Study On On-line Monitoring Of Dynamic Characteristics And Fault Recognition Based On Vibration Signal For DL

The HV breaker is one of the frequent failure equipment in the electric network, its dynamic characteristic can reflect the running failure of the circuit breaker effectively. In view of the facts that current domestic and international dynamic characteristics of circuit breaker on-line monitoring features are single and rarely using advanced method to make an effective diagnosis about the circuit breaker running failure through monitoring the characteristic information, the selection of on-line monitoring dynamic characteristic sensor and characteristic test are studied in this paper. The design of circuit breaker trip characteristic on-line monitoring system and laboratory testing are implemented. According to the fact that the vibration signal—an important parameter reflecting the circuit breaker running failure is difficult to diagnose, a wavelet neural network fault identifying method based on vibration signal is studied.The on-line monitoring displacement sensor of circuit breaker trip characteristic is studied first. According to the feature that the cable-displacement sensor is not only fit for measuring the displacement of linear moving mechanical objects but also for measuring the displacement of curvilinear moving ones, the characteristic testing of cable-displacement sensor is carried out on a real breaker. The effect of different mounting positions to the sensor output characteristic is analyzed. And the best position for testing trip—time characteristic of displacement sensor is confirmed. By analyzing the displacement—time curve, the conclusion that this kind of sensor meets the requirement of monitoring breaker trip—time characteristic can be drawn.An on-line monitoring system of circuit breaker trip characteristic based on the G language graphical programming software LabVIEW is established. Function blocks of system data collection, analysis and storage are designed and debugged. And the monitoring and analysis of breaker trip characteristic is implemented. The stability and accuracy of the system is verified by comparing the trip-time characteristics of 110kV SF6 real breaker.According to the similar characteristics of vibration signal of the same type circuit, the breaker running failure condition can be judged based on the characteristic information in vibration signals. The vibration signal features is sorted and processed by using wavelet energy combined with neural network. By using the wavelet packet decomposition and reconstruction principle, the vibration signal is decomposed to the different frequency bands. Energy of each band is extracted and formed to eigenvector, being the input vector of RBFNN. And the RBFNN is used as a classifier. The breaker fault is identified and sorted based on the vibration signal. The validity of this method is proved by simulating test.