Ice thickness prediction algorithm based on the Makkonen-Transformer model

The thickness of ice around power transmission lines is influenced by numerous parameters. The accurate measurement of key parameters is challenging, which leads to the relatively low prediction accuracy of existing models for ice thickness. A prediction method for ice thickness around power transmission lines based on the Makkonen-Transformer model is proposed. As the most key parameter, freezing coefficient is measured very difficultly. In this paper, the freezing coefficient is predicted by the Transformer model using the historical data related to the ice thickness, and then the ice thickness is predicted with the Makkonen model. This approach combines the advantages of machine learning and physical computational models in the issue of ice thickness prediction, overcoming the strong reliance of machine learning on data quality and quantity, as well as the difficulty in obtaining key parameters in physical computational models. Therefore, the proposed model enhances the prediction accuracy of ice thickness around power transmission lines. The simulation results show that the proposed model has higher accuracy compared with existing single machine learning methods and can provide a scientific basis for grid enterprises to formulate the strategies for anti-icing and disaster reduction.