Research On Name Entity Recognition Model Of Power Equipment Defects Text Based On Deep Learning

With the continuous construction of intelligent information in the power grid,the data in the electric power field is growing explosively,and in-depth mining of valuable data would have a significant impact on the development of market economy and social management in the electric power industry.A lot of information in the production process of the electrical grid is centrally stored and managed in the electric power system,in which a large number of basic information and production process information are recorded in the defective text of electric power equipment.However,in the recording process,the staff in the field will appear repeated expression,unclear logical expression,colloquial and other issues,so that the operation and maintenance staff and quality inspectors cannot accurately,efficiently and structurally manage the logical relationship among the defective text content.Meanwhile,in the process of power equipment operation,if the defect degree of the equipment cannot be judged timely and effectively,the equipment with critical defects will cause a series of cascade faults because it cannot be dealt with in time,which will affect the efficiency of electric power production.Based on the above situation,this thesis aims to focus on the named entity recognition of electric power equipment defect text from two aspects:the defect text cannot be managed structurally and the defect level cannot be judged and processed in time.The research contents of this paper are as follows:(1)A named entity recognition model of BiLSTM-CRF electric power defect text based on selective word vector is proposed based on deep learning.This research uses BiLSTM to learn the global semantic features of the text;add optional location information for the digital information in the text to train new digital word vectors to improve the digital information features;uses the transfer features of CRF to solve the problem of ordering among output tags.The experimental results show that the model can identify the entity category information from a large number of power defect texts,so that a large amount of power defect texts can be managed structurally.(2)This paper propose an electric power defect analysis framework to respond to different degrees of defective equipment in time.Based on this framework,firstly,attention mechanism and enhanced coding are added to improve the named entity recognition model to improve the accuracy of entity recognition,and then combined with the method of knowledge graph technology,this research achieved a function that can quickly be judging the defect degree of the task in the case of the given electric power text is in some extent of lacking description,to achieve an end-to-end defect degree analysis effect of electric power equipment.The experimental results show that the impro ved model can effectively improve the recognition accuracy of named entity information in a defective text.(3)To verify whether the design model of this paper can bring social and economic benefits to the electric power field in the actual electric power engineering scenario this paper raised a framework based on the electric power defect analysis in the way of designing and implementing an electric power defect text named entity recognition prototype system with friendly interface.Through this interface interaction way to help the field staff better understand and make use of the named entity information in the power field.This research work can significantly help the field staff to structurally manage the disordered text data,and quickly and accurately locate,understand and judge the defect degree of power equipment from the defect text records with certain logical relations.Furthermore,it helps the operations engineers to make reasonable treatment and prevention decisions according to the different defect degrees of the equipment.In the meanwhile,this study can lay a foundation for the follow-up intelligent analysis of electric power field and intelligent decision-making recommendation.