Journal Article

A Method for Evaluating the Icing State of Transmission Towers Based on Meteorological Conditions and Random Forest

Mao, Mengting, Qiu, Zhibin, Hu, Fei, Peng, Shiyi, Liu, Chen

2025

Overview

The icing on the towers in transmission lines is mainly affected by meteorological factors such as temperature and humidity. This paper presents an evaluation model for the icing state of transmission towers based on meteorological factors and the random forest (RF) algorithm. Firstly, based on the grid meteorological data and the coordinates of the towers in the transmission lines within Jiangxi Province, China, the meteorological factors corresponding to each tower were matched, and an ice coverage criterion model based on physical conditions was established. The results obtained based on physical criteria were compared with the actual icing monitoring data of transmission towers to verify the validity of the proposed criteria. An RF classifier is adopted to construct a dependency mapping model between meteorological factors and the icing state of towers. Only a small amount of data from seven transmission lines was used to train the model, successfully predicting the icing state of 32,496 data sets with an accuracy and recall of 89.4% and 87.01%, respectively. Its performance was far superior to that of back propagation neural network (BPNN), support vector classifier (SVC) and least squares support vector machine (LSSVM) models, verifying the feasibility of the proposed method in practical applications.

• Publisher: IOP Publishing
• Subject: Back propagation networks / Criteria / Ice cover / Ice formation / Icing state / meteorological condition / Meteorological data / Neural networks / random forest / Support vector machines / Transmission lines / transmission tower / Transmission towers