Enhancing resilience of PV-fed microgrid by improved relaying and differentiating between inverter faults and distribution line faults

Abstract The present load demand and intermittency in the operation of photovoltaic (PV) sources has led to the wide usage of power electronic converters in PV integrated microgrids. A resilient microgrid operation requires early restoration to the healthy state post occurrence of faults in the converter or distribution line. The reported microgrid protection schemes aim at fast and accurate detection and classification of faults. Alongwith the intended protection task of fault detection and classification, the protection scheme designed for the distribution system should be selective to faults in the line only. However, imparting selectivity is quite often challenging because of the similar voltage-current profile during converter switching faults and symmetrical/unsymmetrical line faults. The similarity leads to nuisance tripping and unreliable operation of microgrid with classical protection schemes. In this regard, a Convolutional neural network (ConvNet) based protection scheme has been proposed for islanded microgrids to discriminate between inverter faults in the PV system and symmetrical/unsymmetrical faults in the distribution line, in addition to detecting/classifying the faults and identifying the faulty section. The voltage and current signals recorded at the relaying buses are converted into grayscale images to form the input dataset, which is utilized by ConvNet to derive the required relaying action. The performance of the proposed scheme has been examined and compared with Decision tree (DT) and Support vector machine (SVM) based microgrid protection schemes in terms of reliability indices i.e., dependability, security and accuracy for different scenarios. The proposed scheme has also been substantiated on OPAL-RT digital simulator to examine its feasibility for real-time field applications.