Research On Partial Discharge Monitoring System Of GIS Based On UHF Signal Analysis

Gas insulated switchgear(GIS)has been widely used in power systems of different voltage levels because of its good insulation performance and safety performance.GIS failure may affect the safety of the entire power system,thus it is important to find internal defects and eliminate unsafe factors in time.At present,the most effective way to judge the severity of GIS defects is partial discharge(PD),and ultra-high frequency(UHF)detection technology is a mature technology in the current partial discharge diagnostic application.In view of the above background,this paper proposes a GIS partial discharge online monitoring system based on the UHF signal analysis.The main research contents and achievements of this paper are as follows:(1)The principles of GIS partial discharge and UHF detection are introduced,the application basis of UHF detection technology in the field of GIS partial discharge detection is explained,the characteristics and spectrum characteristics of four typical partial discharge defects of GIS are analyzed,the GIS partial discharge defect model is designed according to the defect characteristics,and the partial discharge spectrum is collected through the partial discharge simulation test.Combined with the partial discharge spectrum collected by the test and the field,the phase resolved partial discharge(PRPD)spectrum data set of GIS is established.The discharge spectrum data can be divided into four categories: corona discharge,insulation discharge,suspension discharge and free particle discharge.(2)Through the research of deep learning method,a GIS partial discharge defect map recognition model based on YOLOv5 algorithm is proposed.the GIS partial discharge map is preprocessed,the GIS partial discharge map recognition model is constructed,and different training methods are used to optimize the model parameters to obtain the optimal detection model.The recognition m AP values of insulation discharge,suspension discharge and free particle discharge were 96.74%,97.01%,93.76% and 96.06%,respectively,higher than those using other algorithms,which prove the advantages of the YOLOv5 recognition model.The PD map recognition model trained in this paper is used to identify GIS multi-source PD maps.This model can accurately identify one or more types of PD faults at the same time,or multiple faults in a short period of time,but there are still problems such as missing detection.(3)A GIS PD online monitoring system based on UHF signal analysis is designed.The system mainly includes UHF intelligent terminal,PD diagnostic system and data access device,which are used to collect PD maps,identify PD types and upload PD data,and proposes a PD online monitoring scheme suitable for this system respectively.The substation GIS was monitored online to find the abnormal signal of partial discharge,the type of partial discharge was determined by the partial discharge diagnosis system,the UHF time difference positioning method was used to determine the partial discharge position.The GIS was disassembled and checked to determine the type of partial discharge and the partial discharge position was consistent with the judgment results of the monitoring system,which proved the reliability and effectiveness of the system monitoring.