Open AccessUHF partial discharge localisation method in substation based on dimension‐reduced RSSI fingerprint
Author(s) -
Li Zhen,
Luo Lingen,
Sheng Gehao,
Liu Yadong,
Jiang Xiuchen
Publication year - 2018
Publication title -
iet generation, transmission and distribution
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.92
H-Index - 110
eISSN - 1751-8695
pISSN - 1751-8687
DOI - 10.1049/iet-gtd.2017.0601
Subject(s) - ultra high frequency , partial discharge , fingerprint (computing) , computer science , interference (communication) , noise (video) , subspace topology , cluster analysis , fingerprint recognition , probabilistic logic , real time computing , pattern recognition (psychology) , artificial intelligence , channel (broadcasting) , engineering , telecommunications , electrical engineering , voltage , image (mathematics)
Ultra‐high frequency (UHF) technique is chosen for partial discharge (PD) detection due to its excellent anti‐interference and stability. The existing UHF PD localisation methods are mainly based on time‐difference technique, that suffer high hardware cost due to the high sampling rate requirement. The wireless UHF sensors together with the scene analysis methods, i.e. received signal strength indicator (RSSI) fingerprint, provide us a low hardware cost and high environmental adaptability solution for PD localisation in substation. The proposed PD localisation method here is divided into two stages. In the offline stage, the RSSI fingerprint map is established by site survey. In the online stage, the position of PD source is estimated by a probabilistic approach. To reduce the influence of noise, the RSSI fingerprints are transformed to a low‐dimensional subspace, while the most of redundant noise is discarded. Furthermore, the PD heterogeneity and interference in the offline stage are also discussed and solved by normalisation and clustering algorithm, respectively. A filed test is performed and the results indicate that the mean localisation error is 1.87 m and 82.6% localisation errors are <3 m. The workload in the offline stage is also reduced ∼50%.