Open Circuit (OC) and Short Circuit (SC) IGBT Switch Fault Detection in Three-Phase Standalone Photovoltaic Inverters Using Shallow Neural Networks

This paper presents a reliable IGBT open-circuit and short-circuit switch fault detection technique for a standalone photovoltaic two-level inverter using a shallow neural network. After applying the extended park vector approach (EVPA) to the three-phase currents at the load terminal, temporal features were extracted. Afterwards, exploratory analysis of these features was conducted via Principal Component Analysis (PCA) technique. Following an ablation study and normalization of the data, the feature-set was split into training, validation, and test sets that were used as the inputs to a variety of neural and non-neural-based classifiers. A shallow neural network with tangent-sigmoid activation that exhibits the property to output probabilities and its class memberships proved to have the best test-set accuracy compared to others.