DC Microgrid Islanding Detection New Approach Based on Multi-Scale Standard Deviation and Optimize Deep Belief Network

The large-scale access of distributed generations (DGs) increases the difficulty of islanding detection of DC microgrids. The DC islanding detection methods are still in their infancy and have low detection accuracy rate in the disturbed environments. This paper applied the concept of deep learning based on Multi scale refined composite standard deviation fuzzy entropy (MRC-SDFE) into islanding detection of DC microgrid. The method combining the Variational Mode Decomposition (VMD) and the MRC-SDFE is used to extract the characteristics of voltage and current signals at PCC, which can effectively filter out interference. On this basis, a Deep Belief Network (DBN) detection model based on the Sea-horses optimizer algorithm (SHO) is proposed for the first time. The SHO is used to automatically optimize and adjust network parameters during the training process without manual adjustments. Finally, islanding and non-islanding disturbance events are tested. Through simulation verification, it is proved that the proposed method can availably guarantee the accuracy and rapidity of islanding and non-islanding disturbance detection in an interference environment.