Evaluation of the reliability of prediction of petrophysical data through imagery and pore network modelling

Prediction of petrophysical data by pore network imaging and modeling has recently received a lot of attention. This technique was pioneered by P.E. Øren but several other solutions have now been proposed that incorporate different imaging technologies, different methods for obtaining detailed pore space models, and/or different algorithms for extracting simplified networks for multi-phase flow simulations. However, very few comparisons between these techniques have been made to date. This paper compares the approaches proposed by two research teams: The Australian National University (X-ray imagery reconstruction) and Numerical Rocks (Digitized thin section and geologically based reconstruction) through a blind test. For this blind test, very different samples with a large range of petrophysical properties were shipped to the collaborating teams. Then they constructed a 3D digital model of the rocks, as well as simplified pore networks, with their own techniques. AU the multiphase flow simulations were performed on the simplified networks with the flow simulators from P. Valvatne, M. Piri and Numerical Rocks (eCore). Finally all the simulated results were compared to experimental inhouse data measured in the Total laboratory. According to this comparison, we conclude that the porosity and classical capillary pressure (i.e. L-shaped) are correctly predicted when the clay content and/ or the micro-porosity are well captured. In addition, uncertainty in absolute permeability can be attributed to voxel precision for ANU and the capability to capture heterogeneities for both approaches. Regarding flow simulations, the gas/oil k_rs are reliable when all the previous parameters are well reproduced, however it is impossible to predict residual gas saturation to water. In summary, the ability to capture micro-porosity and heterogeneities seems to limit the capacity of prediction. Consequently the combined techniques of imagery and PNM cannot currently be considered as an industrial tool on the whole range of rock investigated in petroleum engineering.