Open AccessEnergy Storage State Identification Of Circuit Breaker Based On Fast Extraction Of Interval Feature Of Vibration Signal
Author(s) -
Xiaofei Xia,
Yufeng Lü,
Yi Su,
Jian Yang,
Xiajin Rao,
Zecheng Lu
Publication year - 2020
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/508/1/012177
Subject(s) - circuit breaker , feature extraction , energy (signal processing) , feature (linguistics) , support vector machine , feature vector , pattern recognition (psychology) , vibration , signal (programming language) , computer science , interval (graph theory) , engineering , artificial intelligence , mathematics , acoustics , electrical engineering , statistics , linguistics , philosophy , physics , combinatorics , programming language
Aiming at the problems of slow feature extraction and poor real-time diagnosis of some mainstream state identification methods in actual operation, a fast feature extraction method of vibration signal interval is proposed to identify the energy storage state of circuit breaker. Firstly, according to the kurtosis wavelet modulus maxima, the starting point of the energy storage state of the circuit breaker is detected. Then, the signal envelope sum is extracted as the feature vector by marking the obvious range of envelope amplitude difference through KS test, and the feature is filtered and dimensionally reduced by the releiff-sfs method to get the optimal feature subset. At last, the fuzzy c-means clustering (KFCM) is used to pre classify the features to obtain the optimal hyperplane with the least risk, and support vector machine (SVM) is used to establish the training model for state identification. The experimental results show that the energy storage state identification algorithm proposed in this paper only needs 0.2S to extract features on the premise of ensuring the accuracy, which is of great research value for on-line monitoring of circuit breakers.