Prediction of the jump height of transmission lines after ice-shedding based on XGBoost and Bayesian optimization

Abstract

The maximum jump height after ice-shedding must be determined to avoid jumps in transmission lines caused by ice-shedding, which can result in interphase flashovers and collisions that endanger the safety of transmission engineering. In this work, a numerical study of transmission lines under different structural, ice and wind parameters is carried out to create a dataset with a total of 1980 data for 11 input features that have a significant effect on the jump height. A data-driven model BO–XGBoost combined with Bayesian optimisation (BO) and Extreme Gradient Boosting (XGBoost) algorithm is proposed to predict the jump height of transmission lines after ice-shedding. Analysis results indicate that the application of the BO algorithm to the hyperparameter optimisation of the XGBoost model can improve the prediction accuracy whilst maintaining high efficiency. Meanwhile, the proposed BO–XGBoost model is superior to other benchmark models in various performance indicators, and strong correlations are discovered between the predicted and the target values. In addition, the proposed model has the advantages of high reliability and interpretability and can rapidly and accurately predict the maximum jump height of a transmission line after ice-shedding, which provides an effective and convenient means for the electrical insulation clearance design of transmission lines.