A Coupled Simulator for Stability and Geochemical Analysis of Deep Geological Carbon

A Coupled Simulator for Stability and Geochemical Analysis of Deep Geological Carbon Sequestration

-----Brief Introduction of Background Information for the Field Simulation Case

Zhenze Li, Mamadou Fall

Email:

Department of Civil Engineering, University of Ottawa

Abstract: Carbon capture and sequestration (CCS) is one of the most promising and ambitious project for carbon management. Injection of CO2 into underground involves thermo-mechanical-hydro-chemical (THMC) couplings, which requires comprehensive theoretical studies with respect to its long-term stability and environmental risks. CO2 can react with rock and minerals either by dissolving out excessive porosity or by producing secondary precipitations. Sharp reduction in pH is ubiquitous, along with remobilization of concentrated toxicants. This study first developed a numerical simulator with THMC coupling capacity, which is able to predict the consequences of CO2 leakage. Then, the prediction capability of the simulator is verified against experimental results. Finally, it is applied to study the impact of CO2 leakage on a carbon sequestration site.

1. Site characterizations for QUEST CCS project

The exploration of oil and gas reservoir has been originated since 1980s in Redwater oil fields using acid gases (CO2 and SO2) to enhance the recovery of crude oil (Bachu et al. 1994). Such practices later revolved into geological storage for CCS purpose. Shell Canada Limited (Shell) applied in 2010 to construct, operate and reclaim the Quest Carbon Capture and Storage (CCS) Project with the goal of permanent storage of carbon dioxide (CO2). The CO2 that is to be captured from the Scotford Upgrader, about 5 km northeast of Fort Saskatchewan, Alberta (see Figure 1), will be transported to the storage area, where 3 to 10 wells are required for injecting the CO2 into the BCS for storage. The BCS is reported to be targeted at the formation layer with depth >2000 m that is overlain by a number of formations which provide containment for the CO2 (Stantec 2010). This proposal has been approved by Canada Federal Government in 2012. The lifespan of the Project is considered to be greater than 25 years (Stantec 2010).

The QUEST site doesn’t have the Paskapoo formation that is widely observed in southern Alberta plain (Barker et al. 2013). Instead, the main bed rock geology near the ground surface is indicated as the Belly river group. The figure 2 clearly indicates the profile stratification of the QUEST site.

Figure 1 Cross section through the Beaverhill Lake-Mannville sedimentary in the Edmonton region near the proposed QUEST CCS project (Bachu et al. 2008)

Figure 2 The geological section across the Western Canada Sedimentary basin (Deschamps et al. 2012)

2. In-situ geological parameters

2.1 Ground stress

According to the geological survey report of Bachu et al. (2000), the level of ground stress at the QUEST site is predicted as the following equations,

where Sv is the vertical stress, SHmin is the minimum horizontal stress and z is the depth. The maximum horizontal stress, although provided with information on its trajectory, has not been mentioned in the original report.

The medium horizontal stress is assumed to be in the level of

In the numerical simulations, we hypothesized that the minimum horizontal stress is in X-X direction, while the medium horizontal stress is in Y-Y direction. The vertical stress is regarded as the maximum ground stress.

2.2 Geothermal gradient

At Quest site, the geothermal investigation showed that the temperature at 1200 m is about 31.1 oC, suggesting a geothermal gradient with depth at 25.9 oC/km. (Hitchon 1993) summarized the paleogeothermal gradient for Alberta shales, as distributed in the range of 23-27 oC/km.

2.3 Belly River aquifer

The Belly River aquifer outcrops at the QUEST CCS injection site, and is the aquifer that is in vicinity of ground surface. The overall thickness is about 100 m, consisting of Oldman Formation and Dinosaur Park Formation. This formation layer was predominantly interbedded mudstones to very fine-grained sandstones with subordinate, but prominent coarser grained sandstone beds. Bentonite, coal and concretionary beds are minor constituents. The dominant colors are shades of grey and green. A predominantly sandstone unit is present in the basal 10 to 30 m, generally very fine- to medium-grained, with an overall upward increase in grain size. These more recessive beds often constitute a large proportion of the entire section (Canada 2013).

Meyer and Krause (2006) investigated the permeability and its anisotropy in a study area that is located on the western margin of the Alberta Foreland Basin about 100 km east of the present outcrop edge of the Mesozoic foreland belt, along both margins of the Milk River valley. It is believed that the corresponding permeability distributions are a good analogue for broadly similar lithofacies and marginal marine successions known to make-up important petroleum reservoirs in Alberta and worldwide (e.g. Belly River Formation, Misoa Formation, Brent Group). The horizontal permeability kh was found to be 2 times of the vertical permeability kv in this study area.

2.4 Geochemical information

Apps et al. (2010) reported an inspiring study on a comprehensive geochemical model of aquifers throughout the US, which evaluates the abundance, distribution and mode of occurrence of various hazardous trace elements in potable groundwater. Similar efforts could be made to compile a similar database for Canada, or at least for the analysis region.

Detailed mineral compositions in the QUEST site remain obscure at present. Approximately 100 samples were collected from the cores of 11 wells in the studied area that covers the Edmonton region (Deschamps et al. 2012). The sampling depth ranges from 970 to 2500 m. The shallower region is absent from this study. A detailed report on chemical composition of the Belly River formation has been published by Alberta Geological Survey (Lemay 2003; Lemay and Konhauser 2006). Table 1 shows the typical data about the geochemical characterizations of the groundwater as observed in Belly River formation in Southeastern Alberta plain.

The formation waters near the ground surface are characterized by relatively low pH (5.22-6.03) and hence low alkalinity (2-90 mg/L). The pH calculated based on the assumption that pH is controlled by the carbonate alkalinity and calcite solubility, varies between 4.38 and 5.57 (Morad et al. 1994). Another similar study reported the natural formation water pH as 6.7 (Kharaka et al. 2006). The intrusion of CO2 into the formation water was observed to cause a rapid pH decline, which is consistent to our prediction.

Table 1 Geochemistry of groundwater in both the Belly River formation (Lemay and Konhauser 2006) and Alberta basin (Cheung and Mayer 2009)

Note: a indicates the field test results for dissolved Oxygen, which is abnormally high for the geochemical modeling and thus leads to convergence difficulties. Here a much less value in DO is assumed, at 1E-17 M, to keep a smooth running of simulator.

Figure 3 Spatial contour plotting of equilibrium pH in groundwater

Figure 3 shows the predicted water pH as 7.8 in equilibrium state. This is exactly the average pH value of the observed formation water for Belly River formation as shown in Table 2. It is indicated that the mineral compositions as well as the chemical speciation of the current study are consistent to the field conditions.

Table 2 Mineral components and the abundance in volume

Note: a indicates the source of data is from (Eberth and Hamblin 1993); b indicates the source is from (Lerbekmo 1963) and c means the data about clay minerals are from (Longstaffe and Ayalon 1991).

3. THMC fully coupled simulator

FLAC3D is a robust finite-difference program for mechanics computation, and is especially professional and convenient in light of its capability to simulate the behavior of various geomaterials that undergo plastic flow when yielding has been reached (ITASCA 2009). An elasto-plastic damage model was developed and then implemented into FLAC3D. TOUGHREACT is based on the TOUGH code, considers thermal, hydraulic and chemical couplings and permeability-porosity coupling as well as geochemical reactions, and transport that result from mineral dissolution and precipitation. Chemical transport is allegedly solved on a component-by-component basis (Xu et al. 2012). TOUGHREACT has proven to be efficient and reliable in the modeling of the time dependent transport of non-isothermal multiphase/unsaturated flow and reactive transportation of chemical species in porous media.

As proposed by Rutqvist (2002), two numerical softwares, FLAC3D (designed for mechanical (M) analysis) and TOUGH (a code released by Berkeley for THC coupled modeling), have been coupled together to act as a robust THM coupled simulator for the assessment of geothermal exploration, excavation induced damage, long-term safety of nuclear waste repositories, volcano phenomenon, etc. (Rutqvist 2011; Rutqvist and Tsang 2002). Recently, we have further expanded on this proven method by coupling Flac3D with TOUGHREACT and used this new numerical tool to investigate the potential groundwater contamination in response to CO2 leakage (Li and Fall 2013). Furthermore, a novel algorithm has been proposed for more efficient 3D gridding establishment, data transfer and exchange between softwares. This new feature of the simulator enables us to make a thorough numerical investigation into the cases that resemble field conditions to the largest extent.

4. Modeling scenarios

4.1 Schematic of the conceptual model

Focus of our modeling would be placed on the most conservative cases, as investigated by researchers from Berkeley (Apps et al. 2010; Birkholzer et al. 2008). Therefore the injection phase for CCS project will be omitted; we will skip directly to the leakage phase as shown in the following sketch, which has been used as a base model for the analysis of reactive solute transport by Apps et al. (2010). However, a more realistic geological stratification representing the QUEST site has been incorporated into this model. Double layers of sedimentary and rocks of sufficient data support have been carefully selected. The dimension of the modelling region is 1000*100*100 m3 for the X*Y*Z coordinating directions which cover a wide range of typical aquifers. A 3-dimensional gridding of the analysis zones has been achieved in this study following a recently proposed algorithm for the correlation and sorting of relevant zones from both Flac3D and TOUGHREACT softwares. The conceptual model set up and 3-D gridding graphs are shown here.

Figure 4 Model setup for CO2 leakage into shallow aquifer (not to scale)

Figure 5 3D gridding for the modeling zones created with finite difference method (left: Flac3D; right: TOUGHREACT)

4.2 Target trace elements

Apps et al. (2010) evaluated the potential changes in groundwater quality in response to CO2 leakage from deep geological storage site, and showed that the hazardous trace elements of greatest concern are arsenic and lead (As, Pb). These two elements are therefore taken into account in our simulations. Galena is reported to be the predominant source for Pb in groundwater, while arsenopylite and arsenolite correlate primarily to dissolvable As (Apps et al. 2010).

These minerals are commonly present in geothermal springs or mining site in mountain area with mineral veins observed. In the QUEST site, no information about the distribution or abundance of these minerals is available. However, the both toxicants have been detected at various concentrations in some samples from groundwater and formation water. In order to be conservative, both galena and arsenolite were addressed in this study, with an assumption of very limited abundance at 0.05% for each of the minerals. Galena is a reductive mineral of sulphide which can easily be oxidized into mobile elements. Therefore the redox potential of the groundwater is critical to the stability and migration of Pb.

Since the Belly River formation inclined towards the much deeper geological strata, which is the upstream of the groundwater flow path, the redox potential of the groundwater in the formation is expected to be in reductive state. The redox reactions of the both minerals are in the form of