Research On Prediction Of Contamination State Of Insulator On Catenary Based On Fuzzy Neural Network

Electricity of the electrified railway is transmitted by overhead linesystem,and the insulator is an important part.With the development of economic construction in our country,impacts of high-speed railway on the operating environment,factory distribution and other factors are more and complicated.High-speed railway places heavy demands on electrical insulation strength of catenary insulators these days.Studies have shown that accumulation of contaminants in the insulator surface under the influence of certain external conditions can cause pollution flashover,affecting the normal operation of the railway system and power supply.Therefore,predicting contamination status of operating insulators in a timely manner ensures staffs to take measures to deal with security in time to avoid a potential flashover incidents is important.This thesis is based on a large number of domestic and foreign references,summarizing the results of previous studies and conclusions.According to laboratory test data,the change law of insulator leakage current is analyzed in detail,and fuzzy neural network is used to detect the contamination status of the contact network insulator.It is proposed to improve the early warning capability of insulator state detection.Firstly,in order to explore relationship and function relations between leakage current and foul degrees of insulators.This thesis takes the FQBSG-25/12-970 P rod-type composite insulator as the research object,and obtains leakage current data under different levels of contamination and humidity through tests.Characteristic quantities of leakage current is extracted from the artificial pollution laboratory,and three effective characteristic quantities are selected: leakage current effective Ie,maximum value Im and standard deviation ?,and obtain their relationships with degrees of pollution.Then,characteristic quantities and humidity are used as input variables.BP neural network model is used to predict the contamination degree of insulators,and the results are compared with the experimental results.The BP neural network can determine the weights and thresholds through self-learning.The prediction results are objective and accurate,and the prediction results are compared with the test results.The results show that: characteristic quantities are effective in predicting the degree of pollution.Finally,according to the requirement of the intelligent monitoring system for insulators,a fuzzy neural network-based method for predicting the state of insulator contamination is proposed.Considering the contamination degrees of insulators predicted by BP neural network,various operating environment parameters and many other influencing factors,a network model for judging contamination status is established to comprehensively assess the contamination status of operating insulators.Results show that the fuzzy neural network is effective to assess contamination states of insulators.By using this method,the contamination status of insulators in different operating environments and geographical environments can be obtained.According to results,warnings can be given to maintenance personnel of overhead linesystem so that they can accurately grasp real-time contamination status of insulators and provide engineering personnel with tasks of cleaning and contaminating insulators.