Multiple partial discharge source discrimination with multiclass support vector machines Articles uri icon

publication date

  • August 2016

start page

  • 417

end page

  • 428

volume

  • 55

international standard serial number (ISSN)

  • 0957-4174

electronic international standard serial number (EISSN)

  • 1873-6793

abstract

  • The costs of decommissioning high-voltage equipment due to insulation breakdown are associated to the substitution of the asset and to the interruption of service. They can reach millions of dollars in new equipment purchases, fines and civil lawsuits, aggravated by the negative perception of the grid utility. Thus, condition based maintenance techniques are widely applied to have information about the status of the machine or power cable readily available. Partial discharge (PD) measurements are an important tool in the diagnosis of power systems equipment. The presence of PD can accelerate the local degradation of insulation systems and generate premature failures. Conventionally, PD classification is carried out using the phase resolved partial discharge (PRPD) pattern of pulses.

keywords

  • support vector machine; partial discharges; electric maintenance; machine learning; condition monitoring; risk assessment; pattern-recognition; neural-network; pd sources; classification; separation; apparatus; voltage