Abstract: Aiming at the problems of high acquisition cost, insufficient coverage of working conditions, and difficulty in rapid prediction of aerodynamic parameters of iced quad-bundled conductors, a machine-learning-based prediction method using wind tunnel test data is proposed. Based on wind tunnel test data of iced quad-bundled conductors, bundle spacing, wind attack angle, wind speed, ice thickness, conductor diameter are selected as the original input parameters, while Reynolds number and relative ice thickness are introduced as physics-enhanced features. Prediction models for the drag coefficient and lift coefficient are established separately. Random forest, extremely randomized trees, gradient-boosted decision trees, and Gaussian process regression models are compared, and their performance is evaluated using the coefficient of determination, mean absolute error, root mean square error, and mean square error. The results show that the Gaussian process regression model achieves the best prediction performance, with testing-set coefficients of determination of 0.992 7 for the drag coefficient and 0.977 9 for the lift coefficient, outperforming the other models. The aerodynamic curve reconstruction results under typical working conditions indicate that the proposed model can effectively capture the overall trends and local fluctuations of the drag and lift coefficients as a function of wind attack angle. The proposed method can serve as a reference for rapid prediction of aerodynamic parameters, galloping analysis, and engineering risk assessment of iced quad-bundled conductors.

Key words: quad-bundled conductor, aerodynamic parameters, wind tunnel test, machine learning