Fault diagnosis of three-level inverter based on wavelet analysis and Bayesian classifier

Abstract

This paper presents a fault diagnosis method for a neutral point clamped (NPC) inverter using multi-resolution wavelet and Bayesian classifier. T All possible open-circuit failures arising from one single power device are considered. Bridge voltages are used as measurement to classify the fault modes. The fault features are extracted from the bridge voltages by wavelet analysis, and then used as the inputs of a Bayesian classifier which identifies the fault modes. Simulation and experiments prove the feasibility of the diagnostic method and its advantages in good classification performance and high reliability.