Fault Diagnosis for Power Electronic Inverters: A Data Driven Approach

Power electronic converters contain multiple semi-conductor switches and are susceptible to various types of faults during operation. Fault diagnosis is a critical task in such systems, particularly where reliability is a key requirement. In this paper, a data-driven fault detection algorithm is developed for open circuit faults, and its performance, design and limitations are analyzed. The proposed classification model is trained using features extracted from three-phase inverters, and a comparative analysis is conducted with various algorithms, including Random Forests (RF) and Extreme Gradient Boosting (XGBoost). The hybrid RF-XGBoost model gives around 99% accuracy on test dataset of 2 Level inverter and around 85% accuracy on test dataset of 3 Level inverter for the fault diagnosis.