Improvement of the Calculation Accuracy of Acid Gas Solubility in Deep Reservoir Brines: Application to the Geological Storage of CO2

摘要

The assessment of the short and long term consequences of CO2 injection in aquifers requires both laboratory experiments and numerical modelling in order to better understand the various physical-chemical processes taking place. Modelling injection in a reservoir, where relatively high temperature (above 50°C), high pressure (several hundreds of bars), and high salinity (greater than that of seawater) conditions are likely to be encountered, thus requires numerical tools able to take into account the specific effects of the various electrolytes dissolved in brines, and the non-ideal behaviour of the CO2 gaseous phase. This study evaluates the consistency of the various corrections (activity, fugacity, influence of pressure on thermodynamic constants) to be taken into account in geochemical models to meet these calculation accuracy requirements. These corrections were implemented in the thermo-kinetic modelling software SCALE2000 (Azaroual et al., 2004a) which was used to check their validity by comparing the calculation results with available experimental observations and other results from CO2 solubility calculation models. An estimation of the relative weight of each of the corrections for a 237 g.l-1 brine (60°C, pCO2 = 200 bar) showed a systematic overestimation (higher than 100%) of CO2 solubility when either salinity (NaCl equivalent) is neglected or gas is considered ideal. The error induced by the NaCl-equivalent approximation compared to real brine is lower (less than 5%). The second part of this study presents an application example of a hypothetical scenario of massive CO2 injection in a carbonated reservoir; data used for the brine composition are actual data (Moldovanyi and Walter, 1992) from the Smackover site (Arkansas, United States). The simulations performed considering a representative elementary volume of saturated bulk rock (porous mineral assemblage saturated with the Smackover brine) with a prescribed constant CO2 pressure of 150 bar, show two distinctively different behaviours whether the system is assumed to be a closed (batch reactor) or an open reactor fed by a constant brine flow rate. In the first case, the calculations performed with SCALE2000 lead to negligible variations in the mineralogy. In the second case, more representative of the dynamical nature of an injection system, the results show major modifications in the mineralogy finally leading to a strong increase in porosity (from 20% initially to 85% after 50 y of simulated time). Further calculations were carried out with SCALE2000, now considering a 1D system constituted of a set of four homogeneous identical reactors connected in series (fluid velocity of 1 m.day-1). With initial and boundary conditions similar to those considered earlier, and prescribing a constant pCO2 in the first reactor only, the results showed that significant dolomite precipitation occurred in the most-downstream reactor hence inducing some CO2 precipitation. Mass balance calculations performed on the four reactors system finally demonstrated a global loss in total mineral carbon with respect to the amounts initially available. However, the evolution trends observed in the most-downstream two reactors indicated that possible trapping might be expected beyond the relatively limited geometrical boundaries considered in the modelled system.