GPU-Accelerated Large-Scale Basin and Reservoir Simulation

Project Description

Historically, accurate modeling of oil and gas reservoirs and basins throughout the world has been limited by computational resources, forcing geoscientists and engineers to “upscale” or decimate their datasets in order to fit within computer and time constraints. In other situations, finer resolution was achieved at the expense of “sectoring” the area of interest into smaller parcels, which imposed artificial flow boundaries to the natural flow of fluids in the subsurface. Modern higher-fidelity data acquisition in the field (seismic data, real-time measurements, intelligent-fields and “autonomous”-fields) implies that lots of high-quality data is often under-utilized due to computational constraints, either because of “upscaling” or “sectoring”. One benefit from the “GPU revolution” is the potential to drastically reduce simulation time by an order of magnitude or more while keeping the size and cost of computational resources and power consumption at reasonable levels. This implies that the oil and gas industry no longer needs to compromise model resolution and can fully take advantage of all modern data sources. This will result in more accurate planning of field developments, better positioning of wells, increased oil recovery, better identification of subsurface oil and gas accumulation and better classification of shale potential in yielding oil or gas via hydraulic fracturing. In addition, since we will be dawning into the era of exaflop computing in a few years, new algorithms that harvest GPUs and other new architectures must be tested, not only for performance and scalability but also for hardware stability and overall system resiliency.