向量索引支持的电力故障知识图谱高效查询方法OA

Knowledge graph embedding(KGE)provides an effective approach for efficiently handling query tasks of power fault knowledge graph by mapping entities and relationships into a low-dimensional continuous vector space.However,existing research mainly focuses on optimizing query accuracy.In the context of industrial intelligence,the search for large-scale power fault knowledge graphs generally suffers from insufficient computational efficiency.Aiming at the above problems,an efficient query method for power fault knowledge graph based on vector indexing is proposed.The method first maps the power fault knowledge graph to a low-dimensional vector space through a semantic matching-based KGE technique.The embedding technique employs the Hadamard product to model complex relations(e.g.,symmetric relations)in the power fault domain.Then,the efficient retrieval of knowledge in the power domain is accomplished by constructing a query algorithm for power fault knowledge graph based on clustered vector index(IndexIVFFlat).The indexing mechanism accelerates the searched process by dividing the vector space into different partitions and solving their approximate nearest neighbors by vector inner product.Finally,comparative experiments in the power fault domain and on benchmark datasets verify the effectiveness of the proposed method.It can effectively improve the query efficiency of the fault knowledge graph while ensuring query accuracy,providing basic technical support for the effective handling of power faults.