Anti-islanding detection methods for grid-tied inverters

Islanding is the condition where a section of distribution network keeps running on local generation after it has been disconnected from the main grid, without anyone intending it to. For a homeowner's PV inverter this sounds almost benign — the lights stay on during a fault — but it is one of the most tightly regulated behaviours in distributed generation because it directly threatens the safety of utility line workers who reasonably expect a de-energised line to stay de-energised. This article explains how inverters detect islanding, why detection is genuinely difficult in some circumstances, and what installers and engineers need to verify to be confident the protection is actually working rather than just present on a datasheet.

By the end you should understand the difference between passive and active detection methods, why every active method involves a trade-off called the non-detection zone, how standard test circuits emulate a worst-case island, and the common commissioning mistakes that let a genuinely non-compliant inverter pass a superficial check.

Why islanding is hard to detect

If a PV inverter's local load and local generation are badly mismatched, islanding is trivial to detect: voltage or frequency drifts rapidly out of bounds the instant the grid connection opens, because there is no longer a stiff external source holding those values steady, and the ordinary overvoltage, undervoltage, over-frequency and under-frequency protections described in the companion article on grid codes and inverter protection settings will trip within their normal time bands.

The hard case is when local generation and local load happen to be closely matched in real power, reactive power, and frequency at the instant of disconnection — for example a small commercial site with rooftop PV sized close to its daytime baseload, on a circuit with power-factor-correction capacitors. In that scenario, voltage and frequency can remain within normal protection limits for a surprisingly long time after the grid connection opens, because nothing is pulling them away from nominal. This is the scenario every dedicated anti-islanding method is designed to catch, and it is why passive protection alone is considered insufficient by most grid codes.

Passive methods

Passive methods monitor existing electrical quantities without injecting any disturbance into the system. They are simple, fast, and free of side effects, but they have blind spots.

• Over/under voltage and frequency (OVP/UVP, OFP/UFP): the baseline protection already described in the grid codes article; effective whenever the island is significantly unbalanced, ineffective in the matched-load case.
• Voltage phase jump detection: monitors the instantaneous phase angle between voltage and current; when the grid disconnects, the phase relationship the inverter was tracking can jump abruptly because the load impedance, not the grid, now sets the phase reference. Effective against some load types, weak against resistive loads.
• Harmonic distortion monitoring: watches for a step change in total harmonic distortion of voltage, since the grid's low source impedance normally suppresses inverter-injected harmonics, and losing that stiff source lets harmonic voltage distortion rise. Sensitive to background harmonic noise from other equipment.
• Rate of change of frequency (ROCOF): measures how fast frequency is moving, not just its absolute value; a sudden loss of grid stiffness often produces a fast frequency swing even when the instantaneous value stays within static limits. Widely used but can nuisance-trip during legitimate grid frequency events on weak networks.

Active methods

Active methods intentionally perturb the inverter's output in a small, usually imperceptible way and watch how the system responds. If the grid is present, its low impedance absorbs the perturbation with negligible effect. If the island has formed, the perturbation has nothing to work against but the local load, and its effect accumulates and becomes detectable.

Most commercial inverters combine a frequency-shift based active method with the standard passive protections, because the active method closes most of the non-detection zone left by passive protection alone, at the cost of a very small, generally unnoticeable degradation in power quality.

The non-detection zone concept

Every anti-islanding method has a non-detection zone (NDZ): a region defined by the mismatch between local load and local generation, in both real and reactive power, within which the method fails to detect islanding within the required time. Standards such as IEC 62116 define a specific worst-case test circuit — a parallel RLC load tuned to resonate at the nominal frequency and sized to closely match the inverter's rated output — precisely to probe the edges of this zone rather than test only easy, obviously unbalanced cases.

Worked example: reading a compliance test result

A test lab configures the IEC 62116 test circuit with a resistive load matched to the inverter's rated real power output and reactive elements tuned so that the RLC circuit's resonant frequency sits at the nominal grid frequency, with the quality factor set to the standard's required value (commonly Qf = 1.0, meaning stored reactive energy equals real power dissipated per radian).

1. The utility breaker opens, isolating the RLC-load-plus-inverter combination from the grid.
2. The lab measures the time from breaker opening to the inverter ceasing to energise the local circuit.
3. IEC 62116 requires this to occur within 2 seconds under the worst-case tuned condition.
4. Suppose the recorded trip time is 0.6 seconds. This passes comfortably, indicating the active method is engaging well before the passive protections would have needed to intervene.
5. If the recorded time were, say, 2.4 seconds, the unit fails, even though 2.4 seconds may sound short in absolute terms — the standard's whole purpose is to force a decisive result inside the specified worst-case window, and any excess is a fail regardless of magnitude.

What goes wrong in practice

The most frequent real-world problem is not a single non-compliant inverter but multiple compliant inverters on the same feeder interacting badly. Some active anti-islanding schemes that rely on positive feedback can partially cancel each other's perturbation signal when several units of different vintages or manufacturers share a circuit, widening the effective non-detection zone beyond what any single unit's test certificate would suggest. This is difficult to catch in single-unit laboratory testing and is an active area of grid integration research referenced by IEA PVPS studies on high-penetration PV feeders.

A second common issue is retrofitted power-factor-correction capacitor banks on commercial sites, installed for reasons unrelated to PV, which happen to shift the site's load impedance toward the tuned RLC condition the standard tests against. An inverter that has operated for years without incident can develop a genuine islanding risk after unrelated equipment changes elsewhere on the site.

A third issue is confusing "certified to IEC 62116" on a datasheet with "correctly configured for this installation." Firmware settings that disable or relax active anti-islanding — sometimes offered as an option for isolated microgrid or off-grid-capable hybrid inverters — must never be left enabled on a grid-tied installation, and this is exactly the kind of setting that a rushed hybrid-inverter commissioning can leave misconfigured.

Finally, testing anti-islanding on a live site by simply opening the main breaker and eyeballing the inverter's response is not equivalent to a laboratory RLC test and should not be treated as compliance evidence; it confirms the easy passive case, not the tuned worst case the standard actually targets.

Standards and references

IEC 62116 is the primary international test method for anti-islanding of grid-connected PV inverters and defines the tuned RLC test circuit and the maximum trip time discussed above. IEEE 1547 incorporates equivalent unintentional-islanding requirements for the North American context, and UL 1741 references IEEE 1547 for product safety certification of inverters sold in that market. EN 50549-1 references anti-islanding requirements as part of the broader European connection requirements for generators up to low voltage, tying protection settings and anti-islanding together in a single compliance package that installers should treat as inseparable.

For how anti-islanding sits alongside the wider connection process, see grid connection and export rules explained and the grid application checklist.

Frequently asked questions

Can islanding happen with a single small residential inverter?

It is much less likely than with commercial-scale systems, because a residential inverter's output rarely matches the site's instantaneous load closely enough, in both real and reactive terms, to sit inside the non-detection zone for long. It is not impossible, which is why anti-islanding protection is mandatory regardless of system size rather than being scaled back for small installations.

Does a battery inverter need anti-islanding protection too?

Yes, whenever it is capable of exporting to or supporting the public grid. A hybrid inverter that can also run an off-grid backup circuit needs a clear, tested transition between grid-tied mode, where anti-islanding must be active, and an intentional islanding mode with proper isolation, where it is switched off by design because the system is deliberately separated from the grid, not accidentally so.

Why do some inverters trip faster than the standard requires?

Manufacturers often build in margin, tripping well inside the maximum allowed 2 seconds because faster disconnection reduces the risk of edge cases the standard test circuit does not fully capture, and because a faster response looks better in comparative third-party testing. Tripping fast has essentially no downside for the installation as long as it does not cause nuisance trips during normal grid events.

Is active anti-islanding noticeable in the power output?

In a correctly functioning grid-tied installation, no. The perturbations used by active methods are small enough that the grid's stiff voltage source absorbs them without any measurable effect on power quality or harmonic content at the point of connection, well within the limits discussed in the companion article on harmonics and power quality.

How is anti-islanding tested after installation, not just in the factory?

Most jurisdictions accept the inverter's type-test certificate, issued by an accredited lab against IEC 62116 or the equivalent regional standard, as sufficient evidence at the individual installation level; they do not require a site-specific RLC test. What is checked on site is that the correct, unmodified firmware and settings are in force, typically via the settings printout described in the companion article on grid codes and inverter protection settings.

What is the relationship between anti-islanding and curtailment?

They are unrelated in mechanism but sometimes confused because both can stop an inverter from exporting. Anti-islanding is a safety protection triggered by loss of the grid connection; curtailment is an operational instruction from the network operator or a market signal, discussed in the companion article on curtailment and negative electricity prices, that reduces or stops export while the grid connection remains live and healthy.