The author developed and applied a simple numerical scheme for simulating a two-phase hydrothermal system by using the finite volume method. This scheme consists of two simple procedures. In the first procedure, two governing equations representing the conservation of mass and energy are integrated explicitly with time, yielding the total fluid mass
M and total enthalpy
H in the next time step. In the second procedure, pressure
P, temperature
T, and liquid saturation
S in the next time step are calculated by solving two nonlinear equations with the total fluid mass
M and total enthalpy
H obtained in the first procedure. These two nonlinear equations define
M and
H under the assumptions of thermal equilibrium among water, steam, and rock and phase equilibrium between water and steam.
Numerical simulations for one- and two-dimensional problems were performed. The results of these simulations were consistent with those obtained using the HYDROTHERM numerical simulator. For speeding up the simulations, parallel computing was carried out using a PC cluster containing eight CPU cores. If the two-phase region, for which a larger calculation load than that for the single-phase region was required, was evenly distributed, the efficiency of parallelization increased with a smaller overhead. Otherwise, the efficiency of parallelization tended to decrease by the use of a domain decomposition method under a distributed memory environment. Shared memory computers are useful because not only they are advantageous in that they can be easily programmed but also they can achieve higher efficiency of parallelization.
The upper limit of the length of the time step for maintaining numerical stability was experimentally investigated. The result of the numerical experiment showed that numerical stability was maintained if the dimensionless parameter
PckΔt/μcφΔx2 , where
Pc is the critical pressure;
k, the permeability;
Δt, the length of the time step;
μc, the critical viscosity;
φ, the porosity; and
Δx, the length of the spatial step, was equal to or less than 10
-2.
The important difference between the scheme developed in this study and a conventional scheme using implicit time integration is the most costly process involved from programming to calculations. In the developed scheme, calculations are the most costly process because they require a long time. In the conventional scheme, on the other hand, programming is the most costly process because of the complicated algorithm. Researchers intending to perform numerical simulation should decide which scheme to adopt for achieving the desired cost-benefit balance.
View full abstract