부산대학교 도서관

Islanded microgrids play a crucial role in decentralized, managed power supply plans, providing an independent energy source separated from the main grid. To ensure their safe and steady operation, reliable and practical security components are essential. This paper focuses on the detection and classification of faults in an islanded microgrid using long short-term memory (LSTM) technique. The study involves the creation of an islanded microgrid comprising of solar PV, fuel cell, and battery storage systems. The LSTM model utilizes voltage and current data from the microgrid and to accomplish detection and classification, a network is employed, trained using symmetrical components of current and voltage as input. Based on this analysis, a suggested protection strategy is proposed. The investigation demonstrates the successful classification and detection capabilities of the LSTM methods within the considered system. In comparison to traditional protection tactics, the suggested method has a number of advantages and adaptability to changing system conditions. The results are validated in OPAL RT’s real time simulators. It also reveals that the results observed in MATLAB and Real time simulator are nearly same.