Power grid knowledge graph construction method fusing hybrid expert network model and semantic partitioning

By integrating a hybrid expert network model with semantic segmentation, dynamic segmentation and entity verification of power documents are performed, solving the efficiency and accuracy problems in the construction of existing power grid knowledge graphs and achieving efficient and accurate construction of power grid knowledge graphs.

CN122334438AActive Publication Date: 2026-07-03STATE GRID JIANGXI ELECTRIC POWER CO LTD
8 Cites 0 Cited by

Patent Information

Application Number
CN202610816354.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-06-08
Publication Date
2026-07-03
Estimated Expiration
2046-06-08

AI Technical Summary

Technical Problem

Existing methods for constructing power grid knowledge graphs suffer from problems such as low efficiency in entity relationship extraction, insufficient understanding of power industry texts, and waste of vectorized embedding resources, resulting in slow construction speed, low accuracy, and low resource utilization efficiency.

Method used

A hybrid expert network model and semantic segmentation method are used to perform semantically aware segmentation of power documents, set dynamic overlapping areas, extract entities and relationships using the hybrid expert network model, perform type validity verification and filter invalid entities in conjunction with the power grid schema, omitting the secondary review step, and finally perform vectorized embedding.

Benefits of technology

It significantly improves the accuracy and speed of entity relationship extraction, reduces the number of large model calls, saves computing and storage resources, and improves the construction efficiency of power grid knowledge graph.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a power grid knowledge graph construction method fusing a hybrid expert network model and semantic partitioning, and the method comprises the steps: 1, carrying out the semantic perception partitioning of a power document, setting an overlapping region between adjacent document blocks, and obtaining a plurality of document blocks; step 2, inputting the document blocks into a hybrid expert network model, extracting an entity-relation-entity triple set by the hybrid expert network model, and omitting a secondary rechecking link; and step 3, before vectorization embedding is carried out on the triple sets, type legality verification containing constraint rules is carried out according to the power grid graph, triads containing invalid entities are filtered out, and vectorization embedding is carried out on the remaining triple sets to construct the knowledge graph. Through cooperation of semantic perception blocking, the hybrid expert network model and verification filtering, the problems of low generation speed and inaccurate content understanding in the prior art are solved, the construction efficiency and quality of the power grid knowledge graph are remarkably improved, and the resource cost is reduced.
Need to check novelty before this filing date? Find Prior Art