融合知识库和深度学习的电网监控告警事件智能识别  

Intelligent recognition of power grid monitoring alarm event combining knowledge base and deep learning

作　　者：孙国强[1] 沈培锋 赵扬 朱红勤 丁小柳 卫志农[1] 臧海祥[1] SUN Guoqiang;SHEN Peifeng;ZHAO Yang;ZHU Hongqin;DING Xiaoliu;WEI Zhinong;ZANG Haixiang(College of Energy and Electrical Engineering,Hohai University,Nanjing 210098,China;Nanjing Power Supply Company of State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 210019,China)

机构地区：[1]河海大学能源与电气学院,江苏南京210098 [2]国网江苏省电力有限公司南京供电分公司,江苏南京210019

出　　处：《电力自动化设备》2020年第4期40-47,共8页Electric Power Automation Equipment

基　　金：国家电网公司科技项目(SGJSNJ00FCJS1800810)。

摘　　要：电网监控告警信息是监控人员进行监控事件识别的重要数据基础。针对当前人为处理海量监控告警信息效率低的现状和电网智能技术深化应用的需求,提出一种融合知识库和深度学习的电网监控告警事件自主识别方法。基于自然语言处理技术中的Word2vec模型对监控告警信息进行向量化建模,基于卷积神经网络建立监控告警事件识别模型,通过算例对比验证所建模型的有效性和实用性。提出融合知识库与所建模型的应用方法,实现电网监控告警事件的智能感知和可靠识别。Power grid monitoring alarm information is an important data base for monitoring personnel to identify monitoring events.In view of the low efficiency of artificial processing of mass monitoring alarm information and the demand for deepening the application of power grid intelligent technologies,an automatic recognition of power grid monitoring alarm events based on knowledge base and deep learning is proposed.Based on Word2vec model in natural language processing technology,the vectorized modeling of monitoring alarm information is carried out,and the recognition model of monitoring alarm event is established based on convolutional neural network.The effectiveness and practicability of the proposed model are verified by comparison of examples.The application method of integrating knowledge base and the proposed model is proposed,which realizes the intelligent perception and reliable recognition of power grid monitoring alarm events.

关 键 词：电网监控 告警信息 Word2vec 卷积神经网络 事件识别 知识库 深度学习 

分 类 号：TM73[电气工程—电力系统及自动化] TP3[自动化与计算机技术—计算机科学与技术]