Event Analysis in Transmission Systems Using Spatial Temporal Graph Encoder Decoder (STGED)

Phasor Measurement Units (PMUs) are located at different geographic positions in the transmission system, generating measurement data that can be analyzed to monitor and control the power grid. One utilization for such measurement data is monitoring based on machine learning based event analysis in...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on power systems Jg. 38; H. 6; S. 1 - 12
Hauptverfasser: Ahmed, Arman, Sadanandan, Sajan K., Pandey, Shikhar, Basumallik, Sagnik, Srivastava, Anurag K., Wu, Yinghui
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York IEEE 01.11.2023
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Schlagworte:
Algorithms
Classification
Coders
Data analysis
Deep learning
Event Classification
Event detection
Event localization
Feature extraction
Localization
Location awareness
Machine learning
Measuring instruments
Phasor measurement units
Phasors
Power grids
Power measurement
Power system dynamics
Transmission System and Machine Learning
ISSN:0885-8950, 1558-0679
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Phasor Measurement Units (PMUs) are located at different geographic positions in the transmission system, generating measurement data that can be analyzed to monitor and control the power grid. One utilization for such measurement data is monitoring based on machine learning based event analysis in the transmission system, which includes event detection, localization, and classification . The approach developed in this work exploits the latent spatial and temporal features of PMU measurement data for event analysis. (1) For event detection, this research proposes a novel unsupervised "Spatial Temporal Graph Encoder Decoder" (STGED) deep learning model that concurrently/jointly exploits the spatial and temporal features of PMU measurements. STGED further supports downstream unsupervised event localization and classification. (2) For event localization, events are localized by estimating Turbulence and Proximity statistical scores over predicted measurements/output from STGED. (3) For event classification, an unsupervised algorithm is developed with a classification scoring metric that leverages physics informed rules based on the fundamentals of power system. The proposed approach is evaluated on the IEEE test systems and other benchmark systems for diversified event scenarios. Additionally, performance of the developed approach has been compared with other techniques, and validated using real-world industry data. Experimental results show that the proposed approach outperforms other existing techniques for event analysis.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0885-8950
1558-0679
DOI:10.1109/TPWRS.2022.3226209