A Multi Feature Based Islanding Classification Technique for Distributed Generation Systems

Now a days, distributed generations (DGs) are widely adapted for effectively lowering environmental impacts, improve the safety of power supply and decrease the pressure on power supply during peak hours. However, islanding operation is the most important concern when using the DGs. In order to effectively detect islanding after DG disconnects from main source an efficient islanding detection technique is required. To develop such algorithm, the author observed the various issues associated with single phase grid connected PV systems during islanding condition and their detection methods. The main objective is to develop a trained data set, which is capable of detecting all the possible scenarios of islanding. An accurate representation of the single-phase grid connected PV system is simulated in MATLAB/Simulink environment to determine the islanding. Wavelet transform and Machine learning techniques were adapted to develop the trained data set. The results depicted a 100% accuracy in both training and testing conditions with 18.9 seconds training time and 20 milli seconds detection time.