Predicting the Distribution of Petrophysical Properties in Carbonate Reservoirs Using Process-Based Forward Modeling

Abstract In carbonate reservoir studies, the results of traditional geostatistical distribution of petrophysical properties in reservoir models remain highly uncertain. Process-based forward modelling is a promising way of taking into account all the geological processes forming and then affecting the pore network, and thus of predicting the distribution of petrophysical properties. Here is proposed a global and efficient forward modeling workflow that jointly simulates the genesis of both geological and petrophysical properties of carbonate reservoirs, from deposition to the present day. Computational efficiency is ensured by focusing on the processes that mostly affect the heterogeneity of petrophysical properties. Given the high impact of early diagenesis, depositional processes and synsedimentary diagenetic processes are simulated jointly rather than sequentially. Finally, later diagenesis processes, such as karst formation, which require consideration of fluid flows in the reservoir, itself deformed and fractured during its structural history, are simulated. Initial simulations have yielded realistic and promising results. Simulating the various synsedimentary processes makes it possible to produce models whose overall organization of petrophysical properties is linked not only to sedimentary facies, but also to changes in environmental conditions, such as hydrological and diagenetic zoning. As well, the simulation of dissolution processes linked to correctly represented paleofluid circulations makes it possible to produce realistic models with heterogeneous triple-medium characteristics. Because it preserves geological causality and not simply interpolates between conditioning data, the proposed software enables the present days petrophysical properties of carbonate reservoirs to be actually predicted. It therefore disrupts conventional workflows and allows the data to be used to falsify modeling assumptions, reduce significant initial uncertainty and calibrate predictions.