Detectilfirt-ircuitfault Detection and classification of short-circuit faults onaransmissonnusncurrentsigna on a transmission line using current signal


Coban M., TEZCAN S. S.

BULLETIN OF THE POLISH ACADEMY OF SCIENCES-TECHNICAL SCIENCES, vol.69, no.4, 2021 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 69 Issue: 4
  • Publication Date: 2021
  • Doi Number: 10.24425/bpasts.2021.137630
  • Journal Name: BULLETIN OF THE POLISH ACADEMY OF SCIENCES-TECHNICAL SCIENCES
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aerospace Database, Communication Abstracts, Compendex, Metadex, zbMATH, Directory of Open Access Journals, Civil Engineering Abstracts
  • Keywords: transmission line, fault detection, fault classification, support vector machine, transmissionine, faultdetection, faultclassification, supportvectormachine, FEATURE-EXTRACTION TECHNIQUE, DISTANCE PROTECTION, LOCATION, MORPHOLOGY, ALGORITHM
  • Gazi University Affiliated: Yes

Abstract

Abst.ract. hisThis studystudy ooffersfers twotwo suportsupport vectorvector machinemachine (SVM)(SVM) modelsmodelsfor orfault faul detectiondetectionand andfaul fault classificaton,classification, respectively.respectively.Different Diffshort-erent circuitshort-circuitevents eventswere wr gene ratedgeneratedusing ausing 154a 154kV kV transmisiontransmissioline line modeledmodeledin in MATLAB/SimulinkMATLAB/Simulink sofsoftware.tware. iscreteDiscrete ltwavelet transform (DWT) is performed on the measured single terminal current signals before fault detection stage. Three level wavelet energy values obtained for each of three-phase currents were used as input features for the detector. After fault detection, half cycle (10 ms) of three-phase current signals was recorded by a 20 kHz sampling rate. The recorded currents signals were used as input parameters for the multi-class SVM fclassifiier. The results of the validation tests have demonstrated that quite a reliable fault detection and fclassifiication system can be developed using SVM. The faults fi. - fi generatedwereusedfortrainingandtestingofSVMclassifiers.AnSVM-basedclassificationanddetectionmodelwasfullyimplementedin MATLABsoftware.Thesemodelswerecomprehensivelytestedunderdifferentconditions.Theeffectsofthefaultimpedance,faultinception angle, mother wavelet and fault location were all investigated. Finally, simulation results fverifiied that the study proposed can be used for fault fiti te transision line. detectionandclassificationonheransmissionline.