Forecast on Failure Characteristics of Transmission Line Galloping Based on Neural Networks Method

This paper conducts analysis and processing on galloping data by utilizing data mining and machine learning technology. Select the BP neural networks algorithm, establish the training model of neural networks, and then conduct forecast and early warning to galloping failure characteristics of the power transmission line. The accurate rate of forecast of the model on testing set is 89.52%. After accessing the forecast model to the galloping early warning platform for application, the line galloping accident in the eastern region of Inner Mongolia on October 26 2018 was forecasted successfully, and the forecast result of failure characteristics was output, which conformed to the actual situation, thus verifying the accuracy of model forecast result. This research achievement enhances the fining degree of galloping forecast and early warning, and the Power Grid Running and Maintenance Department can pre-judge the failure degree of galloping to the line according to galloping failure characteristics , thus adopting differentiated anti-galloping measures for solution, reducing running & maintenance cost, and enhancing the capability of power grid in dealing with galloping disaster.