Intelligent islanding detection of a synchronous distributed generation using governor signal clustering

Detection of intentional and unintentional islanding of distributed generation units is one of the major protection issues of the distribution networks. Regarding the safety and reliable operation of a modern distribution network, an expert diagnosis apparatus is required to distinguish network cut off from variety of normal occurrences. Automatic load-frequency controller (ALFC) is an indispensable component of the synchronous generators. Simulation results show that input signal of the governor includes somewhat singular characteristics for each possible phenomenon or disturbance. Therefore, a new method based on Self-Organizing Map (SOM) neural network is proposed using input signal to the governor to cluster various occurrences into islanding and non-islanding categories. Simulation results presented in this paper shows that the input signal of the governor employed by a SOM can cluster a majority of occurrences of the system and distinguishes the islanding phenomenon with high confidence.