Research On Fault Diagnosis Of Three-Level Inverter

With the comprehensive application of multilevel inverter in high voltage and large power field, its fault problem becomes very distinct. Recently, researches on three-level inverter fault diagnosis of home and abroad are relatively rare.According to theoretical and simulative analysis of power devices' faults in main circuit, two fault diagnosis methods based on neural network and support vector machine (SVM) for three-level inverter are studied respectively, to solve fault diagnosis problem of three-level inverter.Firstly, in terms of three-level inverter's topology and work principle, fault types of inverter's main circuit are analyzed in this thesis. The open circuit faults of power devices in main circuit are separated into nine types. Simulation analysis of these faults is implemented in MATLAB software, and three useful conclusions are made. Secondly, the fault diagnosis method based on neural network is studied. Magnitude and phase of the fault signal's harmonics are obtained as fault characteristic vector, by means of Fourier transform. A back-propagation neural network with three layers is used to identify the atypical faults. The expert rules are introduced into diagnosis algorithm to identify the typical faults, which lightens calculation burden of neural network and enhances speed and veracity of diagnosis. A satisfactory diagnosis effect is received. It is a creative point to apply neural network and expert rules in fault diagnosis of three-level inverter. In addition, the fault diagnosis method based on SVM is studied. According to wavelet packet transform theory, fault signal's energy in different frequency band is extracted as fault characteristic vector of classifier. Based on one-against-all method, the multi-fault classifier model is constructed. The simulation results show that the proposed method has excellent classification effect. Application of SVM method in three-level inverter fault diagnosis hasn't been seen in existing articles. Lastly, by designing an experimental platform of three-level inverter based on TMS320F240 DSP chip, all faults of power devices are simulated in the platform. Some fault waveforms are provided, diagnosis results primarily demonstrate feasibility and correctness of fault diagnosis method based on neural network.