A Bayesian Passive Islanding Detection Method for Inverter-Based Distributed Generation Using ESPRIT

In this paper, a new passive islanding detection method for grid-connected inverter-based distributed-generation (DG) systems is proposed. A statistical signal-processing algorithm known as estimation of signal parameters via rotational invariance techniques is used to extract new features from measurements of the voltage and frequency at the point of common coupling as islanding indicators. The new features are defined based on a damped-sinusoid model for power system voltage and frequency waveforms, and include modal initial amplitudes, oscillation frequencies, damping factors, and initial phases. A set of training cases generated on the IEEE 34-bus system was used to train a naïve-Bayes classifier that discriminates islanding and nonislanding events. Cross-validation was used to evaluate the performance of the proposed islanding detection method. The results show that by using the new features extracted from ESPRIT, the classifier is capable of discriminating islanding and nonislanding events with an accuracy close to 100%.