基于TPE-BIRCH的电网安全隐患分类方法  

A Classification Method of Power Grid Security Hazards Based on TPE-BIRCH

作　　者：杨有慧 董申颂 陈明媛 庞壮 覃芳璐 YANG Youhui;DONG Shensong;CHEN Minyuan;PANG Zhuang;QIN Fanglu(Power Dispatching Control Center of Guangxi Power Grid Co.,Ltd.,Guangxi Nanning 530023,China;Baise Power Supply Bureau of Guangxi Power Grid Co.,Ltd.,Guangxi Baise 533000,China;Nanning Power Supply Bureau of Guangxi Power Grid Co.,Ltd.,Guangxi Nanning 530023,China)

机构地区：[1]广西电网有限责任公司电力调度控制中心,广西南宁530023 [2]广西电网有限责任公司百色供电局,广西百色533000 [3]广西电网有限责任公司南宁供电局,广西南宁530023

出　　处：《广西电力》2022年第6期57-63,共7页Guangxi Electric Power

摘　　要：针对大数据量情况下电网安全隐患快速分类难的痛点,本文提出一种基于层次结构平衡迭代聚类(Balanced Iterative Reducing and Clustering Using Hierarchies,BIRCH)和树形结构核密度评估(Tree-structured Parzen Estimator,TPE)的电网安全隐患分类方法。首先,算法基于TextRank构建电力行业词库,使用JIEBA库完成隐患问题文本切分词,并构建对应的特征矩阵。其次,使用BIRCH对特征矩阵进行聚类,并使用TPE优化BIRCH聚类超参数。最后,根据聚类结果映射得到安全隐患分类。实际算例测试表明,所提算法快速有效,可实现电网安全隐患分类的细颗粒度聚焦和多维度概括。Aiming at the difficulty of rapid classification of power grid security hazards with large amount of data,a classification method of power grid security hazards based on Balanced Iterative Reducing and Clustering Using Hierarchies(BIRCH)and tree structure kernel density assessment(TPE)are proposed in this paper.Firstly,the algorithm builds a power industry vocabulary based on TextRank,and uses the JIEBA library to complete text segmentation for hidden problems,and constructs the corresponding feature matrix.Secondly,BIRCH is used to cluster the feature matrix,and TPE is used to optimize the hyperparameter of BIRCH clustering.Finally,the classification of security hazards is obtained by mapping the clustering results.The practical testing results show that the proposed algorithm is fast and effective,which can achieve fine particle focus and multi-dimensional generalization of power grid security hazards classification.

关 键 词：电网调度 安全隐患 文本聚类 轮廓系数 超参数优化 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]