An efficient islanding classification technique for single phase grid connected photovoltaic system

Integration of renewable energy systems into low voltage buses leads to complexity for protection systems and unplanned islanding. Therefore, a need to reconfigure the typical protection configurations arises. In this paper an efficient islanding classification technique is developed to provide remote monitoring and operation of grid connected distributed generation systems. A simulation is developed by considering a 1kW grid connected photovoltaic system. The voltage and current measurements at the point of common coupling act as a main source of data for developing the classification algorithm. Wavelet transform is utilised for extracting features and creating a feature matrix for all possible scenarios of islanding. The feature matrix is subjected to machine learning approach for creating a trained data set which helps in classifying the islanding scenario. The results depicted 100% training and 97%testing efficiency under 0.2 seconds which is better when compared with the conventional methods and literature.