Improved hybrid islanding detection using data fusion, adaptive back propagation neural network and support vector...

A new hybrid islanding detection method (IDM) is proposed in this study to enhance the precision and effectiveness of islanding detection in hybrid microgrids (HMGs), especially in light of the increasing integration of renewable energy sources (RES) into power networks. The IDM presents a novel data-driven approach that combines data fusion techniques with an adaptive backpropagation neural network (ABPNN) and support vector machine. By utilizing feature datasets such as the rate of change of phase angle difference (ROCPAD), intermittent-bilateral reactive power variation (IBRPV), and frequency, the IDM aims to accurately identify islanding events within HMGs. The approach can be summarized in two main stages: The data set for training the ABPNN is cleaned first by offline data preprocessing, using k-means clustering and logic operation techniques. Subsequently, the trained neural network is used to classify the online testing data into different categories by support vector machines so that islanding and non-islanding events can be identified in real scenarios in real time. The results of the proposed IDM show significant improvements in the accuracy of islanding detection, the rapid identification of islanding events, and the prevention of nuisance tripping occurrences. The IDM achieves a zero non-detection zone (NDZ) and exhibits minimal impact on power quality, making it a highly promising solution for islanding detection in HMG environments.