基于MLP神经网络的架空线路未来24 h覆冰风险预测  

Prediction of icing risk on transmission lines based on MLP neural network

作　　者：孙萌[1] 宋云海[1] 梁苇 周震震[1] 王奇[1] 张厚荣 张武英 姚钊泓 郝艳捧[2] 阳林[2] SUN Meng;SONG Yunhai;LIANG Wei;ZHOU Zhenzhen;WANG Qi;ZHANG Hourong;ZHANG Wuying;YAO Zhaohong;HAO Yanpeng;YANG Lin(Maintenance and Test Center of Ultra High Voltage Transmission Company,China Southern Power Grid,Guangzhou 510663,China;School of Electric Power,South China University of Technology,Guangzhou 510640,China;Guangzhou Cheng Yun Technology Limited Company,Guangzhou 511446,China)

机构地区：[1]中国南方电网超高压输电公司检修试验中心,广州510663 [2]华南理工大学电力学院,广州510640 [3]广州城云科技有限公司,广州511446

出　　处：《自动化与仪器仪表》2019年第7期165-168,共4页Automation & Instrumentation

摘　　要：架空线路覆冰威胁电网的安全稳定运行,预测架空线路覆冰风险能为电网及时开展抗冰工作提供决策参考。本文对架空线路覆冰在线监测及微气象数据预处理,利用SPSS数据分析平台,建立MLP神经网络模型,用微气象参数判别架空线路覆冰,进而预测架空线路未来24 h的覆冰风险。研究结果表明:未覆冰线路未来24 h覆冰风险预测的绝对误差不超过7%,该模型能准确预测当前未覆冰的架空线路未来24 h覆冰风险。Ice on overhead lines threatens safe operation of the power grid.Prediction of icing risk on overhead lines can provide decision-making reference for timely anti-icing work.In this paper,we preprocess online icing and micrometeorological monitoring data,using SPSS as data analysis platform,and establishing MLP neural network model to discriminate icing overhead lines with micrometeorological parameters and then predict the icing risk at 24 hour later.The results show that the absolute error of the icing risk prediction on over lines without ice do not exceed 7%,which is accurately for forecasting the icing risk after 24 hour of overhead lines that have neither ice.

关 键 词：MLP神经网络 微气象 架空线路 覆冰风险 预测 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]