Reduced Dimensional Chebyshev-Polynomial Chaos Approach for Fast Mixed Epistemic-Aleatory Uncertainty Quantification of Transmission Line Networks

Abstract

This paper presents a hybrid Chebyshev-polynomial chaos (CPC) metamodel for the uncertainty quantification of multiconductor transmission line networks. The key feature of this metamodel is its capacity to account for two different types of uncertainty simultaneously present in a network-uncertainty arising from imprecise knowledge regarding the network parameters (epistemic uncertainty) and uncertainty arising from the random variability in the network parameters (aleatory uncertainty). Unfortunately, when investigating such mixed epistemic-aleatory problems, the CPU costs to construct the hybrid CPC metamodel becomes exorbitantly large. To address this issue, a new sensitivity sweeping-based dimension reduction algorithm especially tailored for mixed epistemic-aleatory problems has been developed in this paper. Numerical techniques based on iterative model enrichment and priority-based sampling have also been developed to further enhance the efficiency of the algorithm.