Monitoring-Modelling loop – experience from the Ketzin site Stefan Lüth, Thomas Kempka, Alexandra Ivanova GFZ German Research Centre for Geosciences Combined Meeting of the IEAGHG Modelling and Monitoring Networks Edinburgh, 5-8 July 2016
Monitoring-Modelling loop –experience from the Ketzin site
Stefan Lüth, Thomas Kempka, Alexandra Ivanova
GFZ German Research Centre for Geosciences
Combined Meeting of the IEAGHG Modelling and Monitoring NetworksEdinburgh, 5-8 July 2016
Agenda
COP21
Reservoir Modelling and Simulation
Simulated and observed Plumes
Performance criteria
Conclusion and outlook
Project designRisk
assessment
Enhancement of process
understandingForecasting
Modeling and simulation started before injection and accompany entire operation
Bielinski (2007)Kopp et al. (2008)
Frykman (2008)
Lengler et al. (2010)Chen et al. (2014)
Kempka et al. (2010)Wiese et al. (2010)
Kempka and Kühn (2013)Klein et al. (2013)Kempka et al. (2013)Kempka et al. (2014a,b)De Lucia et al. (2015)
Geological model:Norden and Frykman (2013)Kempka et al. (2013a)
Matching simulated and observed pressures –model correctly describes effective hydraulic
properties close to wells
Sleipner: simulated and observed plume areas(top layer)
Chadwick & Noy, 2015
• Shape not exactly reproduced.• Performance criteria defined to describe effective properties.• Convergence to excellent match for, e.g., plume footprint area.
Ketzin: observed and simulated plume geometries
CO2
2009 2012 2015
Seismic
Simulation
0.3
7 m
Quantifying conformity between simulation and observation with uncertainties
Performance parameters
• Geometrical parameters (effective reservoir properties):• Plume footprint area (permeability).• Maximum lateral migration (anisotropy).• Plume volume (permeability, layering, P-T).
• Direct comparison of footprints:• Similarity index (Sørensen-Dice coeffizient, reservoir model
representing „true“ heterogeneity).
AB C
S = 2C / (A + B)
Uncertainties
• „Seismic plume“ • Amplitude (or impedance) anomaly in difference data.• Affected by noise, not related to storage.• Amplitude threshold.
• „Simulated plume“• True reservoir heterogeneity unlikely to be fully described
by the model.• Distribution of CO2 partially goes into very thin layers.
--> Apply performance criteria to range of amplitude thresholdand thickness values.
Plume footprint area 2012 and 2015
2009 2012
Sim 2015
Sim 2012
Seis 2015
Seis 2012
Simulation 2015 Seismic 2015
Seismic 2012Simulation 2012
2015 vs. 2012:
Simulated plumefootprint has grown.Seismic plume footprinthas become smaller.
• Seismic observationrealistic – simulationunderestimatingdissolution…
or
• Simulation realistic –seismic observationhas a detectionissue.
Similarity index
2009 2012
2009 2012 2015
• Similarity index shows decreasing conformity from 2009 – 2012 – 2015.• Reservoir model based on pressure history matching and 2009 seismic data.• Model uncertainty increasing with plume propagating further beyond Ketzin
wells.• Dissolution and diffusion correctly described?
Conclusion & outlook
• Pressure based history matching good for describing efectivehydraulic properties if sufficient pressure observationsavailable.
• High-resolution time-lapse seismic observations and reservoirsimulations allow conformity assessment at reservoir scale.
• Effective performance criteria (e.g. footprint area) andsimilarity demonstrate matching of observations andsimulations.
• Next step forward: coupled inversion of geophysics andreservoir simulations by including performance criteria intoobjective function.
Thank you!