Fault Diagnosis of Distribution Systems Based on Least Squares Support Vector Clustering Algorithm

In most of the fault diagnosis systems, when the diagnosis based on real-time information which in the process of its formation are likely to produce the distortion of information, thus lead to the error results of the fault diagnosis. Support vector clustering (SVC) originated from the concept of s...

Full description

Saved in:
Bibliographic Details
Published inElectrotehnica, Electronica, Automatica Vol. 65; no. 2; p. 133
Main Authors Sun, Sheng, Lu, Chuiwei, Zhang, Guojun
Format Journal Article
LanguageEnglish
Published Bucharest ICPE SA - Electra House of Publishing 01.04.2017
Subjects
Accuracy
Algorithms
Artificial intelligence
Automation
Clustering
Computer engineering
Computer science
Computer simulation
Cybersecurity
Data mining
Distortion
Electrical engineering
Fault diagnosis
Fault tolerance
Learning
Least squares method
Methods
Neural networks
Pruning
Real time
Research methodology
Support vector machines
Training
China
Online AccessGet full text

Cover

More Information
Summary:In most of the fault diagnosis systems, when the diagnosis based on real-time information which in the process of its formation are likely to produce the distortion of information, thus lead to the error results of the fault diagnosis. Support vector clustering (SVC) originated from the concept of support vector machines. SVC has the advantage that the training and test speed, but also a learning accuracy is slightly lower. In order to enhance learning precision of the SVC, the least square method is introduced into the SVC. LS-SVC has the imitation that the important degrees of all the training points are without distinction. In this paper, we propose a novel Weighted Least Squares Support Vector Clustering (WLS-SVC) algorithm which provides a support degree for each training sample. The results show that our WLS-SVC algorithm can improve modelling accuracy. A pruning method is introduced into WLS-SVC to acquire sparseness. This paper, by using our approach for processing the real-time input information distortion, realized the fault diagnosis in distribution systems. Through the simulation test, we proved that our model have high fault tolerance.
ISSN:1582-5175
2392-828X