面向电力变压器规程的知识图谱构建方法研究OA

During the installation,transportation,operation,maintenance,and testing of power transformers,it is necessary to strictly comply with relevant technical requirements and standard specifications.These standard specifications have a large amount of data and are mostly stored in the form of text.When power grid workers query these specification texts,there are problems such as low query efficiency and weak knowledge correlation.As a new way of knowledge organization and management,knowledge graph can achieve intuitive display and efficient retrieval of massive information.Therefore,this article designs a knowledge graph construction method for power transformer regulations.To address the issue of triplet overlap in transformer regulation texts and improve the accuracy of triplet extraction,this paper adopts a CasRel-BiLSTM-MHA entity relationship joint extraction model that integrates multi head attention mechanism and bidirectional LSTM to extract knowledge from texts.The experimental results show that the F1 value based on this model reaches 83.48%,which is significantly improved compared to other comparative models,verifying the superiority of this model in constructing knowledge extraction tasks for power transformer regulations.Finally,applying the constructed knowledge graph to the intelligent Q&A of power transformer regulations can provide strong support for the safe operation and standardized operation of transformers.