Multi-Objective Optimization of Monitoring Well Location for CO2 Leakage Detection in GCS

1. Abstract Capture and sequestration of CO2 in deep geological formations is an attractive option to reduce the amount of CO2 emissions into the atmosphere. The capture of CO2 is mostly done at large stationary point sources, such as power plants and refineries, and then it is injected in deep saline aquifers or old oil and gas fields. One of the environmental and policy concerns with this strategy is the potential of injected CO2 to leak back to the atmosphere through abandoned oil and gas wells or to leak into underground sources of drinking water that can create other environmental problems. A risk management option is to place pressure-monitoring wells in overlying geological formations in order to detect changes in the pressure field that can come to occur due to leakage. The monitoring strategy becomes a question of optimization that tries to answer the question of how many wells to place and where to place them. In order to address this specific problem this study proposes a methodology that incorporates the use o f a computational model that is ran under a Monte Carlo scheme, a multi-objective evolutionary algorithm (MOA) and the use of a discrete static Kalman Filter (KF) to evaluate different objectives to be optimized. The computational model used simulates flow across 10 leaky wells and two geological formations, and the MOA evaluates using the KF the combinations of well number and locations. The two objectives considered in this study are the reduction of (1) the number of wells and (2) the total coefficient of variation (CV). For this study two cases were run: In the first case, the potential leakage well locations are considered to be known but with unknown permeability; the results show that the uncertainty can be completely reduced when using 10 monitoring wells (or more) but a substantial reduction is attainable with only 5 wells – uncertainty reduction of 64%. Simulation results also show that the placement of 10 monitoring wells does not really matter and that any location would reduce the uncertainty due to the high spatial correlation among wells. For the second case, the potential leaky well locations, as well as their permeability, were considered to be unknown. This case showed that the uncertainty is reduced only by 67% with 10 wells and that the placement of the wells follows a more balanced symmetric distribution. In this second case, after the placement of 5 wells the marginal reduction in uncertainty is minimal. Overall, this methodology strengthens the justification for the use of monitoring wells in overlying formation to detect leakage pathways associated with geological carbon sequestration, and it also provides an insight on the appropriate design protocol that can be used.