Research On The Construction Method Of Knowledge Graph For The Health Management Of Main Substation Equipment

The power system is large in scale and complex in structure.With the advancement of informatization and the expansion of power business,the production and service links of power generation,transmissi on,transformation,distribution,and use are generating massive amounts of data and knowledge all the time.As an important part of the substation,the substation plays a vital role.The main substation equipment mainly includes power transformers and gas insulation systems.With the deepening of smart grid and digital grid work,more and more sensors are deployed on the main equipment side of substations,generating a large amount of data all the time.Knowledge graph has good knowledge expression ability and interpretability,and is an important branch in th e field of artificial intelligence.Applying it to the health management of power equipmen t can effectively integrate the data of the whole life cycle of the equipment and form a new knowledge-oriented operation and maintenance management.model.At present,there is no in-depth research on the construction of the power knowledge graph,especially the health management knowl edge graph of the main substation equipment.Although some studies have begun to focus on the construction of the knowledge graph of the main substation equipment health management,there is still a lack of a unified standard and system.This paper first reviews the development of knowledge graph and its application in power system,and then makes a complete exposition of knowl edge graph technology system,including knowledge graph construction method,key technologies,commonly used entity learning models and methods,and analyzes commonly used entity learning The pros and cons of the method.By analyzing the advantages and dis advantages of common entity methods,the residual recurrent neural network entity learning model with attention mechanism proposed in this paper is introduced.First,the traditional recurrent neural network structure is changed,and a residual recurrent n eural network is designed for the field of health management knowledge graph construction.In order to reduce the impact of negative samples on the network and increase the network entity extraction ability,an attention mechanism is introduced in the residual part to capture long-term dependencies,reduce the interference of non-entity information,and enhance the network feature extr action ability.At the end of the paper,through the analysis of examples,the method proposed in this paper is compared with the commonly used methods,and the advantages and feasibility of the method in this paper are verified.Analysis and example analy sis show that the entity learning model proposed in this paper has strong feature extraction ability and can effectively alleviate the disappearance of neural network gradients.Finally,the work of this paper is summarized and prospected.