This work considers methods for uncertainty quantification and data-driven modeling using low-dimensional representations. Specifically, sparse polynomial approximations (or expansions) and active subspaces. Comprised of three main sections, we investigate sparse polynomial chaos expansions via compressed sensing and D-optimal designs, data-driven discovery of dynamical systems via robust project matrix optimization and D-optimal designs, and uncertainty quantification for electrical capacity expansion planning. D-optimal designs, because of their coherence properties, can be used to enhance estimation accuracy with compressed sensing algorithms. We prove that the quality of a D-optimal design can be improved if samples are drawn according to a coherence-optimal distribution. The greatest benefit from this approach is for high-order, low-dimensional models, particularly if the uncertain variables follow normal distributions. We show that accuracy can be improved by precoherencing with robust projection matrix optimization for under-determined linear systems with noise and L-curve optimization. We also demonstrated that D-optimal designs can be used to improve accuracy when dealing with an abundance of data, i.e., over-determined linear systems with noise, and that both robust projection matrix optimization and D-optimal designs can be used in conjunction for further benefit. We apply some of our methods for uncertainty quantification and data-driven modeling, by quantifying the uncertainty in output decisions of a large-scale electrical capacity expansion planning model in collaboration with the National Renewable Energy Laboratory.
|Commitee:||Morrison, Rebecca E, King, Ryan N, Evans, John A, Matsuo, Tomoko|
|School:||University of Colorado at Boulder|
|School Location:||United States -- Colorado|
|Source:||DAI 81/11(E), Dissertation Abstracts International|
|Subjects:||Applied Mathematics, Statistics, Aerospace engineering|
|Keywords:||Active subspaces, Data driven modeling, Machine learning, Polynomial chaos, Sensitivity analysis, Uncertainty cuantification|
Copyright in each Dissertation and Thesis is retained by the author. All Rights Reserved
The supplemental file or files you are about to download were provided to ProQuest by the author as part of a
dissertation or thesis. The supplemental files are provided "AS IS" without warranty. ProQuest is not responsible for the
content, format or impact on the supplemental file(s) on our system. in some cases, the file type may be unknown or
may be a .exe file. We recommend caution as you open such files.
Copyright of the original materials contained in the supplemental file is retained by the author and your access to the
supplemental files is subject to the ProQuest Terms and Conditions of use.
Depending on the size of the file(s) you are downloading, the system may take some time to download them. Please be