Study On The Propagation Characteristics And Pattern Recognition Of The UHF Partial Discharge Signal In GIS

Because of the high reliability and economy of Gas Insulated Switchgear (GIS), it has been widely used in power transmission. So it is important to monitor the insulation condition of the GIS and find out the insulation defects in time, and the Partial Discharge (PD) detection is an effective method. Due to the wave-guide structure of GIS, the ultra-high-frequency (UHF) PD detection technique is more effective than the traditional method for the PD detection. According to detecting electromagnetic signal emitted by PD source, the UHF technique can detect and locate the PD sources in the GIS.In order to get more knowledge of the propagation and attenuation characteristics of the UHF signals in GIS,to economize the UHF sensors in GIS detection, the propagation characteristics of UHF signals emitted from PD are studied theoretically and simulationally. In this paper, GIS straight cavity,L shape corner, insulator are utilized, the tranasmission of PD eleetormagnetic wave in the wave-guide is simulated by FDTD method.Therefore, factors influencing UHF signals are analyzed, including PD current amplitude and pusle width of PD source. Based on the energy received by the probes within 50ns,the results showed that the UHF signals attenuate 4.79dB after a L shape corner,5.50dB after an insulator (ε_r=4.6).Accroding to the performance requests of UHF antenna, based on the study on the characteristics of the UHF signal and noisy signal, a Plane equiangular spiral antenna is designed. The simulational results show that the sensor has superior gain and directional properties.To simulate discharges in GIS, four typical PD models are constructed in a GIS. Using the UHF PD measuring system, the UHF signal emitted from PD and phase resolved PD patterns are studied and the feature parameters are calculated to discriminate between different PD models, it is laid the foundation for PD UHF signal recognition.In this paper, artificial neural network (ANN) is applied to pattern recognition of PD for four types of defect models.In order to improve the reliability and accuracy of fault diagnosis, a new model based on TNFIN(Tsukamoto-Type Neural Fuzzy Inference Network) that is a typical network of Fuzzy Neural Network(FNN) is proposed in this paper. Firstly the network architecture is decided according to the input-output pattern, then network parameters can be updated according to the hybrid algorithm that integrates gradient decent algorithm and LMS algorithm on the basis of supervised learning and error minimization. The results of verification of four typical PD models show that it has high Generalization Ability and recognition capability, in the meanwhile, solve the inherent shortcoming of back propagation neural network (BPNN) such as slow convergence and local minima.