Affiliations: [a] College of Energy and Electrical Engineering, Hohai University, Nanjing, China
| [b] Jiangsu Key Laboratory of Smart Grid Technology and Equipment, Zhenjiang, China
| [c] School of Electrical Engineering, Southeast University, Nanjing, China
Correspondence:
[*]
Corresponding author. Mei Fei, College of Energy and Electrical Engineering, Hohai University, No.8, Fo cheng xi Road, Jiangning District, Nanjing, China. E-mail: meifei@hhu.edu.cn.
Abstract: As an important content of smart grid’s intelligentialize, high voltage circuit breaker (HVCB) can enhance security and stability of power system observably. Therefore, on-line monitoring and fault diagnosis technologies develop significantly. In order to promote diagnosis accuracy and learning ability, an on-line hybrid method is proposed for fault diagnosis of HVCB in this paper, and characteristics of control coil current are used as modeling data. The diagnosis model built by this method includes three modules: fault detection, fault learning and fault recognition. Fault detection is used to inspect monitoring data and find the fault samples using kernel principal component analysis (KPCA). Fault learning is used to judge whether there is new knowledge which is built by kernel fuzzy C-means (KFCM), and the new knowledge can be brought into this diagnosis model to update the fault recognition module subsequently. Fault recognition is used to realize automatic categorizing of fault data to identify the fault type using multi-classification support vector machine (SVM). The diagnosis conclusion will be obtained finally. Perfect results in diagnosing typical failure of HVCB using the proposed method have been proved by data experiment.
Keywords: Fault diagnosis, high voltage circuit breaker, kernel fuzzy C-means, kernel principal component analysis, support vector machine
