GPU-Offloaded General Purpose Simulator for Multiphase Flow in Porous Media

Abstract

Reservoir simulation plays an important role in the petroleum industry. Today, there is a specific demand to run ensembles of megaand even giga-cell models. The iterative solution of the large systems of nonlinear governing equations, which describe the multiphase mass transfer in the subsurface, takes the most of the simulation time. The linearization part of the solution process occupies a significant fraction of that time, especially in compositional models. Moreover, the implementation of the linearization step usually embodies the most substantial, complex, and specific part of the computational loop in modern simulators, defining which physical mechanisms and assumptions are employed. This significantly complicates the implementation of simulation codes for heterogeneous computing hardware, which promises significant improvements in simulation time. In this work, we use the recently proposed Operator-Based Linearization (OBL) approach to develop a general purpose reservoir simulation code aiming to substantially decrease the simulation time. OBL offers a simplified linearization method, enhancing the computational performance of simulation and providing an opportunity of a painless porting to heterogeneous computing architectures. To distinct the contribution of both factors, we developed two versions of the compositional simulation prototype code: for traditional CPU and GPU-accelerated hardware architectures. While the former allowed us to speed the linearization stage up by an order of magnitude in comparison with the conventional approach, based on Automatic Differentiation (AD), the latter improved it further by another order of magnitude. The developed prototype realizes the potential of the OBL approach and GPU computing architecture, proving significant improvement in general purpose simulation performance.

Adaptive Parameterization for Solving of Thermal/Compositional Nonlinear Flow and Transport With Buoyancy

“…The new approach provides almost unlimited flexibility for the extension of the physical model in a simulation framework without the overhead related to the application of the automatic-differentiation technique. The method also constitutes a promising option for implementation of a full nonlinear loop on emerging architectures, for example on GPU, as shown by Khait et al (2017).…”

“…However, the necessity of artificial-data injection (values and derivatives obtained by interpolation) back to automatic-differentiation-based data structures negatively affects the performance. That can be avoided if a standalone simulator is implemented entirely from the OBL perspective, as shown by Khait et al (2017) For all simulations, we used the same discretization operators (Eqs. 6 and 8) derived from the seventh layer of the SPE10 model (Christie and Blunt 2001) shown in Fig.…”

“…days, which corresponds to average Courant-Friedrichs-Lewy (CFL) condition ¼ 5.3. The pressure/volume/temperature properties and relative permeabilities were used from the SPE9 test case (Killough 1995).…”

“…As mentioned previously, the proper implementation of the OBL approach can speed up Jacobian assembly by a factor of 14. Moreover, the migration and optimization of algorithms for emerging architectures (e.g., graphics processing unit or GPU) improve the linearization performance by another order of magnitude (Khait et al 2017).…”

A GPU-Based Integrated Simulation Framework for Modelling of Complex Subsurface Applications

“…In reservoir simulation, the Automatically Differentiable Expression Templates Library (ADETL) was introduced by Younis (2011). Being attractive from the perspective of flexibility, the AD technique by design inherits computational overhead, which affects the performance of reservoir simulation (Khait and Voskov, 2017a).…”