Abstract

In this study, we present a procedure for reservoir property modeling in a channelized carbonate reservoir based on hierarchical geostatistical simulation and seismic data integration. Because soft data integration in facies modeling has always been challenging, we used an innovative approach to incorporate seismic data in facies simulations properly. In this regard, we produced a facies proportion model (FPM) using sequential Gaussian co-simulation of facies proportion as primary data and acoustic impedance as secondary data. The facies proportions were extracted from vertical proportion curves, and the impact of seismic data in facies simulation was determined with the help of correlation coefficient maps. An alternative type of seismic-based soft data was also derived using a supervised neural network to create a facies probability cube (FPC) for each facies. Afterward, pluri-Gaussian simulation (PGS) was implemented to these two prepared soft datasets, and consequently, porosity was simulated twice in each of the models-with and without seismic-derived secondary data. Histogram analysis showed that the facies modeled with the PGS-FPM method reproduced the original well data better than the PGS-FPC method. In addition, blind wells validation showed that the PGS-FPM outputs had up to 79% accuracy, while channel geometries were better constructed using the PGS-FPC approach. The difference between the reservoir quality of channel and background was distinguishable in all hierarchical simulated porosity results. At the same time, the predicted results from simulated porosity in PGS-FPM facies had stronger correlation with true values in blind wells.