Power Electronic Research Of Fault Diagnosis Algorithms Of Convertor Circuits

This paper takes several typical power electronic convertor circuits as models and analyzes different objects with different fault diagnosis algorithms, which optimizing and simplifying every link of the traditional fault diagnosis techniques, completing the fault diagnosis algorithms of power electronic circuits, and improving its practicability.This paper firstly analyzes various algorithms which can be used for diagnosing faults of the convertor circuits and makes a comparison of these algorithms, and then determines the applicable objects of various algorithms according to the characteristics. Secondly, taking the three-phase bridge rectifier as object, this paper uses three sets of diagnosis algorithms to analyze the faults in detail. These three sets of diagnosis algorithms are:the combination of FFT and look-up table, the combination of wavelet pocket transform and logical classification, the combination of input current FFT and SOM network. Taking Dual-in-line12pulse Rectifier as object, this paper makes an analysis of fault diagnosis by using discrete waveform sampling and PNN network. Taking general convertor as object, this paper analyzes the fault diagnosis algorithms by combining the output three-phase current FFT processing and Elman network together. This paper makes an introduction of the application flow of various fault diagnosis algorithms in detail and proves the results by emulation and experiments. At last, this paper summarizes the work that has finished and makes a prospect of the research.The key of this paper is putting forward new methods of signal processing and fault recognition in the fault diagnosis algorithms of power electronic. Moreover, this paper put an emphasis on how to employ the most proper methods to analyze the fault according to its characteristic. This paper proposes the merits and demerits, the universality and the scientificalness of several diagnosis algorithms through a series of theoretical analysis, algorithm simulations and circuit experiments. The results of study have achieved the prospective goal of diagnosis, which can fulfill the real-time and banausic demands in the practical application. The result of this research provide some thoughts for the fields of fault diagnosis, model prediction, signal detection, pattern recognition, etc, which possessing both the theoretical and engineering values is looked forward.