonates, and to develop a new one, if required. This paper presents (i) the results of laboratory Klinkenberg-corrected pulse decay measurements of carbonates with permeabilities in the range 65 nD to 0.7 mD, (ii) use of the data to assess the performance of 16 permeability prediction models, (iii) the development of an improved prediction model for tight carbonate rocks, and (iv) its validation using an independent data set. Initial measurements including porosity, permeability and mercury injection capillary pressure measurements (MICP) were carried out on a suite of samples of Kometan limestone from the Kurdistan region of Iraq. The prediction performance of sixteen different percolation-type and Poiseuille-type permeability prediction models were analysed with the measured data. Analysis of the eight best models is included in this paper and the analysis of the remainder is provided in supplementary material. Some of the models were developed especially for tight gas sands, while many were not. Critically, none were developed for tight gas carbonates. Predictably then, the best prediction was obtained from the generic model and the RGPZ models (R2 = 0.923, 0.920 and 0.915, respectively), with other models performing extremely badly. In an attempt to provide a better model for use with tight carbonates, we have developed a new model based on the RGPZ theoretical model by adding an empirical scaling parameter to account for the relationship between grain size and pore throat size in carbonates. The generic model, the 28 new RGPZ Carbonate model and the two original RGPZ models have been tested against independent data from a suite of 42 samples of tight Solnhofen carbonates. All four models performed very creditably with the generic and the new RGPZ Carbonate models performing well (R2 = 0.840 and 0.799, respectively). It is clear from this study that the blind application of conventional permeability prediction techniques to carbonates, and particularly to tight carbonates, will lead to gross errors and that the development of new methods that are specific to tight carbonates is unavoidable.