基于自适应随机森林算法的极端天气下配电网灾损预测  

Adaptive random forests algorithm for disaster loss prediction of distribution network in extreme weather

作　　者：米昕禾 刘泽宇 王天宇 于振 冯杰 MI Xinhe;LIU Zeyu;WANG Tianyu;YU Zhen;FENG Jie(State Grid Smart Grid Research Institute Co.,Ltd.,Beijing 102209,China;State Grid Key Laboratory of Information&Network Security,Beijing 102209,China;State Grid Corporation of China,Beijing 100031,China)

机构地区：[1]国网智能电网研究院有限公司,北京102209 [2]信息网络安全国网重点实验室,北京102209 [3]国家电网有限公司,北京100031

出　　处：《供用电》2023年第7期57-62,81,共7页Distribution & Utilization

基　　金：国家电网有限公司科技项目“电网灾情智能感知与应急指挥关键技术研究”(5700-202019185A-0-0-00)。

摘　　要：近年来台风、暴雨灾害对配电网基础设施造成了极大破坏,亟须提高配电网灾前损失预测研判水平,因地制宜做好电网差异化应急准备,提升电网防灾减灾水平,减少停电损失。针对台风和暴雨2种易造成配电网大范围损毁的典型极端天气,对灾害天气、配电线路及设备和地理环境3类成灾变量进行数据处理和筛选,并对随机森林算法的决策树构建和投票机制进行改进优化,提出一种双重加权改进的自适应随机森林算法的极端天气下配电网灾损预测方法。以台风“山竹”造成的某地市级配电网杆塔损失为例,验证了该算法模型的运算速度和有效性,综合预测准确率达到90%以上,并根据预测结果分析了配电线路及设备损毁与成灾因素间的作用规律,为电网差异化灾害应对提供理论依据。In recent years,typhoon and rainstorm disasters have caused great damage to the distribution network infrastructure.It is urgent to improve the research and judgment level of distribution network loss prediction before disasters,make differentiated emergency preparations for power grids according to local conditions,improve the level of power grid disaster prevention and mitigation,and reduce power outage losses.Therefore,this paper aims at two typical extreme weather conditions,typhoon and rainstorm,which are easy to cause large-scale damage to the distribution network,and conducts data processing and screening for three types of disaster variables,namely,disaster weather,distribution network lines and equipment and geographical environment,and improves and optimizes the decision tree construction and voting mechanism of Random Forest(RF)algorithm,and proposes a disaster loss prediction method of distribution network under extreme weather conditions based on dual weighted improved adaptive RF algorithm;The calculation speed and effectiveness of the algorithm model are verified by taking the tower loss of a city level distribution network in typhoon“Shanzhu”as an example,and the comprehensive prediction accuracy is more than 90%.Based on the prediction results,the action law between equipment damage and disaster causing factors is analyzed,which provides a theoretical basis for the differential disaster response of the power grid.

关 键 词：配电线路及设备 灾损预测 自适应随机森林算法 极端天气 停电损失 

分 类 号：TM72[电气工程—电力系统及自动化]