Research On Intelligent Recommendation Algorithm For Power Grid Scientific And Technological Resources Based On Knowledge Graph

With the increasing investment in science and technology(S&T)in the power grid field year by year,power grid enterprises have accumulated a large amount of S&T resources.The knowledge value contained therein urgently needs to be exploited and utilized.This paper studies the intelligent recommendation algorithm for power grid S&T resources based on knowledge graph,which is of great significance for improving the level of power grid S&T management.To build a multi-source power grid S&T text corpus,a variety of data including power grid S&T papers,patents,projects,and domain knowledge were collected.To improve the processing ability of models for power grid S&T terminology,a power grid S&T terminology classification model based on pre-training and fine tuning was proposed to construct a glossary of professional terms.The model uses MacBERT as a language model and deep neural network as a fine-tuned classification model,achieving good results in domain text processing;A power grid S&T knowledge graph was researched and constructed,and a grid technology keyword extraction model and domain multi-label classification model were proposed.Entities and relationships such as power grid S&T achievements,keywords,and fields are described in the form of a knowledge graph.The keywords are extracted from power grid S&T resource texts using an improved TextRank algorithm,which fully considers the impact of word location,frequency,and semantic information,and incorporates the document itself and corpus topic information using the LDA topic model.A multi-label classification model for domain allocation was conducted,which can better enrich the semantics of the power grid S&T knowledge graph;A recommendation algorithm based on TransD+RW was proposed.The algorithm makes full use of the structural information of the power grid S&T knowledge graph,learn knowledge representation from the power grid S&T knowledge graph,and uses the results resource keywords and domain as the recommendation basis.Compared to conventional content-based recommendation algorithms,the proposed algorithm achieves better grid technology keyword differentiation,and thus has better recommendation effects.According to the knowledge management process of data collection,knowledge graph modeling,and intelligent recommendation algorithm construction,the paper implements the construction and verification of intelligent recommendation algorithm for power grid technology resources based on knowledge graph using models such as natural language processing and machine learning,providing reference for power grid technology knowledge management.