Keywords
protection system; deep learning; anomaly detection; ICA-FNN
Abstract
Protection relay system is one of the main defense lines to ensure the stable operation of high-voltage networks. However, within the scenarios with the more complex network topology, the line architecture and the distribution, it is difficult to eliminate the potential operating anomalies or even failures. Also, the diversification of the protection equipment types, functions and locations poses new challenges to the defect management and equipment maintenance. Therefore, the automatic early warning technology of equipment abnormal state risk which considers both the timeliness and comprehensiveness should be studied. To this end, a real-time detection model of abnormal state risk based on data mining is proposed in this paper. Firstly, the independent component analysis is used for mass heterogeneous monitoring data to implement noise reduction. This can effectively improve the computational efficiency under high-dimensional data conditions. Secondly, the feed-forward neural network deep learning method which deploys the end-to-end training process to achieve time series anomaly detections is utilized. This can effectively alleviate the multi-category timing conditions of computational complexity. Finally, the protection system equipment in one area is exploited as empirical study, the results verify the abnormal detection performance of the designed model, which can promote the automatic identification and timely response of the protection relay system.
DOI
10.19781/j.issn.1673-9140.2024.04.009
First Page
78
Last Page
83,101
Recommended Citation
WEN, Yu; CHEN, Yanxia; LI, Jing; SUN, Bolong; LI, Xinming; and JIANG, Jianlin
(2024)
"An ICA‑FNN‑based multi‑model early warning approach for the abnormal state risks in high‑voltage network protection devices,"
Journal of Electric Power Science and Technology: Vol. 39:
Iss.
4, Article 9.
DOI: 10.19781/j.issn.1673-9140.2024.04.009
Available at:
https://jepst.researchcommons.org/journal/vol39/iss4/9