A Risk Assessment Modeling Tool to Quantify Leakage Potential ...

A Risk Assessment Tool to QuantifyCO 2 Leakage Potential through Wellsin Mature Sedimentary BasinsMichael A. Celia (Princeton University)Stefan Bachu (Alberta EUB)Jan Nordbotten (U. Bergen and Princeton U.)Dmitri Kavetski (Princeton U.)Sarah Gasda (Princeton U.)

Outline• Leakage and Existing Wells• A Modeling Framework for LeakageEstimation• Example Application• Conclusions and Future Work

Worldwide Density of Oil and Gas WellsFrom IPCC SRCCS, 2005End of 2004

Potential CO 2 Migration and Leakage Paths

Components of the Semi-analytical Model• Injection plume evolution– Similarity solution (Captures full buoyant drive)– Radial Buckley-Leverett type solution (Viscous domination)– Includes drying fronts– Captures essential physics while being simple enough for analytical solution• Leakage Dynamics• Post-injection Redistribution• Upconing around Leaky WellsH

Distribution of Existing Wells in the Wabamun Lake Area

Down-Hole Stratigraphic Model for the Wabamun Lake Area,Alberta, Canada – Pre-Cretaceous Unconformity to Surface

Down-Hole Stratigraphic Model for the Wabamun Lake Area,Alberta, Canada – Precambrian to Pre-Cretaceous Unconformity

Probability Distribution for WellPermeabilities

Monte Carlo Simulations4:1Bimodal Gaussian⎧ N ( −20,1)with plog10[ K] ~ ⎨⎩N( −16,2) 1−p111:11:4Intact cement: K ~ 10 -20 m 2Degraded cement: K ~ 10 -16 m 2

Plume spread – Single realizationLayer above injection, after 50 yrs

Plume spread – AveragedLayer above injection, after 50 yrs

Mean leakage: Time profileFirst two layers above injection over 50 yrs

Leakage statistics after 50 years4:1 Gaussian mixture (10 -20 m 2 : 10 -16 m 2 )Mean ~ -3.1 [0.08%]P(>1%)=8%

Leakage statistics after 50 years1:1 Gaussian mixture (10 -20 m 2 : 10 -16 m 2 )Mean ~ -2.4 [0.4%]P(>1%)=25%

Leakage statistics after 50 years1:4 Gaussian mixture (10 -20 m 2 : 10 -16 m 2 )Mean ~ -2.1 [0.8%]P(>1%)=44%

Maximum Leakage (at 95% Confidence) Above Injection Layer>50%

Maximum Leakage (at 95% Confidence) Above Second Layer

How can we identify this distribution?We need a targeted field campaign:• Re-enter existing (abandoned) wells.• Use statistical mix of wells (location, depth, age, …).• Simple pressure tests to identify K eff in well segments.