Fault Diagnosis of Three-Phase T-Type Multilevel Inverter Based on Variance Analysis

Multilevel inverters (MLIs) are extensively used in modern power conversion systems owing to their superior output quality and reduced harmonic distortion. However, the presence of a large number of power switches increases their susceptibility to faults, which can severely affect system reliability. This paper presents a fault detection and diagnosis framework for a three-phase, three-level T-type MLI (3 varphi-3L-T² MLI) based on Variance (Var) analysis of phase voltages and currents. The Var is first calculated from the raw signals and then normalized with respect to healthy conditions to minimize the influence of varying operating conditions. These normalized Var features are subsequently employed to detect faults and identify the corresponding faulty switch. Unlike computationally intensive artificial intelligence (AI) or model-based approaches, the proposed method offers a simple and efficient diagnostic scheme. The effectiveness of the technique is validated through MATLAB/Simulink simulations, demonstrating accurate and reliable fault detection and diagnosis.