Study On The Flashing Warning Method Of Polluted Insulators Based On The Leakage Current Characteristics

With the construction of smart grid in China, increasing high and extra high voltage lines have been put into practice. Pollution flashover accident has become a serious threat to the safe operation of the external insulation, and it has brought huge economic losses. Recently, domestic and overseas scholars have conducted a lot of theoretical exploration and experiment, and have made some achievements. However, pollution flashover accident of insulators still occurs frequently, so the mechanism and method of pollution flashover need to be further studied. With the exploring of the pollution flashover, a large number of data indicate that the surface leakage current is closely related to pollution flashover. Therefore, using the leakage current to diagnosis insulator contamination state is an effective method. After extensive literature review, a fuzzy neural network model based on variable weight method is proposed. Online assessment of pollution degree can be achieved through the real-time monitoring of insulators and some guidance can be provided for the power sector to clean the insulators.The pollution flashover of insulators is an intricate process of electric, thermal and chemical phenomena, and its insulation performance is finally reflected in the current magnitude of the surface discharge. So, firstly, the discharge mechanism is studied in this thesis. Then, a fuzzy neural network model based on variable weight is proposed. Data is collected by online sensors and is trained by MATLAB. The test shows that this model can effectively evaluate the contamination degree of insulators. Moreover, corresponding warning information can be obtained based on the evaluation results, resulting in effective prevention of the occurrence of pollution flashover.This thesis introduces basic theory of the artificial neural network, membership function of the fuzzy logic, and variable weight theory. Then the parameters that can reflect the whole process of flashover are selected from existing pollution assessment methods, avoiding the problem of single input parameter. In the training process, the number of hidden layers is determined according to the past experience. The weight of fuzzy logic has some shortcomings such as the mutation parameters may be ignored and its strong subjectivity. To overcome these drawbacks, variable weight method is used to determine the weight of each fuzzy input parameter in this thesis. The simulation results show the feasibility of this method. At the same time, in order to improve the accuracy of the evaluation system, a large number of field data is collected and used in the training system. After the training is completed, the weight matrix of the model will be saved and used to realize the pollution assessment of on-line insulators. The whole system has advantages of high degree of automation, strong learning ability, and objective and accurate evaluation.