An Online Deep Learning Approach Toward the Prediction of Power System Stresses Using Voltage Phasors

The outage of a transmission line may change the system phase angle differences to the point that the system experience stress conditions. Hence, the angle differences for post-contingency condition of a transmission lines should be predicted in real time operation. However, online line-based phase...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2019-10
Main Authors: uzan, Elham, Abedi, Sajjad, Lin, Jeremy, Asgarpoor, Sohrab, Bernabeu, Emanuel
Format: Article
Language:English
Subjects:
Adaptive structures
Adaptive systems
Artificial neural networks
Contingency
Decision trees
Deep learning
Learning theory
Machine learning
Measuring instruments
Modules
Neural networks
On-line systems
Phase shift
Phase transitions
Phasors
Predictions
Real time operation
Transmission lines
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The outage of a transmission line may change the system phase angle differences to the point that the system experience stress conditions. Hence, the angle differences for post-contingency condition of a transmission lines should be predicted in real time operation. However, online line-based phase angle difference monitoring and prediction for power system stress assessment is not a universal operating practice yet. Thus, in this paper, an online power system stress assessment framework is proposed by developing a convolutional neural network (CNN) module trained through Deep Learning approach. In the proposed framework, the continuously streaming system phase angle data, driven from phasor measurement units (PMUs) or a state estimator (SE), is used to construct power system stress indices adaptive to the structure parameters of the CNN module. Using this approach, any hidden patterns between phase angles of buses and system stress conditions are revealed at low computation cost while yielding accurate stress status and the severity of the stress. The effectiveness and scalability of the proposed method has been verified on the IEEE 118-bus and more importantly, on the PJM Interconnection system. Moreover, outperformance of the proposed method is verified by comparing the results with artificial neural network (ANN) and decision tree (DT).
ISSN:2331-8422