配电线路跳闸填报文本智能挖掘方法

Inteligent Mining Method for Line Trip Filling Texts in Distribution Systems

线路跳闸作为配电网的一种频发故障,其所积累的大量跳闸填报文本目前主要采用人工处理方式,效率低下且主观因素强。针对这一问题,以构成填报文本因果关系的故障现象及故障原因为挖掘目标,提出一种配电线路跳闸填报文本智能挖掘方法,该方法利用配电跳闸填报文本叙述的逻辑特点,提出了融合分词、词性及句法分析结果的跳闸填报文本抽取策略;在此基础上,提出了一种2阶段筛选方法,首先利用分布式高维向量相似度实现初步筛选,而后基于文本编辑相似度确定文本挖掘最终结果。基于某省的案例分析表明,所提出的文本智能挖掘方法准确率可达72%以上,显著提高了文本处理效率,已能初步满足实际需求。

As a frequent fault in distribution systems,line trip has accumulated a large number of filling texts,which are processed by manual methods.To overcome the disadvantiages of the inefficiency and subjectivity of manual methods,an intelligent mining method for a filling text of line trip of distribution system is proposed to automatically extract the fault phenomena and causes of a filling text,which provides the necessary conditions for the cause analysis of next line trip and the strategy of security control.In this method,the logical characteristics of the filling text narration of distribution system line trip are utilized,and a text extraction method of trip filling texts is proposed based on word segmentation,part-of-speech tagging,and syntactic structure parsing.Thereby,a two-stage text selection method is presented,in which the word vectorization model(word2 vec)is first used to perform the distributed representation of the filling text,then the filling text is filtered according to cosine similarity.After that,the character editing distance is further used for the second-stage filtering to determine the final results of text mining.The case analysis based on a distribution system in a province shows that the proposed text intelligent mining method can achieve an accuracy rate of more than 72%,the text processing efficiency is significantly improved,and it can initially meet the actual needs.