Construction and optimization technology of power grid dispatching knowledge graph based on multi-modal data fusion

Read full text Summarise with VIRAT
Authors: Weiwei Li, Chao Hu, Siqi Shen, Zhangguo Chen
Cited by: 0
References: 20

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The data in the field of power grid dispatching has multi-modal characteristics such as text, images, and time series. How to deeply integrate these heterogeneous data is a key challenge in building an intelligent knowledge graph. The research aims to construct and optimize a power grid dispatching knowledge graph based on multi-modal data fusion. To this end, a unified framework integrating text, images, and time series data is proposed. This framework first uses joint extraction technology to extract entity relationships from text; subsequently, an improved RESCAL model (fusing L2 regularization and data augmentation) is introduced for knowledge embedding to enhance generalization ability; for the multi-modal association problem, a cross-modal transformation network (CMTN) is designed to map different modal data to a shared semantic space to achieve precise retrieval. At the system level, the perceptual hashing algorithm is integrated for fast similarity matching, and a distributed storage architecture is adopted to ensure the efficient processing and dynamic update of massive multi-modal data. Experimental results show that the joint extraction technique achieves high accuracy and recall in entity recognition and relationship extraction tasks, with F1-scores of 0.82 and 0.86 on the PowerGraph and OmniCorpus datasets, respectively. The CMTN exhibits superior performance in cross-modal retrieval, with mean inter-modal similarities of 0.72 and 0.75 and Top-1 alignment accuracies of 0.85 and 0.88 on the two datasets. The constructed knowledge graph effectively supports intelligent power grid dispatching by accurately representing and managing complex multi-modal data. The research provides an effective solution for enhancing the representation, association, and update capabilities of grid dispatching knowledge through multi-modal data fusion technology. At the same time, it can focus on the lightweighting and real-time optimization of the model to promote the integration of this technology into online intelligent dispatching systems.

Fields of Study & Subjects

Advanced Graph Neural Networks Multimodal Machine Learning Applications Semantic Web and Ontologies

Publication details

Journal / source
EAI Endorsed Transactions on Energy Web
Volume
12
Publication date
23 March 2026
Language
EN