Prediction of Insulators Ice Types Driven by Icing Images and Micrometeorology Time Series Fusion Data

Ice accumulation on overhead transmission lines poses significant risks to the stability of power grids. Precise ice types prediction can help power grid assess future freezing risk and prevent icing incidents. This study introduces a data-driven method for insulator ice types prediction, utilizing a fusion of images and micrometeorological data of the previous three days. Based on the icing monitoring data from China Southern Power Grid from 2014 to 2021, the approach fused images with micrometeorological time series for the first 3 days at 6 intervals to create a data fusion dataset. The insulator ice type prediction model, based on eXtreme Gradient Boosting (XGBoost) technology, was trained and tested using 4,503 and 1,931 samples, respectively. Its precision and recall reached 97.2%, which verified the accuracy of the proposed method and predicted the future insulator ice types.