Application of Computer Intelligence Method to Detect Islanding Phenomena in Microgrid Network

Islanding is common operational issues experienced by the electrical microgrids, which can be perceived as a situation in which distributed generators (DGs) continue to supply local loads, while larger part of the electric power system loses power or when the grid power is unavailable. This can lead to abnormal operating voltage and frequency within the microgrid and can also prove to be dangerous for the operating personnel. The IEEE 1547 standard requires that such loss of grid connection be detected within two seconds and immediately trip the DGs to avoid the risk of such voltage and frequency limit violations. This paper presents a simple yet practical scheme for islanding detection in microgrid systems harnessing the capability of machine learning (ML) and computational intelligence (CI). The proposed scheme utilizes important features extracted from the microgrid's operating variables using Principal Component Analysis (PCA) followed by Logistic Regression based pattern classification method to differentiate between islanded and non-islanded operating conditions. Case study on a 3-DG based microgrid system demonstrates that the proposed scheme is simple and yet effective in quickly detecting islanding situation in practical microgrids.