Enhancement of Simulation CPU Time of Reactive-Transport Flow in Porous Media: Adaptive Tolerance and Mixing Zone-Based Approach

In recent decades, advances in aqueous geochemical description incorporated with petrophysical characterization allowed researchers to improve reactive-transport flow modeling significantly. However, detailed simulations are usually associated with an increase in computational cost. This work aims to propose a method to decrease the CPU time of reactive-transport simulations in porous media. Conventionally, reactive-transport reservoir simulators perform geochemical computation in all discretized gridblocks at each time step. We hypothesize that if geochemical calculations can be ignored for regions under equilibrium conditions (far from mixing zone), then the computational performance would be improved without compromising simulation results. To test this supposition, we propose a new speedup approach that can be used in conjunction with parallel simulation. The scheme consists of evaluating relative changes in aqueous molar concentration over time. Hence, if the changes are below a certain tolerance, we assume that the reactive task’s effect is negligible. As a consequence, fewer geochemical computations are performed. Moreover, we propose an adaptive tolerance approach to improve simulation accuracy for cases with sharp concentration changes. We test the speedup scheme for three simulation cases with different complexities (e.g., heterogeneity, rock-fluid interactions, gas solubilization). Simulation results indicate a significant reduction in run time for all tested simulations with speedup improvement from 3 to 15 times and excellent accuracy. Furthermore, we recommend that a reliable reference tolerance is approximately 5.0%. In conjunction with parallel simulation, the proposed scheme significantly decreases the required number of processors needed to reduce the CPU time (from hundreds to below ten cores).