Study Of Discharge Performance Of DC Insulator And Artificial Neural Network Application In Complex Air Condition

With the large-scale development of electric power industry, the DC transmission technology has become more and more important. But there still aren't any regulations can refer to follow in the insulation selection of complex air condition of DC transmission lines at present. In practice, it is still based on the experience and technology of AC transmission.According to the analyzing of the corresponding experiment results of insulation discharge, we can get the performance of insulation discharge in contamination icing low air pressure condition, the thesis collected the data of experiment, and analyzed the result of experiment by building the ANN, and figure out the flashover voltage of insulator in complex air condition; According the character of ANN structure and stylebook, the paper put forward a way to analyze the error of result of ANN. The main works and conclusions in this paper are as follows:(1) According the sediment of contamination on the surface of FC210 insulator of transmission line, the uneven phenomena of contamination distributing would be saturated with the contamination keeps increasing; At the same time, a linear Lagrange calculation had been put forward to estimate the flashover voltage of unknown contamination grade, then, the error had been analyzed.(2) According to the results of experiment of FC210, the paper analyzed the influence of flashover voltage which influenced by contamination, icings air pressure, the flashover voltage would be saturated with the contamination and ice-amount keep increasing. At the same time, the performance of discharge in contamination, icing, low air pressure air condition to be analyzed, the relationship of contamination amount, icing amount and air pressure to the flashover voltage had been gotten.(3) Combined with the experiment data and chose the suitable ANN structure in the Matlab software, the BP ANN model of the relationship of contamination, altitude, ice-amount and flashover voltage had been built as the ANN verified by single factor function. The contamination, altitude were input vectors, flashover voltage was output vector. After the stylebook data trained and learned by ANN, the herald of flashover voltage in complex air condition had be gotten.(4) The main factors of error of ANN are its structure and stylebook; Different stylebook needs different structure. For the ice-amount and air pressure experienceformulas of flashover voltage had been verified by engineering, the error of simulation results of ANN had been analyzed according to those experience formulas and experiment data. Compared with the experience formulas, the ANN had the same definition of calculation. The average of error in the ANN of this paper was not overrun 0.007, and the maximal error was not overrun 0.02, the result of the error could be accepted by engineering.