Passive islanding detection in microgrids using artificial neural networks

This research focuses on modeling and simulating voltage control of passive islanding detections with distributed generation. This research presents how reactive power generation and/or absorption can be utilized to partake voltage control in medium voltage distribution through multi-microgrids for passive islanding detection with rule-based mathematical model and artificial neural network (ANN). With the increased emphasis on renewable energy, modern power grids have become more reliant to smaller distributed generation units. Unlike traditional power grids that rely on larger centralized units, the detection of islanding events in these grids is more complex. However, it is still important to maintain the connection to the power grid to maintain high stability of the system and is also important to deenergize the grid when an islanding event happens, to protect the workers that may be working on the grid to clear the cause of the islanding. Islanding detection methods are categorized into passive, active and communication based. Active methods detect islanding by measuring the influence of noise they add to the grid, to predict the size of the grid, which reduces the quality of the power provided on that grid. Communication-based methods are expensive and highly reliable on the communications infrastructure, which limits their application. Unlike existing methods, the proposed method of ANN relies on the instantaneous current and voltage values to detect any islanding with an accuracy of 97.89% precisely. These values are directly fed to the ANN, i.e., without applying any feature extraction, so that, faster and more accurate decisions are made. The proposed method is expected to combine the high accuracy of ANNs with the faster change in instantaneous values, so that, faster and more accurate detection can be achieved.