islanding detection of synchronous generator in distribution network based on machine learning methods

In this paper, a novel approach for detecting islanding events in distribution networks special for synchronous generator type is presented. the proposed method leverages information derived from negative sequence voltage components, synchronous generator field voltage, positive sequence impedance variation rate, voltage harmonic distortion factor, and features extracted through wavelet transform applied to voltage waveforms. in order to establish a robust classification system without the necessity of explicit threshold determination, a pattern recognition method is employed. the dataset derived from these characteristics undergoes training using multi-layer support vector machines and a random forest optimization algorithm, resulting in five distinct classes. the study incorporates experimental samples encompassing various scenarios such as symmetric and asymmetric fault occurrences, load variations at different points, capacitor bank switching, variable load switching, nonlinear load switching, and islanding on a modified 34-bus ieee network. the proposed islanding detection method demonstrates its effectiveness in distinguishing electrical islanding from power quality phenomena such as voltage oscillation, voltage sag, voltage swell, and dynamic voltage changes. conducted simulations in matlab validate the efficacy of the proposed method.