4.5 Reactive transport modelling in the Sleipner storage project

Audigane et al., 2007 presented a 2D reactive transport model of long-term geological storage of carbon dioxide in the Utsira Formation in Sleipner. Although their results discussed here were not obtained during the site selection phase, they provide a good impression of what should be anticipated when planning a CO2 geological storage. Audigane et al., 2007 used the reactive transport code TOUGHREACT and perform a 25 year injection scenario followed by a 10,000 year storage period. They presented detailed information on the different numerical modelling approaches, as well as a complete reference list dealing with reactive transport modelling. In particular, the numerical method and its implementation in TOUGHREACT are introduced. Details on equations, conditions on the dissolution of CO2 and kinetics of mineral dissolution and precipitation are well described.

The site itself, the Sleipner West natural gas field, is located in the centre of the North Sea. CO2 is injected into the Utsira Formation, which is located above the gas reservoir. The Utsira Formation is a large sandy aquifer extending over 26,100 km2 located at a depth from 700 to 1,000 m (Audigane et al., 2007). Details on Utsira Formation mineral composition and formation waters are described in the paper as well as comparison between batch geochemical modelling and batch simulations are compared.

Audigane et al., 2007 have approximated the Utsira Formation geometry by a vertical 2D mesh with a cylindrical geometrical configuration, centred on an injection point located 155 m beneath the top of the 184 m thick formation. The mesh contains 22 layers in the vertical direction and 52 in the radial direction. The first cell has a radius of 10 m, and is followed by 20 cells with radial increments increasing in logarithmic progression out to 1 km. Beyond 1 km, 15 cells are present with another logarithmic progression to 3 km. Beyond 3 km, 10 more increments are present with logarithmic progression to a radial distance of 10 km, and finally 5 cells extend out 100 km from the injection point; an additional cell with a large volume allows the model to act as an infinite medium (Audigane et al., 2007). Audigane et al., 2007 justify the use of radial mesh geometry by the approximate circular shape of the CO2 plume observed from seismic surveys.

Hydraulic initial and boundary conditions used in the simulations are shown in Tab. 4-1. Audigane et al., 2007 consider homogeneous sand and shale formations, with relative permeability derived from core measurements. They simulate CO2 injection at a rate of 30 kg/sec for 25 years while accounting for an initial pressure of 90 bars at the top of the formation.

Tab. 4-1: Hydrogeological parameters for the 2D model used to simulate CO2 injection at Sleipner with TOUGHREACT (Audigane et al., 2007).

Tab . 4-1

Geochemical initial conditions for sand and shale mineralogies and for formation waters for the 2D model are identical to the batch system considered earlier in Audigane et al., 2006. A low salinity value of 32 g/l is assumed for the formation water in both the sand and the shale.

The short-term simulations, i.e., 25-year injection period, show that the supercritical CO2 plume extends laterally about 300 m away from the injection point, which is consistent with seismic observations (Audigane et al., 2007). The presence of four intra-shale aquifers gives rise to CO2 accumulations at four different depths and slows the upward CO2 migration. CO2 dissolution slightly increases brine density and gives rise to a negative (downward) buoyancy force. At the end of the 25 year injection period, a slight downward migration of the brine enriched in CO2 can be observed. Dissolution of CO2 makes the brine more acidic and short-term acidification occurs mainly in the area where supercritical CO2 is present. Changes in porosity are minor. For more detailed short-term results, the reader should refer to Audigane et al., 2007.

The long-term simulations, i.e., 10 000 years, show the extent of the CO2 plume and dissolved CO2 (Fig. 4-4). The reader should refer to Audigane et al., 2007 for details. At the end of CO2 injection, supercritical CO2 migrates quickly upwards, and most of the supercritical CO2 accumulates just below the cap rock, except residual CO2 that is trapped in sediments. The CO2 plume extends to a maximum radius of 2,000 m around the injection point. CO2 starts to dissolve in the brine, and the free gas is completely dissolved after 6,000 years. The brine containing dissolved CO2 tends to migrate downward as it is approximately 10 kg/m3 denser than brine without CO2. Molecular diffusion from the gas plume to the brine induces a hydrodynamically unstable layering leading to the development of convective currents in the formation. The brine containing dissolved CO2 is carried downward and is progressively replaced by brine with less CO2. Streamlines of fluid migration showing convective cells are represented in Fig. 4-4 after 2,000 years of simulation. The slow brine convection accelerates CO2 dissolution. After 10,000 years, a large volume near the bottom of the formation contains brine with dissolved CO2 out to a radius of 4,000 m.

Key results of Audigane et al., 2007 are the following:

  • The role of convective mixing is crucial for long term CO2 dissolution;

  • The process of gas dissolution and subsequent buoyant convection and mixing of brine involves a range of spatial scales;

  • The interaction between flow and geochemical reactions and its impact on the overall reactivity can be accurately assessed only through coupled modelling and;

  • For long-term reactivity, the evolution of the acidity is crucial; it is directly linked to the amount of CO2 dissolved in the brine.

Audigane et al., 2007 identified four main types of geochemical interactions:

  • Calcite dissolution and precipitation;
  • Albite alteration;

  • Muscovite alteration.

The assessment of the amount of CO2 stored and the induced porosity changes are presented as a mass balance of carbon dioxide in mineral, supercritical and aqueous phases (Fig. 4-5). Mineral tTrapping obviously plays a minor role, although it increases slowly with time and therefore contributes to long-term stability of the storage process. Regarding the spatial distribution of CO2 stored in minerals, the simulated mineral storage occurs mainly at the top of the reservoir and in the major downward convection zone above the injection point. The induced porosity changes are minor: a decrease of less than 2.5 % in the sands, and an increase of up to 15 % in the shales (Fig. 4-6). Audigane et al., 2007 point out that the difference in the porosity changes between sands and shales illustrate the importance of coupled transport and chemical reaction modelling: in the shales the porosity change is the inverse of what was seen in the batch modelling (Audigane et al., 2007).

E. Fig . 4-4

Fig. 4-4: Supercritical CO2 gas phase (SG) migration and mass fraction of the dissolved CO2 in the brine (XCO2L) simulated from 50 years after injection until 10 000 years (Audigane et al.,. 2007).

E. Fig . 4-5

Fig. 4-5: Total amounts of carbon dioxide present as a free (supercritical) gas phase, dissolved in the aqueous phase, and sequestered in minerals (Audigane et al., 2007)

E. Fig . 4-6

Fig. 4-6: Porosity changes after 10,000 years of simulation (Audigane et al., 2007).

Audigane et al., 2007 conclude that the geochemical reactivity of the Utsira Formation is rather low, so that mineral trapping makes only minor contributions to CO2 storage. Solubility is the dominant long-term storage mechanism, which appears to be essentially complete after 5 000 years. Physical and chemical heterogeneity play important roles in the geochemical evolution and associated changes in porosity. The Audigane et al., 2007 results suggest that the Utsira sand is unlikely to undergo major chemical changes due to CO2 injection. Anticipated porosity changes are relatively minor, with a slight long-term decrease expected in the sands and a more significant increase in the shales. Density differences between brines with different dissolved CO2 concentrations give rise to convective flows that cross the shale layers, mobilizing species that subsequently promote precipitation further downstream. Processes in which dissolution of minerals occurs in one region while precipitation occurs in another region with different mineralogy can only be analysed and modelled by coupling flow and transport with chemical reactions. The strong interplay between multiphase and density dependent flows with rock-fluid interactions makes it difficult to interpret modelled results in terms of a few dominant reactions. Separate batch geochemical modelling provides useful guidance for interpretation, but Audigane et al., 2007 show that in some cases the coupling to transport can give rise to qualitatively different behaviour to that derived from batch models.