How is Multiple-Point Geostatistics of Lithofacies Modeling Assisting for Fast History Matching? A Case Study from a Sand-Rich Fluvial Depositional Environment of Zubair Formation in South Rumaila Oil Field

Abstract

AbstractModeling accurate lithofacies and petrophysical properties is a crucial step in the reservoir characterization as it affects reservoir heterogeneity, fluid flow modeling, and history matching, especially in complex geological structures. In this paper, the multiple-point facies geostatistics (MPFG) and sequential gaussian simulation were integrated as an efficient workflow for lithofacies and petrophysical property modeling of a fluvial sand-rich depositional environment of Zubair formation in South Rumaila oil field, Southern Iraq. The lithofacies features of the upper sandstone member has three main lithotypes derived from the core data analysis of 20 wells: sand, shaly sand, and shale.In the MPFG, the surface map of the fluvial depositional system of the upper sandstone reservoir in Zubair formation was created through a 2D user-defined training image. The training image body and channels were pointed to the three aforementioned lithofacies as an alternative to the variogram to create the 3D facies system. Then, the surface map was sampled and trained by neural networks to create the discrete template of 3D facies distribution pattern into the 3D grid construction. The resulted pattern represents a numerical geomodel that captures all the features of the fluvial depositional environment of the reservoir, which then was adopted for 3D lithofacies modeling.The resulting MPFG-lithofacies model reflected a more reasonable facies architecture than the sequential indicator simulation by preserving the fluvial features of the geosystem. Many realizations were generated and cross-validated to determine the most appropriate lithofacies model, which was considered later for the permeability and porosity modeling by the sequential gaussian simulation. To attain history matching, the resulting MPFG and petrophysical model was upscaled and incorporated into the compositional reservoir flow simulation for history matching. A near-perfect and fast history matching with the least mismatch was obtained with respect to observed and calculated cumulative and rates of oil production and water injection for the entire field in addition to all producers and injectors within the whole production history. The results reflect how is efficient considering multiple-point statistics to reconstruct the complex geological features to capture reservoir heterogeneity and achieve fast history matching.