| Summary: | Carbon capture and storage (CCS) is a technology where CO₂ captured from point sources or the atmosphere is injected underground for permanent storage. CCS is part of a portfolio of technologies aimed at enabling the transition to a sustainable energy system with net-zero CO₂ emissions. This Thesis focuses on geologic CO₂ storage (GCS) in faulted siliciclastic sedimentary basins, and addresses the impact of uncertainty sources on forecasts of CO₂ migration made with physics-based numerical models. A key contribution of this work is the quantification of uncertainty on fault petrophysical properties and modeling of flow within clay-smeared faults, which play a central role on CO₂ storage effectiveness and safety.
We start by extending a reservoir simulator to increase fidelity in 3D numerical models of GCS. Extensions include a thermodynamic model to calculate PVT properties of CO₂-brine mixtures and relative permeability hysteresis. We then conduct numerical simulations and experiments of CO₂ injection and migration at the meter scale. Our experiments use tanks with transparent panels that allow recreating realistic basin geometries and CO₂ injection protocols. Using direct observations, we find that local measurements reduce model calibration time and that accurate deterministic forecasts are challenging. Next, recognizing the impact of faults on subsurface flow and shortcomings of previous models of fault architecture and hydraulic properties, we propose a probabilistic method to estimate the directional components of the fault permeability tensor. We extend previous efforts by modeling the fault core in 3D, using flow-based upscaling, and quantifying uncertainty. We then move to the field scale and apply this method to forecast fault CO₂ migration in the Miocene section offshore Texas. Faults are common in this area, and it is crucial to understand how they may limit storage capacity. Our modeling results indicate that, due to subsurface structure, stratigraphy and fault hydrogeology, updip CO₂ migration in listric growth faults is unlikely.
Our findings show that quantitative forecasts are uncertain due to limited subsurface knowledge and modeling choices. Efforts to quantify parameter uncertainty and its impact on modeling forecasts appear necessary, a task that requires development of reduced order models accounting for the limitations of simulation data. Forecasts based on numerical models will benefit from history-matching and updating as field data becomes available.
|
|---|