Automatic load shedding (UFLS/UVLS) — how a grid protects itself against a cascading blackout

The frequency-protection (ANSI 81) and ROCOF guide covers how underfrequency protection (81U) and the rate of change of frequency (ROCOF, 81R) detect that a grid has lost part of its generation. This article covers what typically happens next, at the level of a whole grid's defence plan: automatic load shedding, split into under-frequency load shedding (UFLS) and under-voltage load shedding (UVLS) — the last automatic line of defence before a local imbalance turns into a cascading, wide-area blackout.

Why load shedding exists at all

A power system stays in balance only as long as generation and demand match, second by second. When a large generation deficit occurs (loss of a large power station, loss of an interconnector), frequency starts to fall across the entire synchronously coupled grid. If nothing intervenes, falling frequency eventually causes generators themselves to trip on their own under-frequency protection — removing even more generation and accelerating the decline further, a cascade that can end in a total blackout. Automatic load shedding intervenes before that point, by deliberately disconnecting a controlled amount of load so the remaining generation and remaining demand can rebalance.

Under-frequency load shedding (UFLS)

UFLS relays are distributed across a grid's load feeders and trip in stages: each stage sheds a defined percentage of load once frequency drops below its assigned threshold, usually with a short, deliberate time delay so a brief, self-recovering dip does not trigger unnecessary shedding.

Note: the exact thresholds, number of stages, and shed percentage per stage are set by the grid operator's system defence plan (in Europe, following the ENTSO-E emergency and restoration principles; in North America, NERC standard PRC-006), not by this article — the table below illustrates the pattern, not a universal setting.

Illustrative stage	Frequency threshold	Typical shed per stage
1	≈ 49.0 Hz	5–10 % of load
2	≈ 48.8 Hz	5–10 % of load
3	≈ 48.6 Hz	5–10 % of load
4 (last resort)	≈ 48.0 Hz	remaining defined block

Each successive stage sheds more load at a lower frequency, so the response is proportional to the severity of the deficit rather than an all-or-nothing action.

Under-voltage load shedding (UVLS)

UVLS addresses a different failure mode: voltage instability, which mainly threatens grids that are constrained in reactive power rather than active power. As voltage sags under heavy reactive loading (motor load stalling, tap-changers hunting for more voltage and inadvertently pulling more reactive current), the system can enter a self-reinforcing voltage collapse that frequency protection does not see coming — the frequency itself may still be close to normal while voltage is disintegrating locally. UVLS relays shed load once voltage stays below a set threshold for a set time, deliberately using a time delay long enough to ride through a transient voltage dip from a nearby fault without shedding load unnecessarily.

UFLS/UVLS versus ROCOF: absolute threshold versus rate of change

Ordinary UFLS/UVLS reacts only once frequency or voltage has actually crossed an absolute threshold — inherently reactive. ROCOF (as covered in the linked guide) reacts to the speed of frequency change and can therefore trigger earlier for a severe, fast-developing deficit. In a well-designed defence plan the two work together: ROCOF (or a fast first UFLS stage) catches a severe, rapidly developing deficit early, while the remaining UFLS/UVLS stages handle a slower, developing imbalance in a graduated way.

Why an islanded or small grid needs different settings

A small, islanded grid (a factory microgrid, a greenhouse combined-heat- and-power installation temporarily separated from the public grid, or an industrial site running on backup generators) has far less rotating inertia than a large, interconnected national grid. The same relative generation deficit therefore causes a much faster frequency decline on a small island than on a large grid — copying UFLS/UVLS settings tuned for a large synchronous grid onto a small island typically reacts far too late, because the island's frequency has already fallen well past the first threshold before the intended stage even has time to measure a stable value.

Automatic restoration: not simply the reverse of shedding

After a shedding event, reconnecting the shed load automatically and immediately is not simply "undoing" the shedding action: restoring a large block of load at once can itself cause a renewed frequency dip (from the sudden new demand) or excessive inrush current from simultaneously re-energising many loads. Restoration is therefore typically staged, delayed, and in many defence plans deliberately left to manual dispatcher action rather than fully automatic reconnection.

Practical relevance

When designing the protection settings of an islandable installation (on-site generation with mains parallel operation and an island mode, or a private grid with its own generation), the UFLS/UVLS thresholds and delays must be derived from the installation's own inertia and load profile, not simply copied from the public grid operator's defence plan — and coordinated with any ROCOF-based islanding detection already in place, so the two functions do not conflict or duplicate the same shedding action.

Common mistakes

Applying public-grid UFLS/UVLS settings unchanged to a small, low-inertia island — the frequency (or voltage) decline is typically much faster there, so the same thresholds react too late.
Setting UVLS time delays too short — this causes nuisance load shedding on an ordinary, self-clearing voltage dip from a nearby fault rather than an actual developing voltage collapse.
Automatically and immediately restoring all shed load in one step — this risks a renewed frequency dip or excessive inrush, undoing the benefit that staged shedding just achieved.
Treating ROCOF and UFLS as redundant duplicates instead of complementary functions — without coordination, both can shed load for the same event, removing more load than the defence plan intended.