How to load test a generator
Almost every standby generator in the country is exercised monthly with nothing connected to it. That run proves the set starts, the battery is charged and the fuel has not gone off. It proves nothing at all about whether the set can carry your building, and on a diesel it quietly makes matters worse.
There are two questions about a standby generator and they are not the same question. Will it start? And will it hold the load once it has? A no-load exercise run answers the first. Only a load test answers the second, and it is the second one that matters on the night the supply fails.
What light running does to a diesel
A diesel engine is designed to run hot. Under proper load the combustion chamber reaches a temperature at which fuel burns completely and the rings seal against the bore as they were meant to. Off load, or on a token load, it never gets there. Fuel passes through unburnt, condenses in the exhaust and turbo, and glazes the cylinder walls with a hard varnish that stops the rings bedding.
That is wet stacking, and the name comes from the wet, oily deposit that appears around the exhaust stack. It is progressive, it is caused by exactly the monthly ritual most sites believe is maintenance, and its symptom is the cruellest one available: the set still starts perfectly. It simply cannot make rated power any more, and nobody finds out until the day it is asked to.
Loaded running is both the prevention and, caught early enough, the cure — the deposits burn off once the engine is held hot for long enough. Left for years it becomes an engine overhaul instead.
Why the building load is not the test
The obvious alternative is to transfer the building onto the set and let the real load do the work. Two problems with that, and the first is arithmetic. Standby sets are sized for a future the site has not reached yet, plus a margin, plus whatever the consultant rounded up to. The actual connected load is very often well under half the nameplate. A set that carries it has proved it can do the easy part.
The second problem is that the test is now being run on live equipment. The entire point of a capacity test is to find the failure. Finding it with a ward, a data hall or a production line hanging off the set is an expensive way to learn something a resistor could have told you.
A load bank removes both objections. It presents a load you choose, at the level you choose, held for as long as you want, with nothing important connected — the source is worked properly and the consequences of a failure are a tripped breaker and a wasted afternoon.
| The run | What it proves | What it does not |
|---|---|---|
| Monthly no-load exercise | It starts; battery, fuel and controls are alive | Anything about capacity — and it causes wet stacking |
| Transfer to the building load | The changeover works end to end | Performance above whatever the site happens to draw |
| Load bank test | Output, stability and thermal behaviour at rated load | The changeover, unless the test is staged through it |
| Battery discharge test | The real remaining capacity of the bank today | The state of the charger, which is a separate check |
How the test is staged
A load test is not a single switch. The load is applied in steps, the set is allowed to settle at each one, and readings are taken before the next step goes on. A quarter, a half, three quarters, full is the usual progression, and the value is as much in what happens between the steps as at the top.
At each stage you are watching voltage and frequency recover after the step is applied, and how quickly; coolant and exhaust temperatures stabilising rather than climbing; oil pressure holding; and the exhaust itself, which should clear as the engine comes up to proper load. The final step is held long enough for temperatures to plateau — a short burst at full load tells you the alternator can make the power, not that the cooling system can survive making it.
Step loading also matters because it is how the set will actually be used. Real load arrives in blocks as motors and chillers pick up, and a set that holds a smoothly applied full load can still stumble on a single large step. If your site has a big motor that starts on the generator, ask for the test to include a step that size.
Take the readings, and keep them
The value of a load test compounds only if the sheet is filed. Three years of readings at the same load points show a set drifting long before it fails, and that is the whole argument for testing on a schedule rather than after a bad night. One test in isolation is a pass or a fail; a series is a trend.
How often, and at what load
The standard usually written into contracts and insurance schedules is a monthly exercise run at a meaningful load, plus an annual test at or near full rating. Where a monthly run genuinely cannot reach a meaningful load from the building, a load bank is brought in to make up the difference — which is the arrangement most hospitals and data centres end up with.
Check what your own contract says before designing a regime around a rule of thumb. Insurers, hospital accreditation schemes and data centre uptime standards all specify this, they do not all specify the same thing, and the one that governs you is the one in writing.
For a new set, the test that matters most is the one before you accept it. Commissioning at rated output, witnessed, with the readings signed, is the last moment the supplier has an obligation and the cheapest moment to find a problem.
Resistive or reactive
A resistive load bank loads the set at unity power factor. It works the engine, the fuel system and the cooling — the mechanical half of the machine — and for most standby diesel testing it is what is being asked for.
A reactive load bank adds a lagging power factor, which loads the alternator and the automatic voltage regulator rather than the engine. That matters when the real load is motor-heavy, when the specification calls for a test at the set’s rated power factor, or on acceptance testing where the alternator is part of what is being proved. Combined resistive-reactive banks do both, stepped independently.
The short version: if you want to know the engine can do the work, resistive. If you want to know the whole set can do the work at the power factor it will actually see, reactive as well.
The same tool, two other jobs
A load bank is not only for generators. A UPS bought on a runtime figure has that figure verified in exactly one way — by discharging the bank against a known load and timing it. Battery capacity falls from the day of installation, and the decline is invisible until the discharge that counts. The commissioning certificate in the file records what the system did on one day, several years ago, at an ambient temperature nobody wrote down.
The same applies after any major repair. A machine that powers up on the bench and a machine that holds rated load for an hour are different claims, and only one of them is worth putting back into service.
Hire one or own one
For a single commissioning, hire. The machine is idle for the other three hundred and sixty four days and somebody else can own that problem.
Owning starts to make sense once testing is routine rather than occasional — a generator dealer proving every set before handover, a rental fleet checking machines between jobs, a UPS or panel builder load-testing its own output, a facilities team with several sets on an annual cycle. At that point the hire invoices, the scheduling and the waiting overtake the cost of the machine. We build load banks up to 1000 kW, specified against the plant they will be used on.
Questions we get asked
Anything not covered here, call us — most of it is faster to answer on the phone.
01How often should a standby generator be load tested?
The regime most contracts and insurance schedules specify is a monthly exercise run at a meaningful load plus an annual test at or near full rating. Where the building cannot supply a meaningful load, a load bank makes up the difference. Check the wording that actually governs your site — insurers, hospital accreditation schemes and data centre standards all require this and they do not all require the same thing.
02What is wet stacking?
It is what happens to a diesel that spends its life on light load. The engine never reaches proper combustion temperature, so fuel passes through unburnt, collects in the exhaust and turbo, and glazes the cylinder bores. The deceptive part is that the set still starts perfectly — it just cannot make rated power. Running it under real load is both the prevention and, caught early, the cure.
03How long does a load bank test take?
It depends on how many steps are being taken and how long the top step is held, and the holding is the part that should not be shortened. Temperatures have to plateau for the reading to mean anything. Budget for a planned morning rather than a quick visit, and agree the step schedule and the hold time in writing before anyone arrives.
04Can a load bank test a UPS and its batteries as well?
Yes, and for a UPS bought on a runtime figure it is the only way to verify that figure. The bank is discharged against a known load and timed. Battery capacity declines from installation onward and the decline does not show until the discharge that matters, so a commissioning certificate several years old is a record of the past rather than a statement about today.
05Do I need a resistive or a reactive load bank?
Resistive loads the engine, the fuel system and the cooling at unity power factor, and for most standby diesel testing that is the test being asked for. Reactive adds a lagging power factor and loads the alternator and the AVR instead — worth it when the real load is motor-heavy or when a specification calls for a test at the rated power factor. Tell us what has to be proved and we will confirm which applies.