A combined noise-rejection method for UHF PD detection on-site

An important condition monitoring technology of power apparatus is ultra-high-frequency (UHF) partial discharge (PD) detection. This has been widely used in gas insulated substations, power transformers, etc. It benefits from its high sensitivity and low overlap with noise frequency. However, lots of field practices indicate that some interference, such as the mobile phone communication signals, the PD signals propagated from adjacent equipments, or some pulse-type interferences, do mix with the to-be-detected PD signals and generate great difficulties in UHF diagnosis. Based on various field experiences, this paper presents a combined UHF noise rejection method, which integrates the waveform correlation analysis, the time domain windowing method based on external noise signals, and the clustering method based on the magnitude ratio from two consecutive sensors. Experimental results show that this method can distinguish real PD signals from environmental noises from various sources effectively, and can separate different PD signals within the same apparatus in test.