The neural network-based diagnostic method for atypical faults in NPC three-level inverter

Abstract

This paper presents a fault diagnosis method for a neutral point clamped (NPC) three-level inverter using BP neural network. As the fault mode occurring in one single power device has got a lot of attention, this paper focuses on the study of atypical faults. All possible open-circuit failures arising from two power devices on two cross bridge arms (atypical faults) are considered. Output voltages are used as measurement to classify the fault modes. The fault features are extracted from the output voltages by Fourier transform method and then used as the inputs of BP neural network which identifies the fault modes. Simulation results prove the feasibility of the diagnostic method and its good classification performance.