Reservoir Simulation using two different computer systems is investigated:
Wizard Adapter - an Intel i860 implemented on a micro channel card. The adapter was installed in an IBM PS/2 Model 80.
RISC System/6000 - IBM's second generation RISC workstation series.
A reservoir simulator with demonstrated high performance is used to test these machines.
Both systems can perform reservoir simulations at more than ten times the speed of an Intel 80486 and at a significant fraction of the speed of a supercomputer. With an efficient simulator and enough memory, both of these processors can solve compositional problems with 40,000 grid points and black oil problems with 200,000 grid points. With starting prices of less than $15,000 for a complete machine, these systems provide an attractive alternative to mainframes and supercomputers.
This York uses a reservoir simulator with demonstrated high performance on a number of vector computers1 and current state-of-the-art scalar microcomputers.2 Reservoir simulations were conducted using the two computer systems. Both systems use RISC (Reduced Instruction Set Computer) design concepts. Although both the i860 and RS/6000 are based on RISC technology, the processors obtain high performance in fundamentally different ways. The i860 emulates the design of the CRAY - l supercomputer.3 The Rs/6000 is based on IBM's second generation "super-scalar" RISC technology.
Both equation-of-state (EOS) compositional simulations and black oil simulations are used in this work. Compositional simulators model reservoir processes where mass transfer between the oil and gas phases is significant. Compositional effects are important in the recovery of gas condensates and volatile oils and in most enhanced recovery processes. The injection of carbon dioxide, nitrogen or lean natural gas inevitably cause significant component interchange between phases. Due to the importance of these processes, there is an increasing need for compositional reservoir simulation.
Compositional simulators represent the hydrocarbon fluids by a number of components, typically 6 to 10. Equilibrium between the hydrocarbon gas and liquid phases is calculated using an EOS such as the Peng-Robinson5 or Redlich-Kwong.6 At each time step, the phase equilibria calculations are performed at every grid point infinite difference grid and are coupled to the discretized material balances which govern fluid flow. Due to the large number of grid points and the complexity of the EOS, the calculations are computationally intensive.
In situations where the complexity of an EOS is not required, a Black Oil simulator can be used. A black oil simulator represents the hydrocarbon fluid by only two components, namely stock tank oil and separator gas. The solubility of the gas component in the oil phase is defined by the solution gas oil ratio. The formation volume factors specify the volumetric properties of the fluid. Formation volume factors, solution gas oil ratio and viscosities are all tabular functions of pressure. When the oil is under saturated, the formation volume factor and viscosity are treated as linear functions of pressure.
In the past, reservoir simulation studies have required so much computation time that only simple fluid descriptions, coarse grids or small reservoir segments could be considered.