A Method for Determining Partial Discharge Defect Sources in GIS Online Monitoring

This paper designs a method for identifying the local defect homology in ultra-high frequency(UHF) on-line monitoring of gas insulated switchgear(GIS) equipment, and uses the time correlation and data similarity of the measured points to find out the source of local discharge defect in the substation. It further combines the energy characteristics of the local discharge signals at each station with the physical topological relationship of each station to identify whether the local discharge signals are internal signal or external interference. This method solves the problem of how many discharge signals there are and whether these discharge sources are internal or external when a large number of local measurement and point alarms are on site. In addition, this method uses Jaccard distance to measure temporal correlation of time series data based on clustering characteristics. Multidimensional local discharge eigenvalues are constructed from very high frequency phase resolved pulse sequence(PRPS) data, and the data similarity between different measured points is measured based on weighted Euclidean distance. Combines the topological characteristics of the deployed points with the energy decay characteristics of the measured point signals to identify whether the homologous signals are internal or external interference. This method has the features of high recognition accuracy and high practicability.