A Scalable Massively Parallel Dual-Porosity Dual-Permeability Simulator for Fractured Reservoirs with Super-K Permeability

Abstract

Abstract Modeling Middle East giant carbonate reservoirs with fractures and super-k conductivity present unique challenges to conventional simulation approaches. Fractures in these giant oil fields can exist as fracture swamp neighboring fault zones. These regional/local fractures can interact with Super-K layers (high permeability stratiform) which form extremely high conductivity to fluid flow. Typically, meaningful simulation for such an oil field requires the use of millions of grid cells involving thousands of wells. These reservoirs are not the classical dual porosity where matrix permeability is poor and does not form a part of the global flow path. Dual porosity dual permeability (DPDP) formulation allowing a flexible regional representation of both single porosity and dual porosity behavior on a cell by cell basis is more suitable. A parallel dual porosity dual permeability simulator has been developed to efficiently solve multi-million grid cell fractured reservoirs problems with super-k fracture permeability. To be effective, the simulator has to engage tens to hundreds of processors in a highly scalable manner. This paper discusses the formulation, the data design and solution procedures involved in this development. The parallelization paradigm is the mixed approach with MPI (Message Passing Interface) and OpenMP (Open Message Passing). These parallelization approaches are open standards supported by most major hardware vendors. This allows for easy portability among various hardware architectures. However, the data design and solution methods discussed herein are amenable to MPI only or other parallelization approaches as well. The mixed paradigm approach is flexible and allows various combination and permutation of MPI processes and OpenMP threads to be used to solve a problem. Results are presented for a multi-million cell fractured reservoir simulation with super-k involving thousands of wells and over 60 years of history match. We present simulation results, computational efficiency and parallel scalability on the IBM Nighthawk II, as well as on the PC Xeon Linux cluster hardware.