Reducing Uncertainty in Reservoir Model Predictions: from Plume Evolution to Tool Responses

Abstract

With the transition of geological carbon dioxide (CO2) storage from pilot to industrial scale, risk analysis has become pivotal for assessing the suitability of a storage site in satisfying regulatory requirements during and after injection. Risk analysis is inherently related to uncertainty quantification. This paper extends our previously developed workflows to quantify and propagate uncertainty from geological, geophysical and petrophysical data to project-level performance metrics including capacity, injectivity, and containment characteristics. We perform global sensitivity analysis to quantitatively link uncertainty in the computed performance metrics to the uncertainties in the underlying reservoir parameters. This enables one to identify measurements to reduce uncertainty estimates of reservoir performance metrics and predicted monitoring tool responses. We illustrate this workflow in a pre-injection uncertainty study for the Illinois Basin Decatur Project (IBDP). The reservoir model is built using geostatistical approach based on all available geophysical and petrophysical data including 3D seismic interpretation, results of special core analysis and injectivity tests. We focus on three groups of performance predictions: spatial extent of the CO2 saturation profile, partitioning of CO2, and predicted responses seen with the Westbay* multilevel groundwater characterization and monitoring system installed in the monitoring well. Based on multiple realizations of the reservoir model we evaluate the uncertainty range in pressures and CO2 saturations at the Westbay monitoring zones during the life of the project. Predicted measurements at the monitoring zones covering a depth interval between 1950 m and 2120 m are analyzed. Global sensitivity analysis is used to identify the key petrophysical properties of the reservoir whose uncertainties have the most effect on the uncertainty of the reservoir performance and measurement predictions. Results of this analysis provide valuable insight for targeted site characterization and design of the monitoring program.