Search for: All records

Abstract Localization of faults in a large power system is one of the most important and difficult tasks of power systems monitoring. A fault, typically a shorted line, can be seen almost instantaneously by all measurement devices throughout the system, but determining its location in a geographically vast and topologically complex system is difficult. The task becomes even more difficult if measurements devices are placed only at some network nodes. We show that regression graph neural networks we construct, combined with a suitable statistical methodology, can solve this task very well. A chief advance of our methods is that we construct networks that produce localization without having being trained on data that contain fault localization information. We show that a synergy of statistics and deep learning can produce results that none of these approaches applied separately can achieve. 

The concepts of physical dependence and approximability have been extensively used over the past two decades to quantify nonlinear dependence in time series. We show that most stochastic volatility models satisfy both dependence conditions, even if their realizations take values in abstract Hilbert spaces, thus covering univariate, multi‐variate and functional models. Our results can be used to apply to general stochastic volatility models a multitude of inferential procedures established for Bernoulli shifts. 

We develop statistical methodology for the quantification of risk of source-destination pairs in an internet network. The methodology is developed within the framework of functional data analysis and copula modeling. It is summarized in the form of computational algorithms that use bidirectional source-destination packet counts as input. The usefulness of our approach is evaluated by an application to real internet traffic flows and via a simulation study. 

Abstract. Comparisons of observed and modeled climate behavior often focus on central tendencies, which overlook other important distributional characteristics related to quantiles and variability. We propose two permutation procedures, standard and stratified, for assessing the accuracy of climate models. Both procedures eliminate the need to model cross-correlations in the data, encouraging their application in a variety of contexts. By making only slightly stronger assumptions, the stratified procedure dramatically strengthens the ability to detect a difference in the distribution of observed and climate model data. The proposed procedures allow researchers to identify potential model deficiencies over space and time for a variety of distributional characteristics, providing a more comprehensive assessment of climate model accuracy, which will hopefully lead to further model refinements. The proposed statistical methodology is applied to temperature data generated by the state-of-the-art North American Coordinated Regional Climate Downscaling Experiment (NA-CORDEX).