Research On Intelligent Recognition And Application Of Alarm Information In Power Grid Based On Text Mining

With the continuous development of the power grid,the scale of the power grid continues to expand,and the smart grid is integrated with informatization.The data contained in the power grid is increasing explosively.When the power grid occurs fault,a large amount of alarm information data will be uploaded to the monitoring alarm window through the monitoring system.The alarm information is huge and difficult to use effectively.If text mining is applied to the power grid alarm information,we can extract the required and valuable information from the alarm information,and the information from the substation side and the dispatch side can be organically combined.It is conducive to the operation and maintenance end of the substation to quickly handle the fault and accurately identify the cause of the fault.This paper aims to realize the "self-healing" of the smart grid.In view of the large amount of power grid alarm information and it is difficult to obtain valuable information,we have carried out research on the text mining technology of alarm information.The process of text preprocessing,text classification,information extraction and information storage is carried out on the alarm information.And put forward the concept of fault tracing,using alarm information to reverse the cause of relay protection fault.In summary,this paper proposes an intelligent recognition and automatic classification method of power grid alarm information based on text mining,and mines valuable alarm information for application research of fault trace.The specific work and achievements are as follows:(1)According to the characteristics of alarm information,this paper develops methods and procedures for text mining of alarm information.Through the process of Chinese word segmentation and stop words removal,the text cleaning is achieved.Then the alarm information is trained by word vector of distributed representation based on skip gram model of word2vec.It realizes the semantic feature extraction of the text and the transformation from unstructured data to structured data.This paper converts the Chinese text of alarm information that is difficult for the computer to recognize into a text representation method that the computer can recognize intelligently.(2)Information storage of alarm information is carried out by constructing a semantic model.The establishment of the semantic model uses the pattern matching algorithm-Horspool algorithm to extract information from the alarm information.In order to realize automatic text classification of alarm information,the extracted alarm information is classified according to the tree structure standard through tree structure matching.Then,the TextCNN classification model based on deep learning is constructed.The model is trained by selecting historical alarm information as corpus to realize automatic text classification of alarm information.Experiments in the Python environment verified the classification effect of the classifier.(3)The alarm information after preprocessing and classification is used for fault diagnosis and fault trace application research.Using alarm information to reversely track the cause of the incorrect action of the relay protection device,the screening rules for the alarm information required for fault diagnosis and fault trace have been formulated.Combining FSM model with alarm information,FSM model based on alarm information is constructed.Through text mining of massive alarm information,the cause of incorrect action of relay protection device is reflected,and fault tracking of fault equipment is realized.Finally,the effectiveness of the fault trace method is verified in a practical case.