Adaptive Local-Global Multiphase Upscaling for Improved Coarse-Grid Modelling of Condensate Banking Phenomena

Abstract

In this paper, we show a new upscaling process to accurately predict the production performance of gas-condensate reservoirs, accounting for condensate banking phenomena, using a coarsely-gridded black-oil simulation model. In reservoir simulation, it has been necessary to conduct time-consuming fine-grid compositional simulation to reasonably predict the reservoir pressure profile and condensate banking near the wellbore. A newly-developed in-house program utilizes the Adaptive Local-Global Multiphase Upscaling (ALGMP) methodology developed by Nakashima et al. (2012)1. The main feature of the program is that both the entire model domain (global boundary) and local domain (well vicinity) are taken into account in the upscaling process to retain the accuracy of prediction within a relatively short computation time. The entire model domain was reflected in the form of a coarse-grid full-field model, from which a near-wellbore area of interest was extracted for fine-grid compositional modelling (using velocity-dependent relative permeability (VDRP) to represent relative permeability improvement around wellbore caused by high velocity and/or low interfacial tension (IFT)), applying common flux boundary conditions at each time-step. In the coarse-grid full-field model, mobility change is mainly due to the near-wellbore condensate drop-out observed in the fine-grid compositional model. This was reflected by an iterative upscaling of the gas-condensate relative permeability and fluid properties, using an optimization loop over the local domain, to match the well phase flow rates in the coarse-grid black-oil model with that in the fine-grid compositional model, within an acceptable margin. The developed program was applied to a real gas-condensate field model and demonstrated that the ALGMP procedure can provide well performance predictions close to the reference fine-grid simulation results with reduced computational time. The accuracy of the ALGMP procedure is higher than that of a standard single-phase upscaling approach.