Research On Tested Signal Processing And Fault Identification Method Of Circuit Breaker

Circuit breaker is one of the most important control and protection equipment in the power grid, judge the state change quickly and accurately is particularly important, non-contact measurement technology has unique advantages in signal acquisition and fault status recognition of the running equipment, being more widely used. However, the non-contact measurement process susceptible to interference, in which sound signal would be effected most serious by noise, such as: vehicle, trumpets and thunder, etc. Therefore, to accurately identify the operating status and failure risks of circuit breaker, the original signal should be pre-processed.Firstly, designed the overall circuit breaker test system based on non-contact idea, and proposed the requirement to the function of hardware and software. Then, analyze the transmission of two signals, vibration signal is affected by environmental slightly, use the appropriate filtering technology can achieve the purpose of de-noising. But for sound signal with the same route of transmission of noise effected seriously, simple filtering technology can not meet the needs, even affected subsequent analysis of the characteristics and identify of the fault condition seriously, cause wrongful convictions, miscarriage of justice. Therefore, this paper process sound signal of the breaker by blind source separation. That first estimated the number of source signals of the collected sound signal by K-means algorithm based on BWP indicators. And then combined with improved decomposition method none-underdetermined problem. Finally, process the signal with blind source separation by FastICA to obtain the breaker sound signal separated.Finally, analysis the collected signals in time-domain, calculate the spectral kurtosis after wavelet packet decomposition. Combined with wavelet packet decomposition,use ELM to classify the characteristics value of the signal. Through comparison of the classify results of several typical failure in laboratory, which before and after the blind source separation, the method proved it’s feasibility and effectiveness.Through the study of different signal preprocessing and fault identification method, we designed a new status recognition system of circuit breaker based on non-contact measurement.The system can acquisite several test signal, and can correctly identify the operating status of the circuit breaker, which lay the foundation for the circuit breaker fault diagnosis and condition maintenance.