DPMU-based Event Classification in Microgrids Using Time Domain and Spectral Features of Limited Measurements

A microgrid can be subjected to various unexpected events, such as sudden loss of generation/ load demand, faults, maloperation of capacitor banks, etc. To take appropriate control or remedial actions, the microgrid management system (MGS) should detect and classify an event. Due to economic reasons, all buses and lines in a microgrid are not equipped with measuring devices. Thus, detecting and classifying an event in a sparsely monitored microgrid is a challenging task. This paper proposes an event detector and classifier using measurements collected from few distribution phasor measurement units (DPMUs) connected only at the generator buses. The proposed method is a multi-step process. Firstly, a DPMU measurement based linear state estimator is used to estimate the voltage and current phasors for all the buses in the network. The second step uses the change in the estimated voltage phasors during an event to select a few candidate buses which are further analyzed to classify the event. An offline process is used for training the event classifier, where several time domain and spectral features are extracted from the estimated change in voltage and currents phasors at the candidate buses. In the third step, a neural network is trained using the selected features, which is used for event classification. The proposed method is validated on a 13-bus microgid system simulated using OPALRT hypersim real-time digital simulator.