Insulation fault identification of vacuum circuit breakers based on improved MFCC and SVM

Insulation faults in vacuum circuit breakers can produce physical phenomena such as visible sound, ultrasound, and electromagnetic waves. Through the real-time collection and analysis of the discharge sound during the internal insulation failure of the circuit breaker, the insulation state of the circuit breaker can be judged. This article proposes a method. The improved MFCC algorithm is used to extract the characteristic parameters of the discharge signal. The recognition of the flashover discharge sound signal is achieved by one-class SVM. The one-class SVM is constructed, so one-class SVM can be used to identify whether the signal is a surface discharge or not. The characteristic vectors of the audio signal of discharge have intrinsic similarity, and the distribution can be concentrated, which is significantly different from other abnormal sounds. The conclusion is verified by experimental data. The calculation results show that the it can effectively identify the insulation state of vacuum circuit breaker by the method of using of MFCC feature extraction based on Fisher criterion and one-class support vector machine.