Machine Learning Based Protection Strategy of Microgrid Using Synchrophasor Measurement

Microgrid protection suffers from several challenges when conventional protection schemes of distribution networks are used. To overcome the challenges, an intelligent microgrid protection scheme using an explainable machine learning (ML) model is proposed by processing the reduction of positive sequence impedance and change of Angle related to the direction of active power flow. Phasor measurement units (PMU)s are used to collect the voltage and current data at the line ends. Microprocessor-based relays utilize this data from the PMUs to compute the change of positive sequence impedance and angle related to the direction of active power flow due to a perceived fault. Based on the above data, a decision tree (DT) is then used to identify the section of the microgrid where the fault has occurred. Further, measurement is made of Euclidean distance, over a three-cycle sliding window, between the rms value of the rated line current and the square of the instantaneous line current. This distance and the magnitude of the zero-sequence current are fed to the machine learning (ML) algorithm to classify the fault. Extensive data collection is done for different sections of a sample microgrid, for a fault at both upstream & downstream of the DGs connected, at different distances on a line and with different fault impedances.