Synchrophasor Data Driven Islanding Detection, Localization and Prediction for Microgrid Using Energy Operator

A new method is suggested for islanding detection in microgrids with huge penetration of non-dispatchable renewable energy sources. Time synchronised data is collected in the form of variables like voltage phasors, frequency and rate of change of frequency using phasor measurement units from various locations. The method then uses weighted difference principal component analysis for considering nonlinearities among variables and for data condensation. Multi-resolution Teager energy operator (MTEO) is employed to Q statistics obtained to detect islanding events. Further, to eliminate false alarms a difference of the obtained signal with its mean (MTEOD) is computed. When it is compared with a threshold which is derived based on upper control limit of Q, can successfully identify islanding events with no spurious detections. The method implemented in DIgSILENT/Power Factory and MATLAB achieved 100% accuracy, precision and reliability with zero non detection zone. Moreover, islanding location has been precisely identified to attain situational awareness of the microgrid. Immunity to noisy conditions in the microgrid has been demonstrated using a four-variable mathematical model. For a faster online prediction decision tree classifier is utilized. Real time performance and prediction capability was evaluated with decision tree technique which shows promising results.