mathematical models and numerical simulation tools play an important role in evaluating the feasibility of Co2 storage in subsurface brine reservoirs.
The basis of geochemical modeling of aqueous electrolyte solutions is the computation of activity coefficients, and hence activities, of relevant solutes and the solvent water through maintenance of mass and charge balances and equilibrium relations. The Debye–huckle method (Debye and Hückel, 1923a,b) of calculating activity coefficients provided a theoretical understanding and the hKF model (helgeson et al., 1981) and the software program suPCRT92 (Johnson et al., 1992) provided a way to calculate the thermodynamic properties of minerals, gases, and aqueous species at elevated temperature and pressure. early conceptual and numerical models, for example: SOLMINEQ (Kharaka and Barnes, 1973), EQ3/EQ6 (Wolery, 1979, 1992), PhReeQe (Parkhurst et al., 1980) and SOLVEQ (Reed, 1982), utilized these principles to calculate the distribution of aqueous species and mineral equilibria in a variety of natural and experimental solutions. a theoretical model of concentrated electrolytes based on a virial coefficient approach was developed by Pitzer (1973, 1975), implemented by harvie and Weare (1980) and software created by Kharaka et al. (1988) (SOLMIN 88), Plummer et al. (1988) (PhRQPITZ), and Parkhurst (1995) (PhReeQC) to describe solution properties and mineral solution equilibria at high ionic strength. Later, the Pitzer equations were extended to aqueous solutions of naCl and Co2 at elevated temperature (Corti et al., 1990).
During Co2 injection, geochemical processes are strongly affected by physical processes such as multiphase fluid flow and solute transport.
accurate geochemical simulations therefore require a computational capability that couples multiphase-flow processes with kinetically controlled geochemical processes. state of the art modeling has advanced from separate thermodynamically-derived equilibria-based numerical models for batch process simulation to fully coupled simulation programs for non-isothermal reactive geochemical transport in variable saturated geologic media. Integrated reaction path and solute transport models such as TouGhReaCT (Xu and Pruess, 1998; Xu et al., 2004b) and updated versions of eQ3/6 (Wolery and Daveler, 1992) couple complex geochemical, hydrological, and mechanical interactions following Co2 injection (marini, 2007, p. 349). models that include physico-chemical processes, chemical reaction, fluid flow, heat transfer, and mechanical properties vastly improve predictive abilities of reservoir simulations.
These and other models have been utilized to investigate Co2 injection, storage, and sequestration processes and analyze the coupled mechanisms that lead to structural, residual, solubility, and mineral trapping. The following are representative examples of reactive transport modeling of different reservoir types.
3.4.2 Sedimentary basins Sandstone aquifers
The Utsira Sand, Sleipner, northern North Sea. One of the first modeling exercises simulated Co2 injection into the utsira formation at statoilhydro’s north-sea sleipner facility (Johnson et al., 2001) where the subsurface is well characterized. Johnson et al. (2001) used an integrated approach combining reactive transport simulators nuFT (nitao, 1998) and GImRT/os3D (steefel and Yabusaki, 1996; steefel, 2001, 2008), supporting geochemical software suPCRT92 (Johnson et al., 1992; shock, 1998), thermodynamic-kinetic databases GEMBOCHS (Johnson and Lundeen, 1994a,b), and a graphic utility Xtool (Daveler, 1998). They created numeric simulations of three distinct sequestration scenarios, differing in the extent of intra-aquifer shale units and lateral permeability. Results of this study showed the relationship of intra-aquifer permeability to solubility and mineral trapping and the importance of the overlying shale caprock to long-term storage. Gaus et al.
(2005) also performed PhReeQC (Parkhurst, 1995) 1D diffusive reactive transport simulations of dissolved Co2 in the caprock of the utsira aquifer (sleipner project) and concluded that plagioclase feldspar in the shale alters to dawsonite, disordered dolomite, and calcite.
Glauconitic sandstone of the Alberta Basin. Two contrasting reaction path models of the potential Co2 sequestration in the glauconitic sandstone aquifer of the alberta basin, Western Canada have been proposed (marini, 2007, p. 396). In fluids super-saturated with CO2, Gunter et al. (1997), utilizing PaThaRCh (Perkins and Gunter, 1995a) and anorthite/albite and muscovite proxies for oligoclase and illite, modeled the alteration of primary glauconite (as annite) to siderite. At a fixed under-saturated PCo2, with the same proxies, Gunter et al. (2000), again utilizing PaThaRCh, modeled the almost complete consumption of Co2 in the precipitation of siderite, calcite, and dolomite, showing the high reactivity of Fe+2-bearing silicates to Co2. In contrast, Xu et al. (2000, 2004a), utilizing TouGhReaCT and the actual mineral assemblage of this formation, modeled precipitation of the Co2-bearing phases as mainly siderite and ankerite with some dawsonite and dolomite.
Later, Strazisar et al. (2006) investigated the periphery of this system at the Co2 reaction front away from the injection site. These investigators point out that precipitation reactions are more likely to occur downstream of the injection site, where the mineral assemblage buffers the ph at higher levels.
utilizing PhReeQC, they found that Co2 was trapped initially in siderite from the ph-dependent dissolution of annite and coupled with kaolinite dissolution and K-feldspar precipitation. as the Co2 front migrated further, K-feldspar dissolved and calcite and dolomite precipitated. They concluded
that most of the Co2 is trapped in siderite, consistent with previous studies (Gunter et al., 1997, 2000; Xu et al., 2000, 2004a).
Deep sand aquifers of the Powder River Basin of Wyoming. mcPherson and Lichtner (2001), utilizing TOUGH2 (Pruess et al., 1999), carried out a numerical model that included multiphase fluid flow of sustained injection of Co2 into the deep sand aquifers of the Powder River basin of Wyoming.
Their calculated Co2 residence time and migration rates show the unintended impact of wide-scale brine displacement out of the target aquifer and potentially spreading into adjacent sealing layers.
Tertiary Gulf Coast sediments: The Frio, Vicksberg and Wilcox formations.
apps (1996) first carried out batch geochemical modeling of Gulf Coast sediments as a potential repository basis for the deep injection disposal of hazardous and industrial wastes. Xu et al. (2004a,b), utilizing TouGhReaCT, constructed a reactive transport model of the Frio Formation for potential Co2 sequestration. These investigators incorporated in their model the high organic matter (kerogen) and salt content from diapirs that are characteristic of these sediments, along with a representative mineral assemblage at a Co2 injection pressure of 260 bars. Dawsonite and ankerite were the primary Co2-bearing products; calcite and siderite were modeled to initially precipitate then dissolve. Xu et al. (2005), again utilizing TouGhReaCT, focused on the diffusion of Co2 and acidity into the caprock and found carbonate precipitation extending into the shale. Knauss et al. (2005), utilizing the geochemical software CRunCh, another computer program for simulating multicomponent multidimensional reactive transport in porous media (steefel, 2001), investigated the effects of ancillary contaminant gases in the Frio Co2 injection stream. This work concluded that only so2 might have an impact on reaction processes due to the resulting extremely low ph.
Permian White Rims sandstone. White et al. (2005), utilizing ChemTouGh (White, 1995), simulated Co2 injection into this reservoir rock, situated beneath the hunter Power plant in central utah. These investigators found that calcite and dolomite precipitated but predicted that after 1000 years about 17 % of the Co2 had leaked to the ground surface.
Carbonate aquifers
Tuscan Nappe limestone formation. Cantucci et al. (2008) utilized a modified version of the PRheeQC (V2.11) software package to investigate the short- and long-term consequences of Co2 storage in an offshore Italian porous carbonate saline aquifer, the Tuscan nappe limestone formation. numeric
simulations of the fate of Co2 injected into the saline aquifer suggest that solubility trapping prevails within the first 100 years.
Nisku carbonate aquifer of the Alberta Basin. Gunter et al. (2000) modeled the interaction of under-saturated Co2 charged fluid with carbonate (calcite, dolomite) rocks of the nisku and found rapid dissolution of calcite and precipitation of dolomite.
Serpentine aquifers
Serpentinites of the Gruppo di Voltri (Genova, Italy). The high mg content of serpentinites has the potential for fixing CO2 as magnesite. Cipolli et al.
(2004) carried out reaction path modeling of potential Co2 sequestration in deep aquifers hosted by the serpentinized ultramafic rocks of the southern Piedmont, Italy. These investigators concluded that the capacity for Co2 sequestration is high through dissolution of serpentine and precipitation of magnesite and chalcedony but cautioned about the progressive loss of porosity, especially if amorphous silica precipitates instead of chalcedony.
They also point out that reactant armoring may occur during precipitation reactions. not often noted in the literature, this phenomenon is seen in experimental systems.
3.4.3 Model inter-comparison
Pruess et al. (2001a,b, 2002, 2004; Pruess, 2005) initiated and carried out an inter-comparison study of reactive transport models on test sets of representative Co2 sequestration in potential reservoirs. Results were quantitatively similar, indicating broad agreement among the models. allen et al. (2005) conclude that models are only as reliable as the data and reaction scheme upon which they are based and emphasize the importance of pressure corrections to thermodynamic data. model inter-comparisons indicate that failure to adjust all equilibrium constants to account for elevated Co2 pressures results in significant errors in both solubility and mineral formation estimates (allen et al., 2005). moreover, bateman et al. (2005) concluded that model predictions tend to overestimate the degree of reaction compared with experimental results. For example, some mineral phases such as dawsonite that are predicted to form in large quantities by models are not observed in the experimental system. These authors highlight the need for appropriate thermodynamic and kinetic data to address these discrepancies. The most robust analyses of Co2 sequestration in potential reservoirs incorporate a combination of experimental, model, and field evidence that often requires large-scale pilot project support (Fig. 3.10).
79dioxide (CO2) sequestration in deep saline aquifers
© Woodhead Publishing Limited, 2010
1 2 Other 3 Unknown
3 1
1
3
3 3 3 3
3
1
3
2
3.10 Geologic storage and related projects in operation or proposed around the world. Most are research, development or demonstration projects. Some are commercial operations (image source: CO2CRC).