Type Identification of GIS Partial Discharge Based on SF<inf>6</inf> Decomposition Characteristics under Negative DC Voltage

Xu Yang,Yi Liu,Yi Jiang,Hao Wen,Jing Zhang,Jia Chen
DOI: https://doi.org/10.1109/ICHVE49031.2020.9279820
2020-01-01
Abstract:In order to use SF <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">6</sub> decomposition characteristics to identify faults of DC gas-insulated switchgear (GIS) under partial discharge (PD), the author chooses 6 experimental voltages to represent different PD stages, and carries out SF <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">6</sub> decomposition experiments. The experimental results show that SF <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">6</sub> decomposition produces include five stable components of CF <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">4</sub> , CO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> , SO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> , SOF <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> and SO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> , among which SOF <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> is the most important decomposed product, and the concentration of sulfur-containing components is higher than that of carbon-containing components. Finally, a feature set consisting of 21 concentration ratios is constructed, and the maximum relevance minimum redundancy criterion is used for the feature selection. BP neural network and support vector machine are used for fault diagnosis, and the accuracy rate is higher than 88%. The research work lays the foundation for the on-line monitoring and insulation state evaluation of DC GIS based on SF <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">6</sub> decomposition characteristics in the future.
What problem does this paper attempt to address?