Special Connections for Representing Multiscale Heterogeneities in Reservoir Simulation

Abstract

Abstract The significant quantities of oil contained in fractured karst reservoirs in Brazilian pre-salt fields adds new challenges to the development of upscaling procedures to reduce time on numerical simulations. This work aims to represent multiscale heterogeneities in reservoir simulators based on special connections between matrix, karst, and fracture mediums, both modeled in different grid domains within a single porosity flow model. The objective of this representation is a good balance between accuracy and simulation time. The principle combines the Embedded Discrete Karst Model (EDKM) and the Special Connection Fracture Model (SCFM), both developed by Correia et al, 2019. Therefore, this work extends the approach of special connections to integrate both karst and fracture mediums modeled in different grid domains and block scales. For a proper validation and numerical verification, we applied the methodology to a complex case study. The reference model is a refined grid model with: (1) an arrangement of large conduits (karsts), which are poorly connected, (2) a well-connected and orthogonal system of fractures and (3) a background medium (matrix). For the simulation model, the matrix grid domain has a grid block size of roughly 25 × 25 × 5 meters. The fracture domain has a grid block size of roughly 50 × 50 × 5 meters and the karst grid domain has the same block size as the reference model. Well-connected fractures are upscaled to the fracture domain by applying the Oda (1985) method. The results show a significant performance improvement regarding a dynamic matching response with the reference model, within a suitable simulation time and maintaining the dynamic resolution according to the representative elementary volume of heterogeneities, without using an unconstructed grid. Therefore, this work introduces a new method that can be applied to commercial flow simulators concerning the use of special connections to represent multiscale heterogeneities in flow simulators, without disregarding the accuracy over upscaling and an adequate simulation time.