Measuring partial discharges in electrical insulation systems is becoming a standard procedure for quality control, qualification and commissioning of electrical apparatus. Even more important, partial discharges are the main property to be monitored to assess the health conditions of any organic insulation system. However, large part of the potentiality offered by partial discharge measurements is hindered by the need of experts to record, process and interpret data, which delays and even prevent from the diffusion of this diagnostic property. This paper presents a new algorithm which seems to be very effective in providing automatic separation of partial discharges from noise, which is the first step to develop a fully automatic and unsupervised diagnostic approach. Multi-dimensional signal decomposition is achieved resorting to various transformation applied to recorded pulses, including time and frequency domains, and entropy. Clusters thus obtained are separated by statistical and artificial intelligence algorithms. Applications to AC sinusoidal voltage supply are presented, highlighting that the proposed approach is valid also under DC and power electronics supply.

A Track towards Unsupervised Partial Discharge Inference in Electrical Insulation Systems

Ghosh R.;Seri P.;Montanari G. C.
2020

Abstract

Measuring partial discharges in electrical insulation systems is becoming a standard procedure for quality control, qualification and commissioning of electrical apparatus. Even more important, partial discharges are the main property to be monitored to assess the health conditions of any organic insulation system. However, large part of the potentiality offered by partial discharge measurements is hindered by the need of experts to record, process and interpret data, which delays and even prevent from the diffusion of this diagnostic property. This paper presents a new algorithm which seems to be very effective in providing automatic separation of partial discharges from noise, which is the first step to develop a fully automatic and unsupervised diagnostic approach. Multi-dimensional signal decomposition is achieved resorting to various transformation applied to recorded pulses, including time and frequency domains, and entropy. Clusters thus obtained are separated by statistical and artificial intelligence algorithms. Applications to AC sinusoidal voltage supply are presented, highlighting that the proposed approach is valid also under DC and power electronics supply.
2020
2020 IEEE Electrical Insulation Conference, EIC 2020
190
193
Ghosh R.; Seri P.; Montanari G.C.
File in questo prodotto:
Eventuali allegati, non sono esposti

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11585/774728
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 8
  • ???jsp.display-item.citation.isi??? 41
social impact