Multiscale Modeling of Meandering Fluvial Reservoir Architecture Based on Multiple-Point Geostatistics: A Case Study of the Minghuazhen Formation, Yangerzhuang Oilfield, Bohai Bay Basin, China

Abstract

Meandering river reservoirs are essential targets for hydrocarbon exploration, although their characterization can be complex due to their multiscale heterogeneity. Multipoint geostatistics (MPS) has advantages in establishing reservoir architectural models. Training image (TI) stationarity is the main factor limiting the uptake of MPS modeling algorithms in subsurface modeling. A modeling workflow was designed to reproduce the distribution of heterogeneities at different scales in the Miocene Minghuazhen Formation of the Yangerzhuang Oilfield in the Bohai Bay Basin. Two TIs are established for different scales of architecture. An initial unconditional model generated with a process-based simulation method is used as the megascale TI. The mesoscale TI of the lateral accretion layers is characterized by an uneven spatial distribution of mudstone in length, thickness, frequency, and spacing. Models of different scales are combined by the probability cube obtained by lateral accretion azimuthal data as an auxiliary variable. Moreover, the permeability function sets are more suitable than the porosity model for collaboratively simulating the permeability model. Model verification suggests this workflow can accurately realize the multiscale stochastic simulation of channels, point bars, and lateral accretion layers of meandering fluvial reservoirs. The produced model conforms geologically realistically and enables the prediction of interwell permeability variation to enhance oil recovery.