ANN-based automatic contingency selection for electric power system

K.L. Lo, W.P. Luan, M.J. Given, M. Bradley, H. Wan

Research output: Contribution to journalArticle

Abstract

Automatic contingency selection aims to quickly predict the impact of a set of next contingencies on an electric power system without actually performing a full ac load flow. Artificial neural network methods have been employed to overcome the masking effects or slow execution associated with existing methods. However, the large number of input features for the ANN limits its applications to large power systems. In this paper, a novel feature selection method, named the Weak Nodes method, based on a heuristic approach is proposed for an ANN-based automatic contingency selection for electric power system, especially for the voltage ranking problem. Pre-contingency state variables of weak nodes in the power system are adopted as input features for the ANN. The method is tested on the 77 busbar NGC derived network by Counter-propagation Method and it is proved that it reduces the input features for ANN dramatically without losing ranking accuracy.
LanguageEnglish
Pages193-207
Number of pages15
JournalCOMPEL: The International Journal for Computation and Mathematics in Electrical and Electronic Engineering
Volume21
Issue number2
DOIs
Publication statusPublished - 2002

Fingerprint

Electric Power System
Electric power systems
Busbars
Feature extraction
Neural networks
Power System
Ranking
Electric potential
Masking
Vertex of a graph
Feature Selection
Artificial Neural Network
Voltage
Heuristics
Propagation
Predict

Keywords

  • artificial neural networks
  • power devices
  • voltage
  • electrical engineering
  • computer science

Cite this

@article{019f6a01ff514965a7fa23a8ed33154c,
title = "ANN-based automatic contingency selection for electric power system",
abstract = "Automatic contingency selection aims to quickly predict the impact of a set of next contingencies on an electric power system without actually performing a full ac load flow. Artificial neural network methods have been employed to overcome the masking effects or slow execution associated with existing methods. However, the large number of input features for the ANN limits its applications to large power systems. In this paper, a novel feature selection method, named the Weak Nodes method, based on a heuristic approach is proposed for an ANN-based automatic contingency selection for electric power system, especially for the voltage ranking problem. Pre-contingency state variables of weak nodes in the power system are adopted as input features for the ANN. The method is tested on the 77 busbar NGC derived network by Counter-propagation Method and it is proved that it reduces the input features for ANN dramatically without losing ranking accuracy.",
keywords = "artificial neural networks, power devices, voltage, electrical engineering, computer science",
author = "K.L. Lo and W.P. Luan and M.J. Given and M. Bradley and H. Wan",
year = "2002",
doi = "10.1108/03321640210416304",
language = "English",
volume = "21",
pages = "193--207",
journal = "COMPEL: The International Journal for Computation and Mathematics in Electrical and Electronic Engineering",
issn = "0332-1649",
publisher = "Emerald Publishing Limited",
number = "2",

}

TY - JOUR

T1 - ANN-based automatic contingency selection for electric power system

AU - Lo, K.L.

AU - Luan, W.P.

AU - Given, M.J.

AU - Bradley, M.

AU - Wan, H.

PY - 2002

Y1 - 2002

N2 - Automatic contingency selection aims to quickly predict the impact of a set of next contingencies on an electric power system without actually performing a full ac load flow. Artificial neural network methods have been employed to overcome the masking effects or slow execution associated with existing methods. However, the large number of input features for the ANN limits its applications to large power systems. In this paper, a novel feature selection method, named the Weak Nodes method, based on a heuristic approach is proposed for an ANN-based automatic contingency selection for electric power system, especially for the voltage ranking problem. Pre-contingency state variables of weak nodes in the power system are adopted as input features for the ANN. The method is tested on the 77 busbar NGC derived network by Counter-propagation Method and it is proved that it reduces the input features for ANN dramatically without losing ranking accuracy.

AB - Automatic contingency selection aims to quickly predict the impact of a set of next contingencies on an electric power system without actually performing a full ac load flow. Artificial neural network methods have been employed to overcome the masking effects or slow execution associated with existing methods. However, the large number of input features for the ANN limits its applications to large power systems. In this paper, a novel feature selection method, named the Weak Nodes method, based on a heuristic approach is proposed for an ANN-based automatic contingency selection for electric power system, especially for the voltage ranking problem. Pre-contingency state variables of weak nodes in the power system are adopted as input features for the ANN. The method is tested on the 77 busbar NGC derived network by Counter-propagation Method and it is proved that it reduces the input features for ANN dramatically without losing ranking accuracy.

KW - artificial neural networks

KW - power devices

KW - voltage

KW - electrical engineering

KW - computer science

U2 - 10.1108/03321640210416304

DO - 10.1108/03321640210416304

M3 - Article

VL - 21

SP - 193

EP - 207

JO - COMPEL: The International Journal for Computation and Mathematics in Electrical and Electronic Engineering

T2 - COMPEL: The International Journal for Computation and Mathematics in Electrical and Electronic Engineering

JF - COMPEL: The International Journal for Computation and Mathematics in Electrical and Electronic Engineering

SN - 0332-1649

IS - 2

ER -