| dc.description.abstract | The international commitments for carbon capture will require a rapid increase in
carbon capture and storage (CCS) projects. The key to any successful carbon sequestration
project lies in the long term storage and prevention of leakage of stored
carbon dioxide (CO2). In addition to being a greenhouse gas, CO2 leaks reaching the
surface can accumulate in low-lying areas resulting in a serious health risk. Among
several alternatives, some of the more promising CSS storage formations are the hundreds
of thousands of depleted oil and gas reservoirs, whereby definition the reservoirs
had good geological seals prior to hydrocarbon extraction. With more CSS wells coming
online, it is imperative to implement permanent, automated monitoring tools. In
this study, we applied machine learning models to automate the leakage detection
process in carbon storage reservoirs using rates of supercritical (CO2) injection and
pressure data measured by simple pulse tests. To validate the promise of this machine
learning-based work
ow, we implemented data from pulse tests carried out in
the Cran eld reservoir, Mississippi, USA. The data consist of a series of pulse tests
conducted with baseline parameters and with an artificially introduced leak. Here,
we pose the leakage detection task as an anomaly detection problem where deviation
from the predicted behavior indicates leaks in the reservoir. The results obtained
show that different machine learning architectures such as multi-layer feed-forward
network, Long Short-Term Memory, convolutional neural network are able to identify leakages and can act as an early warning. These warnings can then be used by human interpreters to take remedial measures. | en_US |
