Preemptive mode-change triggering for inverters in response to unintentional islanding situations in high solar PV penetration distribution networks

Abstract

Unintentional islanding is one of the major concerns associated with integration of large solar photovoltaic generation with the power distribution network. However the issue of islanding needs to be comprehensively understood and managed given a challenging scenario ahead. This demands transition towards predictive approaches rather than the prevalent reactive ones for island detection and system management. The present work describes a machine learning based preemptive islanding detection module designed on a Raspberry Pi microcomputer for issuing advance alert to grid-interfaced solar photovoltaic inverters in the event of an imminent islanding situation and triggering the subsequent action. The module can activate a timely change in the inverter's operating mode by detecting precursors to accidental islanding, before the utility relay can trip a circuit breaker. Thus an intentional island can be maintained with acceptable power quality after the modechanging operation. The module can also issue a shut-down command to the inverter's control in order to completely avoid an unintentional island formation. Preliminary tests on an islanding precursor, discovered in simulation and on real laboratory-based network, have been presented. The precursor is simulated on a modified IEEE feeder and the online classification accuracy and speed indicate promising potential as a field-level prototype.