Authors:
Eric Klinginsmith
;
Richard Barella
;
Xinghui Zhao
and
Scott Wallace
Affiliation:
Washington State University Vancouver, United States
Keyword(s):
Smart Grid, Phasor Measurement Unit (PMU), Synchrophasors, Machine Learning, Clustering, Event Characterization.
Related
Ontology
Subjects/Areas/Topics:
Architectures for Smart Grids
;
Energy and Economy
;
Energy Monitoring
;
Energy-Aware Systems and Technologies
;
Evolutionary Algorithms in Energy Applications
;
Smart Grids
Abstract:
In the past decade, with the world-wide initiative of upgrading the electrical grid to smart grid, a significant amount of data have been generated by the grid on a daily basis. Therefore, there has been an increasing need in handling and processing these data efficiently. In this paper, we present our experience in applying unsupervised clustering on PMU data for event characterization on the smart grid. We show that although the PMU data are time series in nature, it is more efficient and robust to apply clustering methods on carefully selected features from the data collected at certain instantaneous moments in time. These features are more representative at the moments when the events have the most impact on the grid. Experiments have been carried out on real PMU data collected by Bonneville Power Administration in their wide-area monitoring system in the pacific northwest, and the results show that our instantaneous clustering method achieves high homogeneity, which provides gre
at potentials for identifying unknown events in the grid without substantial training data.
(More)