Carbon Sequestration: Uncertainty and Parameter Estimation

Abstract

Carbon Capture and Sequestration (CCS) appears to be a practical technology for large-scale storage of CO2 to reduce anthropogenic CO2 emissions. Risk is an inevitable component of any geological project with the aim of storing CO2, and thus, is a concern to the public, policy makers, and scientists. Uncertainty that arises in the application of mathematical Carbon Sequestration (CS) models has a negative impact on the quality of risk assessment. Parameter uncertainty is believed to play a dominant role in the uncertainty of the outputs of the CS system models. However, reducing parameter uncertainty in CS models involves a trade-off between accuracy and computational efficiency of the model calibration methodology. The goal of this thesis is to reduce the trade-off between accuracy and computational efficiency when calibrating CS models. This is accomplished by, one, reducing the dimensionality of the parameter space; two, developing efficient calibration algorithms; and, three, reducing the computational cost of model simulation during calibration. The primary contributions of this thesis are: 1. The development of a sensitivity analysis to identify which parameters contribute the most to the uncertainty of the CS system model output, accounting for both parameter uncertainty and model structure. 2. The development of a computationally efficient and flexible Bayesian Importance Sampling (IS) method for continuous calibration of CS models using noisy monitoring data collected during the injection phase. 3. The development of the Response Surface Methodology (RSM) in a novel adaptive way to mitigate the computational demand of CS model calibration with negligible effect on the accuracy of the results. The methodologies and results presented in this thesis contribute to efficient calibration of CS models by identifying the most influential parameters in uncertainty of CS model outputs and calibrating those models accurately and efficiently.