Pressure monitoring data assimilation to locate and quantify leaks in carbon storage projects.

Abstract:

We investigate the use of pressure data from monitoring wells overlying a carbon storage
aquifer with uncertain geology, in order to locate and quantify leakage as quickly and inexpensively as
possible. Formal data assimilation using the Karhunen-Loeve expansion reduces the optimization
variable space, while enabling candidate solutions to implicitly honor both hard (well) data and prior
geologic characterizations. Minimization of the mismatch in pressure response between synthetic
`true' and history matched models is performed using particle swarm optimization. Our results
indicate that, given the prior geologic characterization, as little as six to 12 months of pressure
monitoring data may be sufficient to reasonably locate a leak and to quantify leakage, including future
leakage over extended periods. We find that the accuracy in predicting leak locations improves with
additional monitoring wells and suggest that three to four wells may be sufficient for reasonable
location estimates. No significant benefit is seen in the cases considered when using multilevel versus
single-level wells for monitoring in the overlying aquifer. Finally, we find that adding white noise, with
magnitude consistent with current pressure monitoring techniques, generally reduces error in
solutions, which is likely due to a regularization effect. Taken in total, the results and procedures
introduced in this study should be of use in designing monitoring strategies for large-scale carbon
storage projects.