Research On The Key Technologies Of The Power Knowledge Platform Based On Knowledge Graph

Knowledge graph,also known as smart database,is an intelligent knowledge management system that combines artificial intelligence technology with traditional databases.Introducing knowledge graph into power knowledge management platform will enable the platform to gain the ability to mine useful knowledge from a large amount of text,thereby consolidating fragmented knowledge in the power sector,improving knowledge utilization,and serving users within the platform.The paper focuses on the key technologies of power knowledge platform based on knowledge graph.It has important engineering practice significance and engineering application value.This article studies the key technologies,related design and construction schemes of the intelligent power knowledge management platform.It mainly explains the concept pair generation algorithm,entity pair generation model,platform system function design and benefit analysis.Firstly,for the problem of common-sense extraction,a concept-pair sorting algorithm based on mutual information is studied.This algorithm can achieve concepts that satisfy the typical prioritization of a given verb/relationship by using the knowledge base and concept classification system.Then,from the similarity point of view,the use of the Markov cluster compression algorithm can generate several concept-pair clusters from concept pairs which have been achieved.The algorithm introduces the concept of mutual information,which solves the problem that the concept or the entity itself has a high frequency and thus affects the typicality of the evaluation concept.The algorithm is tested in the knowledge base and obtain a high correlation accuracy(91%).Secondly,this article studies the active knowledge base updating technology,which can generate entity-pairs from the collection of the extracted concepts and topic patterns.Based on the probability graph model,the unified generation mechanism under the topic mode and the concept mode is studied.In order to obtain a unified entity pair generation model under two mechanisms,a hidden variable is introduced;and the hidden variable is taken into account when solving the model parameters.An EM algorithm is deduced for solving the model.This model is tested on the knowledge base Google Syntactic N-Grams7.Compared with other entity generation models,this model can obtain better accuracy and recalling rate.Thirdly,based on the knowledge extraction and knowledge updating algorithm,this paper designs the system architecture of the power knowledge management platform and analyzes its system characteristics.Based on the characteristics,the main functional design of the platform is completed,and the non-functional parameters of the system,including reliability,security,scalability parameters,etc.are configured.Finally,the intelligent power knowledge management platform was evaluated and assessed for benefits,including the benefits of society and power supply companies.The evaluation focuses on the analysis of the effectiveness of platform construction in improving the quality of power marketing services,optimizing the quality of employee training,and promoting the knowledge-creating model.