Fault Detection For DC/DC Converters Based On HMM

Switching power devices are widely used in modern electrical equipment. As a core component of the devices, DC/DC converter plays an important role in maintaining the reliability of the equipment. It is demanded that multiple faults can be accurately detected and diagnosed in a converter.In the thesis, hidden Markov model (HMM) based fault diagnosis technique is proposed for the converter, in particular, for the degeneration of a boost converter and the failure of a full bridge converter. By comparing of the four commonly used diagnosis methods, namely HMM, wavelet-HMM, neural networks and support vector machine (SVM), it is shown that the proposed method needs less training samples, while has higher identification accuracy. Therefore it has great potential for practical applications,The thesis is organized as follows. The basic principle of the DC/DC converter is first introduced, including the classification of the DC/DC converter and the commonly used identification method of fault diagnosis. The basic theory and algorithms of the HMM are then reviewed. The material steps of modeling and faults diagnosis of HMM methods are propounded. The application results of the proposed method are demonstrated for diagnosis of the degeneration of a boost converter and the failure of a full bridge converter. Then the above mentioned four techniques are compared and analyzed. Finally showed the DC/DC converter fault diagnosis system based on HMM. The overall design and the main function of the system are introduced.