Generator load bank testing field guide for electrical crews

Load bank testing, and what the nameplate does not prove

Load bank testing applies a controlled artificial load to a generator so the set runs at or near its rated output, instead of whatever light load the building happens to draw that day. The load bank turns the generator's power into heat and throws it away. The point is not the heat. The point is the readings you take while the set is working: voltage on every phase, frequency, kilowatts, and the engine temperatures and pressures under real strain.

A nameplate is a promise, not a measurement. A 500 kW set is rated 500 kW because the factory said so, but the set that has sat in a basement for two years lightly loaded, with a marginal cooling system and an air filter nobody changed, may not hold 500 kW for an hour without overheating or dropping frequency. The only way to know is to make it carry the load and watch.

That is the whole job. You build a load the building rarely produces, hold it in steps, and record how the machine behaves. The companion guides on standby generator and transfer switch installation and on emergency and standby power systems under NEC 700 cover getting the set in and classifying the load. This guide covers proving it works after it is in.

Why load test a standby generator?

A standby generator almost never sees its rated load in a real outage. The connected load is a fraction of the nameplate, the outage is short, and the set idles along at 20 or 30 percent and shuts down before it ever proves anything. Load testing exists because the real duty cycle does not test the machine. You have to test it on purpose.

Three things the test gives you. It proves the set can carry the rated load while holding voltage and frequency, which is the only honest answer to whether it will survive the worst outage instead of the average one. It runs the engine hot enough and long enough to find the cooling and fuel problems that a short light run hides. And on a diesel, it burns off the unburned fuel and carbon that light running deposits in the exhaust, the wet stacking problem in the next section.

Skip load testing and you are running on faith. The set might start every month, run for thirty minutes at a trickle, log a clean exercise, and still fail the day the building actually needs full power. A set that starts is not a set that performs. Starting is the easy part.

What is wet stacking?

Wet stacking is what happens to a diesel that runs too lightly for too long. The engine never reaches its design operating temperature, the fuel does not fully burn, and the unburned diesel and soot collect in the combustion chambers, on the injectors and valves, and in the exhaust. The name comes from the wet, oily, sooty residue you find at the exhaust stack. Black, greasy, and dripping in a bad case.

The mechanism is temperature. A diesel is built to run hot, and combustion is only complete when cylinder temperatures are high enough. Run the set at 20 or 30 percent load and the cylinders stay cool, so a slug of fuel goes out the exhaust unburned every cycle. Over months that residue glazes the cylinder bores, fouls the injector tips, gums the rings, and cakes the turbo and exhaust. The set loses power, smokes, and runs rough.

The fix and the prevention are the same. Make the engine work. A load test that holds the set at a high percentage of its rating for an hour or more brings the cylinders up to temperature, completes combustion, and burns the deposits back out of the exhaust. This is the single biggest reason a lightly loaded diesel needs a periodic load test, and why ignoring it is a slow path to a rebuild.

The load bank: what the device actually is

A load bank is a device that creates an electrical load on demand and dissipates the generator's output as heat. In its simplest form it is a bank of resistor grids, the same idea as a giant toaster, with a cooling fan moving air across the elements to carry the heat away. You connect it to the generator, dial in a load in kilowatts, and the grids absorb that power and blow it off as hot air.

The controls let you step the load up and down, usually in fixed increments, so you can hold 25, 50, 75, and 100 percent of a target and read the set at each level. A good load bank has its own metering, but the readings that matter are taken at the generator, because that is the machine under test.

The cooling air is the part people underestimate. A load bank rejecting a few hundred kilowatts of heat moves a lot of very hot air, and it has to have clear intake and a clear path for the exhaust. Park it against a wall or under a low soffit and it recirculates its own heat, the elements overheat, and the test stops itself. The bank needs room to breathe as much as the generator does.

What is the difference between a resistive and reactive load bank?

A resistive load bank creates a load at unity power factor, pure kilowatts, and loads the engine, the prime mover, to its full rated kW. A resistive plus reactive load bank adds inductive load measured in kVAR, so it can run at a lagging power factor, commonly 0.8, the way real motor and transformer loads behave. The difference decides what actually gets tested.

Here is the trap. A resistive-only test loads the engine fully but loads the alternator to only about 80 percent of its rated current, because gensets are rated at 0.8 power factor and a unity-PF resistive load draws less current for the same kW. So a resistive-only full-kW test gives the engine a real workout but never pushes the alternator and the automatic voltage regulator to their rated limit. The part of the machine that holds your voltage is the part the resistive test undertests.

For burning off wet stacking and confirming the engine makes power, a resistive bank is enough and it is what most routine testing uses. For commissioning a new set, or any test where you need to prove the alternator and the AVR hold voltage at the rated current, you want the reactive bank so you can load the machine at its actual 0.8 power factor. Resistive-only on an acceptance test is a common and expensive shortcut.

Why the 0.8 power factor matters

Most gensets are rated at 0.8 power factor, which means the engine and the alternator are sized for a load that draws more current than its kilowatts alone would suggest. A 500 kVA set rated at 0.8 PF makes 400 kW. Load it with a resistive bank and you can take it to 400 kW, the full engine rating, but the alternator current at unity PF is lower than it would be at 0.8, so the alternator coasts at around 80 percent of its rated amps.

The current is what heats the alternator windings and what the AVR has to regulate against. Test only at unity power factor and you never find the alternator that overheats at rated current, the AVR that cannot hold voltage when the reactive load comes on, or the cooling on the alternator end that is marginal. Those are exactly the failures a real inductive building load will expose on the worst day.

So the rule for a meaningful full test is to load the set at its rated power factor with a reactive bank, not just to its rated kW with resistors. The resistive test answers can the engine make the power. The reactive test answers can the whole machine deliver it the way the building will draw it.

The stepped load test: 25, 50, 75, 100 percent

A proper load test is run in steps, not slammed straight to full load. The common sequence walks the set up through roughly 25, 50, 75, and 100 percent of the target, holding at each step long enough for the temperatures to stabilize and for you to take a full set of readings. ISO 8528-6, the genset performance standard, frames its test points around these levels. The exact durations follow the standard and the manufacturer.

Stepping up does two things. It lets the engine and cooling system come up to temperature in stages instead of being shocked, and it gives you a curve, not a single point. You see how voltage, frequency, and the temperatures move as the load climbs, so a set that is fine at 50 percent but loses frequency at 90 percent shows you where it falls apart. A single full-load reading would hide that.

A typical commissioning sequence holds the lower steps for a shorter time and the top step longest, often the full-load step for the better part of an hour, because that is where the cooling and the wet-stacking burn-off really happen. Hedge the specific minutes per step to the project commissioning spec, the ISO test schedule, and the engine manufacturer. They do not all agree, and the spec controls.

What does NFPA 110 require for generator testing?

NFPA 110, the standard for emergency and standby power systems, drives the testing schedule for most code-required generators. It calls for a monthly operational test and, for diesels that cannot reach a load threshold on the monthly run, an annual test under supplemental load, which in practice means a load bank. The exact figures live in the standard and shift between editions, so confirm them against the adopted edition.

The monthly test, as commonly written, runs the set for at least 30 minutes under load, either loaded to maintain the minimum exhaust gas temperature the manufacturer specifies, or under operating temperature conditions at not less than 30 percent of the nameplate kW rating. The annual supplemental test, for diesels that miss the monthly threshold, is commonly written as not less than 50 percent of nameplate kW for 30 minutes and not less than 75 percent for one hour, a total of at least 1.5 continuous hours.

Treat those numbers as the framework, not gospel for your job. NFPA 110 has Level 1 and Level 2 systems, the AHJ adopts a specific edition with local amendments, and healthcare and other occupancies layer their own testing rules on top through NFPA 99 and the accreditation bodies. Read the adopted edition and confirm with the AHJ before you write the test plan.

Monthly exercise versus annual load test

These are two different tests with two different jobs, and people conflate them. The monthly exercise is a frequent, short run to prove the set starts, transfers, and runs, and to keep the engine exercised. The annual load test is a longer run at a higher load to prove rated performance and, on a diesel, to burn off accumulated wet stacking. One keeps the set limber. The other proves it can lift the weight.

The monthly run usually rides on whatever load the building presents, supplemented only if the connected load is too light to hit the threshold. The annual test on a diesel that runs light all year is where the load bank earns its keep, because the building will not produce 50 or 75 percent of nameplate on its own.

If your building load reliably carries the set above the monthly threshold and exercises it hard enough, you lean on the real load and the load bank shows up mainly for commissioning and the annual proof. If the set spends its life lightly loaded, the load bank becomes a regular tool, not an occasional one. Know which building you have.

What is the 30 percent rule?

The 30 percent rule is the common shorthand for the NFPA 110 monthly loading threshold: run the diesel at not less than 30 percent of its nameplate kW rating, or at a load that maintains the manufacturer's minimum exhaust gas temperature. It exists because a diesel loaded below roughly that level does not get hot enough to burn its fuel cleanly, which is the doorway to wet stacking.

The exhaust temperature is the real target and the percentage is the proxy. The standard gives the manufacturer's minimum exhaust gas temperature as the primary criterion and the 30 percent figure as the alternative when you are running on operating temperature conditions. If the engine maker publishes a minimum exhaust temp, that number is the more honest measure of whether combustion is complete.

The practical consequence is this: when the connected building load cannot push the set to 30 percent of nameplate, the monthly run on real load is not enough, and the code path is to make up the difference with a load bank or supplemental load, and to run the annual supplemental test. A standby set that is badly oversized for its actual load is a wet-stacking machine waiting to happen, and the 30 percent rule is the line that flags it. Hedge the exact percentage and the exhaust criterion to the adopted NFPA 110 edition and the engine manufacturer.

The factory and acceptance load test

A new set gets load tested twice in its early life, and the two are not the same. The factory acceptance test, the FAT, runs the set on the manufacturer's load bank before it ships, to prove it makes rated power and to catch a defect before it travels. The field acceptance or commissioning load test runs the installed set on site, to prove it performs after the trip, the install, the fuel, the exhaust, and the cooling air the real location gives it.

The site test is the one that matters most, because the factory test cannot see your engine room. A set that passed at the factory can fail on site because the radiator discharge recirculates in a tight enclosure, the fuel polishing was never done, the exhaust backpressure is too high from a long run, or the intake air is starved. The commissioning load test is where those install-specific problems surface, which is exactly why the companion install guide treats it as the last real step, not a formality.

A commissioning load test is typically a stepped run at the rated power factor with a reactive bank, held at the levels long enough for temperatures to stabilize, with a full data set logged at each step. The duration, the steps, and the power factor follow the project commissioning specification. That spec, not habit, defines what acceptance means for this set.

The readings: what data the test collects

The load test is only as good as the data you log off it. At each load step, once temperatures have settled, you record the electrical output and the engine readings, and you record them on a schedule, commonly every 5 to 15 minutes, so you have a trend and not just a snapshot.

On the electrical side: voltage on all phases, frequency, current, kilowatts, kVA, and power factor. On the engine side: coolant temperature, oil pressure, exhaust gas temperature, fuel pressure, and the ambient temperature in the space, because a hot engine room changes the result. The trend across the test is the real product. A coolant temperature still climbing at the end of the top step is a different story than one that stabilized.

The acceptance criteria a clean set holds are tight. Voltage and frequency should stay within a defined band, commonly on the order of plus or minus a few percent on voltage and a fraction of a hertz on frequency, with the exact limits set by the manufacturer and the project spec. A set that holds rated load but wanders on frequency, or sags on voltage when the load steps up, is telling you the governor or the AVR needs attention. Record the numbers, not impressions.

Setting up and connecting the load bank

The load bank ties to the generator at a defined connection point, and where you tie in decides what you are testing. Connect at the generator output terminals or a dedicated load bank tap and you test the set. Connect downstream at the transfer switch or the distribution and you can test the set plus part of the chain, which is what a fuller commissioning test wants. Cam-lok connectors and a load bank tap on the switchgear make this clean on jobs that test regularly.

Cabling has to be sized for the test current and routed so it is not a trip hazard or a heat trap, and the connections have to be tight, because a loose high-current connection arcs and heats under load. Place the load bank with its cooling air in mind: clear intake, clear hot-air discharge, away from the generator's radiator discharge and exhaust so the two are not fighting each other or feeding each other hot air.

Plan the connection as a de-energized, locked-out task. You are landing high-current conductors on live-capable gear, and the connection goes on with the source off and verified dead, the same discipline as any other termination. The companion install guide covers the permanent tap and the switched-versus-solid neutral that the connection point interacts with.

Safety on a load bank test

A load bank test stacks several hazards in one place, and the heat is the one people forget. A load bank rejecting hundreds of kilowatts blows air hot enough to burn, the elements glow, and the cabinet stays hot after the test ends. Keep people and combustibles clear of the discharge, and do not assume cool just because the fan stopped.

Then the exhaust. A diesel run hard for an hour and a half produces a large volume of exhaust, and carbon monoxide kills quietly. The exhaust has to be ducted or vented to clear air, never into an enclosed space where the crew is working, and a portable set on a load test needs the same exhaust discipline as a permanent install.

The electrical side is high current at the connection and live gear during setup. Land and remove the load bank connection with the source de-energized and verified dead, lock and tag it, and treat the cabling as energized once the test is on. None of this is exotic. It is the same hazard control as any high-current work, applied to a job where the heat and the exhaust raise the stakes.

Portable versus permanent load banks

Most load testing is done with a portable load bank: a trailer-mounted or roll-up unit brought to the site for the test and taken away after. It is the right tool for an annual test, a commissioning run, or a wet-stacking burn-off, because the equipment is expensive to own for a job you do a few times a year. A service contractor rolls it in, connects at the tap, runs the test, and rolls it out.

A permanent load bank is installed and wired in at facilities that test often enough to justify it, or that need to dump load automatically. Data centers and other critical sites mount permanent banks so they can load-test the gensets on a routine schedule without waiting on a rental, and so the test can run as part of an automated sequence. The permanent bank also gives a place to send load during a test that cannot use the building's real load.

The decision is frequency and criticality. Test once a year and a rental portable is cheaper and simpler. Test monthly or quarterly on a critical system, and the permanent bank pays for itself in availability and in not depending on a truck showing up.

Load banks in data-center and large-facility commissioning

On a data center or a large critical facility, the load bank is central to the integrated systems test, the IST, where the whole power chain is proven under load before the building goes live. You cannot wait for real IT load to test the gensets, so banks simulate the design load, and the test proves the generators, the transfer scheme, the switchgear, and the cooling all behave together when the utility is dropped.

The IST is bigger than any single load test. It drops normal power, watches the gensets start and accept load on the schedule the design requires, holds the simulated load through the steps, and confirms the cooling and the controls ride through. The emergency and standby systems guide covers the classification and selective coordination that the IST has to satisfy. The load bank is how you put the weight on the system to find out.

The AI buildout has pushed this hard. Large data center campuses carry generator fleets in the tens of megawatts, and proving that fleet, the paralleling controls, and the load-sharing under a realistic load is a load-bank exercise at a scale that used to be rare. The physics is the same as a single set. The orchestration is the harder part, and the load bank is still the only honest way to load it.

Testing on the building load versus the load bank

There are two ways to put load on a standby set for a test: transfer the real building load onto it, or apply an artificial load with a load bank. Each has a place, and they are not interchangeable. Transferring the building load proves the set carries what it is actually there to carry, with the real motor inrush and the real power factor, but it only loads the set to whatever the building draws, which on a standby set is usually well short of nameplate.

The load bank proves rated capacity and burns off wet stacking, because you can dial the load to 100 percent regardless of what the building needs. What it does not exercise is the transfer switch and the real load behavior, unless you connect downstream and combine the two.

The fuller commissioning approach uses both: transfer the real load to prove the transfer scheme and the live behavior, then add load bank on top to bring the set up to its rated load for the high-percentage steps. A closed-transition transfer can let you load-test against the utility in parallel on systems built for it, which avoids dropping the building, but that is a designed capability, not something to improvise. Confirm the transfer scheme before planning a parallel test.

What does a passing load test look like?

A passing load test is a set that takes each load step, holds voltage and frequency inside the specified band at every step including the top, and keeps its engine temperatures and oil pressure within limits through the full duration without tripping, derating, or smoking past the point a warm diesel should clear. The numbers, not the fact that it ran, are the pass.

The failures cluster. A set that cannot hold the top step, or that overheats before the full-load step ends, is usually a cooling problem: an undersized or fouled radiator, recirculating discharge air, or a hot engine room. A set that loses voltage when the reactive load steps on is an alternator or AVR problem the resistive-only test would have missed. A set that wanders on frequency under load is a governor or fueling problem. A diesel that smokes heavily and will not clear is showing you the wet stacking the test is supposed to burn off.

Pass or fail, the test is only finished when the data is written down and the deficiencies are listed with what each one needs. A test that finds a marginal cooling system and gets it fixed is a successful test, even though the set did not pass clean. A test that passes clean and gets no record is half a test.

What a load test commonly reveals

Run enough load tests and the same findings come up. Wet stacking on a lightly loaded diesel, showing as heavy smoke and oily exhaust residue that a hard run begins to clear. Cooling that cannot hold the top step, from a fouled radiator core, recirculating air in a tight enclosure, or an engine room that runs too hot. A real derate, where the set simply will not make its nameplate because the install conditions, altitude, or temperature take capacity off the top.

On the electrical side, the AVR that cannot hold voltage at the rated current, which only shows up under a reactive load near full current, and the governor or fuel system that lets frequency sag or hunt when the load steps. Fuel problems show up as a set that starts fine and then stumbles under sustained load: a clogging filter, contaminated or aged fuel, or air in the fuel system that a light run never stresses.

These are the reasons the test exists. None of them show on a monthly thirty-minute trickle run. All of them show when you make the machine work, which is the entire argument for load testing instead of trusting the start.

Load testing as part of generator maintenance

Load testing is one piece of a generator maintenance program, not a thing you do once and forget. It belongs alongside the oil and filter changes, the coolant and fuel service, the battery checks, and the fuel polishing that keep a standby set ready. NFPA 70B and NETA frame the maintenance and acceptance testing of electrical gear, and the engine manufacturer sets the engine service intervals. The load test proves the maintenance worked.

The records are the part that gets neglected and the part that matters most. Each load test should leave a dated report with the load steps, the readings at each step, the deficiencies found, and what was corrected, kept with the maintenance history for the set. That history is what an inspector, an insurer, or the next service tech reads to know whether the machine has actually been proven or only started.

A set with a clean maintenance history and no load-test record is an unknown. You know it has been serviced. You do not know it performs. The load test is what closes that gap, and the record is what makes the closed gap mean something a year later.

Field checklist

What to document

A load test that leaves no report is a test nobody can stand behind. The record is what proves the set was carried to its rated load and held, and it is what the next person reads to know the machine performs and not just starts.

Capture the set identity and nameplate, the load bank type and power factor, the steps and durations, the full readings at each step, the acceptance criteria used, the deficiencies and corrections, the test date, and who ran it. The table below is what a useful report shows at a glance.

Common mistakes

• Never load testing at all, so the set wet stacks and its real capacity is unknown until the outage.
• Running a resistive-only test on a commissioning job, loading the engine to full kW but the alternator and AVR only to about 80 percent.
• Letting the monthly run sit below 30 percent of nameplate or below the manufacturer's minimum exhaust temperature.
• Skipping the annual supplemental load test on a diesel that the building load cannot exercise hard enough.
• Taking a single snapshot instead of trending readings across the steps and the full duration.
• Staging the load bank against a wall or near the radiator discharge so it recirculates hot air and stops itself.
• Venting diesel exhaust into the work space, or working near the load bank discharge heat.
• Ignoring the wet-stacking signs, the smoke and the oily exhaust, instead of running the set hard enough to clear them.

Standards and references

NFPA 110, the standard for emergency and standby power systems, drives the testing schedule for code-required generators, including the monthly operational test and the annual supplemental load test for diesels that cannot reach the loading threshold on real load. It defines Level 1 and Level 2 systems and the maintenance and testing each requires. The exact percentages, durations, and exhaust criteria shift between editions, so confirm them against the edition the AHJ has adopted and any local amendments.

ISO 8528, particularly Part 6 on test methods, frames generator set performance and the stepped load test points the acceptance and factory tests use. NFPA 70B and the NETA acceptance and maintenance testing specifications frame the broader electrical maintenance and testing program the load test sits inside. The engine and generator manufacturer's instructions set the service intervals, the minimum exhaust temperature, and the voltage and frequency limits the set has to hold, and those instructions govern where they are more specific.

For the install and the load classification that sit on either side of the test, the companion guides cover standby generator and transfer switch installation and the NEC 700, 701, and 702 emergency and standby system classes. Cite the standard that controls the point, hedge the percentages and durations to the adopted NFPA 110 edition and the manufacturer, and let the project commissioning specification define what acceptance means for the set in front of you.

Units, terms, and conversions

Load testing crosses electrical and engine units, and the same set is described in kW, kVA, and a power factor that ties them together, so the vocabulary is worth pinning down.

Real power is kilowatts, kW, what the engine produces and the resistive load bank absorbs. Apparent power is kilovolt-amperes, kVA, what the alternator is rated for. Reactive power is kilovolt-amperes reactive, kVAR, what the reactive load bank adds. Power factor is the ratio of kW to kVA, commonly 0.8 lagging on a genset rating. Exhaust gas temperature is often abbreviated EGT, and the automatic voltage regulator is the AVR.

kW / kVA / kVAR

Real power, apparent power, and reactive power; the genset rating ties them through power factor

Power factor (PF)

Ratio of kW to kVA; gensets are commonly rated at 0.8 lagging, which the reactive load bank simulates

Wet stacking

Unburned fuel and soot building up in a diesel run too lightly to reach full combustion temperature

Load bank

Device that applies an artificial electrical load and dissipates the generator's output as heat

Resistive vs reactive bank

Resistive loads at unity PF (engine only); reactive adds kVAR to test the alternator and AVR at rated current

EGT

Exhaust gas temperature, the truer measure of whether a diesel is loaded enough to burn fuel completely

AVR

Automatic voltage regulator, which holds alternator voltage and is only fully tested under a reactive load

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What size standby generator do I need?
• What is the difference between emergency and standby power?
• How long is a generator load bank test?
• What is an NFPA 110 acceptance test?
• What are the conditions to parallel two generators?