摘要
针对电力系统调度员在OMS调度日志中记录的电力设备缺陷数据繁杂,冗余等问题,提出了一种基于文本挖掘的电力设备缺陷用户画像构建新方法。首先在分析电力设备缺陷文本数据的基础上,对集合样本进行标签化处理,以电力调度员的录入缺陷数据及运检人员在现场实际发现的缺陷数据为输入,运用改进的聚类算法对电力设备缺陷标签进行了定义及归类,最后结合构建的文本预处理模型以及卷积神经网络分类器,对电力设备缺陷文本进行了分类,并根据分类文本构建了电力设备缺陷的用户画像系统。实验结果表明,该方法有效的解决了电力调度员标签语义化问题,将需要调度人员关心的设备缺陷信息进行智能提取,实现了从PMS系统所有缺陷数据中自动推荐调度员关心的缺陷。
Aiming at the redundancy and other problems of power equipment defect data in the OMS scheduling recorded by power dispatchers,this paper presents a new method for constructing user portraits of power equipment defects based on text mining.Firstly,the text data of power equipment defects are analyzed,the set samples are labeled.The defect data recorded by the power dispatcher and the actual defect data found by the operator are taken as input,the defect labels of power equipment are defined and classified by the improved clustering algorithm.Finally,combined the text preprocessing model and the convolutional neural network classifier,classified the defect text of power equipment.,a user portrait system of power equipment defects is constructed according to the classified text.Experimental results show that this method can solve the problem of semanticization of power dispatcher labels effectively,the equipment defect information concerned by the dispatcher is extracted intelligently,and can recommend defects to the dispatcher from all defect data of the PMS system automatic.
作者
张鹏
王玮
赵德伟
司晓峰
Zhang Peng;Wang Wei;Zhao Dewei;Si Xiaofeng(State Grid Gansu Electric Power Co.Ltd Lanzhou,GanSu 730030)
出处
《科技风》
2019年第33期177-180,共4页
关键词
电力调度
设备缺陷
聚类算法
用户画像构建
文本挖掘
electric power dispatching
defective equipment
clustering algorithm
user portrait
text mining