Every grid-connected inverter has to disconnect when the grid goes away. The requirement is universal, the reasoning is sound, and the settings that implement it cause a surprising amount of operational trouble.
An island forms when a section of network becomes isolated from the main supply but remains energised by a local generator. If your solar continues supplying a circuit the utility believes is dead, several bad things follow.
The safety case is the primary one: line workers may approach an isolated conductor expecting it to be de-energised. This alone justifies the requirement.
There are technical reasons too. An island has no reference for voltage and frequency control, so both drift. Equipment on the island can be damaged. And reconnection to the main grid out of synchronism can cause severe mechanical stress on rotating plant and damage to equipment.
Passive methods monitor voltage and frequency at the connection point and trip outside defined windows. Straightforward, but there is a detection blind spot: if local generation happens to closely match local load, voltage and frequency may stay within limits for a period even after the grid is gone.
Active methods address that. The inverter continuously injects a small perturbation – a slight frequency shift, a small reactive power variation – and observes the response. Connected to a stiff grid, the perturbation is absorbed with no measurable effect. In an island, it causes a detectable drift, which trips the inverter.
Certified inverters implement these to standards such as IEEE 1547 and UL 1741 in North America, or EN 50549 and national requirements in Europe. Testing verifies detection within a specified time, typically around two seconds.
Protection settings are a compromise. Tight settings detect genuine islands quickly but trip on normal grid disturbances. Loose settings ride through disturbances but risk slow island detection.
Common causes of repeated tripping:
Get the inverter event log. Modern units record the trip reason and usually the measured value that caused it. That single step distinguishes overvoltage from underfrequency from a genuine loss-of-mains event, and the remedies are entirely different.
If it is overvoltage, and it correlates with high generation and low site load, the cause is usually voltage rise from your own export. Options include tap changes on the transformer, reactive power absorption by the inverters to counteract the rise, or negotiating a revised set point with the network operator.
Do not simply widen the trip settings to stop the alarms. Those settings are usually a condition of your connection agreement, and altering them without approval can breach it. Where a change is genuinely appropriate, the operator can authorise it.
Recording this at handover costs nothing. Reconstructing it three years later, when nobody remembers who configured what, costs a service visit and an argument.