Research On Fault Diagnosis Methods Of Multi-level Inverter IGBP

In the high-voltage、high-power applications, the multi-level inverter has been widelyapplied. Compared with the traditional two-level inverter, the multi-level inverter has thecomplex structure, the low reliability. In order to reduce the losses caused by the faults, ithas important significance to carry out multi-level inverter faults detection and diagnosisresearch.At present, multi-level fault diagnosis problems is less of a concern in the domesticand foreign countries. In this paper, using the diode-clamped (NPC) three-level inverter asthe example, studied power device of the multi-level inverter faults detection and diagnosistheory and method. Works is the analysis of the main IGBT switch open state operation andfaults modes, proposing the IGBT open circuit fault diagnosis method of the three-levelinverter that based on genetic algorithms and neural networks.In this paper, three-level inverter topology and working principle of the powerswitching device IGBT open circuit faults are classified.It uses the MATLAB platform tobuild a fault model; using the fourier transform to extract fault signal amplitude and phasecharacteristics to structure fault feature vectors; studying law expert rules, BP neuralnetwork, GA-BP neural network for the fault diagnosises and the discussion about thegroup’s advantage. Experimental results show that the diagnosis of the GA-BP neuralnetwork performs better than the traditional BP neural network such as convergence andaccurate diagnosis. Design the real-time online fault diagnosis system based on DSP(TMS320LF2407) for the control chip, to achieve fault diagnosis system hardware andsoftware circuit, and give the experimental verification programs for the fault diagnosissystem.