Canadian Unconventional Resources and International Petroleum Conference,
19-21 October 2010,
Calgary, Alberta, Canada
The geological storage of carbon dioxide (CO2) provides the possibility of
maintaining access to fossil energy, while reducing emissions of CO2 to the
atmosphere. One of the essential concerns in geologic storage is the risk of
CO2 leakage from the storage formations. The leakage occurs through possible
pathways in the seal. Characterization of the CO2 leakage pathways from the
storage formations into overlying formations is required. The aquifer cap-rock
may be characterized before CO2 storage. This will allow for the determination
of proper storage aquifers and locations for the injection wells. In a
companion paper, a flow and pressure test has been suggested for
characterization of leakage pathways in aquifer cap-rock. Water is injected in
the target aquifer, and the pressure is observed in an overlying aquifer. The
pressure data are analyzed to characterize the leakage pathways in the
cap-rock. In this work, design considerations to maximize the capability of
leakage characterization are presented.
A leakage pathway can be characterized by the leak transmissibility and
location parameters. A successful test should be able to provide sufficient
information to evaluate the leakage parameters. In this work, different
strategies are evaluated in order to achieve a successful test. The strategies
include increasing the sampling frequency, use of pulsing, increasing the
number of monitoring/injection wells and utilization of prior information.
Prior information on the leak is provided through analysis of the pressure
Estimation of the leakage parameters is actually an inverse problem that is
generally ill-conditioned and very sensitive to noise. The information provided
by different strategies is evaluated, based on their effects on well-posing the
inverse problem. The effects are studied based on information and correlation
matrices, as well as the confidence interval.
One of the main challenges facing development and deployment of CO2 capture and
storage in deep saline aquifers is the risk of leakage. The injected buoyant
CO2 may leak through the leakage pathways in an otherwise sealing cap-rock. In
a companion paper (Zeidouni and Pooladi-Darvish, 2010), a pressure and flow
test has been introduced to detect and characterize the leak based on pressure
data measured in an overlying aquifer. The leak is characterized based on
location and transmissibility parameters. In dimensionless form, these
parameters are the dimensionless abscissa of the leak (xlD), the dimensionless
ordinate of the leak (ylD), and the leakage coefficient α (klrl 2/(2kshshl)).
For convenience, the origin of the coordinate system is considered to be at the
(first) injection well, while the (first) monitoring well is considered to be
at (L,0) coordinates.
Applying the inverse methodology to a base case, it is shown that, due to
relatively linearly dependent sensitivity coefficients and a high correlation
between the location parameters, it may be difficult to obtain the leak
parameters through pressure information at a single monitoring well. The
parameter estimation process may be very unstable, causing the leakage
parameters inferred from pressure data to vary over a very large confidence
interval. It is noted that the solution to the leakage inverse problem is
non-unique. At least two and at most four leaks with the same transmissibility
but different locations exist that equally fit the data. Obtaining one set of
parameters the remaining possible sets can be calculated.
In this work, we investigate different strategies to maximize the information
that can be obtained in order to characterize the leak. The strategies include
increasing the number of data samples, use of injection pulses, considering
multiple monitoring wells, considering multiple injection wells, and
regularization using information extracted from derivative analysis. The
information added through these procedures are investigated based on
information and correlation matrices, as well as confidence intervals.