Identification of potential islanding initiators in radial feeders with high solar PV penetration for preemptive preparedness

Reverse power flow is one of the anticipated implications of integrating large amount of distributed solar photovoltaic power with distribution feeders. Such an occurrence can transform into anomalies that can potentially affect the integrity of the network. Unintentional islanding is one such issue that has not been comprehensively addressed, with the consensus of all the stakeholders of distribution networks. It still remains a cause of concern among utilities, arising out of large penetration of distributed solar power. In this work, an anomalous over-current spike, attributed to reverse power flow was discovered on a radial feeder modeled with emulator hardware. Load-inverter interaction alongside grid-end disturbances in a high penetration scenario resulted into the anomaly that was found to be an islanding initiator. This paper uses suitable machine learning models to detect such occurrences so as to preemptively prepare and respond to any imminent islanding condition. Results of offline testing on a dedicated Raspberry Pi microcomputer have been presented. This paper presents a preliminary analysis of the feasibility of such a framework to be carried forward for online testing on hardware-generated data.