An Intelligent Protection Scheme for Microgrids based on S-transform and Deep Belief Network

This paper develops a data-mining based intelligent protection scheme for fault detection in a microgrid with multiple distributed generations. The protection scheme uses one cycle of the three-phase post-fault currents at both ends of the feeders to derive some effective features. The retrieved current samples are pre-processed using S-transform to obtain statistical features, such as energy, standard deviation (STD), and median absolute deviation (MAD), which are further used to build a deep belief network (DBN) based model. The model is validated on the unseen data set for fault detection in the microgrid. The proposed relaying scheme is developed on a real time digital simulator (RTDS) platform which is integrated with Matlab. Using extensive test results, the performance of the proposed intelligent relaying scheme is evaluated for a microgrid with wide variations in operating conditions for radial and mesh topologies in grid-connected and islanded modes of operations. It is also shown that the proposed method outperforms the existing methods in the literature.