Research On Intelligent Diagnosis Method Of Trip Event Based On Substation Alarm Information

The electric power control system will receive a large amount of alarm information from the substation during a tripping accident.If the regulator cannot make an effective decision in a short time,it may cause serious consequences.The alarm information of the substation contains a lot of valuable information.The alarm content in the alarm information is a short Chinese text in a specific form,which is unstructured data.It contains many professional vocabularies in the electric power field,and is mixed with many irregular elements such as numbers,separation marks,and special symbols.Therefore,it is of great significance to construct an event-based analysis method for accident trips based on big data technology,which can mine the value of alarm information and provide decision-making guidance for regulators in accident handling.In this regard,this paper studies the intelligent diagnosis method of trip events based on alarm information,and proposes two trip event diagnosis methods based on pattern matching and natural language processing technology to realize the diagnosis of accidental trip events in substations.In this paper,the alarm information and its pattern matching method are deeply studied,the characteristics of the alarm information are excavated,and the concept of the accident alarm information set is proposed.After that,based on the pattern matching method of substation alarm information and the idea of improved prefix tree,by summarizing the template alarm information sets of different accident trip types and using the shared level of the shared alarm information keywords as the sorting basis of the tree nodes,a diagnosis method for trip events is established.This method has the advantages of saving space and facilitating update.Aiming at the problem of low fault diagnosis accuracy caused by signal missed and wrongly sent,this paper uses Chinese text natural language processing technology to vectorize the alarm information text,and proposes a trip event diagnosis method based on vectorized alarm data.First,the vectorization method of alarm information is studied,and the ontology dictionary in the field of substation accident tripping is summarized.The hidden Markov model and the Viterbi algorithm in it are used to realize the word segmentation of the alarm information text.The bag of words model is used to map the alarm information to the vector space,which realizes the vectorization of the text of the alarm information.After that,for the vectorized alarm information,a two-layer architecture algorithm based on support vector machine classification and k-means clustering was designed.The first-layer classification algorithm was used to determine the trip event,and the second-layer clustering algorithm was used to judge the type of trip event.The diagnostic accuracy of this method is higher,and it can effectively avoid problems such as signal missing and signal mistransmission.