In short. Load shedding is the deliberate disconnection of lower-priority loads so that the generators keep the loads that matter when capacity falls short. It is triggered by a generator trip, a reserve margin falling below a threshold or under-frequency, and it acts in stages from the least important load upwards. Restoration is a separate, slower sequence that returns load only when capacity is confirmed. A plant without shedding logic loses everything when the last set trips.
Why a generator plant needs it
A diesel plant has a fixed amount of capacity online at any moment. If a set trips, that capacity falls instantly and the remaining sets are asked to carry a load they may not be able to hold. Frequency falls within a second, the governors go to full fuel, and if the overload exceeds what the engines can absorb the remaining sets trip one after another. The whole facility goes dark, including the loads that justified the generators. Load shedding interrupts that sequence by removing enough lower-priority load within a few hundred milliseconds that the remaining sets stay within their capability. In a paralleled plant it is as important as the protection. The critical load shedding and priority control page sets the function in the context of the complete critical-power design.
Classifying the loads first
No shedding scheme can be designed until the facility has decided which loads can be interrupted, for how long and in what order. That classification is an operational decision for the plant owner, and it should be written down and signed. A kW figure says nothing about importance; a 15 kW effluent pump may matter more to a pharmaceutical licence than a 300 kW chiller. The usual outcome is three to five priority groups, each fed from feeders that can be tripped together, which often means re-arranging the switchboard so that priorities and feeders coincide. The critical load classification and prioritisation page offers a method and the load shedding priority planning tool turns the result into a stage table.
| Priority | Typical loads in an industrial plant | Shed | Restore |
|---|---|---|---|
| 1, never shed | Life safety, process safety interlocks, control power, UPS input for critical IT | Never | Not applicable |
| 2, shed last | Core production line, cold storage compressors, critical pumps | Only on second contingency | First, after capacity confirmed |
| 3, shed second | Secondary production, air compressors, process chillers | On reserve shortfall after stage 4 | Second, in blocks |
| 4, shed first | Comfort air conditioning, general lighting, canteen, water heating | Immediately on any trigger | Last, after all higher priorities are stable |
Triggers and how fast each must act
Generator breaker trip
The fastest and most reliable trigger is a contact from the generator breaker or controller showing that a set has left the bus. The master controller knows the load that set was carrying and the capacity that remains, and can trip the required feeders within 100 to 200 milliseconds, before frequency has fallen far. This event-based shedding should be the primary mechanism in any paralleled plant.
Reserve margin
Slower, measurement-based shedding acts when the spinning reserve, the difference between online capacity and actual load, falls below a set figure for a few seconds. It covers load growth and derating rather than sudden loss of a set, and should start another generator first, shedding only if none is available or the start fails. The load shedding automation and priority logic page describes how event-based and reserve-based logic are combined.
Under-frequency
Under-frequency relays on the feeders are the last line of defence, acting on the physical symptom when the controller has failed or the event was not anticipated. Stages are typically set between 49 Hz and 47.5 Hz with short definite-time delays, dropping the lowest priority first. Because a diesel plant has little inertia the stages must be tight; 48.5 Hz with 0.2 seconds for the first stage and 48 Hz for the second is a common starting point, to be confirmed against the engines’ capability.
Sizing the reserve (illustrative)
Reserve = Online capacity minus Actual load
Online capacity is the sum of the ratings of the sets on the bus and actual load is the measured kW. Consider three 800 kW sets carrying 1,900 kW, a reserve of 500 kW. If one set trips, capacity falls to 1,600 kW against 1,900 kW of load, an overload of 300 kW or 19 percent. Two sets can absorb a brief overload of that order but not for long, so the shedding scheme must drop at least 300 kW, and in practice around 400 kW to restore a working margin, within a fraction of a second. The priority 4 group totalling 450 kW is therefore the first stage. If the plant were instead carrying 2,300 kW, the loss of one set would leave an overload of 44 percent, beyond what any diesel engine can ride through, and shedding would need to remove both priority 4 and priority 3 in the same action. Working through these cases for every combination of sets online is how the stage table is built.
Restoration
Returning load is more dangerous than shedding it, because every reconnection is a load step and motors that were running now start from standstill at locked-rotor current. Restoration should wait until the standby set has synchronised or measured reserve has exceeded a threshold for a sustained period, then return load in blocks smaller than the shed stages with tens of seconds between them so frequency and voltage recover. Chillers and compressors with anti-recycle timers will not restart immediately anyway. Automatic restoration suits lighting and comfort loads; for production equipment many plants prefer the operator to confirm each block. The automatic load restoration and restart control page details the timing logic, and the black start guide covers the related problem of building load from a dead bus.
Testing and keeping the scheme current
A shedding scheme should be tested at commissioning by tripping a set under real load and confirming that the correct feeders open within the intended time, with frequency recorded throughout. Repeat the test annually and review the stage table whenever a feeder is added, a load changes priority or a generator is replaced. Settings and priorities belong in a controlled document with the same change authorisation as protection settings. A scheme unreviewed for five years is likely to shed the wrong loads.
Common mistakes
- Relying on under-frequency relays alone, which act after the frequency has already collapsed and often too late to save the remaining sets.
- Arranging priorities by kW rather than by consequence, so a small but licence-critical load is shed with the comfort loads.
- Restoring everything at once when the standby set comes online, producing a load step that trips the plant a second time.
- Never testing the scheme by tripping a loaded set, so the first real test is the first real outage.
What to send us
Send the load list grouped by feeder with the owner’s priority against each group, the number and rating of sets and the usual number online, a recent load profile or maximum demand record, the controller make and model, and the single-line diagram of the main switchboard. Note any loads with restart constraints such as chillers or furnaces. Use the application engineering enquiry and we will propose stages, triggers and a restoration sequence.
Questions engineers ask
How fast must load shedding act on a diesel plant?
Event-based shedding triggered by a generator breaker opening should trip the first stage within 100 to 200 milliseconds. Diesel sets have little inertia, so frequency falls quickly after a loss of capacity, and shedding that waits for a frequency measurement to confirm the event is often too late to prevent the remaining sets tripping.
What frequency should the under-frequency stages be set at?
Typical first stages sit around 48.5 to 49 Hz with a delay of 0.1 to 0.3 seconds, with further stages down to about 47.5 Hz. The exact values depend on the engines’ overload capability and on the frequency tolerance of the loads, and should be confirmed with the generator supplier and recorded in the setting file.
Does a single generator installation need load shedding?
Usually a simpler form of it. A single set cannot lose a partner, but it can be overloaded by load growth or by too many motors starting together. A priority arrangement that keeps non-essential feeders off until the set is running and stable, and an overload stage that trips them if load exceeds a threshold, protects the set and the essential loads.
Can the load shedding scheme also start a standby generator?
Yes, and it should. The master controller that detects a reserve shortfall should call the next available set before shedding wherever time allows, and should shed only when no set is available or the start fails. Restoration is then tied to that set successfully synchronising.
Technical content reviewed by Srinivas Reddy, Managing Partner, Reddy Generators. Last updated 19 October 2026.