Research On Ultra High Voltage Engineering Data Mining Based On Knowledge Graph Technology

UHV project is an indispensable part of promoting the coordinated development of power grid and power supply,in which there exists a huge amount of knowledge,and using this knowledge can provide guarantee for the construction of new UHV projects.As a form of knowledge embodiment of UHV engineering,the text of UHV engineering archives has the problem of difficult knowledge acquisition.Knowledge mapping technology,with the research in recent years,is able to make a perfect structured representation of knowledge and has been successfully applied in various fields.At present,there is still less research on the construction of knowledge mapping for power engineering,especially for extra-high voltage engineering,and the related knowledge base has not been constructed yet.In order to fully exploit the knowledge in UHV engineering,this paper studies the knowledge graph construction method for UHV(AC)engineering archives.By analyzing the research and application of knowledge mapping in various fields at home and abroad as well as the principle of relevant deep learning models,this paper studies the mapping construction technology based on multi-label neural network and proposes the framework of knowledge mapping construction for extra-high voltage engineering.On the basis of extracting the knowledge of "AC Transmission and Transformation Engineering Construction Management Process Manual"and forming a preliminary knowledge map of extra-high voltage engineering,the loss function of multi-label neural network is improved to achieve batch training and more stable convergence,so as to better realize the automatic classification of engineering archives,and then add the information of extra-high voltage engineering archives to the constructed knowledge map of extra-high voltage engineering automatically to form a The complete knowledge graph of UHV engineering archives is formed to explore the association relationship between UHV engineering archives.Finally,the extracted knowledge is displayed in the form of visualization using the Neo4j graph database and analyzed for application.The improved multi-label text classification model in this paper is validated by simulation with a unified evaluation standard,and the performance is improved based on the original model.The constructed knowledge graph can clearly show the individual entities and the association between them through visualization,which demonstrates the effectiveness of the model and method proposed in this paper.It demonstrates the knowledge representation capability of knowledge mapping in the field of extra-high voltage engineering.