- Email: [email protected]
Quantitative Interpretation of Trapping Mechanisms of CO2 at Nagaoka Pilot Project a History Matching Study for 10-Year post-injection –
Available online at www.sciencedirect.com
ScienceDirect Energy Procedia 114 (2017) 5058 – 5069
13th International Conference on Greenhouse Gas Control Technologies, GHGT-13, 14-18 November 2016, Lausanne, Switzerland
Quantitative interpretation of trapping mechanisms of CO2 at Nagaoka pilot project – A history matching study for 10-year post-injection – Hajime Yamamotoa*, Takahiro Nakajimab, Ziqiu Xue b a
b
Taisei Corporation,344-1, Nase-cho, Totsuka-ku, Yokohama 245-0051, Japan Research Institute of Innovative Technology for the Earth (RITE), 9-2, Kizugawadai, Kizugawa, Kyoto 619-0292, Japan
Abstract Demonstrating the permanence of CO2 storage is an important task of pilot projects. In the Nagaoka project, Japan’s first pilottest, a stable containment of CO2 in a reservoir has been successfully demonstrated by kept monitoring the CO 2 behavior even after the end of injection during more than 10 years. Systematic and continuous data acquisition of time-lapse well loggings illustrated the detailed nature of CO2 migration at intra-reservoir resolution. In this study, a three-dimensional reservoir model with sub-meter spatial resolution has been developed that involves coupled process of two-phase fluid flow and geochemical transport. The model was history-matched against a set of monitoring data including pressure, well loggings, and fluid samplings over the post-injection period. From the history matching study, the following insights into the trapping processes of CO2 can be drawn: 1) The uneven arrival times of CO2 to the well-depths are well explained by, and consistent with the non-uniform permeability distribution measured at wells; 2) Slow or even negligible vertical migration of free CO2 inside the reservoir suggests that even a thin intra-reservoir muddy-layer behaves like an impermeable flow barrier to trap CO2; 3) Pressure-drivenflow during the injection squeezed the formation water out of the reservoir, and subsequent slow diffusive transport of CO2 promoted dissolution of rock minerals including dissolution of calcite and almino-silicates, suggesting precursor of mineral trapping by the precipitation of carbonates such as calcite in the future in this site. © 2017 2017The TheAuthors. Authors. Published by Elsevier © Published by Elsevier Ltd. Ltd. This is an open access article under the CC BY-NC-ND license Peer-review under responsibility of the organizing committee of GHGT-13. (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the organizing committee of GHGT-13. Keywords: Nagaoka; post-injection; monitoring; simulation; trapping process; hystory matching; resistivity log; neutron log; fluid sampling
* Corresponding author. Tel.: +81-45-814-7237; fax: +81-45-814-7253. E-mail address: [email protected]
1876-6102 © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the organizing committee of GHGT-13. doi:10.1016/j.egypro.2017.03.1659
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
5059
1. Introduction Long term stability of CO2 stored in reservoir is of intrinsic importance for ensuring the viability of geologic sequestration of carbon dioxide. Demonstrating the permanence of storage is an important task of pilot projects. The Nagaoka project is the Japan’s first pilot-test of geological CO2 storage that injected about 10k tonnes of CO2. A stable containment of CO2 in a reservoir has been successfully demonstrated by kept monitoring the CO 2 behavior even after the end of injection for more than 10 years. Systematic and continuous data acquisition of time-lapse well loggings (e.g., resistivity, neutron, and sonic velocity) well illustrated the detailed nature of CO2 migration at intrareservoir resolution, including uneven CO2 arrivals to the well-depths and the prevention of vertical movement of buoyant CO2 inside the reservoir by a potential seal layer. The time evolutions of CO2 saturation were quantitatively evaluated under sub-meter resolutions by neutron logs. Resistivity logs even enable us to infer the dispersed area of CO2 dissolved in water in the reservoir. Fluid samplings at multiple depths also provided an insight into geochemical processes among CO2 and water. However, in order to better understand the trapping mechanisms from these detailed observations at intra-reservoir scale, the numerical modeling with sub-meter spatial resolution is necessary, which has not been performed yet in the project. In this study, a three-dimensional reservoir model with sub-meter spatial resolution has been developed that comprehensively involves coupled process of two-phase fluid flow and geochemical transport. The model was history-matched against a set of monitoring data acquired during the post-injection period including pressure, well loggings, and fluid samplings. The calibration of a large model is computationally demanding, hence we employed a parallel version of coupled fluid flow and geochemistry code TOUGHREACT V2.0/ECO2N with MPI parallelism, which has also been newly developed in-house. The new code also features hysteretic effect in relative permeability and capillarity which was not implemented in the original TOUGHREACT V2.0. 2. Post-injection monitoring at Nagaoka Site 2.1. Geological setting and well placement The Nagaoka CO2 storage site is located near the Nagaoka city in Niigata prefecture in Japan. From 2003 to 2005, approximately 10 kilo-tonnes of CO2 were injected and stored in an onshore aquifer. The storage aquifer is the early Pleistocene Haizume sandstone formation at 1,100m below surface, which is about 60m-thick, monoclinal structure inclined about 15° to the east-northeast. The reservoir can be subdivided into five intervals (Zone-1 to 5) based on lithofacies identified through well testing, core analysis, well-loggings, and correlation between these wells [1]. The location of an injection well (IW-1) and three observation wells (OB-2, OB-3, and OB-4) are shown in Fig. 1. OB-2 is located 40m down-dip from IW-1, while OB-3 and OB-4 are located in up-dip direction 120m and 60m away from IW-1 respectively. In these wells, a variety of CO2 monitoring has been conducted including continuous measurements of pressure and temperature, crosswell seismic tomography, well logging and in situ fluid sampling. The monitoring was continued even after the end of injection for more than 10 years. The details of the project are summarized in Sato et al. [1].
Fig. 1 Locations of injection well and observation wells [1].
5060
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
2.2. CO2 injection and arrival at observation wells In the storage aquifer, a thin sandy layer with 12 m thickness (Zone-2, average permeability 7 mD) was selected as the reservoir. In March 2004 (247 days after the start of the injection), the sonic, neutron, and resistivity loggings at observation wells OB-2 all detected significant changes, which are attributed to CO2 arrival [1, 2]. CO2 also later arrived at OB-4 in July 2004 (367 days), while any indication of CO2 arrival has not been detected at OB-3. The well-logging and fluid sampling indicated that CO2 arrived through only two highly permeable sub-intervals in Zone-2 (Zone-2a and 2b) as described in detail in the following sections. 2.3. Stratigraphy in Zone-2 Fig. 2 shows the stratigraphic profile of Zone-2 at the injection well IW-1 along with the vertical distribution of injectivity, mud content [3], and permeability. The geology of Zone-2 consists of alternating sand and silt/shale. The mud content and permeability distribution were determined by core analyses. The well injectivity was measured at the early period of the CO2 injection, and indicated that significant injectivity occurred in permeable interval inside Zone-2. Zone-2 can be further divided into three sub-intervals: Zone-2a, 2b, and 2c. The sub-intervals Zone-2a and 2b are bound by a thin muddy layer with high mud content nearly 100% as shown in Fig. 2.
A
Zone-2c Zone-2b
B
C D E 0
20 40 Permeability, md
Zone-2 (Injection Zone)
Thin muddy layer Zone-2a
Low injectivity
60
Permeability (md) 0 20 40 60 80 100
Injectivity (%)
Mud content (%)
Fig. 2 Stratigraphic profile of injection well IW-1 at intra-reservoir scale [3]
2.4. Post-injection monitoring During and after the injection for more than 10 years, time-lapse well loggings including induction resistivity, neutron, and sonic velocity have been repeatedly performed 44 times so far, and are still on going on an annual basis [1]. Formation water was collected twice in 2005 and 2011 at some depths in OB-2 by Cased Hole Dynamic Tester (CHDT). Among the monitoring data, in Fig. 3, the time evolution of resistivity change from baseline value obtained from resistivity logs for 12 years (about 4500 days) since the injection started (7 July, 2003) was depicted as a contour map [2]. This suggested uneven CO2 arrivals and subsequent saturation changes in Zone-2b. More importantly the
5061
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
vertical movement of buoyant CO2 has been prevented at the top of Zone-2b, where the potential seal layer with high mud content was identified as seen in Fig. 2. The resistivity logs even enable us to infer the dispersed area of dissolved CO2 in water, by the negative resistivity changes gradually spread over and beneath Zone-2b. In addition to this resistivity result, neutron logs quantitatively evaluated CO2 saturation under sub-meter resolutions, and fluid samplings at multiple depths also provided an insight into geochemical processes among CO 2, water, as discussed in following chapters. As described above, the post-injection monitoring depicted the detailed nature of CO2 migration at the intrareservoir scale. However, in order to better understand the trapping processes from these detailed observations, numerical modeling with sub-meter spatial resolution is necessary, which has not been performed yet in the project. Post-Injection
Injection
CHDT (1108.6m)
A CHDT (1112m)
Resistivity Change (%) 20 10
B
CHDT (1114m)
C
0
D
-10
E
CHDT (1118m)
CHDT (1118m)
-20
CHDT (1119.5m)
Fig. 3 Temporal change of resistivity from baseline along the observation well OB-2 [2].
3. Model Setup 3.1. Numerical code TOUGHREACT V2.0 [4] is a coupled fluid flow and geochemistry code developed by introducing reactive chemistry into TOUGH2 [5], which is a three-dimensional, fully implicit model of non-isothermal, multi-component and multi-phase flow and heat in porous and fractured media. For applications to geologic storage of CO2 in saline aquifers, a fluid properties module ECO2N [6] can be used together. In this study, we newly developed a parallel version of coupled fluid flow and geochemistry TOUGHREACT V2.0 with MPI parallelism, which run on multiple CPU/core computer systems [7]. The parallelization scheme is similar to that of TOUGH2-MP [8]. It solves sparse linear systems arising from discretization of the partial differential equations for mass and energy balance. The code uses MPI for parallel implementation, the METIS software package [9] for simulation domain partitioning, and the iterative parallel linear solver package Aztec [10] for solving linear equations by multiple processors. This in-house code also features hysteretic effect in relative permeability and capillarity developed by Doughty [11], which has not been implemented in the original TOUGHREACT V2.0.
5062
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
3.2. Reservoir model The reservoir is geologically interpreted as deltaic or coastal plain deposits characterized by up-ward-shallowing successions [3]. A heterogeneous distribution of permeability and porosity in three-dimension was realized on the basis of geological interpretation by the concept of sequential stratigraphy and sequential Gaussian simulation on the Petrel software [3, 12]. The Petrel model was converted to integral finite difference grids for TOUGHREACT simulations by utilizing a workflow & tools developed by BRGM [13]. The three dimensional reservoir model used in this study is shown in Fig. 4, which has been developed by Nakajima et al.[12]. Local grid refinement (LGR) was applied in the vicinity of the CO2 injection area. The grid-sizes in horizontal direction inside the outer and inner LGR region are 10m and 5 m respectively. In this study, we performed numerical modeling to interpret the detailed observations of well loggings acquired under sub-meter vertical resolutions, whereas the original model was vertically discretized Zone-2 with only 3 layers (i.e., 5.5 m grid-spacing for Zone-2a and 2b, and 1m grid-spacing for Zone 2c). Hence the original model was further refined with a sub-meter grid spacing of 50 cm in vertical direction, and inhomogeneous distribution of porosity and permeability along depths arising from alternating thin layers of mud, sand, and conglomerate was directly represented in the model. Based on the cross-well correlation of facies and permeability, we assumed that the vertical inhomogeneous pattern of permeability identified in Zone-2 at IW-1 (Fig. 2a, rightmost) can be extrapolated laterally throughout the Zone-2. Fig. 2b shows the resulted distribution of permeability on the east-west vertical cross-section through IW-1. Additionally we inserted thin mud layers, which have lower permeability (0.01 mD), and high capillarity, at the upper and bottom boundary of Zone-2b. This operation is based on the facts that mud content around the top of Zone-2b is nearly 100 % and no injectivity below Zone-2b were found, as seen in Fig. 2.
a
OB-4 OB-3
IW-1 OB-2
b
Fig. 4 Reservoir model with permeability distribution: (a) Perspective view; (b) E-W vertical cross-section through IW-1
5063
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
3.3. Coupled flow and geochemical model The settings of fluid flow and geochemical model are summarized in Table 1 and Table2, respectively. The fluid flow model involved three components (H2O, CO2, and NaCl) and two-phase (CO2 and water) flow [6], which was further coupled with chemically reactive transport with 11 primary aqueous species and 7 minerals including some carbonate, almino-silicate, and clay minerals. The geochemical model was basically taken from Mito et al. [15], and SOLTHERM thermodynamic database [16] was used. Table 1 Settings of fluid flow model Settings Iso-thermal Three components (H2O, CO2, and NaCl) Two phases (scCO2 and brine˅ Pressure: about 10 MPa Temperature: 48°C Salinity: 8h10-3 (NaCl mass fraction in aqueous phase) Sequential Gaussian simulation Permeability was calibrated against temporal pressure changes and CO2 arrival time at OB-2. Lateral extraporation of vertical inhomogeneous pattern of permeability in Zone-2b at IW-1 Thin muddy layers at the top and bottom of Zone-2b Curve fitting using laboratory tests (core flooding[14]) Parker and Lenhard model Curve fitting using laboratory tests (mercury injection)[12] van Genuchten model Two phase diffusion by Millington-Quirk ఉ ݀ =2h10-9 m2/s
Fluid flow model
Initial reservoir condition
Permeability Porosity
Relative permeability Capillary pressure Diffusion model
Boundary condition Injection rate Simulation duration
Quartz K-Feldspar Anorthite Albite-low Ca-montmorillonite Calcite Kaolinite
Anisotropy: ky=6hkx, kz=kx/10
Hysteretic effects [11] are incorporated. Hysteretic effects [11] are incorporated.
All boundaries closed (top, bottom, lateral) Measured flowrate data (temporally changes) 12 years
Table 2 Settings for geochemical model [15] (a) Initial mineral composition and kinetic parameters Mineral
Remarks ECO2N fluid property module [6]
Intial Vol% 29.3 13.5 9.8 20.2 29.2 0.2 0.0
RSA (m2/m3 mineral) 2.40E+04 2.40E+04 2.40E+04 2.40E+04 1.50E+06 2.40E+04 1.50E+06
(b) Primary aqueous species Rate Constant (mol/m2 s) 1.26E-14 1.00E-10 1.60E-12 1.00E-11 4.00E-14 1.60E-09 1.00E-13
Activation Energy (kJ/mol) 87.5 41.9 18.4 67.8 48.0 42.0 62.8
Component pH HCO3SO42ClNa K Mg Ca Si Al
Initial Concentration mmol/kg 6.4 3.9 0.8 96 74 6.4 0.7 10 1.8 0.004
5064
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
4. Results and Discussion 4.1. General behavior of CO2 Distribution of CO2 saturation (free CO2) and CO2 dissolved in water on E-W vertical cross-section through IW-1 are shown in Fig. 5. Free CO2 preferentially invades in high permeable layers of Zone-2a and 2b. The front of CO2dissolved formation water proceeds ahead of the free CO2 front. Inside the free CO2 zones, buoyancy gradually drives the portion of high CO2 saturation in the up-dip direction, leaving the footprint of CO2 plume down-dip, where residual trapping takes place. Pressure transient at IW-1 and OB-2 for 800 days is shown in Fig. 6. Due to the history matching efforts previously performed by Nakajima et al. [12], simulated pressure transient at OB-4 matched well with measured one for this refined-grid model. At IW-1, the pressure obtained from simulation generally exceeds the observation, probably due to acid treatment of IW-1 for increasing injectivity performed prior to the injection. The acid treatment can greatly increase permeability around the well and result in a large negative skin effect [14], which is not accounted for in this simulation.
a
b
Fig. 5 Time evolution of CO2 migration on E-W vertical cross-section through IW-1: (a) 1.6 years (just after injection terminated); (b) 10 years (8.5 years after injection terminated)
a
b
Fig. 6 Temporal pressure change at wells: (a) IW-1; (b) OB-4
4.2. Neutron logging Fig. 7 shows changes in CO2 saturation at OB-2 and OB-4 over time for three depth-intervals in Zone-2b (B, C, D in Fig. 3) at OB-2, and middle part of Zone-2 at OB-4. The CO2 saturation was evaluated from neutron loggings
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
5065
repeatedly conducted [1], and averaged values within each interval are shown in the figure. The figure also includes the simulation results, which are fairly matched with the observation, although it needed to assume a strong anisotropy of permeability to the north-south direction (kx:ky = 1:6) to explain the early arrival time of CO2 at OB-2. The direction of the anisotropy is almost corresponding to the direction of sedimentary sequences. The uneven arrival times of CO2 at the three depth intervals (B, C, D) in Zone-2b at OB-2 are well explained by, and consistent with the non-uniform permeability distribution shown in Fig.2: Permeability of C interval is obviously lower than that of B and D. Subsequent temporal changes in CO2 saturations are also well simulated by taking the hysteretic effect of relative permeability into account in the current model. At the down-dip OB-2, CO2 saturation generally dropped from 30 - 40 % to 15 - 20 % between about 1500 to 2000 days, and then the saturation levels were stabilized. This behavior is attributed to buoyant flow and residual trapping effect. In fact, at OB-4 which is located at the up-dip direction of IW-1, neither simulation nor observation results showed such reduction of CO2 saturation so far. This is presumably because the effect of buoyant flow on CO2 saturation has not been still prominent.
Fig. 7 CO2 saturation changes at observation wells OB-2(down-dip) and OB-4(up-dip). B, C, and D are sub-intervals inside Zone-2b: see Fig. 3.
4.3. Resistivity logging Time evolution of CO2 saturation and dissolved-CO2 in aqueous phase at OB-2 obtained from the simulation are shown in Fig. 8. Comparing with the resistivity result shown in Fig. 3, the following interpretations can be drawn at present. (1) Vertical migration of free CO2 In Fig. 3, the top of red-colored region (positive change of resistivity) at 1113 m depth (Zone-2b top) has stayed the same depth throughout the monitoring period for more than 10 years. This suggests that the vertical migration of free CO2 has been prevented at the depth, which is consistent with the simulation result (Fig. 8a): In the model, the vertical invasion of free CO2 from Zone-2b to 2a is prevented by a combined effect of lower vertical permeability and high capillarity of a thin muddy layer assumed at the top of Zone-2b. Our history-matching efforts indicated that the containment of free CO 2 in Zone-2b cannot be reproduced without the thin muddy layer. This suggests the continuity of the thin muddy layer over Zone-2b from IW-1 to OB-2, and that even a thin, intra-reservoir muddylayer behaved like an impermeable flow barrier to trap free CO2 during the decade. (2) Buoyancy effect of free CO2 The bottom boundary of free CO2 is slowly moves upward in the simulation (Fig. 8a). This is attributed to the buoyancy flow due to lower density of supercritical CO2 than water. However, it is not prominent in the observation
5066
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
of resistivity change in Fig. 3. Practically, the buoyancy flow in the vertical direction can be regarded as being negligible in the result. (3) Early arrival of CO2 dissolved in water When injecting CO2 into a virgin aquifer, it is reasonable that water dissolving CO2 migrates ahead of free CO2 plume. The negative resistivity change observed across Zone-2a after the termination of the injection (Fig. 3) is attributable to early arrival of dissolved CO2 in water prior to the arrival of free CO2. According to the simulation, while free CO2 arrives at OB-2 through Zone-2b, only water dissolving CO2 reaches in Zone-2a. Fig. 8b presents the vertical distribution of dissolved CO2 in aqueous phase at OB-2 over time, which shows arrival of CO2 dissolving water in Zone-2a at about 500 days, while no free CO2 arrival in the zone until 12 years (about 4500 days). (4) Dispersion/diffusion of CO2 dissolved in water It is also seen in Fig. 3 that negative resistivity change occurred just over and beneath Zone-2b in Fig. 3. The region with negative change of resistivity slowly spread above and below even during the post-injection. Regarding this behavior, the current simulation provides the following interpretation (Fig. 8b): 1) During the injection, pressure-driven-flow squeezed the formation water out of Zone-2b, which resulted in a rapid hydrodynamic dispersion of CO2 in water; 2) After the injection terminated, the reservoir pressure recovered to the initial and water ceased to flow. But still slow diffusion process gradually transported CO2 through formation water. The spreading speed of dissolved-CO2 water is almost comparable with that of negative resistivity change in Fig. 3 (about 2-3 meters in 10 years). Here we may try to calculate the resistivity change from the simulation result. Archies’s equation relates formation resistivity R and CO2 saturation Sco2 as below: ܴ ൌ Ƚି ሺͳ െ ܵைଶ ሻି ܴ௪
(1)
where, n is the saturation exponent (usually equal to 2), Rw is resistivity of formation water. From an experimental study [17], we employed α = 1, m = 1.59, and Rw = 1. By taking the resistivity change due to CO2 dissolution also into account, we roughly estimated the resistivity change from the simulated result as shown in Fig. 9.
a
b
Fig. 8 Time evolution of CO2 saturation and dissolved-CO2 in formation water at OB-2 obtained from the simulation. The amount of dissolvedCO2 in water is represented as mass fraction of CO2 in aqueous phase (XCO2aq): (a) Saturation of gaseous CO2; (b) Dissolved CO2 in formation water
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
a
-20
Resistivity Change (%) -10 0 10
20
5067
b
A
B C D
E
Fig. 9 Comparison of resistivity changes estimated from simulation with observation: (a) Resistivity logging; (b) Simulation
4.4. Fluid sampling The formation water was collected from OB-2 by Cased Hole Dynamic Tester (CHDT) at 1108.6 m, 1114 m, and 1118 m depths in December 2005 (1st CHDT), and then at 1112 m, 1118 m and 1119.5 m depths in September 2011 (2nd CHDT) [15]. The location and timing of the samplings are shown in Fig.3. Clear increases in concentrations of HCO3-, Ca, Fe, Mg, and SiO2 have been detected after the injection (1st CHDT), suggesting significant geochemical reactions had occurred in the formation water [15]. These increases in ion concentration are consistent with the decrease in resistivity at the sampling depths (circles) shown in Fig. 3. At 1118 m depth, where 1st and 2nd CHDT were both carried out, further increase in concentrations of HCO 3- and SiO2 associated with slight decrease in Ca concentration have been observed during the post-injection period. The observed increase in HCO3- concentration outside of the free CO2 zone can be attributed to the dispersive and diffusive transport of CO2 as discussed above. This should promote dissolution of minerals including calcite and almino-silicates, which increases such as Ca, and SiO2 concentrations. The history matching of the geochemical model is challenging and still ongoing in the study. This paper only presents a result from a base model [15]. Fig. 10 includes simulated temporal changes in pH and aqueous concentrations of HCO3, Ca, and SiO2 at 1118m depth. Although the concentrations of each component obtained from the simulation are also almost comparable with the fluid sampling result. However, while the geochemical model contains a lot of uncertain parameters, the constraints for history matching of the geochemical model are insufficient. Fig. 11 shows the time evolution of mineral dissolution and precipitation simulated by the current model. Just after the CO2 arrives, pH dropped quickly and the dissolutions of calcite and feldspar (albite) proceed quickly. After injection stopped, pH gradually recovers and precipitation of calcite occurs slowly. The decrease in Ca concentration between 1st and 2nd CHDT may be attributed to calcite precipitation as suggested by Mito et al. [15]. This result can be regarded as a precursor of mineral trapping by the precipitation of carbonates such as calcite in the future in this site. However, the current geochemical model does not include full set of potential carbonate minerals such as dawsonite, siderite, maginesite, so further improvement will be needed in the next step.
5068
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
Fig. 10 Temporal changes in pH and aqueous concentrations at 1118 m depth
Fig. 11 Time evolution of mineral dissolution and precipitation simulated by the current model
5. Conclusion The modeling study successfully reproduced the behavior of CO2 observed at sub-meter scale in the field during the 10-year post-injection period. The current model fairly explains observation data from pressure measurement, neutron and resistivity logs and geochemical fluid samplings. From the result, the following insights into the trapping processes of CO2 at the project can be drawn at present. x During the injection, free CO2 migrated preferentially through higher permeable layers inside the reservoir. The uneven arrival times of CO2 to the well-depths are attributable to non-uniform permeability distribution
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
x
x
x
5069
as measured at wells. Subsequent temporal changes in CO2 saturation was due to the effect of buoyant CO2 flow followed by residual gas trapping. The current model well simulated the temporal changes quantitatively by taking the hysteretic relative permeability and capillarity into account. Slow or even negligible migration of free CO2 to the vertical direction inside the reservoir, suggested by time-lapse resistivity logs, is attributed to a thin, intra-reservoir muddy-layer that behaved like an impermeable flow barrier to trap CO2, by a combined effect of low vertical permeability and high capillarity. Pressure-driven-flow during the injection resulted in hydrodynamic dispersion of dissolved-CO2 in water through lower permeable layer above and beneath the free CO2 zone. This was followed by slow diffusive transport during the post-injection. This interpretation is highly consistent with the negative resistivity changes found in the result from time-lapse logs in the post-injection period more than 10 years. The dissolved-CO2 in formation water increased bicarbonate concentration in formation water above and beneath the free CO2 zone, and promoted dissolution of rock minerals including dissolution of calcite and almino-silicates resulting in increase of calcium and silica concentrations, as suggested in previous studies. The coupled geochemical simulation suggests that these are precursor of mineral trapping by the precipitation of carbonates such as calcite in the future in this site.
Acknowledgements This work is part of an R&D project “the Development of Safety Management Technology for Large-Scale CO2 Geological Storage", commissioned to the Geological Carbon Dioxide Storage Technology Research Association by the Ministry of Economy, Trade and Industry (METI) of Japan. References [1] Sato, K., S. Mito, T. Hoie, H. Ohkuma, H. Saito, J. Watanabe, T. Yoshimura, 2011, Monitoring and simulation studies for assessing macroand meso-scale migration of CO2 sequestered in an onshore aquifer: Experiences from the Nagaoka pilot site, Japan. Int. J. Greenhouse Gas Control, doi:10.1016/j.ijggc.2010.03.003 [2] Nakajima, T., Xue, Z., 2016, Trapping mechanisms in field scale: Results from Nagaoka geologic CO2 storage site, GHGT13. [3] Ito, T., Nakajima, T., Chiyonobu, S., Xue, Z., 2015, Geostatistical modelling for the spatial mud content: an application to the Nagaoka CO2 storage site, Japan, Journal of Geological Society of Japan, Vol. 121, No. 9, p. 311-323 (in Japanese with English abstract) [4] Xu, T., Spycher N., Sonnenthal E., Zheng L., Pruess K, 2012, TOUGHREACT User’s Guide: A Simulation Program for Non-isothermal Multiphase Reactive Geochemical Transport in Variably Saturated Geologic Media, Version 2.0., Lawrence Berkeley National Laboratory [5] Pruess, K., Oldenburg, C., Moridis, G., 1999, TOUGH2 User’s Guide, Version 2.0, Rep. LBNL-43134. Lawrence Berkeley National Lab. [6] Pruess, K., 2005, ECO2N : A TOUGH2 Fluid Property Module for Mixtures of Water, NaCl, and CO2, Rep. LBNL-57952. [7] Yamamoto, H., Nakajima, K., Zhang, K., Nanai, S., 2014, Performance of Parallel Simulators on Peta-scale Platforms for Coupled Multiphysics Modelling of CO2 Geologic Sequestration, Energy Procedia, 63, 3795-3804. [8] Zhang, K., Wu, Y.S., Pruess, K., 2008, User’s Guide for TOUGH2-MP, Rep. LBNL-315E. Lawrence Berkeley National Lab. [9] Karypsis, G., Kumar, V., 1998, METIS V4.0, Technical Report, University of Minnesota [10] Tuminaro, R.S., Heroux, M., Hutchinson, S.A., Shadid, J.N.: Official Aztec user’s guide,Ver 2.1, Sandia National Laboratories (1999) [11] Doughty, C., 2009, User̓s Guide for Hysteretic Capiillary Pressure and Relative Permeability Functions in TOUGH2, Lawrence Berkeley National Laboratory [12] Nakajima, T. Ito, T., Xue, Z., 2016, Numerical simulation of the CO2 behaviour to obtain a detailed site characterization: A case study at Nagaoka pilot-scale injection site, GHGT13. [13] Audigane, P., Chiaberge, Ch., Mathurin, F., Picot-Colbeaux, G., Lions, J., 2011. A workflow for handling heterogeneous 3D models with the TOUGH2 family of codes: Applications to numerical modeling of CO2 geological storage. Computers & Geosciences, 37, 610-620. [14] Ohtake, 2013, Evaluation of CO2 inderground behavior from injector̓s time-lapse pressure fall off analysis: A case study of CO2 aquifer storage project. Energy Procedia, 37, 3307-3318. [15] Mito, S., Z. Xue, T. Sato, 2013, Effect of formation water composition on predicting CO 2 behavior: A case study at the Nagaoka postinjection monitoring site, Applied Geochemistry 30, pp.33-40. [16] Reed M.H. and Palandri J., 2006. SOLTHERM.H06, a database of equilibrium constants for minerals and aqueous species. Available from the authors, University of Oregon, Eugene, Oregon. [17] Jongwook, K., Matsuoka, T., Xue, Z., 2011, Monitoring and detecting CO2 injected into water-saturated sandstone with joint seismic and resistivity measurements, Exploration Geophysics (Butsuri-Tansa), 42, 53-68.
ScienceDirect Energy Procedia 114 (2017) 5058 – 5069
13th International Conference on Greenhouse Gas Control Technologies, GHGT-13, 14-18 November 2016, Lausanne, Switzerland
Quantitative interpretation of trapping mechanisms of CO2 at Nagaoka pilot project – A history matching study for 10-year post-injection – Hajime Yamamotoa*, Takahiro Nakajimab, Ziqiu Xue b a
b
Taisei Corporation,344-1, Nase-cho, Totsuka-ku, Yokohama 245-0051, Japan Research Institute of Innovative Technology for the Earth (RITE), 9-2, Kizugawadai, Kizugawa, Kyoto 619-0292, Japan
Abstract Demonstrating the permanence of CO2 storage is an important task of pilot projects. In the Nagaoka project, Japan’s first pilottest, a stable containment of CO2 in a reservoir has been successfully demonstrated by kept monitoring the CO 2 behavior even after the end of injection during more than 10 years. Systematic and continuous data acquisition of time-lapse well loggings illustrated the detailed nature of CO2 migration at intra-reservoir resolution. In this study, a three-dimensional reservoir model with sub-meter spatial resolution has been developed that involves coupled process of two-phase fluid flow and geochemical transport. The model was history-matched against a set of monitoring data including pressure, well loggings, and fluid samplings over the post-injection period. From the history matching study, the following insights into the trapping processes of CO2 can be drawn: 1) The uneven arrival times of CO2 to the well-depths are well explained by, and consistent with the non-uniform permeability distribution measured at wells; 2) Slow or even negligible vertical migration of free CO2 inside the reservoir suggests that even a thin intra-reservoir muddy-layer behaves like an impermeable flow barrier to trap CO2; 3) Pressure-drivenflow during the injection squeezed the formation water out of the reservoir, and subsequent slow diffusive transport of CO2 promoted dissolution of rock minerals including dissolution of calcite and almino-silicates, suggesting precursor of mineral trapping by the precipitation of carbonates such as calcite in the future in this site. © 2017 2017The TheAuthors. Authors. Published by Elsevier © Published by Elsevier Ltd. Ltd. This is an open access article under the CC BY-NC-ND license Peer-review under responsibility of the organizing committee of GHGT-13. (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the organizing committee of GHGT-13. Keywords: Nagaoka; post-injection; monitoring; simulation; trapping process; hystory matching; resistivity log; neutron log; fluid sampling
* Corresponding author. Tel.: +81-45-814-7237; fax: +81-45-814-7253. E-mail address: [email protected]
1876-6102 © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the organizing committee of GHGT-13. doi:10.1016/j.egypro.2017.03.1659
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
5059
1. Introduction Long term stability of CO2 stored in reservoir is of intrinsic importance for ensuring the viability of geologic sequestration of carbon dioxide. Demonstrating the permanence of storage is an important task of pilot projects. The Nagaoka project is the Japan’s first pilot-test of geological CO2 storage that injected about 10k tonnes of CO2. A stable containment of CO2 in a reservoir has been successfully demonstrated by kept monitoring the CO 2 behavior even after the end of injection for more than 10 years. Systematic and continuous data acquisition of time-lapse well loggings (e.g., resistivity, neutron, and sonic velocity) well illustrated the detailed nature of CO2 migration at intrareservoir resolution, including uneven CO2 arrivals to the well-depths and the prevention of vertical movement of buoyant CO2 inside the reservoir by a potential seal layer. The time evolutions of CO2 saturation were quantitatively evaluated under sub-meter resolutions by neutron logs. Resistivity logs even enable us to infer the dispersed area of CO2 dissolved in water in the reservoir. Fluid samplings at multiple depths also provided an insight into geochemical processes among CO2 and water. However, in order to better understand the trapping mechanisms from these detailed observations at intra-reservoir scale, the numerical modeling with sub-meter spatial resolution is necessary, which has not been performed yet in the project. In this study, a three-dimensional reservoir model with sub-meter spatial resolution has been developed that comprehensively involves coupled process of two-phase fluid flow and geochemical transport. The model was history-matched against a set of monitoring data acquired during the post-injection period including pressure, well loggings, and fluid samplings. The calibration of a large model is computationally demanding, hence we employed a parallel version of coupled fluid flow and geochemistry code TOUGHREACT V2.0/ECO2N with MPI parallelism, which has also been newly developed in-house. The new code also features hysteretic effect in relative permeability and capillarity which was not implemented in the original TOUGHREACT V2.0. 2. Post-injection monitoring at Nagaoka Site 2.1. Geological setting and well placement The Nagaoka CO2 storage site is located near the Nagaoka city in Niigata prefecture in Japan. From 2003 to 2005, approximately 10 kilo-tonnes of CO2 were injected and stored in an onshore aquifer. The storage aquifer is the early Pleistocene Haizume sandstone formation at 1,100m below surface, which is about 60m-thick, monoclinal structure inclined about 15° to the east-northeast. The reservoir can be subdivided into five intervals (Zone-1 to 5) based on lithofacies identified through well testing, core analysis, well-loggings, and correlation between these wells [1]. The location of an injection well (IW-1) and three observation wells (OB-2, OB-3, and OB-4) are shown in Fig. 1. OB-2 is located 40m down-dip from IW-1, while OB-3 and OB-4 are located in up-dip direction 120m and 60m away from IW-1 respectively. In these wells, a variety of CO2 monitoring has been conducted including continuous measurements of pressure and temperature, crosswell seismic tomography, well logging and in situ fluid sampling. The monitoring was continued even after the end of injection for more than 10 years. The details of the project are summarized in Sato et al. [1].
Fig. 1 Locations of injection well and observation wells [1].
5060
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
2.2. CO2 injection and arrival at observation wells In the storage aquifer, a thin sandy layer with 12 m thickness (Zone-2, average permeability 7 mD) was selected as the reservoir. In March 2004 (247 days after the start of the injection), the sonic, neutron, and resistivity loggings at observation wells OB-2 all detected significant changes, which are attributed to CO2 arrival [1, 2]. CO2 also later arrived at OB-4 in July 2004 (367 days), while any indication of CO2 arrival has not been detected at OB-3. The well-logging and fluid sampling indicated that CO2 arrived through only two highly permeable sub-intervals in Zone-2 (Zone-2a and 2b) as described in detail in the following sections. 2.3. Stratigraphy in Zone-2 Fig. 2 shows the stratigraphic profile of Zone-2 at the injection well IW-1 along with the vertical distribution of injectivity, mud content [3], and permeability. The geology of Zone-2 consists of alternating sand and silt/shale. The mud content and permeability distribution were determined by core analyses. The well injectivity was measured at the early period of the CO2 injection, and indicated that significant injectivity occurred in permeable interval inside Zone-2. Zone-2 can be further divided into three sub-intervals: Zone-2a, 2b, and 2c. The sub-intervals Zone-2a and 2b are bound by a thin muddy layer with high mud content nearly 100% as shown in Fig. 2.
A
Zone-2c Zone-2b
B
C D E 0
20 40 Permeability, md
Zone-2 (Injection Zone)
Thin muddy layer Zone-2a
Low injectivity
60
Permeability (md) 0 20 40 60 80 100
Injectivity (%)
Mud content (%)
Fig. 2 Stratigraphic profile of injection well IW-1 at intra-reservoir scale [3]
2.4. Post-injection monitoring During and after the injection for more than 10 years, time-lapse well loggings including induction resistivity, neutron, and sonic velocity have been repeatedly performed 44 times so far, and are still on going on an annual basis [1]. Formation water was collected twice in 2005 and 2011 at some depths in OB-2 by Cased Hole Dynamic Tester (CHDT). Among the monitoring data, in Fig. 3, the time evolution of resistivity change from baseline value obtained from resistivity logs for 12 years (about 4500 days) since the injection started (7 July, 2003) was depicted as a contour map [2]. This suggested uneven CO2 arrivals and subsequent saturation changes in Zone-2b. More importantly the
5061
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
vertical movement of buoyant CO2 has been prevented at the top of Zone-2b, where the potential seal layer with high mud content was identified as seen in Fig. 2. The resistivity logs even enable us to infer the dispersed area of dissolved CO2 in water, by the negative resistivity changes gradually spread over and beneath Zone-2b. In addition to this resistivity result, neutron logs quantitatively evaluated CO2 saturation under sub-meter resolutions, and fluid samplings at multiple depths also provided an insight into geochemical processes among CO 2, water, as discussed in following chapters. As described above, the post-injection monitoring depicted the detailed nature of CO2 migration at the intrareservoir scale. However, in order to better understand the trapping processes from these detailed observations, numerical modeling with sub-meter spatial resolution is necessary, which has not been performed yet in the project. Post-Injection
Injection
CHDT (1108.6m)
A CHDT (1112m)
Resistivity Change (%) 20 10
B
CHDT (1114m)
C
0
D
-10
E
CHDT (1118m)
CHDT (1118m)
-20
CHDT (1119.5m)
Fig. 3 Temporal change of resistivity from baseline along the observation well OB-2 [2].
3. Model Setup 3.1. Numerical code TOUGHREACT V2.0 [4] is a coupled fluid flow and geochemistry code developed by introducing reactive chemistry into TOUGH2 [5], which is a three-dimensional, fully implicit model of non-isothermal, multi-component and multi-phase flow and heat in porous and fractured media. For applications to geologic storage of CO2 in saline aquifers, a fluid properties module ECO2N [6] can be used together. In this study, we newly developed a parallel version of coupled fluid flow and geochemistry TOUGHREACT V2.0 with MPI parallelism, which run on multiple CPU/core computer systems [7]. The parallelization scheme is similar to that of TOUGH2-MP [8]. It solves sparse linear systems arising from discretization of the partial differential equations for mass and energy balance. The code uses MPI for parallel implementation, the METIS software package [9] for simulation domain partitioning, and the iterative parallel linear solver package Aztec [10] for solving linear equations by multiple processors. This in-house code also features hysteretic effect in relative permeability and capillarity developed by Doughty [11], which has not been implemented in the original TOUGHREACT V2.0.
5062
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
3.2. Reservoir model The reservoir is geologically interpreted as deltaic or coastal plain deposits characterized by up-ward-shallowing successions [3]. A heterogeneous distribution of permeability and porosity in three-dimension was realized on the basis of geological interpretation by the concept of sequential stratigraphy and sequential Gaussian simulation on the Petrel software [3, 12]. The Petrel model was converted to integral finite difference grids for TOUGHREACT simulations by utilizing a workflow & tools developed by BRGM [13]. The three dimensional reservoir model used in this study is shown in Fig. 4, which has been developed by Nakajima et al.[12]. Local grid refinement (LGR) was applied in the vicinity of the CO2 injection area. The grid-sizes in horizontal direction inside the outer and inner LGR region are 10m and 5 m respectively. In this study, we performed numerical modeling to interpret the detailed observations of well loggings acquired under sub-meter vertical resolutions, whereas the original model was vertically discretized Zone-2 with only 3 layers (i.e., 5.5 m grid-spacing for Zone-2a and 2b, and 1m grid-spacing for Zone 2c). Hence the original model was further refined with a sub-meter grid spacing of 50 cm in vertical direction, and inhomogeneous distribution of porosity and permeability along depths arising from alternating thin layers of mud, sand, and conglomerate was directly represented in the model. Based on the cross-well correlation of facies and permeability, we assumed that the vertical inhomogeneous pattern of permeability identified in Zone-2 at IW-1 (Fig. 2a, rightmost) can be extrapolated laterally throughout the Zone-2. Fig. 2b shows the resulted distribution of permeability on the east-west vertical cross-section through IW-1. Additionally we inserted thin mud layers, which have lower permeability (0.01 mD), and high capillarity, at the upper and bottom boundary of Zone-2b. This operation is based on the facts that mud content around the top of Zone-2b is nearly 100 % and no injectivity below Zone-2b were found, as seen in Fig. 2.
a
OB-4 OB-3
IW-1 OB-2
b
Fig. 4 Reservoir model with permeability distribution: (a) Perspective view; (b) E-W vertical cross-section through IW-1
5063
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
3.3. Coupled flow and geochemical model The settings of fluid flow and geochemical model are summarized in Table 1 and Table2, respectively. The fluid flow model involved three components (H2O, CO2, and NaCl) and two-phase (CO2 and water) flow [6], which was further coupled with chemically reactive transport with 11 primary aqueous species and 7 minerals including some carbonate, almino-silicate, and clay minerals. The geochemical model was basically taken from Mito et al. [15], and SOLTHERM thermodynamic database [16] was used. Table 1 Settings of fluid flow model Settings Iso-thermal Three components (H2O, CO2, and NaCl) Two phases (scCO2 and brine˅ Pressure: about 10 MPa Temperature: 48°C Salinity: 8h10-3 (NaCl mass fraction in aqueous phase) Sequential Gaussian simulation Permeability was calibrated against temporal pressure changes and CO2 arrival time at OB-2. Lateral extraporation of vertical inhomogeneous pattern of permeability in Zone-2b at IW-1 Thin muddy layers at the top and bottom of Zone-2b Curve fitting using laboratory tests (core flooding[14]) Parker and Lenhard model Curve fitting using laboratory tests (mercury injection)[12] van Genuchten model Two phase diffusion by Millington-Quirk ఉ ݀ =2h10-9 m2/s
Fluid flow model
Initial reservoir condition
Permeability Porosity
Relative permeability Capillary pressure Diffusion model
Boundary condition Injection rate Simulation duration
Quartz K-Feldspar Anorthite Albite-low Ca-montmorillonite Calcite Kaolinite
Anisotropy: ky=6hkx, kz=kx/10
Hysteretic effects [11] are incorporated. Hysteretic effects [11] are incorporated.
All boundaries closed (top, bottom, lateral) Measured flowrate data (temporally changes) 12 years
Table 2 Settings for geochemical model [15] (a) Initial mineral composition and kinetic parameters Mineral
Remarks ECO2N fluid property module [6]
Intial Vol% 29.3 13.5 9.8 20.2 29.2 0.2 0.0
RSA (m2/m3 mineral) 2.40E+04 2.40E+04 2.40E+04 2.40E+04 1.50E+06 2.40E+04 1.50E+06
(b) Primary aqueous species Rate Constant (mol/m2 s) 1.26E-14 1.00E-10 1.60E-12 1.00E-11 4.00E-14 1.60E-09 1.00E-13
Activation Energy (kJ/mol) 87.5 41.9 18.4 67.8 48.0 42.0 62.8
Component pH HCO3SO42ClNa K Mg Ca Si Al
Initial Concentration mmol/kg 6.4 3.9 0.8 96 74 6.4 0.7 10 1.8 0.004
5064
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
4. Results and Discussion 4.1. General behavior of CO2 Distribution of CO2 saturation (free CO2) and CO2 dissolved in water on E-W vertical cross-section through IW-1 are shown in Fig. 5. Free CO2 preferentially invades in high permeable layers of Zone-2a and 2b. The front of CO2dissolved formation water proceeds ahead of the free CO2 front. Inside the free CO2 zones, buoyancy gradually drives the portion of high CO2 saturation in the up-dip direction, leaving the footprint of CO2 plume down-dip, where residual trapping takes place. Pressure transient at IW-1 and OB-2 for 800 days is shown in Fig. 6. Due to the history matching efforts previously performed by Nakajima et al. [12], simulated pressure transient at OB-4 matched well with measured one for this refined-grid model. At IW-1, the pressure obtained from simulation generally exceeds the observation, probably due to acid treatment of IW-1 for increasing injectivity performed prior to the injection. The acid treatment can greatly increase permeability around the well and result in a large negative skin effect [14], which is not accounted for in this simulation.
a
b
Fig. 5 Time evolution of CO2 migration on E-W vertical cross-section through IW-1: (a) 1.6 years (just after injection terminated); (b) 10 years (8.5 years after injection terminated)
a
b
Fig. 6 Temporal pressure change at wells: (a) IW-1; (b) OB-4
4.2. Neutron logging Fig. 7 shows changes in CO2 saturation at OB-2 and OB-4 over time for three depth-intervals in Zone-2b (B, C, D in Fig. 3) at OB-2, and middle part of Zone-2 at OB-4. The CO2 saturation was evaluated from neutron loggings
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
5065
repeatedly conducted [1], and averaged values within each interval are shown in the figure. The figure also includes the simulation results, which are fairly matched with the observation, although it needed to assume a strong anisotropy of permeability to the north-south direction (kx:ky = 1:6) to explain the early arrival time of CO2 at OB-2. The direction of the anisotropy is almost corresponding to the direction of sedimentary sequences. The uneven arrival times of CO2 at the three depth intervals (B, C, D) in Zone-2b at OB-2 are well explained by, and consistent with the non-uniform permeability distribution shown in Fig.2: Permeability of C interval is obviously lower than that of B and D. Subsequent temporal changes in CO2 saturations are also well simulated by taking the hysteretic effect of relative permeability into account in the current model. At the down-dip OB-2, CO2 saturation generally dropped from 30 - 40 % to 15 - 20 % between about 1500 to 2000 days, and then the saturation levels were stabilized. This behavior is attributed to buoyant flow and residual trapping effect. In fact, at OB-4 which is located at the up-dip direction of IW-1, neither simulation nor observation results showed such reduction of CO2 saturation so far. This is presumably because the effect of buoyant flow on CO2 saturation has not been still prominent.
Fig. 7 CO2 saturation changes at observation wells OB-2(down-dip) and OB-4(up-dip). B, C, and D are sub-intervals inside Zone-2b: see Fig. 3.
4.3. Resistivity logging Time evolution of CO2 saturation and dissolved-CO2 in aqueous phase at OB-2 obtained from the simulation are shown in Fig. 8. Comparing with the resistivity result shown in Fig. 3, the following interpretations can be drawn at present. (1) Vertical migration of free CO2 In Fig. 3, the top of red-colored region (positive change of resistivity) at 1113 m depth (Zone-2b top) has stayed the same depth throughout the monitoring period for more than 10 years. This suggests that the vertical migration of free CO2 has been prevented at the depth, which is consistent with the simulation result (Fig. 8a): In the model, the vertical invasion of free CO2 from Zone-2b to 2a is prevented by a combined effect of lower vertical permeability and high capillarity of a thin muddy layer assumed at the top of Zone-2b. Our history-matching efforts indicated that the containment of free CO 2 in Zone-2b cannot be reproduced without the thin muddy layer. This suggests the continuity of the thin muddy layer over Zone-2b from IW-1 to OB-2, and that even a thin, intra-reservoir muddylayer behaved like an impermeable flow barrier to trap free CO2 during the decade. (2) Buoyancy effect of free CO2 The bottom boundary of free CO2 is slowly moves upward in the simulation (Fig. 8a). This is attributed to the buoyancy flow due to lower density of supercritical CO2 than water. However, it is not prominent in the observation
5066
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
of resistivity change in Fig. 3. Practically, the buoyancy flow in the vertical direction can be regarded as being negligible in the result. (3) Early arrival of CO2 dissolved in water When injecting CO2 into a virgin aquifer, it is reasonable that water dissolving CO2 migrates ahead of free CO2 plume. The negative resistivity change observed across Zone-2a after the termination of the injection (Fig. 3) is attributable to early arrival of dissolved CO2 in water prior to the arrival of free CO2. According to the simulation, while free CO2 arrives at OB-2 through Zone-2b, only water dissolving CO2 reaches in Zone-2a. Fig. 8b presents the vertical distribution of dissolved CO2 in aqueous phase at OB-2 over time, which shows arrival of CO2 dissolving water in Zone-2a at about 500 days, while no free CO2 arrival in the zone until 12 years (about 4500 days). (4) Dispersion/diffusion of CO2 dissolved in water It is also seen in Fig. 3 that negative resistivity change occurred just over and beneath Zone-2b in Fig. 3. The region with negative change of resistivity slowly spread above and below even during the post-injection. Regarding this behavior, the current simulation provides the following interpretation (Fig. 8b): 1) During the injection, pressure-driven-flow squeezed the formation water out of Zone-2b, which resulted in a rapid hydrodynamic dispersion of CO2 in water; 2) After the injection terminated, the reservoir pressure recovered to the initial and water ceased to flow. But still slow diffusion process gradually transported CO2 through formation water. The spreading speed of dissolved-CO2 water is almost comparable with that of negative resistivity change in Fig. 3 (about 2-3 meters in 10 years). Here we may try to calculate the resistivity change from the simulation result. Archies’s equation relates formation resistivity R and CO2 saturation Sco2 as below: ܴ ൌ Ƚି ሺͳ െ ܵைଶ ሻି ܴ௪
(1)
where, n is the saturation exponent (usually equal to 2), Rw is resistivity of formation water. From an experimental study [17], we employed α = 1, m = 1.59, and Rw = 1. By taking the resistivity change due to CO2 dissolution also into account, we roughly estimated the resistivity change from the simulated result as shown in Fig. 9.
a
b
Fig. 8 Time evolution of CO2 saturation and dissolved-CO2 in formation water at OB-2 obtained from the simulation. The amount of dissolvedCO2 in water is represented as mass fraction of CO2 in aqueous phase (XCO2aq): (a) Saturation of gaseous CO2; (b) Dissolved CO2 in formation water
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
a
-20
Resistivity Change (%) -10 0 10
20
5067
b
A
B C D
E
Fig. 9 Comparison of resistivity changes estimated from simulation with observation: (a) Resistivity logging; (b) Simulation
4.4. Fluid sampling The formation water was collected from OB-2 by Cased Hole Dynamic Tester (CHDT) at 1108.6 m, 1114 m, and 1118 m depths in December 2005 (1st CHDT), and then at 1112 m, 1118 m and 1119.5 m depths in September 2011 (2nd CHDT) [15]. The location and timing of the samplings are shown in Fig.3. Clear increases in concentrations of HCO3-, Ca, Fe, Mg, and SiO2 have been detected after the injection (1st CHDT), suggesting significant geochemical reactions had occurred in the formation water [15]. These increases in ion concentration are consistent with the decrease in resistivity at the sampling depths (circles) shown in Fig. 3. At 1118 m depth, where 1st and 2nd CHDT were both carried out, further increase in concentrations of HCO 3- and SiO2 associated with slight decrease in Ca concentration have been observed during the post-injection period. The observed increase in HCO3- concentration outside of the free CO2 zone can be attributed to the dispersive and diffusive transport of CO2 as discussed above. This should promote dissolution of minerals including calcite and almino-silicates, which increases such as Ca, and SiO2 concentrations. The history matching of the geochemical model is challenging and still ongoing in the study. This paper only presents a result from a base model [15]. Fig. 10 includes simulated temporal changes in pH and aqueous concentrations of HCO3, Ca, and SiO2 at 1118m depth. Although the concentrations of each component obtained from the simulation are also almost comparable with the fluid sampling result. However, while the geochemical model contains a lot of uncertain parameters, the constraints for history matching of the geochemical model are insufficient. Fig. 11 shows the time evolution of mineral dissolution and precipitation simulated by the current model. Just after the CO2 arrives, pH dropped quickly and the dissolutions of calcite and feldspar (albite) proceed quickly. After injection stopped, pH gradually recovers and precipitation of calcite occurs slowly. The decrease in Ca concentration between 1st and 2nd CHDT may be attributed to calcite precipitation as suggested by Mito et al. [15]. This result can be regarded as a precursor of mineral trapping by the precipitation of carbonates such as calcite in the future in this site. However, the current geochemical model does not include full set of potential carbonate minerals such as dawsonite, siderite, maginesite, so further improvement will be needed in the next step.
5068
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
Fig. 10 Temporal changes in pH and aqueous concentrations at 1118 m depth
Fig. 11 Time evolution of mineral dissolution and precipitation simulated by the current model
5. Conclusion The modeling study successfully reproduced the behavior of CO2 observed at sub-meter scale in the field during the 10-year post-injection period. The current model fairly explains observation data from pressure measurement, neutron and resistivity logs and geochemical fluid samplings. From the result, the following insights into the trapping processes of CO2 at the project can be drawn at present. x During the injection, free CO2 migrated preferentially through higher permeable layers inside the reservoir. The uneven arrival times of CO2 to the well-depths are attributable to non-uniform permeability distribution
Hajime Yamamoto et al. / Energy Procedia 114 (2017) 5058 – 5069
x
x
x
5069
as measured at wells. Subsequent temporal changes in CO2 saturation was due to the effect of buoyant CO2 flow followed by residual gas trapping. The current model well simulated the temporal changes quantitatively by taking the hysteretic relative permeability and capillarity into account. Slow or even negligible migration of free CO2 to the vertical direction inside the reservoir, suggested by time-lapse resistivity logs, is attributed to a thin, intra-reservoir muddy-layer that behaved like an impermeable flow barrier to trap CO2, by a combined effect of low vertical permeability and high capillarity. Pressure-driven-flow during the injection resulted in hydrodynamic dispersion of dissolved-CO2 in water through lower permeable layer above and beneath the free CO2 zone. This was followed by slow diffusive transport during the post-injection. This interpretation is highly consistent with the negative resistivity changes found in the result from time-lapse logs in the post-injection period more than 10 years. The dissolved-CO2 in formation water increased bicarbonate concentration in formation water above and beneath the free CO2 zone, and promoted dissolution of rock minerals including dissolution of calcite and almino-silicates resulting in increase of calcium and silica concentrations, as suggested in previous studies. The coupled geochemical simulation suggests that these are precursor of mineral trapping by the precipitation of carbonates such as calcite in the future in this site.
Acknowledgements This work is part of an R&D project “the Development of Safety Management Technology for Large-Scale CO2 Geological Storage", commissioned to the Geological Carbon Dioxide Storage Technology Research Association by the Ministry of Economy, Trade and Industry (METI) of Japan. References [1] Sato, K., S. Mito, T. Hoie, H. Ohkuma, H. Saito, J. Watanabe, T. Yoshimura, 2011, Monitoring and simulation studies for assessing macroand meso-scale migration of CO2 sequestered in an onshore aquifer: Experiences from the Nagaoka pilot site, Japan. Int. J. Greenhouse Gas Control, doi:10.1016/j.ijggc.2010.03.003 [2] Nakajima, T., Xue, Z., 2016, Trapping mechanisms in field scale: Results from Nagaoka geologic CO2 storage site, GHGT13. [3] Ito, T., Nakajima, T., Chiyonobu, S., Xue, Z., 2015, Geostatistical modelling for the spatial mud content: an application to the Nagaoka CO2 storage site, Japan, Journal of Geological Society of Japan, Vol. 121, No. 9, p. 311-323 (in Japanese with English abstract) [4] Xu, T., Spycher N., Sonnenthal E., Zheng L., Pruess K, 2012, TOUGHREACT User’s Guide: A Simulation Program for Non-isothermal Multiphase Reactive Geochemical Transport in Variably Saturated Geologic Media, Version 2.0., Lawrence Berkeley National Laboratory [5] Pruess, K., Oldenburg, C., Moridis, G., 1999, TOUGH2 User’s Guide, Version 2.0, Rep. LBNL-43134. Lawrence Berkeley National Lab. [6] Pruess, K., 2005, ECO2N : A TOUGH2 Fluid Property Module for Mixtures of Water, NaCl, and CO2, Rep. LBNL-57952. [7] Yamamoto, H., Nakajima, K., Zhang, K., Nanai, S., 2014, Performance of Parallel Simulators on Peta-scale Platforms for Coupled Multiphysics Modelling of CO2 Geologic Sequestration, Energy Procedia, 63, 3795-3804. [8] Zhang, K., Wu, Y.S., Pruess, K., 2008, User’s Guide for TOUGH2-MP, Rep. LBNL-315E. Lawrence Berkeley National Lab. [9] Karypsis, G., Kumar, V., 1998, METIS V4.0, Technical Report, University of Minnesota [10] Tuminaro, R.S., Heroux, M., Hutchinson, S.A., Shadid, J.N.: Official Aztec user’s guide,Ver 2.1, Sandia National Laboratories (1999) [11] Doughty, C., 2009, User̓s Guide for Hysteretic Capiillary Pressure and Relative Permeability Functions in TOUGH2, Lawrence Berkeley National Laboratory [12] Nakajima, T. Ito, T., Xue, Z., 2016, Numerical simulation of the CO2 behaviour to obtain a detailed site characterization: A case study at Nagaoka pilot-scale injection site, GHGT13. [13] Audigane, P., Chiaberge, Ch., Mathurin, F., Picot-Colbeaux, G., Lions, J., 2011. A workflow for handling heterogeneous 3D models with the TOUGH2 family of codes: Applications to numerical modeling of CO2 geological storage. Computers & Geosciences, 37, 610-620. [14] Ohtake, 2013, Evaluation of CO2 inderground behavior from injector̓s time-lapse pressure fall off analysis: A case study of CO2 aquifer storage project. Energy Procedia, 37, 3307-3318. [15] Mito, S., Z. Xue, T. Sato, 2013, Effect of formation water composition on predicting CO 2 behavior: A case study at the Nagaoka postinjection monitoring site, Applied Geochemistry 30, pp.33-40. [16] Reed M.H. and Palandri J., 2006. SOLTHERM.H06, a database of equilibrium constants for minerals and aqueous species. Available from the authors, University of Oregon, Eugene, Oregon. [17] Jongwook, K., Matsuoka, T., Xue, Z., 2011, Monitoring and detecting CO2 injected into water-saturated sandstone with joint seismic and resistivity measurements, Exploration Geophysics (Butsuri-Tansa), 42, 53-68.