A novel unstructured-grid generator using upscaling applied to a compositional reservoir simulator

This work introduces a novel unstructured grid generator using an upscaling technique and the Element-based Finite Volume Method (EbFVM). The proposed grid generator can either replicate the original structured Corner Point grid as an unstructured mesh or coarsen it by upscaling reservoir properties for computational efficiency. The proposed methodology enables flexible grid coarsening from structured geological models while preserving numerical accuracy in critical flow regions. The approach integrates a user-defined clustering algorithm with a Dykstra-Parsons coefficient to generate hybrid-resolution grids. A modified Cardwell-Parsons method is used for permeability upscaling and volumetric averaging for porosity. An unstructured grid option has been implemented into the UTCOMPRS simulator, and the grid generator is evaluated and validated through five case studies, including realistic reservoir models and industry benchmarks such as UNISIM-I, model 2 of the SPE10 case, and the Sleipner CO 2 storage project using the UTCOMPRS. Results demonstrate the approach’s ability to significantly reduce computational costs while maintaining reliable production forecasts. This work presents a compelling and computationally efficient tool for reservoir simulation.