Research on Power Communication Defect Diagnosis Technology based on Unsupervised Learning

• Published in: 2022 IEEE 4th International Conference on Power, Intelligent Computing and Systems (ICPICS) pp. 28 - 31
• Main Authors: Wang, Muwei, Liu, Yan, Dong, Jiaojiao
• Format: Conference Proceeding
• Language: English
• Published: IEEE 29.07.2022
• Subjects: alarm merging; artificial intelligence; Clustering algorithms; Computer architecture; Data mining; defect diagnosis; Maintenance engineering; Merging; Power systems; Scalability; similarity calculation; unsupervised learning
• DOI: 10.1109/ICPICS55264.2022.9873572

In order to study the power communication defect diagnosis technology based on unsupervised learning, four methods of alarm merging technology, artificial intelligence technology, unsupervised learning technology and graph data mining technology were analyzed. Four technical routes were positioned and graded, and the subject was tested and studied. For alarm data, a self-learning algorithm based on unsupervised clustering and frequent subgraph mining to realize alarm merging and defect pattern discovery is proposed, and a framework for automatic defect diagnosis and disposal is designed. The architecture has good scalability and iterative update ability, and is verified by experiments on real scene datasets, and the results show good performance.