Micro-Topographic Feature Extraction for High Wind Risk Prediction of Transmission Lines

In complex terrain regions, particularly under extreme wind conditions, transmission lines are highly susceptible to significant incidents such as wind-induced flashover and line breakage. However, current meteorological models often lack sufficient accuracy in addressing wind speed variations caused by micro-topography, which hinders precise risk predictions. This study proposes a micro-topographic feature extraction method tailored for high-wind risk prediction of transmission lines. By integrating an improved UNet model with the DBSCAN clustering algorithm, this approach enables accurate extraction and classification of disaster-prone micro-topographies, such as ridges and valleys, thus providing a foundation for high-wind early warnings. The method begins by using an enhanced UNet model to identify key terrain features, including peaks, ridges, and valleys. Then, DBSCAN density-based clustering is applied to refine the segmentation, isolating high-risk micro-topographic areas with increased wind exposure. Additionally, ECMWF wind field data are coupled with terrain information, while local wind field analysis, facilitates high-precision high-wind risk warnings within micro-topographic zones. Enhanced feature extraction precision further supports refined high-wind forecasting capabilities. Experimental results demonstrate that this method achieves outstanding performance in both disaster-prone terrain feature extraction and regional classification. The generated dynamic risk maps provide robust early warning and decision-making support for the safe operation of transmission lines and show promise for broader applications in meteorological risk prediction across other complex terrains.