- Email: [email protected]
Coupled flow-geomechanical performance assessment of CO2 storage sites using the Ensemble Kalman Filter
Available online at www.sciencedirect.com
ScienceDirect Energy Procedia 63 (2014) 3475 – 3482
GHGT-12
Coupled flow-geomechanical performance assessment of CO2 storage sites using the Ensemble Kalman Filter Jorge de la Torre Guzman*, Masoud Babaei, Ji-Quan Shi, Anna Korre and Sevket Durucan Department of Earth Science and Engineering, Royal School of Mines, Imperial College London, London SW7 2BP, United Kingdom
Abstract A methodology for coupled flow-geomechanical CO2 storage site performance assessment using the ensemble Kalman filter (EnKF) is developed. The methodology is applied to the In Salah case study dataset in order to validate the developed technique. The EnKF approach is used to estimate the parameters controlling the performance of the CO2 storage site by assimilating data from multiple sources. The controlling parameters in this study are fault/fracture transmissibility and anisotropic permeability at the injection wells; and the data sources are the injection wells’ flowing bottom-hole pressure (FBHP) and Interferometric Synthetic Aperture Radar (InSAR) surface deformation. The methodology is shown to be effective in updating the parameters controlling the performance of the CO2 storage site. © 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license © 2013 The Authors. Published by Elsevier Ltd. (http://creativecommons.org/licenses/by-nc-nd/3.0/). Selection and peer-review under responsibility GHGT. of GHGT-12 Peer-review under responsibility of the Organizing of Committee Keywords: In Salah CO2 storage; InSAR surface uplift monitoring; reservoir history matching; coupled flow-geomechanical modelling
1. Introduction The ability to detect CO2 migration through fractures, permeable or semi-permeable faults that are pre-existing or injection induced is a key component in the performance assessment of geological CO2 storage sites [1]. The presence of faults, fractures and related hydro-fractures may lead to enhanced CO2 migration paths and may cause changes in the pore-pressure and permeability distribution within the injected zone [2]. Forward and inverse
* Corresponding author. Tel.: +44-20-7594-7368 E-mail address: [email protected]
1876-6102 © 2014 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/3.0/). Peer-review under responsibility of the Organizing Committee of GHGT-12 doi:10.1016/j.egypro.2014.11.376
3476
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
methodologies using InSAR surface uplift observations have been used independently to gain insights into CO2 storage site performance at In Salah [3, 4, 5, 6, 7]. In the context of oil and gas production, the Ensemble Kalman Filter (EnKF) has been used to integrate forward and inverse coupled flow-geomechanical modelling. Chang et al. [8] tested the applicability of the EnKF to coupled flow-geomechanical phenomena due to fluid production and showed an improved performance of the EnKF by assimilating both surface subsidence and production data. Wilschut et al. [9] have used the EnKF to assimilate surface subsidence and production data to estimate reservoir fault transmissibilities at the Roswinkel field. In this study, a methodology, which uses the EnKF to assimilate observations from multiple sources, including flowing bottom-hole pressure (FBHP) and InSAR deformation data, was developed for coupled flow-geomechanical performance assessment of CO2 storage and applied to the In Salah case study dataset. The methodology involves defining an ensemble of coupled flow-geomechanical models that capture the uncertainty in the flow-geomechanical model parameters. The EnKF is able to simultaneously assimilate observations from different sources. The ensemble of flowígeomechanical models are sequentially updated to honour injection data and geomechanical observations, therefore reducing the uncertainty in the driving parameters controlling the performance of the CO2 storage site. Fig. 1 shows a schematic workflow of the methodology.
Fig. 1.
Schematic workflow for the coupled flow-geomechanical performance assessment of CO2 storage sites using the EnKF.
1.1 The In Salah CO2 storage complex In Salah is an industrial-scale CO2 storage project located in central Algeria. Between August 2004 and June 2011, around 3.8 million metric tonnes of CO2 have been separated from the extracted natural gas from the In Salah gas fields and re-injected into the aquifer leg of the Krechba field Carboniferous reservoir through 3 horizontal
3477
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
injection wells KB-501, KB-502 and KB-503 (Fig. 2). The main CO2 storage aquifer is the 20-25 m thick C10.2 formation with a relatively low permeability at a depth of approximately 1,880 m, which is overlain by the 20 m thick tight sandstone and siltstone C10.3 formation. These are in turn overlain by a 950 m thick formation of mudstone interbedded with thin dolomite and siltstone layers. The C10 formation, together with the lower cap rock (C20.1–C20.3), forms the CO2 storage complex at Krechba. There are three CO2 injection wells at the In Salah storage site (KB-501, KB-502, and KB-503). FBHP estimates and InSAR surface uplift data are available over the entire injection period (2004-2011) at the site. Further details of the field site layout can be found in Mathieson et al. [10]. 2. Methodology The EnKF is a sequential data-assimilation method that updates both model parameters and state variables as data measurements become available [11]. The EnKF is an approximation to the Kalman filter, with the error covariance approximated from a finite ensemble. A detailed description of the EnKF for reservoir monitoring and continuous model updating can be found in Naevdal et al. [12]. The first step in the implementation of the methodology is to define an initial ensemble of coupled flow-geomechanical models consistent with any prior knowledge of the initial state. Each ensemble contains the parameters and state variables that define the coupled flow-geomechanical model. Typical primary parameters for fluid-flow simulations are porosity and permeability. The parameters for elastic geomechanical processes are Young’s Modulus and Poisson’s Ratio. State variables for a coupled model include pressure, stress, strain and displacement. The next stage is the forecast step. The forecast step consists of a coupled flow-geomechanical simulation for each ensemble member. The third stage is the data assimilation or analysis step in which both the model parameters and the state of the system are updated to honour the available observations. The update to each ensemble member is made using an ensemble approximation to the Kalman gain. There are several existing approaches for coupled flow-geomechanical simulations [13]. In this study, a one-way coupling approach between the fluid-flow modelling code ECLIPSE and the geomechanical code FLAC3D was chosen in order to assess the geomechanical changes associated with CO2 injection. In a one-way coupling approach the reservoir model is exported from ECLIPSE to FLAC3D and embedded into a larger geomechanical model with sideburden and underburden added to minimise boundary effects. The two codes are sequentially executed and linked through an external coupling module. The pore pressure predictions from the ECLIPSE code are exported into FLAC3D and act as a load for surface deformation predictions. The reservoir/overburden model consisted of 5 stratigraphic units and a total of 15 layers. The main storage reservoir C10.2 contains three layers and the overlying tight sandstone formation consists of one layer. The next 4 layers represent the C20.1-2 formations. The reservoir is attributed with geostatistically simulated porosity and permeability values. For the geostatistical simulations, the porosity and permeability data were derived from core measurements from 6 wells and were conditioned on seismic porosity surfaces. In this study, one geostatistical simulation is used for all ensemble members. The transmissibility for the fault/fracture zone intersecting KB-502 and the anisotropic permeability multipliers for KB-501 and KB-503 are the uncertain parameters. Expected values for these parameters were taken from a previous study by Shi et al. [6]. Table 1. Flow properties (based on data by the In Salah JIP) Mean porosity (std. dev)
Mean kh, mD
Log10(kh) (std. dev.)
kv/kh
Shallow Aquifer
0.25 (0.2)
1,000
3.0 (0.5)
0.1
C20.3+
0.10 (0.01)
0.001
-3.0 (0.5)
0.1
C20.1-2
0.05 (0.01)
0.1
-1.0 (0.5)
0.1
C10.3
0.01 (0.01)
0.02
-1.7 (0.5)
0.1
C10.2
Numerical grid with stochastically generated permeability and porosity values
The set of elastic rock properties defined for the geomechanical model are the Young’s Modulus E and Poisson’s ratio Ȟ of the reservoir and the over/side/under-burden rocks (Table 2) based on data provided by the In Salah JIP. It
3478
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
has been noted that the variation of rock elastic properties with depth in layered media can have a significant impact on the surface deformation. Vasco et al. [4] have shown that the elastic properties of the lower caprock have a large impact on the surface uplift. The elastic properties for the shallow aquifer, the main caprock and the lower caprock (C20.1-3) are subject to considerable uncertainty, whereas those for the tight sandstone layer (C10.3), the main reservoir (C10.2) and the Devonian underburden are estimated from good quality logs [14]. For this study, the geomechanical parameters are kept constant for the ensemble members. Table 2. Geomechanical model rock elastic properties (based on data by the In Salah JIP) E (GPa)
Ȟ
Shallow Aquifer
3.0
0.25
C20.3+
5.0
0.30
C20.1-2
18.0
0.30
C10.3
20.0
0.25
C10.2
10.0
0.20
Devonian
15.0
0.30
It has been shown previously that important features of the site are the dynamic behaviour of the fracture/fault zone intersecting injection well KB-502 [2, 15] and the anisotropic flow at injection well KB-501 and KB-503 [6]. In an integrated fracture analysis in which fracture network prediction and present-day stress analysis are combined, Bond et al. [16] predicted a fracture dominated anisotropic flow field along the NW-SE maximum stress direction consistent with InSAR and well data. Against this backdrop and building on a previous study by Shi et al. [6], the fracture/fault transmissibility and the anisotropic permeability multipliers were chosen as the parameters to be estimated in this study. Three scenarios are defined: 1) No data assimilation; 2) BHP assimilation; 3) joint BHP and InSAR data assimilation. In this study only the first two cases were considered. The bottom-hole pressures (BHP’s) at the three CO2 injection wells were not measured directly. The BHP is estimated from well head pressure measurements using a well model [17]. There is uncertainty in the BHP conversion with discrepancies ranging up to ~30 bar [7]. A 5% error was assumed for the estimated BHP, to account for the uncertainties in the estimation of the data. Surface deformation and uplift has been observed extending laterally for several kilometres at all three CO2 injection wells (Fig. 2). The Permanent-scatterer InSAR (PSInSAR) processing of Envisat data by Tele-Relivamento Europa (TRE) used during the first monitoring period from 2003 to early 2009 are described in Vasco et al. [3, 4] and Rucci et al. [18]. During the second monitoring period beginning in November 2009, the SqeeSARTM processing technique [19] was employed. In this study, only the BHP estimates are assimilated.
Fig. 2. January 20, 2012 Surface deformation detected over the CO2 injection and production wells using PSInSAR (courtesy of the In Salah Gas Joint Industry Project).
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
3. Results and discussion This section presents the application of the above methodology to the In Salah CO2 storage case study dataset using an ensemble size of Ͳ. In the first case, no data assimilation is performed. In the second case BHP is assimilated on a monthly basis. Fig. 3 shows the match to the estimated BHP for all three injection wells compared to the initial case. The range of fault/fracture transmissibilities and anisotropic multipliers that have been used to initialise the ensemble are shown in Figures 4 and 5.
Fig. 3. Match to estimated BHP data from the initial ensemble and from the updated ensemble using BHP data assimilation.
It is interesting to note the evolution of the BHP sensitivity to the ensemble parameters for the initial case with no data assimilation. Injection well KB-501 displays a weak BHP sensitivity to the initial parameters. For the second half of injection the BHP is over-predicted. The BHP assimilation has a diminished impact in reducing the uncertainty. At injection well KB-502 there is a significant increase in BHP sensitivity to the fault/fracture transmissibility between 2005 and 2007. Afterwards, as expected the sensitivity is gradually reduced throughout the
3479
3480
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
shut-in period between 2007 and 2009. The sensitivity increases once again in the second injection phase starting from the end of 2009 to 2011. BHP assimilation at injection well KB-502 significantly improves the match. The BHP predictions for injection well KB-503 display a more constant sensitivity throughout the entire CO2 injection period for the initial parameters. Data assimilation also produces a better match to the estimated BHP. After several BHP assimilation steps, the uncertainty in the BHP predictions are reduced. Fig. 4 shows the updates for the transmissibility at every data assimilation step. The transmissibility increases considerably starting in November 2005 consistent with the sighting of the two-lobe uplift deformation pattern. The uncertainty increases during the shut in period when injection stopped. Once injection resumes the average transmissibility increases and is tuned by the EnKF, reducing its uncertainty. This is in agreement with previous interpretations and modelling work [2] suggesting hydraulic stimulation starting in November 2005.
Fig. 4. Updated ensemble transmissibility for well KB-502.
Fig. 5. Updated ensemble permeability multipliers in the x- and y- direction for well KB-501 and KB-503.
Fig. 5 shows the updates for the permeability multipliers at every data assimilation step for well KB-501 and KB503. The updated anisotropic multipliers for KB-501 suggest an increase in flow over time but the BHP is overpredicted given the small ensemble variability at this injection well. For KB-503, the average anisotropic multipliers
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
for KB-501 in the x-direction is lower for the initial half of the injection period with an increase observed for the second half. The average permeability in the y-direction remains stable. The observed increase in flow over the second half of injection at the site is consistent with CO2 being diverted to these two wells during the shut-in period at injection well KB-502, increasing the injection rates and enhancing flow. Finally, Fig. 6 shows the predicted surface deformation for the initial ensemble and for the BHP data assimilation case. Injection well KB-501 follows the correct trend but is slightly over-predicted. At injection well KB-502 there is a mismatch starting from the shut-in period. At injection well KB-503 the deformation is under-predicted with a mismatch in trend at the start of injection and from the start of 2009.
Fig. 6. Surface deformation predictions from the initial ensemble and from the updated ensemble using BHP data assimilation.
4. Conclusions In this study, a methodology using the EnKF for assessing the coupled flow-geomechanical performance of CO2 storage sites was tested on the In Salah CO2 storage site. The assimilation of BHP using the EnKF provided an effective method for updating the parameters controlling the performance of a CO2 storage site. In particular at well KB-502, the results were in agreement with previous modelling studies. The BHP was tracked by the EnKF and the dynamic fault/fracture transmissibility was updated accordingly. At injection wells KB-501 and KB-503 the
3481
3482
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
permeability multipliers increased in the second half of injection suggesting enhanced flow due to CO2 injection. An underlying assumption was that the parameters are dynamic. The aim was to understand the dynamic reservoir changes in response to CO2 injection. The well-known problem of inconsistency in the EnKF analysis was not considered and should be further investigated. Strictly speaking, only BHP was assimilated in this study, but insight from InSAR deformation patterns were used to define anisotropic flow directions and fault/fracture zones. The modelled surface deformation suggests it is important to include the uncertainty in the geomechanical parameters in order to improve our geomechanical model. Further studies will aim to include uncertainty in geomechanical properties and the assimilation of InSAR surface deformation in addition to BHP. Acknowledgements The authors wish to thank the In Salah Gas Joint Industry Project (BP, Statoil, Sonatrach) for provision of data and permission to publish this analysis. The ECLIPSE 300 software used in this study is kindly provided by Schlumberger. References [1] Rutqvist J. The Geomechanics of CO2 storage in deep sedimentary formations. Geotechnical and Geological Engineering 2012; 30 (3): 525551. [2] Shi J, Sinayuc C, Durucan S, Korre A. Assessment of carbon dioxide plume behaviour within the storage reservoir and lower caprock around the KB-502 injection well at In Salah. International Journal of Greenhouse Gas Control 2012; 7: 115-126. [3] Vasco DW, Ferretti A, Novali F. Estimating permeability from quasi-static deformation: temporal variations and arrival-time inversion. Geophysics 2008; 73: O37–O52. [4] Vasco DW, Rucci A, Ferretti A, Novali F, Bissell RC, Ringrose PS, Mathieson AS and Wright IW. Satellite-based measurements of surface deformation reveal fluid flow associated with the geological storage of carbon dioxide. Geophysical Research Letters 2010; 37: L03303. [5] Rutqvist J, Vasco DW, Myer L.. Coupled reservoir-geomechanical analysis of CO2 injection and ground deformations at In Salah, Algeria, International Journal of Greenhouse Gas Control 2010; 4: 225-230. [6] Shi J, Smith J, Durucan S, Korre A. A coupled reservoir simulation-geomechanical modelling study of the CO2 injection-induced ground surface uplift observed at Krechba, In Salah. Energy Procedia 2013; 37: 3719-3726. [7] White JA, Chiaramonte L, Ezzedine S, Foxall W, Hao Y, Ramirez A and McNab W. Geomechanical behavior of the reservoir and caprock system at the In Salah CO2 storage project. Proceedings of the National Academy of Sciences 2014; 1-6. [8] Chang H, Chen Y, Zhang D. Data assimilation of coupled fluid flow and geomechanics using the ensemble Kalman filter. SPE Journal 2010; 382-394. [9] Wilschut F, Peters E, Visser K, Fokker PA, van Hooff PME. Joint history matching of well data and surface subsidence observations using the ensemble Kalman filter: a field study. SPE International 2011; 1-12. [10] Mathieson A, Midgley J, Dodds K, Wright I, Ringrose P, Saoula N. CO2 sequestration monitoring and verification technologies applied at Krechba Algeria. The Leading Edge 2010; 29 (2): 216–222. [11] Evensen G. Sequential data assimilation with a nonlinear quasi-geostrophic model using Monte Carlo methods to forecast error statistics. Journal of Geophysical Research 2004; 99 (C5): 10143-10162. [12] Naevdal G, Johnsen LM, Aanonsen SI, Vefring EH. Reservoir monitoring and continuous model updating using ensemble Kalman filter. SPE Journal 2005; 10 (1): 66-74. [13] Samier P, Onaisi A, Fontaine G. Comparisons of uncoupled and various coupling techniques for practical field examples. SPE Journal 2006; 11 (1): 89-102. [14] Gemmer L, Hansen O, Iding M, Leary S, Ringrose P. Geomechanical response to CO2 injection at Krechba, In Salah, Algeria. First Break 2012; 30: 79-84. [15] Rinaldi AP, Rutqvist J. Modeling of deep fracture zone opening and transient ground surface uplift at KB-502 CO2 injection well, In Salah, Algeria. International Journal of Greenhouse Gas Control 2013; 12 (0): 155-167. [16] Bond CE, Wightman R, Ringrose PS. The influence of fracture anisotropy on CO2 flow. Geophysical Research Letters 2013, 40 (7): 12841289. [17] Bissell RC, Vasco DW, Atbi M, Hamdani M, Okwelegbe M, Goldwater MH. A full field simulation of the In Salah gas production and CO2 storage project using a coupled geo-mechanical and thermal fluid flow simulator. Energy Procedia 2011; 4:3290–3297. [18] Rucci A, Vasco DW, Novali F. Monitoring the geologic storage of carbon dioxide using multicomponent SAR interferometry. Geophysical Journal International 2013; 193:197–208. [19] Ferretti A, Fumagalli A, Novali F, Prati C, Rocca F, Rucci A. A new algorithm for processing interferometric data-stacks: SqueeSAR. IEEE Trans. Geosci.Remote Sens. 2011; 49: 3460–3470.
ScienceDirect Energy Procedia 63 (2014) 3475 – 3482
GHGT-12
Coupled flow-geomechanical performance assessment of CO2 storage sites using the Ensemble Kalman Filter Jorge de la Torre Guzman*, Masoud Babaei, Ji-Quan Shi, Anna Korre and Sevket Durucan Department of Earth Science and Engineering, Royal School of Mines, Imperial College London, London SW7 2BP, United Kingdom
Abstract A methodology for coupled flow-geomechanical CO2 storage site performance assessment using the ensemble Kalman filter (EnKF) is developed. The methodology is applied to the In Salah case study dataset in order to validate the developed technique. The EnKF approach is used to estimate the parameters controlling the performance of the CO2 storage site by assimilating data from multiple sources. The controlling parameters in this study are fault/fracture transmissibility and anisotropic permeability at the injection wells; and the data sources are the injection wells’ flowing bottom-hole pressure (FBHP) and Interferometric Synthetic Aperture Radar (InSAR) surface deformation. The methodology is shown to be effective in updating the parameters controlling the performance of the CO2 storage site. © 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license © 2013 The Authors. Published by Elsevier Ltd. (http://creativecommons.org/licenses/by-nc-nd/3.0/). Selection and peer-review under responsibility GHGT. of GHGT-12 Peer-review under responsibility of the Organizing of Committee Keywords: In Salah CO2 storage; InSAR surface uplift monitoring; reservoir history matching; coupled flow-geomechanical modelling
1. Introduction The ability to detect CO2 migration through fractures, permeable or semi-permeable faults that are pre-existing or injection induced is a key component in the performance assessment of geological CO2 storage sites [1]. The presence of faults, fractures and related hydro-fractures may lead to enhanced CO2 migration paths and may cause changes in the pore-pressure and permeability distribution within the injected zone [2]. Forward and inverse
* Corresponding author. Tel.: +44-20-7594-7368 E-mail address: [email protected]
1876-6102 © 2014 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/3.0/). Peer-review under responsibility of the Organizing Committee of GHGT-12 doi:10.1016/j.egypro.2014.11.376
3476
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
methodologies using InSAR surface uplift observations have been used independently to gain insights into CO2 storage site performance at In Salah [3, 4, 5, 6, 7]. In the context of oil and gas production, the Ensemble Kalman Filter (EnKF) has been used to integrate forward and inverse coupled flow-geomechanical modelling. Chang et al. [8] tested the applicability of the EnKF to coupled flow-geomechanical phenomena due to fluid production and showed an improved performance of the EnKF by assimilating both surface subsidence and production data. Wilschut et al. [9] have used the EnKF to assimilate surface subsidence and production data to estimate reservoir fault transmissibilities at the Roswinkel field. In this study, a methodology, which uses the EnKF to assimilate observations from multiple sources, including flowing bottom-hole pressure (FBHP) and InSAR deformation data, was developed for coupled flow-geomechanical performance assessment of CO2 storage and applied to the In Salah case study dataset. The methodology involves defining an ensemble of coupled flow-geomechanical models that capture the uncertainty in the flow-geomechanical model parameters. The EnKF is able to simultaneously assimilate observations from different sources. The ensemble of flowígeomechanical models are sequentially updated to honour injection data and geomechanical observations, therefore reducing the uncertainty in the driving parameters controlling the performance of the CO2 storage site. Fig. 1 shows a schematic workflow of the methodology.
Fig. 1.
Schematic workflow for the coupled flow-geomechanical performance assessment of CO2 storage sites using the EnKF.
1.1 The In Salah CO2 storage complex In Salah is an industrial-scale CO2 storage project located in central Algeria. Between August 2004 and June 2011, around 3.8 million metric tonnes of CO2 have been separated from the extracted natural gas from the In Salah gas fields and re-injected into the aquifer leg of the Krechba field Carboniferous reservoir through 3 horizontal
3477
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
injection wells KB-501, KB-502 and KB-503 (Fig. 2). The main CO2 storage aquifer is the 20-25 m thick C10.2 formation with a relatively low permeability at a depth of approximately 1,880 m, which is overlain by the 20 m thick tight sandstone and siltstone C10.3 formation. These are in turn overlain by a 950 m thick formation of mudstone interbedded with thin dolomite and siltstone layers. The C10 formation, together with the lower cap rock (C20.1–C20.3), forms the CO2 storage complex at Krechba. There are three CO2 injection wells at the In Salah storage site (KB-501, KB-502, and KB-503). FBHP estimates and InSAR surface uplift data are available over the entire injection period (2004-2011) at the site. Further details of the field site layout can be found in Mathieson et al. [10]. 2. Methodology The EnKF is a sequential data-assimilation method that updates both model parameters and state variables as data measurements become available [11]. The EnKF is an approximation to the Kalman filter, with the error covariance approximated from a finite ensemble. A detailed description of the EnKF for reservoir monitoring and continuous model updating can be found in Naevdal et al. [12]. The first step in the implementation of the methodology is to define an initial ensemble of coupled flow-geomechanical models consistent with any prior knowledge of the initial state. Each ensemble contains the parameters and state variables that define the coupled flow-geomechanical model. Typical primary parameters for fluid-flow simulations are porosity and permeability. The parameters for elastic geomechanical processes are Young’s Modulus and Poisson’s Ratio. State variables for a coupled model include pressure, stress, strain and displacement. The next stage is the forecast step. The forecast step consists of a coupled flow-geomechanical simulation for each ensemble member. The third stage is the data assimilation or analysis step in which both the model parameters and the state of the system are updated to honour the available observations. The update to each ensemble member is made using an ensemble approximation to the Kalman gain. There are several existing approaches for coupled flow-geomechanical simulations [13]. In this study, a one-way coupling approach between the fluid-flow modelling code ECLIPSE and the geomechanical code FLAC3D was chosen in order to assess the geomechanical changes associated with CO2 injection. In a one-way coupling approach the reservoir model is exported from ECLIPSE to FLAC3D and embedded into a larger geomechanical model with sideburden and underburden added to minimise boundary effects. The two codes are sequentially executed and linked through an external coupling module. The pore pressure predictions from the ECLIPSE code are exported into FLAC3D and act as a load for surface deformation predictions. The reservoir/overburden model consisted of 5 stratigraphic units and a total of 15 layers. The main storage reservoir C10.2 contains three layers and the overlying tight sandstone formation consists of one layer. The next 4 layers represent the C20.1-2 formations. The reservoir is attributed with geostatistically simulated porosity and permeability values. For the geostatistical simulations, the porosity and permeability data were derived from core measurements from 6 wells and were conditioned on seismic porosity surfaces. In this study, one geostatistical simulation is used for all ensemble members. The transmissibility for the fault/fracture zone intersecting KB-502 and the anisotropic permeability multipliers for KB-501 and KB-503 are the uncertain parameters. Expected values for these parameters were taken from a previous study by Shi et al. [6]. Table 1. Flow properties (based on data by the In Salah JIP) Mean porosity (std. dev)
Mean kh, mD
Log10(kh) (std. dev.)
kv/kh
Shallow Aquifer
0.25 (0.2)
1,000
3.0 (0.5)
0.1
C20.3+
0.10 (0.01)
0.001
-3.0 (0.5)
0.1
C20.1-2
0.05 (0.01)
0.1
-1.0 (0.5)
0.1
C10.3
0.01 (0.01)
0.02
-1.7 (0.5)
0.1
C10.2
Numerical grid with stochastically generated permeability and porosity values
The set of elastic rock properties defined for the geomechanical model are the Young’s Modulus E and Poisson’s ratio Ȟ of the reservoir and the over/side/under-burden rocks (Table 2) based on data provided by the In Salah JIP. It
3478
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
has been noted that the variation of rock elastic properties with depth in layered media can have a significant impact on the surface deformation. Vasco et al. [4] have shown that the elastic properties of the lower caprock have a large impact on the surface uplift. The elastic properties for the shallow aquifer, the main caprock and the lower caprock (C20.1-3) are subject to considerable uncertainty, whereas those for the tight sandstone layer (C10.3), the main reservoir (C10.2) and the Devonian underburden are estimated from good quality logs [14]. For this study, the geomechanical parameters are kept constant for the ensemble members. Table 2. Geomechanical model rock elastic properties (based on data by the In Salah JIP) E (GPa)
Ȟ
Shallow Aquifer
3.0
0.25
C20.3+
5.0
0.30
C20.1-2
18.0
0.30
C10.3
20.0
0.25
C10.2
10.0
0.20
Devonian
15.0
0.30
It has been shown previously that important features of the site are the dynamic behaviour of the fracture/fault zone intersecting injection well KB-502 [2, 15] and the anisotropic flow at injection well KB-501 and KB-503 [6]. In an integrated fracture analysis in which fracture network prediction and present-day stress analysis are combined, Bond et al. [16] predicted a fracture dominated anisotropic flow field along the NW-SE maximum stress direction consistent with InSAR and well data. Against this backdrop and building on a previous study by Shi et al. [6], the fracture/fault transmissibility and the anisotropic permeability multipliers were chosen as the parameters to be estimated in this study. Three scenarios are defined: 1) No data assimilation; 2) BHP assimilation; 3) joint BHP and InSAR data assimilation. In this study only the first two cases were considered. The bottom-hole pressures (BHP’s) at the three CO2 injection wells were not measured directly. The BHP is estimated from well head pressure measurements using a well model [17]. There is uncertainty in the BHP conversion with discrepancies ranging up to ~30 bar [7]. A 5% error was assumed for the estimated BHP, to account for the uncertainties in the estimation of the data. Surface deformation and uplift has been observed extending laterally for several kilometres at all three CO2 injection wells (Fig. 2). The Permanent-scatterer InSAR (PSInSAR) processing of Envisat data by Tele-Relivamento Europa (TRE) used during the first monitoring period from 2003 to early 2009 are described in Vasco et al. [3, 4] and Rucci et al. [18]. During the second monitoring period beginning in November 2009, the SqeeSARTM processing technique [19] was employed. In this study, only the BHP estimates are assimilated.
Fig. 2. January 20, 2012 Surface deformation detected over the CO2 injection and production wells using PSInSAR (courtesy of the In Salah Gas Joint Industry Project).
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
3. Results and discussion This section presents the application of the above methodology to the In Salah CO2 storage case study dataset using an ensemble size of Ͳ. In the first case, no data assimilation is performed. In the second case BHP is assimilated on a monthly basis. Fig. 3 shows the match to the estimated BHP for all three injection wells compared to the initial case. The range of fault/fracture transmissibilities and anisotropic multipliers that have been used to initialise the ensemble are shown in Figures 4 and 5.
Fig. 3. Match to estimated BHP data from the initial ensemble and from the updated ensemble using BHP data assimilation.
It is interesting to note the evolution of the BHP sensitivity to the ensemble parameters for the initial case with no data assimilation. Injection well KB-501 displays a weak BHP sensitivity to the initial parameters. For the second half of injection the BHP is over-predicted. The BHP assimilation has a diminished impact in reducing the uncertainty. At injection well KB-502 there is a significant increase in BHP sensitivity to the fault/fracture transmissibility between 2005 and 2007. Afterwards, as expected the sensitivity is gradually reduced throughout the
3479
3480
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
shut-in period between 2007 and 2009. The sensitivity increases once again in the second injection phase starting from the end of 2009 to 2011. BHP assimilation at injection well KB-502 significantly improves the match. The BHP predictions for injection well KB-503 display a more constant sensitivity throughout the entire CO2 injection period for the initial parameters. Data assimilation also produces a better match to the estimated BHP. After several BHP assimilation steps, the uncertainty in the BHP predictions are reduced. Fig. 4 shows the updates for the transmissibility at every data assimilation step. The transmissibility increases considerably starting in November 2005 consistent with the sighting of the two-lobe uplift deformation pattern. The uncertainty increases during the shut in period when injection stopped. Once injection resumes the average transmissibility increases and is tuned by the EnKF, reducing its uncertainty. This is in agreement with previous interpretations and modelling work [2] suggesting hydraulic stimulation starting in November 2005.
Fig. 4. Updated ensemble transmissibility for well KB-502.
Fig. 5. Updated ensemble permeability multipliers in the x- and y- direction for well KB-501 and KB-503.
Fig. 5 shows the updates for the permeability multipliers at every data assimilation step for well KB-501 and KB503. The updated anisotropic multipliers for KB-501 suggest an increase in flow over time but the BHP is overpredicted given the small ensemble variability at this injection well. For KB-503, the average anisotropic multipliers
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
for KB-501 in the x-direction is lower for the initial half of the injection period with an increase observed for the second half. The average permeability in the y-direction remains stable. The observed increase in flow over the second half of injection at the site is consistent with CO2 being diverted to these two wells during the shut-in period at injection well KB-502, increasing the injection rates and enhancing flow. Finally, Fig. 6 shows the predicted surface deformation for the initial ensemble and for the BHP data assimilation case. Injection well KB-501 follows the correct trend but is slightly over-predicted. At injection well KB-502 there is a mismatch starting from the shut-in period. At injection well KB-503 the deformation is under-predicted with a mismatch in trend at the start of injection and from the start of 2009.
Fig. 6. Surface deformation predictions from the initial ensemble and from the updated ensemble using BHP data assimilation.
4. Conclusions In this study, a methodology using the EnKF for assessing the coupled flow-geomechanical performance of CO2 storage sites was tested on the In Salah CO2 storage site. The assimilation of BHP using the EnKF provided an effective method for updating the parameters controlling the performance of a CO2 storage site. In particular at well KB-502, the results were in agreement with previous modelling studies. The BHP was tracked by the EnKF and the dynamic fault/fracture transmissibility was updated accordingly. At injection wells KB-501 and KB-503 the
3481
3482
Jorge de la Torre Guzman et al. / Energy Procedia 63 (2014) 3475 – 3482
permeability multipliers increased in the second half of injection suggesting enhanced flow due to CO2 injection. An underlying assumption was that the parameters are dynamic. The aim was to understand the dynamic reservoir changes in response to CO2 injection. The well-known problem of inconsistency in the EnKF analysis was not considered and should be further investigated. Strictly speaking, only BHP was assimilated in this study, but insight from InSAR deformation patterns were used to define anisotropic flow directions and fault/fracture zones. The modelled surface deformation suggests it is important to include the uncertainty in the geomechanical parameters in order to improve our geomechanical model. Further studies will aim to include uncertainty in geomechanical properties and the assimilation of InSAR surface deformation in addition to BHP. Acknowledgements The authors wish to thank the In Salah Gas Joint Industry Project (BP, Statoil, Sonatrach) for provision of data and permission to publish this analysis. The ECLIPSE 300 software used in this study is kindly provided by Schlumberger. References [1] Rutqvist J. The Geomechanics of CO2 storage in deep sedimentary formations. Geotechnical and Geological Engineering 2012; 30 (3): 525551. [2] Shi J, Sinayuc C, Durucan S, Korre A. Assessment of carbon dioxide plume behaviour within the storage reservoir and lower caprock around the KB-502 injection well at In Salah. International Journal of Greenhouse Gas Control 2012; 7: 115-126. [3] Vasco DW, Ferretti A, Novali F. Estimating permeability from quasi-static deformation: temporal variations and arrival-time inversion. Geophysics 2008; 73: O37–O52. [4] Vasco DW, Rucci A, Ferretti A, Novali F, Bissell RC, Ringrose PS, Mathieson AS and Wright IW. Satellite-based measurements of surface deformation reveal fluid flow associated with the geological storage of carbon dioxide. Geophysical Research Letters 2010; 37: L03303. [5] Rutqvist J, Vasco DW, Myer L.. Coupled reservoir-geomechanical analysis of CO2 injection and ground deformations at In Salah, Algeria, International Journal of Greenhouse Gas Control 2010; 4: 225-230. [6] Shi J, Smith J, Durucan S, Korre A. A coupled reservoir simulation-geomechanical modelling study of the CO2 injection-induced ground surface uplift observed at Krechba, In Salah. Energy Procedia 2013; 37: 3719-3726. [7] White JA, Chiaramonte L, Ezzedine S, Foxall W, Hao Y, Ramirez A and McNab W. Geomechanical behavior of the reservoir and caprock system at the In Salah CO2 storage project. Proceedings of the National Academy of Sciences 2014; 1-6. [8] Chang H, Chen Y, Zhang D. Data assimilation of coupled fluid flow and geomechanics using the ensemble Kalman filter. SPE Journal 2010; 382-394. [9] Wilschut F, Peters E, Visser K, Fokker PA, van Hooff PME. Joint history matching of well data and surface subsidence observations using the ensemble Kalman filter: a field study. SPE International 2011; 1-12. [10] Mathieson A, Midgley J, Dodds K, Wright I, Ringrose P, Saoula N. CO2 sequestration monitoring and verification technologies applied at Krechba Algeria. The Leading Edge 2010; 29 (2): 216–222. [11] Evensen G. Sequential data assimilation with a nonlinear quasi-geostrophic model using Monte Carlo methods to forecast error statistics. Journal of Geophysical Research 2004; 99 (C5): 10143-10162. [12] Naevdal G, Johnsen LM, Aanonsen SI, Vefring EH. Reservoir monitoring and continuous model updating using ensemble Kalman filter. SPE Journal 2005; 10 (1): 66-74. [13] Samier P, Onaisi A, Fontaine G. Comparisons of uncoupled and various coupling techniques for practical field examples. SPE Journal 2006; 11 (1): 89-102. [14] Gemmer L, Hansen O, Iding M, Leary S, Ringrose P. Geomechanical response to CO2 injection at Krechba, In Salah, Algeria. First Break 2012; 30: 79-84. [15] Rinaldi AP, Rutqvist J. Modeling of deep fracture zone opening and transient ground surface uplift at KB-502 CO2 injection well, In Salah, Algeria. International Journal of Greenhouse Gas Control 2013; 12 (0): 155-167. [16] Bond CE, Wightman R, Ringrose PS. The influence of fracture anisotropy on CO2 flow. Geophysical Research Letters 2013, 40 (7): 12841289. [17] Bissell RC, Vasco DW, Atbi M, Hamdani M, Okwelegbe M, Goldwater MH. A full field simulation of the In Salah gas production and CO2 storage project using a coupled geo-mechanical and thermal fluid flow simulator. Energy Procedia 2011; 4:3290–3297. [18] Rucci A, Vasco DW, Novali F. Monitoring the geologic storage of carbon dioxide using multicomponent SAR interferometry. Geophysical Journal International 2013; 193:197–208. [19] Ferretti A, Fumagalli A, Novali F, Prati C, Rocca F, Rucci A. A new algorithm for processing interferometric data-stacks: SqueeSAR. IEEE Trans. Geosci.Remote Sens. 2011; 49: 3460–3470.