A combined deep belief network and time-time transform based intelligent protection Scheme for microgrids

Abstract An intelligent fault detection and classification scheme for a microgrid (MG) with several distributed generations (DGs) is proposed. Time-time (TT) transform is used to preprocess one cycle of the post-fault current samples measured at both ends of feeders. Several statistical metrics are derived from the processed data and used for training a deep belief network (DBN). The intelligent relaying scheme is implemented on a real time digital simulator (RTDS) which is integrated with Matlab. The performance of the DBN-based protection technique is compared to three other data-mining models and conventional relays under different MG conditions such as grid connected/islanded network and radial/loop topology. Also, the comparison is presented for different fault types including shunt and high impedance faults. A comprehensive study on a 25kV IEC standard MG confirms that the presented relaying scheme has high accuracy, fast response time, and robustness to noisy measurements. The comparative simulations and test results show that the combined TT-DBN scheme is superior to the previous techniques in the literature.