Accelerating Compositional Reservoir Simulation with GPUs and Advanced Solvers

Numerical reservoir simulation is the primary tool a reservoir engineer uses to predict future production from hydrocarbon reservoirs and to develop a plan to maximize production at minimal cost. Compositional reservoir simulation, in particular, is required to have an accurate physical description of the subsurface oil and gas in development situations involving gas injection, condensate production, or the coupling of reservoirs with multiple oil qualities. Because this simulation is extremely computationally intensive, engineers are often forced to simplify the models of their reservoirs, sacrificing accuracy and predictive capacity in order to allow their simulations to run in a practical time frame. To address this performance problem, an extremely high-performance compositional package will be added to the researchers’ existing reservoir simulator, which currently implements a simpler black-oil description of the hydrocarbons. The proposed compositional addition will utilize the exceptional throughput provided by graphics processing units (GPUs) in combination with a novel approach to the implicit solution of the nonlinear partial differential equations to drastically reduce simulation runtime. It is estimated that typical compositional reservoir models will run ten to one hundred times faster than with currently available reservoir simulators, enabling higher accuracy predictions in less time. In Phase I of this project, a compositional equation-of-state module will be developed, validated, optimized for efficient operation on GPUs, and benchmarked against published results. In the first year of Phase II, this module will be integrated into the existing high-performance simulator, and code for the novel nonlinear solution method will be implemented. In the second year of Phase II, the new package will be extensively validated with public benchmarks and proprietary models provided by customers, and the code will be optimized and hardened in preparation for commercial deployment. The advance in speed provided by the compositional reservoir simulator resulting from this project will enable higher quality decisions to be made about optimal field development. This can help increase the ultimate recovery from the nation’s finite petroleum resources. The density and cost effectiveness of the GPU computing platform used for the simulator will also make extremely high performance computing available in a workstation computer. This will make HPC much more accessible to smaller corporations, including engineering consulting firms that cannot afford the capital outlay and maintenance costs of a traditional HPC cluster. New computational methods and software will be developed to allow faster and more accurate simulations of the flow of subsurface water and hydrocarbons in petroleum reservoirs. The resulting reservoir simulation software will enable more efficient production of the nation’s petroleum reserves.