Comparison between the performance of ANN-based loss of mains relays using multi-layer and non-layered perceptron approaches

S. K. Salman, David J. King

Research output: Contribution to journalArticle

Abstract

Loss of mains is a condition whereby an embedded/distributed generator becomes islanded with part of a public load. This is usually caused by disturbances on the associated utility network which results in having an EG (embedded generation) to become “islanded” with a public load. This condition is highly undesirable and should be correctly detected in order to ensure the tripping of the generator affected by such conditions. Currently the two methods most commonly in use to detect LOM (loss of mains) utilise ROCOF (rate of change of frequency) and PDM (vector surge) techniques, either on their own, or in tandem. However, these methods have not proved to be totally reliable. This paper reports the development of a new relay for detecting LOM. It is based on the application of artificial neural network (ANN). Two ANN architectures have been considered; one based on multi-layer perceptron approach while the other based on a non-layered approach using NeuralWorks Predict features.
Original languageEnglish
Pages (from-to)24-28
Number of pages5
JournalThe International Journal of Innovations in Energy Systems and Power
Volume1
Issue number1
StatePublished - Nov 2006

Fingerprint

Neural networks
Multilayer neural networks
Network architecture
Pulse width modulation

Cite this

@article{a67cfd68a9d141e6bce466c69b064c72,
title = "Comparison between the performance of ANN-based loss of mains relays using multi-layer and non-layered perceptron approaches",
abstract = "Loss of mains is a condition whereby an embedded/distributed generator becomes islanded with part of a public load. This is usually caused by disturbances on the associated utility network which results in having an EG (embedded generation) to become “islanded” with a public load. This condition is highly undesirable and should be correctly detected in order to ensure the tripping of the generator affected by such conditions. Currently the two methods most commonly in use to detect LOM (loss of mains) utilise ROCOF (rate of change of frequency) and PDM (vector surge) techniques, either on their own, or in tandem. However, these methods have not proved to be totally reliable. This paper reports the development of a new relay for detecting LOM. It is based on the application of artificial neural network (ANN). Two ANN architectures have been considered; one based on multi-layer perceptron approach while the other based on a non-layered approach using NeuralWorks Predict features.",
author = "Salman, {S. K.} and King, {David J.}",
year = "2006",
month = "11",
volume = "1",
pages = "24--28",
journal = "The International Journal of Innovations in Energy Systems and Power",
issn = "1913-133X",
number = "1",

}

TY - JOUR

T1 - Comparison between the performance of ANN-based loss of mains relays using multi-layer and non-layered perceptron approaches

AU - Salman,S. K.

AU - King,David J.

PY - 2006/11

Y1 - 2006/11

N2 - Loss of mains is a condition whereby an embedded/distributed generator becomes islanded with part of a public load. This is usually caused by disturbances on the associated utility network which results in having an EG (embedded generation) to become “islanded” with a public load. This condition is highly undesirable and should be correctly detected in order to ensure the tripping of the generator affected by such conditions. Currently the two methods most commonly in use to detect LOM (loss of mains) utilise ROCOF (rate of change of frequency) and PDM (vector surge) techniques, either on their own, or in tandem. However, these methods have not proved to be totally reliable. This paper reports the development of a new relay for detecting LOM. It is based on the application of artificial neural network (ANN). Two ANN architectures have been considered; one based on multi-layer perceptron approach while the other based on a non-layered approach using NeuralWorks Predict features.

AB - Loss of mains is a condition whereby an embedded/distributed generator becomes islanded with part of a public load. This is usually caused by disturbances on the associated utility network which results in having an EG (embedded generation) to become “islanded” with a public load. This condition is highly undesirable and should be correctly detected in order to ensure the tripping of the generator affected by such conditions. Currently the two methods most commonly in use to detect LOM (loss of mains) utilise ROCOF (rate of change of frequency) and PDM (vector surge) techniques, either on their own, or in tandem. However, these methods have not proved to be totally reliable. This paper reports the development of a new relay for detecting LOM. It is based on the application of artificial neural network (ANN). Two ANN architectures have been considered; one based on multi-layer perceptron approach while the other based on a non-layered approach using NeuralWorks Predict features.

M3 - Article

VL - 1

SP - 24

EP - 28

JO - The International Journal of Innovations in Energy Systems and Power

T2 - The International Journal of Innovations in Energy Systems and Power

JF - The International Journal of Innovations in Energy Systems and Power

SN - 1913-133X

IS - 1

ER -