广东电力2025，Vol.38Issue(3)：46-54,9.DOI:10.3969/j.issn.1007-290X.2025.03.006
架空输电线路微气象预测的气象站时空融合数据驱动轻梯度提升机模型
LightGBM Model Driven by Temporal and Spatial Fusion Data from Weather Stations for Micrometeorological Prediction of Overhead Transmission Lines
摘要
Abstract
Extreme micrometeorology is prone to cause excessive ice-covering loads on the overhead transmission lines,which seriously threatens the safe operation of the power grid.Micrometeorological prediction of the overhead transmission lines provides meteorological prediction data for ice-covering prediction,improves the early-warning capability of ice-covering risk of transmission lines,and guarantees the safe operation of ice-covering period of the main channel of west-to-east power transmission and the large-scale clean energy bases of the power transmission channel.Aiming at the difficult problem of forecasting micrometeorology of the overhead transmission lines with high variability and fluctuation,this paper studies the prediction of micrometeorology of overhead transmission line terminals with weather forecasts from weather stations,propose the spatio-temporal data fusion method with the closest distance and moment between micrometeorology terminals and weather stations,and establishes a spatio-temporal data-driven micrometeorology prediction light gradient elevator model(LightGBM)for overhead transmission line terminals based on 4 474 pieces of data.A typical terminal is used as an example to study the effects of four prediction sample fusion methods,such as a certain moment of the model,the first 1 h,the first 2 h,and the first 3 h,etc.The results show that the micrometeorological prediction of the first 1h prediction sample fusion method is the most effective,and the average absolute errors of the micro-meteorological temperature,humidity,and wind speed prediction of the test set of 1,295 pieces of data are 0.87℃,3.178％,and 0.986 m·s-1,respectively.The LightGBM model based on the spatio-temporal fusion approach of the first 1h prediction samples predict the micrometeorology of 163 terminals,and the comparison with the monitoring values shows that the mean values of the prediction errors of temperature,humidity,and wind speed are 1℃～3℃,6％～13％,0.5～1.5 m·s-1,respectively,which provide accurate micrometeorological prediction data for the prediction of the overhead transmission line ice-covering.关键词
微气象预测/输电线路/气象站/LightGBM模型/数据融合Key words
micrometeorological forecasts/transmission lines/weather stations/LightGBM model/data fusion分类
动力与电气工程引用本文复制引用
郝艳捧,李鑫贺,黄磊,王黎伟..架空输电线路微气象预测的气象站时空融合数据驱动轻梯度提升机模型[J].广东电力,2025,38(3):46-54,9.基金项目
中国南方电网有限责任公司防冰减灾重点实验室支撑项目(GZKJXM20222180) （GZKJXM20222180）
国家自然科学基金委员会-国家电网公司智能电网联合基金重点项目(U1766220) （U1766220）
