Study On The Catenary Insulator's Security Stage Leakage Current Characteristics And The Prediction Of Pollution Flashover

Insulator is an important part of the electrical and mechanical connection of the electrified railway catenaries. With the rapid development of Chinese high-speed railway and the increasingly complicated operation environment of catenaries, the requirements of electrical insulation performance for insulators are supposed to be much higher in the future. Once the flashover happened, the tripping operation will appear on the lines which did affect the regular work of traction power supply system as well as the safe operation of trains. This thesis took the porcelain insulator XP-70 and the composite insulator FQBSG-25/12-970 P as the research objects, through the artificial pollution tests in the high-voltage laboratory, leakage current data were obtained under different environment and contamination degree. Then the characteristics of leakage current were researched and the insulator pollution flashover was predicted.At present, the exploration of constitutive relations are relatively less among leakage current and other factors for porcelain insulator XP-70. Therefore, two methods to evaluate the leakage current amplitude(Ih) were proposed in this thesis. The genetic algorithm(GA) was used to obtain function relationships of the Ih, the relative humidity density(RH), the equivalent salt deposit density(?ESDD) and the applied voltage(U) as the first method. The genetic algorithm to optimize the BP neural network was proposed to establish the prediction model of the Ih based on the RH, the ?ESDD and the U as the other method. The simulation result shows that the GA can identify the specific function form of them, and improves the accuracy of the estimation. The prediction model of the Ih based on the genetic algorithm to optimize the BP neural network can approach the nonlinear relationship between them accurately and effectively, and the two methods are effective. The research could be used as the reference to estimate the contamination degree of insulators based on evaluating leakage current.In order to get valid parameter about expressing the contamination degree, the leakage current security stage of the composite insulator FQBSG-25/12-970 P was carried out in time domain and frequency domain analysis. The mean value(Iem), maximum value(Iemax) and standard deviation(?) of the root mean square of leakage current are extracted as the time domain characteristics of the leakage current through analyzing the leakage current waveform which varies with RH and ?ESDD. The ratio of amplitude value between 3th harmonic component and the fundamental(K13) and the ratio of high frequency energy to total energy(?) are extracted as the frequency domain characteristics of the leakage current through analyzing each harmonic and energy which vary with RH and ?ESDD after the fast fourier transformation(FFT) and power spectrum estimation. Result shows that these five characteristics could provide useful information to predict the contamination degree and pollution flashover of the insulators in catenaries.At last, the time domain characteristics extracted from the security stage of leakage current are corrected to increase the applicability of the model. Combining with the frequency characteristics of the leakage current, a ?ESDD prediction model based on support vector machines(SVM) and a pollution flashover prediction model of the insulators based on the generalize regression neural network(GRNN) are established, respectively. The simulation results show that the SVM can be used as the prediction of ?ESDD, and provide a reference for determining contamination degree. The method of establishing the pollution flashover prediction model based on the GRNN not only provides guidance in practical engineering for cleaning and maintenance of insulators in catenaries, but also provides a method and preference for flashover warning.