Open Switch Fault Diagnosis in T-Type Multilevel Inverter using Machine Learning for EV Applications

Abstract:
Conventional three-phase three-level T-type inverters (3L-T(exp2)I) are extensively employed in applications such as electric vehicles, PV inverters, active power filters, and UPS. In this work, a 3L-T(exp2)I is considered with a permanent magnet synchronous motor (PMSM) as the load. Due to the use of twelve power switches, these inverters are prone to a higher failure rate, making fault detection and diagnosis essential for reliable operation. This study addresses the diagnosis of open-circuit (OC) switch faults in 3L-T(exp2)I using a machine learning (ML)-based approach. The diagnostic method utilizes output voltage and current signals, from which two key features - root mean square (RMS) value and total harmonic distortion (THD) are extracted and subsequently normalized with respect to the healthy operating condition. Four ML algorithms, namely support vector machines (SVM), random forests (RF), k-nearest neighbors (k- NN), and decision trees (DT), are evaluated for fault diagnosis. The framework is implemented and validated in the MATLAB/Simulink environment. Among the models, the random forest classifier achieved the best performance with an accuracy of 99.89% using a 70:30 training-to-testing data split.