Multilevel pressure measurements for monitoring and prediction of CO2 and displaced brine migration

Advisors

Sally M. Benson, primary advisor
Jef Caers, advisor
Roland N. Horne, advisor

Abstract

The motivation for the current work stems from the recent and unparalleled implementation of multilevel pressure monitoring at the Illinois Basin - Decatur Project (IBDP). The IBDP is a Carbon Capture and Sequestration (CCS) pilot project in Decatur, Illinois, USA, aimed to demonstrate the ability of the Cambrian-age Mt. Simon Sandstone to accept and store one million metric tons of CO₂ over three years. The CO₂ is captured from an ethanol plant owned by the Archer Daniels Midland Company (ADM), and injection into the lower portion of the Mt. Simon Sandstone started in November 2011. As part of an extensive Monitoring, Verification, and Accounting program, the Westbay multilevel groundwater characterization and monitoring system was installed in a deep in-zone verification (monitoring) well (2,000 m) to record the pressure at multiple depths before, during, and after CO₂ injection. With two years of hourly pressure transient data available for analysis, the goal of this work was to establish whether (and to what extent) multilevel pressure transient data could provide valuable information on CO₂ and displaced brine migration, both real-time and for forecasting. Based on a synthetic study and analyses of simulated pressure data, we show that pressure buildups normalized to the pressure buildup at the depth of injection and vertical pressure gradients normalized to the initial hydrostatic pressure gradient are diagnostic of reservoir structure (layering and anisotropy) soon after the start of injection and over time provide information on the height of the CO₂ plume in the reservoir. The diagnostic features in the pressure response pertaining to the height of the CO₂ plume are evident long before the CO₂ arrives at the monitoring well and can be attributed to buoyancy induced and gravity segregated aqueous flows caused by the advancing CO₂ plume. The multilevel pressure transient data acquired at the IBDP have provided a unique opportunity to validate the identified diagnostics for tracking buoyant migration of CO₂ using multilevel pressure transient data. Based on diagnostics alone, the multilevel pressure transient data show that CO₂ plume is confined largely to the injection interval, which is consistent with data from Reservoir Saturation Tool logs and sampling data. Hence, we successfully show that multilevel pressure transient data can be used to determine CO₂ plume migration real-time. A thorough review of local and regional geology at the IBDP site points to a braided river system being the primary depositional environment in the lower portion of the Mt. Simon Sandstone where the CO₂ is injected. Of particular interest is the presence and lateral extents of low-permeability layers that act as baffles and impede upward flow of CO₂ and displaced brine. First, a layer-cake model (effectively 2D with laterally extensive layers and suitable for radial flow) is considered with focus on history matching. By history matching the multilevel pressure transient data at the IBDP at four different locations (injection well and three monitoring zones at the verification well), we show that it is possible to develop a highly resolved hydrogeologic model that in turn can be used to forecast future CO₂ plume migration. Second, 3D models with focus on the uncertainty associated with non-extensive low-permeability layers are considered. Training images are generated to represent a simplified braided river system with sand interbedded with laterally non-extensive low-permeability layers. Conditioned to training images, well log data, and probability maps that capture various plausible configurations (model scenarios) of one specific layer of very low permeability (believed to stem from the deposition of a playa lake), multiple geologic model realizations are generated for each model scenario, and multiple permeability combinations are considered for each realization. At early time (as demonstrated in the synthetic diagnostics study), the multilevel pressure transients at the monitoring well are diagnostic of reservoir structure and insensitive to the type of fluid injected. Hence, early-time water injection (single-phase) simulations are used as proxy simulations in the place of full multiphase flow simulations, and pressure transient responses from the proxy simulations are compared to each other and to the IBDP "truth" using distance-based modeling. The calculated dissimilarities (distances between pressure transient responses) are clustered into groups that are diagnostic of average permeability properties and also provide information on which Playa Lake configurations to disregard based on early-time pressure transient data. Longer-time multiphase flow simulations are run on a few representative models to further constrain the uncertainty associated with low-permeability layers that may or may not restrict upward CO₂ migration depending on their lateral extents. This work has shown that continuous multilevel pressure measurements at a monitoring well within the storage reservoir are useful for monitoring and predicting vertical CO₂ plume migration. At early time, multilevel pressure transients are diagnostic of reservoir structure, which can aid in the prediction of future CO₂ migration. At later times, information on the height of the buoyant CO₂ plume within the storage reservoir is available.