A Model for Predicting the Flashover Voltage of Ice-Covered DC Insulator Strings Based on Extreme Learning Machine Neural Network

Icing tests of insulator string are limited by many factors, such as geographical location, climate conditions and test equipment. Therefore, this paper proposes a new model for applying artificial neural network to select the external insulation on the basis of the experimental data obtained, that is, the mapping relationship between complex environmental conditions and the Flash loop voltage of dc insulator is developed in view of the extreme learning machine neural network. When the neural network is verified and trained, It can be used to forecast ice flash pressure. By comparing the prediction results obtained by extreme learning machine neural network with the experimental results, the relative error between them is less than 4.46%. The method accurately predicts flashover stresses of insulators with ice and reasonably predicts the flashover stresses of icing insulators. The implementation of this project has very important theoretical and practical significance for the prevention and control of flood and drought disasters in our country.