Preparing Islanding Threshold Data for Machine Learning Algorithm

Under the smart grid paradigm, numerous independent operators oversee distribution systems; therefore, for situational awareness and protection, all dispatchers need to have a comprehensive picture of the network. This visualization is made possible by a protection type micro-phasor measurement unit (-PMU) mounted on the relevant distributed generation (DG) bus. The potential for accidental islanding, which jeopardizes utility personnel and hinders orderly reconnection, is one of the biggest issues with DGs. The heavily sampled 'Big Data' captured by phasor measurement units (PMUs) contains important information about the system's health. Most power system challenges, such as voltage stability, power system modelling, fault event monitoring, unintended islanding, state estimate, and so on, may be addressed by efficient and timely analysis of this data. In light of this, this paper recommends using PMU to detect inadvertent islanding in real time and presents a methodology for islanding threshold data preparation for artificial intelligence (e. g. , machine learning) algorithm. The discrete Fourier transform voltage and current phasors obtained from these PMUs are then processed using the Fortescue transform to compute the angle of sequence components. Under islanded conditions, the absolute angle difference between positive and zero components are employed to initiate signal for islanding. A modified IEEE 30 bus system is simulated by using Power World Simulator for data generation.