Prediction model for aerodynamic coefficients of iced quad bundle conductors based on machine learning method.

The lift, drag and torsional moment coefficients, versus wind attack angle of iced quad bundle conductors in the cases of different conductor structure, ice and wind parameters are numerically simulated and investigated. With the Latin hypercube sampling and numerical simulation, sampling points are designed and datasets are created. Set the number of sub-conductors, wind attack angle, bundle spacing, ice accretion angle, ice thickness, wind velocity and diameter of the conductor as the input variables, a prediction model for the lift, drag and moment coefficients of iced quad bundle conductors is created, trained and tested based on the dataset and extra-trees algorithm. The final integrated prediction model is further validated by applying the aerodynamic coefficients from the prediction model and numerical simulation, respectively, to analyse the galloping features. The developed efficient prediction model for the aerodynamic coefficients of iced quad bundle conductors plays an important role in the quick investigation, prediction and early warning of galloping.