Research On Photoelectric Combined Detection Technology Of Partial Discharge Of GIS

Partial discharge is a common problem in high-voltage equipment.which can lead to insulation deterioration,equipment damage,power supply interruption,and even serious accidents.Gas Insulated switchgear(GIS)is an important component of high-voltage transmission systems and is prone to partial discharge.Currently,the Ultra High Frequency(UHF)detection method is a relatively mature technology for partial discharge detection,but there is electromagnetic interference in the field environment of substations,and the detected UHF signals have noise.Therefore,this paper proposes a GIS optoelectronic combined detection technology that combines UHF detection with fluorescence fiber detection.The statistical characteristics of partial discharge signals extracted from fluorescence fiber signals are used to replace the statistical characteristics of partial discharge signals extracted from traditional UHF signals.First,by analyzing the gas discharge mechanism and surface discharge mechanism,this paper determines the three typical defect models to be studied:the tip discharge model,the free metal particle model,and the surface discharge model.The electric field distribution of GIS in three defect states is verified through electric field simulation calculation,which provides a physical model reference for the optoelectronic combined detection experiment of GIS partial discharge.Second,a GIS partial discharge optoelectronic combined detection experimental platform is established,and partial discharge experiments are carried out under three defect models.The original calculation formula for statistical characteristics is modified,and the statistical characteristics of fluorescence fiber signals are used to replace the statistical characteristics of UHF signals.A total of 15 feature quantities are extracted from the frequency feature,energy feature,and statistical feature perspectives to form a defect recognition feature vector and establish a defect recognition database.Finally,for the problem of GIS partial discharge defect recognition,the Back Propagation Neural Network(BP Neural Network)is improved,and the BP neural network structure and transfer function are determined.Due to the randomness of BP neural network weight distribution,it may fall into local optimal solutions.Therefore,genetic algorithm is used to optimize the BP neural network by adjusting the weight and network accuracy.The results show that using genetic algorithm to optimize the BP neural network can improve the defect recognition rate.