2009 · 11 citations · 5 references
ComputingComputer ScienceElectrical EngineeringMachine LearningPower ElectronicsTime Series AnalysisDiagnosisDiagnosticsElectrical System DiagnosticsFault DetectionFault DiagnosisSupport Vector MachinesElectrical Fault DiagnosisFault ClassificationFault Diagnosis AlgorithmsFault IdentificationMachine Learning For Fault DetectionMachine Learning-based Fault DiagnosisSingular Value DecompositionWavelet Transform
The paper based on the singular value decomposition and support vector machines, a new fault diagnosis of commutation failures method in HVDC system was proposed. The coefficient matrix acquired from wavelet package transform is first decomposed on singular value, by which fault current are mapped to different time-frequency sub-space. Then the singular value is put into support vector machines to carry out the SVM training and fault type identification. The new method in this paper has high recognition rate, identification speed and stability. It can solve the fault type classification well.
5