A Convolutional Hierarchical Generative Model for Microgrid Fault Diagnosis

This paper aims to develop an intelligent protection scheme for microgrids with a number of distributed generation units considering different modes of operation. The conventional computational intelligencebased shunt fault detection and classification approaches have shallow architecture and involve a huge number of trainable parameters that restrains the effective feature extraction. In this work, a hierarchical generative model is developed that fuses the benefit of the convolutional operation and the weight sharing mechanism which improves the feature extraction process as well as reduces the trainable parameters. Also, the fault data in transmission line domain is limited. The proposed method can able to dig out the most efficient feature from the limited training dataset. The results presented in this study confirm the high performance of the proposed framework.