In short. A load-bank test applies a controlled electrical load to a generator so its performance can be measured independently of the facility. It proves that the engine delivers rated power, that voltage and frequency hold through load steps and that cooling copes at full load, and it burns off the deposits left by light running. A test proves only what it was designed to prove, so the objective, steps and measurements are fixed before the load bank is connected.
What a load-bank test proves that a no-load run cannot
Most standby generators run weekly for fifteen minutes on no load. That confirms the engine starts and the alternator produces voltage, and nothing else. It does not show whether the engine can deliver rated power, whether the radiator and ventilation hold coolant temperature on a 42 degree afternoon, or whether the governor and AVR hold frequency and voltage when a block of load is thrown on. It also does harm: a diesel run for long periods below about 30 percent load never reaches operating temperature, unburnt fuel and carbon accumulate in the exhaust and turbocharger, and the engine develops wet stacking, which shows as oil weeping from exhaust joints and heavy smoke at the next real load. A loaded test proves rated capability, and an hour or two at high load is the cure for wet stacking. The load-bank sizing guide covers how much load bank a given set needs.
Define the objective first
Different objectives call for different tests, and a load bank on site does not by itself constitute an acceptance programme. The main objectives and what each requires are set out below.
| Objective | Typical profile | What must be measured | What it does not prove |
|---|---|---|---|
| Commissioning a new or relocated set | Steps of 25, 50, 75 and 100 percent, then 100 percent for one to two hours, often a short period at 110 percent where the manufacturer permits | Voltage, frequency, current, kW, coolant and oil temperature, exhaust temperature, fuel consumption, transient response at each step | Automatic transfer, behaviour with the facility’s motors and UPS |
| Annual proof test for a standby set | Build to 75 to 100 percent and hold for one to two hours | Same readings at intervals, with emphasis on temperatures reaching steady state | Transfer switch operation, which is tested separately |
| Wet-stacking correction | 75 percent or more for two to four hours until exhaust clears | Exhaust condition, oil pressure, temperatures | Anything about transient performance |
| Post-repair verification | Chosen to reproduce the original fault condition | The parameter that failed, recorded through the condition that caused it | Performance in modes not tested |
| Paralleling or transient investigation | Step loads applied and rejected, reactive load where kVAr behaviour matters | Recorded voltage and frequency traces, load sharing between sets | Steady-state capacity alone |
The percentages are of the set’s rated kW at the rating being tested. A standby-rated set tested at its standby rating is not being proven for prime duty, and the record should say which rating was used.
Resistive or reactive load
A resistive load bank loads the engine fully in kW at unity power factor but barely loads the alternator’s excitation system, because little reactive current flows. That is enough for engine capability, cooling and wet-stacking work. Where the objective includes the alternator and AVR, such as after a rewind, for a kVAr sharing problem on a paralleled plant, or for a facility whose load runs at 0.8 power factor with large motors, a combined resistive and reactive load bank is needed so the alternator is loaded at rated kVA and rated power factor. The motor starting on alternators guide explains why reactive load is where alternators are actually stressed.
Planning the connection and the site
During the test the load bank is the largest consumer on the generator, and the temporary connection has to be engineered like a permanent one. The cables are sized for the test current and the run length, the connection point is chosen so the set’s own breaker and protection are in circuit, and the cable terminations are made onto proper lugs and not clamped onto busbar edges. A 500 kVA set at full load draws around 700 A per phase at 415 V (illustrative), which is not a connection to make with jumper leads. The load bank itself rejects its full rating as heat, so it is positioned outside, clear of the generator’s air intake and radiator discharge, away from anything that cannot tolerate hot air, and with its own fan supply confirmed. The test window is agreed with the site because a sustained full-load run is loud.
Where the set is tested in its installed room, the test also proves the installation: record whether doors and louvres were in their normal operating position, because a test done with the room doors propped open proves nothing about ventilation in service. Where the load bank is connected ahead of the facility distribution, the test proves the machine but not the downstream cables, breakers or transfer switch; those are proven separately by an ATS test or a facility load test. See load bank rental for the equipment range.
Running the test and recording it
With the set warmed at no load, load is applied in the planned steps, and at each step the transient dip and recovery in voltage and frequency are noted before readings are taken at steady state. At full load, readings are taken every ten to fifteen minutes: phase voltages and currents, frequency, kW and power factor, oil pressure, coolant temperature, exhaust temperature if instrumented, battery charging voltage, ambient temperature at the air inlet, and fuel consumed. Coolant temperature is the figure to watch, because it keeps rising for thirty to forty-five minutes before it settles, and a test stopped at twenty minutes has not proved cooling. On completion, load is removed in steps rather than all at once, the set is cooled at no load and the exhaust is inspected. The report states the objective, the rating tested, the profile actually achieved, every reading with its time, any alarm or deviation, and what was outside the scope of the test.
Related photographs and videos
Common mistakes
- Running a weekly no-load exercise for years and calling it testing, which never proves capacity and leaves the engine wet-stacked for the day it is needed.
- Ending a full-load test after twenty minutes because the readings look fine, before the coolant temperature has reached steady state, so a cooling deficiency is missed.
- Connecting the load bank with undersized temporary cables or poor terminations, which overheat during the run and turn the test into an incident.
- Testing with the plant-room doors open and the louvres removed for access, then treating the result as proof that ventilation is adequate in service.
- Accepting a resistive test as verification of an alternator rewind or a kVAr sharing repair, which only reactive load can prove.
What to send us
Send the generator nameplate with its standby and prime ratings, the location and access for a load bank and cables, the available connection point and its breaker rating, the objective of the test (commissioning, annual proof, wet stacking or post-repair), and any previous test report. For paralleled sets, include the single-line diagram. Use the service and AMC enquiry form, or the generator rental enquiry if you only need the load bank supplied.
Questions engineers ask
How often should a standby generator be load-bank tested?
Annually is common practice for sets that otherwise run only on no load or light facility load, with a test of one to two hours at 75 percent load or more. Sets that regularly carry a substantial facility load during outages or scheduled tests may not need a separate load-bank test, provided the readings from those runs are recorded.
What load level fixes wet stacking?
Running at 75 percent of rated load or higher for two to four hours brings the exhaust system up to a temperature that burns off the accumulated deposits. Heavy smoke at the start of the run is expected and clears as the deposits burn. Severe cases may need a repeat run or turbocharger and exhaust inspection.
Can the facility load be used instead of a load bank?
If the facility can provide a known, stable load close to the set’s rating and the test can be run safely on generator for an hour or more, yes, and it has the advantage of testing the real distribution. Most facilities cannot provide full rated load on demand or hold it steady, which is why a load bank is used for capacity and cooling proof.
Should the test go above 100 percent load?
Some commissioning specifications include a short run at 110 percent of prime rating where the manufacturer’s data permits it, to demonstrate overload capability. Standby-rated sets are not designed for overload and should not be tested above their standby rating. Follow the manufacturer’s rating definitions rather than a generic specification.
Technical content reviewed by Srinivas Reddy, Managing Partner, Reddy Generators. Last updated 19 October 2026.