Prediction method for galloping features of transmission lines based on FEM and machine learning

Abstract Parameter study on galloping of iced transmission lines with different structural, ice and wind parameters is carried out by means of the finite element method (FEM), and a dataset is then created based on the numerical results. The random forests algorithm is used to set up a classification model to predict the galloping mode and vertical vibration mode. Using the span length, initial conductor tension, ice thickness, initial wind attack angle, wind velocity and the galloping mode as well as the vertical vibration mode obtained by the classification model as the input variables, a back-propagation (BP) neural network model is created to predict the galloping features including the frequencies, vibration amplitudes and the maximum conductor tension. The dataset created based on the parameter study is divided into two sub-datasets for training and testing of the two machine learning models respectively. The effects of the input variables on the output variables are quantified with a feature importance analysis for the classification model and a global sensitivity analysis for the feature prediction model respectively. The two models are finally integrated into one to predict galloping features efficiently and quickly. Making use of the proposed method, an early warning system for galloping of transmission lines may be able to be set up in the future.