Protection · ROCOF / Loss of Mains

ROCOF & Loss-of-Mains Protection in a Low-Inertia Grid: The Nuisance-Trip vs Missed-Islanding Trade-off

As grids lose inertia, frequency protection is squeezed between two failure modes: nuisance-trip your generation, or miss a genuine island. How the elements work, why GB moved to 1.0 Hz/s, and where the line sits.

Published 2026-08-04 · EI Portal

Frequency protection used to have plenty of margin. On a low-inertia grid, it doesn't. As synchronous generation is displaced by inverter-based wind, solar and storage, the system's inertia falls — and frequency moves faster after any disturbance. That squeezes ROCOF and loss-of-mains protection between two failure modes, and understanding the squeeze is now core protection work.

Inertia sets the nadir

After a sudden generation loss, the frequency falls at a rate set, to first order, by the power imbalance and the system inertia:

RoCoF ≈ ΔP · f₀ / (2H)
      

Less inertia → a steeper initial rate of change of frequency → a deeper, faster nadir → less time for governors, fast-frequency response or under-frequency load shedding to arrest it. Everything downstream — your settings, your shedding scheme, your margin — is governed by that curve.

The scale of the effect is easy to see in a live model. In the Frequency Protection Suite's default islanding study (a mixed synchronous-and-inverter grid, 32% severity), dropping the inertia class from low to very low deepens the nadir from 48.83 Hz to 48.24 Hz and steepens the measured peak df/dt from 0.88 to 1.32 Hz/s — enough to carry the frequency into the load-shedding region. Raising it to high, the same event barely dents the trace: 49.47 Hz at 0.40 Hz/s. Same disturbance; the only change is stored rotating energy.

The protection elements

A frequency-protection scheme typically combines:

• ROCOF (df/dt) — trips on a fast rate of change of frequency; the classic loss-of-mains element.
• Underfrequency (UF) stages — e.g. 49.0 Hz with a 1 s delay, 47.5 Hz with 0.5 s.
• Overfrequency (OF) — e.g. 51.5 Hz.
• Under-frequency load shedding (UFLS) — staged load disconnection (a common reference around 48.8 Hz) to arrest a decline before it reaches the underfrequency trip stages.
• Loss-of-mains / vector shift — detecting that embedded generation has been islanded with a section of network.

The measurement chain matters as much as the pickups: df/dt measured over too short an averaging window is noise; over too long, the element is slow. Typical practice uses a window of a few hundred milliseconds with smoothing on top.

Failure mode 1 — too sensitive (nuisance tripping)

Set ROCOF too sensitively and a normal grid disturbance looks like a fault. On a low-inertia system the real df/dt during ordinary events is higher than it used to be, so an old-style sensitive setting picks up — and multiplied across thousands of distributed generators, a small grid event becomes a large, self-reinforcing loss of generation. This is exactly why Great Britain raised the RoCoF setting from 0.125 Hz/s to 1.0 Hz/s (with a time delay), through the Accelerated Loss of Mains Change Programme: the sensitive setting had become a system risk.

Worked live: set the pickup to 0.18 Hz/s and apply a benign weak-grid event — the kind that should ride through — and the element trips in 1.425 s, flagged by the model as a possible nuisance trip.

Failure mode 2 — too insensitive (missed islanding)

Push the settings the other way and you create the opposite problem. Loss-of-mains protection exists to detect islanding — a safety and power-quality requirement. A very insensitive ROCOF and low underfrequency stages can leave a genuine island with a small power imbalance undetected.

Worked live: at a pickup of 8 Hz/s with the underfrequency stage dragged down to 45 Hz, a genuine islanding event sails through 7.12 Hz/s below pickup — neither element asserts, and the model flags a possible failure to detect islanding. That's the security-versus-dependability tension: every loss-of-mains setting is a deliberate position between these two failures.

The source mix matters too

"Low inertia" isn't a single derating factor — the type of source changes the shape of the response:

• Grid-following inverters largely ride the excursion, and under a voltage fault they hit their current limit and saturate (a synchronous machine pushes through). In the worked model, a grid-following-dominated mix crashes the nadir to 47.73 Hz with the security margin flagged as tight.
• Grid-forming inverters and battery fast-frequency response actively arrest the fall: the same event holds at 49.45 Hz, restrained.

Reasoning about a real scheme means modelling these separately, not lumping them into one number.

See the trade-off

The Frequency Protection Suite simulates generation-loss, islanding and weak-grid events across inertia classes and source models, then evaluates the ROCOF, UF, OF, UFLS and loss-of-mains elements — flagging both the nuisance-trip and the missed-islanding cases on the same model, with the frequency trace, the df/dt slope and each element's decision. It also reads the same scenario across four relay vocabularies — ABB Relion, MiCOM, Siemens 81R and SEL 81D — in each manual's own terms.

The dynamics are an illustrative scenario model (showing how the frequency response moves with inertia and source mix), paired with rigorous protection-element logic. It is not a calibrated EMT/RMS dynamic simulation and not a grid-code compliance verdict; the engineering judgement remains the engineer's. RoCoF and loss-of-mains practice follows the applicable grid code (e.g. UK G99).

Frequently asked questions

What is ROCOF protection?

ROCOF (rate of change of frequency, df/dt) protection trips when frequency moves faster than a set rate for longer than a set delay. It is the classic loss-of-mains element for embedded generation: a genuine island usually has a power imbalance, so its frequency runs away faster than the interconnected grid's ever would.

Why did Great Britain change the RoCoF setting from 0.125 to 1.0 Hz/s?

On a lower-inertia system, ordinary grid disturbances produce a higher df/dt than they used to, and the legacy 0.125 Hz/s setting was tripping distributed generation during normal events — turning small disturbances into large generation losses. The Accelerated Loss of Mains Change Programme moved the setting to 1.0 Hz/s with a time delay to make the element secure while keeping islanding detection.

What is under-frequency load shedding (UFLS)?

Staged, automatic disconnection of load blocks as frequency falls (a common reference around 48.8 Hz on a 50 Hz system), designed to arrest a decline before it reaches the underfrequency trip stages and cascades. Lower inertia gives UFLS less time to act, which is why the inertia debate and the load-shedding debate are the same conversation.

Why does low inertia make frequency protection harder?

Because RoCoF ≈ ΔP·f₀/(2H): halve the inertia and the same power imbalance moves the frequency twice as fast, deepening the nadir and shrinking every response window. The gap between a setting that nuisance-trips and one that misses an island narrows with it.

What is the difference between grid-following and grid-forming inverters here?

Grid-following inverters synchronise to the grid's waveform and largely ride a frequency excursion — and under a voltage fault they hit their current limit and saturate. Grid-forming inverters (and battery fast-frequency response) actively construct the waveform and arrest the fall, lifting the frequency nadir. The same event can trip on one source mix and stay restrained on the other.