Research On Fault Diagnosis Method Of Power Electronic Rectifier

Power electronic components have been developed rapidly and used widely,electronic and electronic devices now has a profound impact in all aspects of industrial production and social life.The power electronic rectifier is the key part of the whole power electronic equipment.Once the power electronic rectifier fails,it may cause damage to the system and even threaten the safety of life and property.consequently,it is of great practical and economic significance to apply the automatic fault diagnosis technology to the intelligent and precise positioning.In this thesis,optimization of two key technologies for power electronics rectifier device feature extraction and fault identification.In the first place,the fault modes of three-phase fully controlled rectifier circuit are studied,and the failure modes and types are comprehensively predicted and analyzed.The fault simulation model is built through the MATLAB platform,and the common fault states are simulated to get the fault waveform.Secondly,using the Fourier transform and the three layer wavelet packet analysis respectively for feature extraction,the rectifier output voltage fault signal shows that the three layer wavelet packet decomposition to get the energy distribution of each frequency contains a large number of fault information,the fault feature extraction method is reasonable.In the end,extract the fault signal energy feature as the input vector of BP neural network and Elman neural network,the identification and diagnosis of various faults,namely fault classifier results show that Elman neural network diagnosis method can improve the training speed of the network based on the fault recognition accuracy,to achieve the desired results.In this thesis,a fault diagnosis method based on wavelet packet and Elman neural network is used to diagnose the fault of thyristor controlled circuit breaker.In summary,the combination of wavelet packet analysis and Elman neural network to extract the feature information of fault signal and fault classification is an innovation of this thesis.The experimental results show that the method has faster training speed and smaller test error.This is a new attempt in the application of fault diagnosis in power electronic rectifier,The generality of this method has some theoretical significance and practical value.