Fault Diagnosis Research Of Power Electronic Circuits Based On The Wavelet Transform Signal Processing

With the extensive application of the power electronic technology,the structure of power electronic equipment is more and more complicated,the scale is also more larger.At the same time,the fault problems of power electronic equipment are more and more prominent.Power electronic circuit as the core component of power electronic equipment,Its failure may lead to the failure of power electronic equipment and even the whole system,causing great damage and severe loss.In order to ensure the safe and reliable operation of the system,the fault circuit should be effectively diagnosed in time.The study of power electronic circuit fault diagnosis theory and methods receive more and more attention.The key problems of fault feature extraction and fault identification of power electronic circuits are studied in this paper,including the follows:(1)The fault forms of power electronic circuit are analyzed,and the simulation model of each fault is established.According to the characteristics of power electronic circuit fault and the noise interference in actual operating conditions,all fault information is superimposed noise to simulate the actual operation condition.(2)This paper proposed a method for fault feature extraction based on wavelet packets transform.The fault signals were denoised by wavelet optimized forecasting variable threshold method,and then the original fault feature vectors were extracted by wavelet energy spectrum.The dimension of the original fault feature vectors were reduced by the principal component analysis and the characteristics of new fault feature vectors were highlighted.Then the fault identification method based on improved BP neural network,which can identify the new fault feature,the ideal results of fault diagnosis are obtained,the effectiveness of the proposed method is proved by the simulation results.(3)Taking into account the complexity of the fault feature extraction based on wavelet packets transform.This paper proposed an novel method for fault feature extraction based on cross-wavelet transform in order to improve the fault feature recognition.The cross-wavelet transform method has the capability of anti-noise,thus it can extract the fault feature without denoising operation,It also analyzes two signal directly to obtain the amplitude and phase of the cross-wavelet spectrum,and combined them into the original fault feature vectors,then the final fault vectors areobtained by principal component analysis.Improved BP neural network identified the final fault vector to finish the fault diagnosis.The accuracy of the diagnosis was compared with the above-stated method,simulation results show the effectiveness of the proposed method and it has a higher diagnosis accuracy.