The goal of CO2 storage is to efficiently and safely store large amounts of CO2 in subsurface formations, without risk for leakage to groundwater resources, and with minimum leakage to the atmosphere. Decisions must be made on what formations to use for CO2 storage and how to monitor the process for control and safety. Geological measurements are important but not sufficient and should be complemented with realistic computer models based on geological data. When simulations are used to inform the decisions on CO2 storage, limited understanding of the underlying physics and lack of data will be a big problem. A few spread-out measurements over several square kilometers do not contain all relevant information, and the result is significant model uncertainty.
The uncertainty in the geological properties will lead to uncertainty in the transport of CO2 over time. The CO2 plume moves only a few meters per year, but over the centuries the uncertainty in the extent of the CO2 plume implies risk for leakage. There may also be risk for leakage through fractures caused by high pressure as a result of excessive injection of CO2.
Storage facilities are very expensive, and uncertainty quantification is a key technology for realizing large-scale CO2 storage and minimizing unnecessary costs due to uninformed decisions. The bottleneck is the lack of reliable tools for quantifying uncertainty due to limited data.
In this project some of these knowledge gaps have been bridged by the development of simulation tools for uncertainty quantification. We focus on uncertainty in three important processes that may lead to practical constraints on storage: increasing pressure during injection of CO2 into a subsurface formation, the movement of CO2 over time, and the risk for leakage through the caprock covering the storage formation.
A mathematical framework for simulating uncertain transport of CO2 after injection into a reservoir has been developed. The effect of uncertain physical parameters on the long-term CO2 plume propagation has been investigated. The model has been extended to handle dissolution of CO2 and methods for more efficient quantification of uncertainty have been developed. Work on efficient representation of highly heterogeneous domains (e.g., channels) has been performed, using a data driven method for optimal fit with empirical data. Identification of the location of interfaces between channel and surrounding matrix is essential for accurate computations.
In order to increase the interest and awareness of uncertainties related to CO2 storage, a popular science article was made accessible via the website of the newspaper Bergens Tidende. The article (only available in Norwegian) has the title 'Hvordan håndtere usikkerhet ved CO2-lagring'.
A reduced-dimensional model for coupled two-phase flow and geomechanical deformation within the context of CO2 storage has been derived. Dimension reduction is essential to make stochastic simulations meaningful, otherwise the task requires too much computing time. The reduced-order model simplifies the complex flow and interaction within thin storage units, while retaining the full-dimensional poroelastic equations for the over- and under-burden that confine the aquifer.
A Matlab simulation tool for poro-mechanics, compatible with the Matlab Reservoir Simulaiton Toolbox developed by Sintef, has been released as open source (see https://github.com/keileg/fvbiot). This methodology forms the basis for stochastic simulations in WP3.
When storing large quantities of CO2 in groundwater aquifers, the capacity will generally be limited by pressure build-up in the aquifer. In certain cases, the pressure can be reduced because of leakage of formation water through the reservoir cap rock. This reduction in pressure build-up is very uncertain, as it is partly determined by the properties of the reservoir where only few measurements exist. We have conducted an analysis of leakage through the cap rock and determined a threshold for when the leak can be neglected. Furthermore, we have derived a method that facilitates calculation of the pressure build-up when leakage must be considered. In this method the leakage is calculated through a convolution instead of an extension of the computational grid of the reservoir.
During density driven dissolution of CO2 in oil, the speed of the process will depend strongly on the density difference between oil with dissolved CO2 and oil without CO2. The density of the mixture is a strongly non-monotone function of the mole fraction of CO2. We have demonstrated a significant uncertainty in the density function and its propagation to dissolved CO2.
The methodology for uncertainty quantification developed in CONQUER will be used in future CCS projects, but also in other applications, e.g., spreading of microplastics in the oceans.
As an efficient means to reduce CO2 in the atmosphere, large-scale storage of CO2 in subsurface formations is today a necessary technology to reduce the effect of global warming. The goal is to efficiently store large amounts of CO2, without significant risk for leakage.
The implementation of large-scale CO2 storage requires decisions on injection sites and strategies. When simulations are used to inform the decision making, lack of geological data such as permeability, caprock topography, location of fractures/faults, and stress in the formation will be sources of significant uncertainty. This uncertainty is transferred throughout the storage operation, via capacity estimates, operational constraints and potential creation of leakage, to legal questions about operational licenses and CO2 crossing international borders. Considering the costs of storage facilities, uncertainty quantification is a key enabling technology for realizing large-scale storage. The bottleneck is the lack of reliable tools for quantifying uncertainty due to limited data.
In this project the knowledge gaps will be bridged by simulation tools for uncertainty quantification, tailored for large-scale storage. We focus on three key processes that may lead to storage constraints: pressure buildup during injection, long-term migration of CO2, and leakage through the caprock overlying the formation. We devise a model based on interconnected modules, each consisting of a tailored numerical method for one of the problems (geomechanics model, pressure problem and transport problem). Data from pilot studies such as Sleipner and Snøhvit will be used.
Due to the prohibitive numerical cost of a direct implementation (complex physics and many sources of uncertainty), we use reduced order models and simplified physics. The expected outcome is a set of open source simulation tools where the errors from lack of data, modeling errors and numerical errors are systematically treated within a unified framework.