Clustering governor signal of distributed generation for islanding detection

An important part of synchronous Distributed Resources (DR) is automatic load-frequency controller (ALFC). This paper introduces a new intelligent approach based on clustering of ALFC response with regard to input signal to governor. Self-organizing map (SOM) neural network is applied to identify and classify islanding and non-islanding phenomena. Output of SOM adds an extra binary control signal to the automatic voltage regulator (AVR) system to trip under voltage relay. Based on MATLAB/SIMULINK environment, performance of islanding detection is evaluated considering some common disturbances. The study results show that SOM neural network is able to identify and classify phenomena satisfactorily. In addition, the proposed technique eliminates negative effects of active methods on normal operation of DR.