Prediction Model For Dynamic Responses Of Transmission Lines After Ice-shedding Based On Numerical Simulation And Machine Learning

The transmission lines are covered with ice in ice zones,and ice will shedding from conductors when the temperature rises or the wind blows.The falling ice can produce vertical vibrate and horizontal swing,which will cause the reduction of the insulation gap between lines and may lead to electrical accidents.In addition,the ice-shedding breaks the balance of the horizontal tension of adjacent conductors,and the excessive unbalanced tension will be applied to the hanging point of the insulator string on the tower,for severe cases,it will destroy the conductor and even break down the transmission tower,and resulting in huge losses.There is no effective method to estimate the horizontal unbalanced tension in existing research results,and the calculation for jump height and horizontal amplitude is over complicated.Therefore,it is significant to establish a model to determine the dynamic responses of transmission lines after ice shedding in a fast and accurate way.This paper introduces the numerical simulation method of the ice-shedding process of the transmission line,and simulates the dynamic response of the transmission lines after ice-shedding with different parameters.Some typical parameters of lines like span number,bundle number,conductor type,span length,elevation difference,initial stress,ice thickness,ice-shedding rate and wind speed influence on jump height,horizontal amplitude,trajectory and unbalance tension are analyzed.The prediction model is built based on machine learning theories,and the finite element numerical(FEM)results of the dynamic responses of transmission lines after ice shedding are sorted to be data samples.The line structure parameters,ice thickness,iceshedding rate and wind speed are used as the input variables of the prediction model.Based on BP neural network and random forest algorithm,the prediction models of the jump height of the transmission line after ice shedding are constructed.These two different prediction models can both give fast and accurate predictions.Compared with each other,the prediction model using random forest algorithm is more efficient,the model based on BP neural network is more accurate.The jump height,horizontal amplitude and horizontal unbalanced tension are simultaneously used as outputs,and multi-output prediction models are established based on the BP neural network and the Extra-Tree algorithm,respectively.The results of the prediction model based on the Extra-Tree algorithm are better than the model which based on BP neural network in efficiency and accuracy,especially for predicating the horizontal amplitude of transmission lines after ice shedding when wind load exists.With the structural parameters and load parameters of the transmission lines,the models built in this paper can predict the dynamic response of the transmission lines after ice shedding efficiently,which gives a fast and accurate estimation of the maximum jump height,horizontal amplitude and horizontal unbalanced tension,and provide an important basis for the design of transmission lines in ice zones.