Datacenter
Load bank test acceptance criteria for data center commissioning
Apply a measured load, hold it, watch the dips and recovery, and sign the record that proves the source carries rated capacity before IT load arrives.
Direct answer
A load bank applies a controlled, measurable electrical load to a power source so it can be proven at rated capacity without waiting for real IT load. In data center commissioning it confirms a generator, UPS, or power chain carries full kW and kVA and recovers from block load, with the project spec and manufacturer controlling the pass criteria.
Key takeaways
- Load bank acceptance is judged on stability and recovery: hold voltage and frequency at each step, recover from block load, and carry full kW and kVA at rated power factor.
- Most gensets are rated at 0.8 power factor, so a resistive-only bank leaves the alternator and voltage regulator about 20 percent untested; use resistive/reactive for rated kVA acceptance.
- A commissioning full-load hold is commonly 2 to 4 hours at 100 percent after a stepped ramp through 25, 50, 75 percent; the hold finds cooling and fuel faults a short power check skips.
- To fully load a 2000 kW, 0.8 PF generator, provide 2000 kW resistive plus reactive to reach about 2500 kVA; a resistive-only bank drives the alternator to only about 80 percent current.
- NFPA 110 sets a monthly exercise (at least 30 percent nameplate or the manufacturer exhaust gas temperature) and an annual load test, separate from one-time commissioning acceptance.
What a load bank is, and why you use one
A load bank is a device that applies a controlled, measurable electrical load to a power source so the source can be run at rated capacity on demand. It turns that load into heat through resistive elements, inductive elements, or both, and dumps the heat to air or water. The point is to prove the source carries its nameplate and behaves correctly under load, without waiting months for real IT load to show up.
That last part is why load banks exist in commissioning. A data center gets built and energized long before the servers arrive, and the day you need the generator and UPS to work is the day the utility drops. You cannot find out then. The load bank lets you put rated load on the generator, the UPS, the busway, and the whole power path while there is still time and budget to fix what fails.
A load bank does not test the IT load. It stands in for it. The closer the bank can mimic real load, in magnitude and in power factor, the more the test actually means, which is why the type and size of bank you bring decides how much the result is worth.
Why diesel generators need a load bank
A diesel generator that only ever idles or runs lightly loaded fails when it is finally called. The engine is built to run hot and worked. Run it light for months and the combustion temperature never gets high enough to fully burn the fuel, and the leftovers do real damage. This is the case for load banking every standby diesel, not just a commissioning formality.
Two problems stack up under light load. The first is wet stacking, covered in its own section, which is unburned fuel and soot fouling the exhaust side. The second is glazing of the cylinder bores, where the rings never seat because cylinder pressures stay low. A glazed engine burns oil and loses compression, and it does not heal on its own.
The generator that has carried at least 30 percent of nameplate once a month and a full load test once a year is the one that picks up the building cleanly. The one that sat and idled is the one that smokes, surges, and trips on its first real call. You pay for the loading now with a load bank, or you pay for it later with an outage.
Why does my generator wet stack?
Wet stacking is unburned fuel, carbon, and lube oil building up on the exhaust side of a diesel engine because it has run too lightly for too long. The tell is black, oily residue weeping from the manifold, turbo, or stack, often with a diesel smell and heavy smoke when load finally comes on.
The mechanism is temperature. A diesel needs high cylinder and exhaust temperature to vaporize and burn the fuel completely. Under light load the engine never reaches that temperature, so some fuel leaves unburned. It coats the valves, the turbocharger, and the stack with soot, raises exhaust backpressure, and drops turbo efficiency. Worse, the cool cylinder lets the rings under-seal, so raw fuel slips past into the crankcase and dilutes the oil, which speeds up wear.
The cure is load and heat. Running the engine at roughly 75 to 100 percent of nameplate for a few hours raises exhaust temperature enough to vaporize the deposits and blow the soot out. A load bank is how you do that on a generator with no real load to give it. Confirm the loading and duration against the engine manufacturer, because a badly fouled engine sometimes needs service before a load test will clean it up.
Resistive or resistive/reactive: which load bank do I need?
A resistive load bank tests kW at unity power factor. A resistive/reactive load bank adds inductive elements to test kVAR and brings the test down to the generator's rated 0.8 power factor, which is where the alternator and the voltage regulator actually get worked. For full generator acceptance, resistive-only is not the whole test.
Here is the number that decides it. Most generator sets are rated at 0.8 power factor, meaning 80 percent real power and 20 percent reactive. A resistive-only load bank loads the engine to full kW but drives the alternator to only about 80 percent of its rated current, and it never makes the excitation system and the automatic voltage regulator hold voltage against a lagging load. You have tested the engine and left the alternator and its controls 20 percent untested.
Resistive-only still earns its keep. It clears wet stacking, it proves engine power and cooling, and it is cheaper and lighter to bring on site. But when the spec says prove the generator to nameplate kVA, you need the reactive bank to load it at 0.8 power factor and watch the regulator hold. Verify the required power factor and kVA against the project specification and the generator data sheet, not against what the rental house had on the yard.
Test setup and connection
The load bank ties to the source through high-current cable on cam-lock connectors, landed at the generator output, the load bank terminals, or a dedicated test tap on the switchgear. Match phase rotation and size the cable for the test current with margin, because a hot connection at full load is a fire, not a footnote. Torque the lugs to spec and check them again under load.
Placement is about heat and air. A load bank rejects its full rating as heat, so a 2 MW bank dumps 2 MW into the air around it. Keep its intake clear, keep its hot discharge from recirculating into the generator radiator or the building intakes, and keep it away from anything the exhaust plume can damage. On a rooftop or in a yard, watch which way the wind and the generator exhaust are going before you set it down.
Coordinate the test with whoever owns the gear. Lockout and tagout the points you are working on, confirm the source breaker and the load bank breaker are both controllable, and brief everyone on how the test gets stopped fast if a reading goes wrong. The person at the load bank controls and the person watching the generator need to be talking the whole run, not checking in at the end.
Step loading and the test profile
Load is applied in steps, not slammed on all at once, so the readings at each level can be recorded and the engine and alternator are not shocked unnecessarily. A common profile ramps through roughly 25, 50, 75, and 100 percent of nameplate, holding at each step long enough for temperatures and readings to stabilize, then holds at full load for the long pull.
Separate two ideas here. Stepped loading is the ramp used to characterize the source and log readings at each level. Block loading is the deliberate large step, often picking up a big chunk of load in one move, used to test how far voltage and frequency dip and how fast they recover. The commissioning spec usually calls for both: the staged ramp for the data, and one or more block-load steps to prove transient response.
The full-load hold is the headline and it is project dependent. Two to four hours at 100 percent is common for a commissioning load test, both to confirm the source holds rated load and to drive enough heat to clear an engine. The duration, the steps, and the power factor are set by the project specification and the manufacturer, so confirm them before the bank shows up, not on test day with a crew standing around.
What are the acceptance criteria for a load bank test?
Acceptance is judged on stability and recovery under load, not just on whether the source reached its rating. The source has to hold voltage and frequency steady at each load step, recover from block load within a defined band, carry full kW and kVA at the rated power factor, and keep engine, alternator, and exhaust temperatures inside the manufacturer's limits for the whole hold.
The numbers in the table below are typical commissioning ranges, not a standard you can cite. Voltage and frequency tolerances, allowable dip, and recovery time all come from the project specification, the generator and UPS data sheets, and the equipment standard the spec invokes. A facility with sensitive load may hold a tighter band than a general standby system. Use the table to know what to watch and roughly where the lines fall, then hold to the contract numbers.
The one criterion people forget is the hold itself. A source that hit 100 percent for five minutes and got shut down did not pass a load test. It passed a power check. The hold is where cooling, fuel, and oil-temperature problems actually show up, and a short test skips exactly the part that finds them.
| Measured at the test | What it proves | Typical range (verify against spec and mfr) |
|---|---|---|
| Steady-state voltage | Regulation holds at each load step | About plus or minus 1 to 2 percent of nominal |
| Steady-state frequency | Governor holds speed under load | About plus or minus 0.5 percent at steady load |
| Voltage dip on block load | Alternator and AVR ride the transient | Often within about 10 to 15 percent dip |
| Frequency dip on block load | Engine and governor ride the transient | Recovery to band commonly within a few seconds |
| kW and kVA at rated PF | Full real and reactive capacity | 100 percent of nameplate at 0.8 PF |
| Engine and exhaust temps | Cooling and combustion under load | Within mfr limits for the full hold |
| Oil pressure and fuel rate | Engine health under sustained load | Stable within mfr range |
Generator, UPS, and integrated load testing
Three different tests get called load bank testing, and they prove different things. Generator load testing proves the engine, alternator, and controls carry rated load. UPS load testing proves the UPS holds its output and the battery delivers its runtime. The integrated systems test proves the whole power chain transfers and rides through together, the way it has to in a real outage.
Done in sequence, they build on each other. You accept the generator on its own bank, you accept the UPS on its own bank, then the integrated test loads the live power path and drops the utility to watch the UPS carry the gap while the generator starts, picks up, and the transfer completes. That last test is the one the building lives or dies on, and it belongs to the power-QA scope of commissioning.
A source that passes alone can still fail integrated. The generator started fine on the bench, but under the real transfer the block load it picks up dips voltage past what the UPS bypass tolerates, and the system trips. You only find that with the integrated test, with load banks standing in for the IT load that is not there yet.
UPS and battery load testing
A UPS load bank test loads the UPS output to rated kW and kVA and confirms three things: the UPS holds output voltage and frequency under load, it transfers between normal, bypass, and battery without dropping the load, and the battery delivers its rated runtime. The runtime is the part that cannot be faked, because a string that has lost capacity reads fine at float and collapses under discharge.
The test discharges the battery into the load bank at rated load and times it to the inverter shutdown or the design end voltage, logging cell or string voltages along the way. A runtime that comes in short of the design number, or a cell that sags well below its neighbors, is the finding. Better to catch the weak string on a load bank than during the first real utility loss with live load on the floor.
Match the load bank power factor to what the spec calls for, since UPS systems are rated in both kW and kVA and a kW-only test can miss the kVA limit. Confirm the discharge load, the runtime target, and the end voltage against the UPS and battery data sheets and the project specification before you start the discharge, because you do not get a clean second run on a freshly drained battery.
Busway, PDU, and RPP load testing and heat rise
Downstream of the UPS, the distribution gets load tested too, and the thing you are watching is heat rise at the connections. Busway, the PDU, and the remote power panel all carry full current to the racks, and a loose or undersized joint shows up as a hot spot under load that was invisible cold. Load the distribution, let it stabilize, and scan the joints.
A thermal camera on the bus joints, the breaker terminations, and the PDU transformer under sustained load is how the heat-rise problem gets caught. A connection running well above its neighbors is a bad joint, and you fix it before it carries IT load, not after it has cooked the insulation. Busway plug-in joints are a common offender, because they get bumped during install and never re-torqued.
The load bank for this stage usually lands at the PDU or RPP output, standing in for the rack load. Confirm the load steps and the acceptable temperature rise against the equipment ratings and the project specification, and record the ambient, because heat rise is read above ambient, not as an absolute number on the screen.
The integrated systems test
The integrated systems test, often called the IST or the pull-the-plug test, loads the live power and cooling systems and then forces failures to prove the building rides through them. Load banks stand in for the IT load while the test drops the utility, fails a UPS module, opens a tie, and watches whether the load stays up and the systems coordinate the way the design says they will.
This is where the power chain meets the cooling and the controls. A generator and a UPS can each pass alone and still fail together, because the real test is the timing: the UPS carrying the load through the gap, the generator starting and accepting the block load, the transfer switches operating in the right order, and the cooling coming back before the room overheats. The Uptime Institute frames this kind of integrated testing as the way you find the management and configuration faults that cause most real outages.
Run the failure scenarios the design claims to survive, one at a time, with the load banks holding rated load. The script comes from the basis of design and the commissioning plan. A scenario that is not in the script does not get tested, so the script is where the rigor lives, and a thin script produces a clean report that proves very little.
Fuel, cooling, and ventilation over a long test
A two to four hour full-load run is also a test of everything that supports the engine, and that is where long tests actually fail. Fuel first: the day tank, the transfer pumps, and the return all have to keep up at full burn for the whole hold, and a day tank sized for short runs can run the engine dry mid-test. Confirm fuel supply and polishing before a long run.
Cooling and ventilation carry the heat. The radiator or remote cooling has to reject full-load heat at the worst-case ambient the room or yard will see, and the louvers, dampers, and intake path have to be open and moving enough air. An engine that climbs toward its high-coolant-temperature trip during the hold is telling you the cooling or the ventilation is short, and that finding is worth more than the kW number.
The load bank adds its own heat on top. Its discharge has to go somewhere that does not feed back into the generator radiator or the building intake. Plan the airflow for the bank and the source together, because two large heat sources in one yard can cook each other when the wind is wrong.
Data logging and the witnessed record
A load bank test that was not logged did not happen, as far as the record is concerned. The acceptance is the trend data and the witnessed sheet, not the memory of the day. Log voltage, frequency, current, kW, kVA, power factor, and temperatures at each load step and across the full hold, on a time base, so the dips and the recovery are captured and not just the steady values.
Most modern load banks data-log automatically, and the generator and UPS controllers trend their own values. Pull all of it, line it up on the same clock, and keep it with the test sheet. The block-load transient especially has to be captured fast enough to show the dip and the recovery time, which a once-a-minute reading will miss entirely.
The witnessed part matters. The commissioning agent, and often the owner and the engineer of record, sign the test record, which ties the result to the people who watched it and the spec it was held to. That signature is what a future operator relies on when the question is whether this generator was ever actually proven, or only run.
What if the generator fails the voltage dip and recovery test?
When a generator dips too far or recovers too slowly on block load, the cause is usually the alternator, the voltage regulator, the governor, or a block step bigger than the machine was sized for. Sort which one before you touch settings, because the fixes pull in opposite directions.
A deep voltage dip that recovers slowly points at the alternator and the automatic voltage regulator. The excitation cannot push voltage back up fast enough against the load step, which can be regulator tuning, an undersized alternator for the reactive load, or a step too large for the machine. A frequency dip that hangs points at the engine and governor: the engine cannot take the load fast enough, which can be governor response, fuel delivery, or turbo lag on a hard step.
Before re-tuning, confirm the test itself is fair. A block step larger than the design call, a reactive load the generator was never rated for, or a power factor below the data sheet will all make a healthy machine look like it failed. If the step matches the spec and the machine still fails, it goes back to the manufacturer's commissioning technician for regulator and governor adjustment, then you re-run the block test and log it again. Do not accept a tuned result you did not witness re-tested.
Sizing the load bank to the source
The load bank has to be big enough to load the source to 100 percent at the rated power factor, and this is where tests quietly fall short. A bank that tops out at 80 percent of the generator nameplate cannot prove the machine, and a resistive-only bank cannot reach rated kVA no matter how many kW it carries. Size the bank to the source and the spec, not to what is on the truck.
Account for power factor in the sizing. To load a 2000 kW, 0.8 power factor generator fully, you need 2000 kW of resistive plus enough reactive to bring the combined load to about 2500 kVA at 0.8 power factor. A resistive bank alone, even a 2000 kW one, leaves the alternator at roughly 80 percent current. That mismatch is the most common reason an acceptance test does not actually accept anything.
For UPS and distribution, size the bank to the rated kVA of the equipment under test, and confirm it can hold the load for the full required duration, including the battery runtime. A bank that trips its own elements on a long hold ends the test early and proves nothing except that the bank was undersized.
NFPA 110, the monthly exercise, and the annual load test
Commissioning load testing is a one-time acceptance. NFPA 110 governs the ongoing testing that keeps a standby system reliable after handover, and the two get confused. The commissioning test proves the new system. The NFPA 110 routine keeps proving it for the life of the plant.
NFPA 110 sets a monthly exercise and an annual load test for emergency and standby diesel systems. The monthly exercise, in the operational testing section, is commonly met one of two ways: run at no less than 30 percent of nameplate kW for a set period, or load enough to reach the minimum exhaust gas temperature the engine manufacturer specifies. The 30 percent figure exists because lighter loading does not get the engine hot enough to keep it from wet stacking. The annual load test runs the engine through load steps for a longer period to confirm capacity and clear deposits.
Cite NFPA 110 by topic, not by a section number you are not sure of, because the loading steps and durations have changed between editions, and recent editions revised the annual step profile. Confirm the current requirement against the edition the jurisdiction has adopted and the authority having jurisdiction. The commissioning spec, the manufacturer, and the AHJ together set what the test has to be.
Field checklist
Want this checklist to run itself on every job — with photo proof and a signed record crews can hand the customer? That's FieldOS.
What to document
The load bank record is what a future operator trusts when the question is whether this source was ever proven. Capture enough that a reviewer who was not there can reconstruct the test and check the result against the spec.
Record the source identification and ratings, the load bank used and its type, the readings at each load step, the block-load dips and recovery times, the temperatures and engine readings across the hold, the fuel used, the total duration, the pass or fail against each criterion, and who witnessed it. If anything was adjusted and re-tested, record the before and after, because the next person needs to see what changed and that it was proven after the change.
| Field to record | Why it matters |
|---|---|
| Source ID and ratings (kW, kVA, PF, voltage) | Ties the test to the specific machine and its nameplate |
| Load bank type, rating, connection | Shows the source was loaded to full kVA at rated PF |
| Readings at each load step | Documents regulation and stability across the ramp |
| Block-load dip and recovery time | The transient result, against the spec band |
| Engine, alternator, exhaust temps over the hold | Proves cooling and combustion held under load |
| Oil pressure and fuel consumption | Engine health and fuel-system adequacy |
| Full-load hold duration | Confirms the hold met the spec, not just a power check |
| Pass or fail per criterion | The verdict, criterion by criterion, not a blanket pass |
| Witnesses and date | Ties the result to the people and the spec edition |
Common mistakes
- Bringing a load bank too small to reach 100 percent of the source at rated power factor.
- Testing resistive-only when the spec calls for reactive, leaving the alternator and AVR at about 80 percent.
- Skipping the full-load hold and calling a five-minute power check a load test.
- Reading steady-state values only and missing the block-load dip and recovery a slow log never caught.
- Letting the load bank discharge recirculate into the generator radiator and skewing the cooling result.
- Undersizing the day tank or cooling for a long hold and ending the test early on a trip.
- Accepting a generator alone and never running the integrated test that proves the transfer.
- Confusing the NFPA 110 monthly exercise with a commissioning acceptance test.
Standards and references
Several bodies govern different parts of this, and naming the right one for the point is the difference between a credible record and a guess. NFPA 110, the standard for emergency and standby power systems, governs the ongoing monthly exercise and annual load test after handover, including the 30 percent nameplate and exhaust-gas-temperature options for the monthly run. Confirm the loading and durations against the adopted edition, because they have changed across cycles.
For acceptance testing of the electrical equipment, ANSI/NETA ATS gives the field test and inspection requirements before energization, and recent editions cover UPS and battery systems alongside the traditional gear. IEEE standards cover generator and UPS testing methods and battery test practice. The Uptime Institute frames integrated systems testing as a reliability practice, the structured way to prove the whole site rides through the failures the design claims to survive.
Above all of these sit the manufacturer's instructions and the project specification, which govern the actual numbers. The generator and UPS data sheets set the voltage and frequency tolerances, the rated power factor, and the temperature limits. The spec sets the load steps, the hold duration, and the acceptance bands. When a standard and the spec disagree, the stricter controlling document wins, and the authority having jurisdiction has the final say on what is enforceable.
Units and terms
The power numbers on a load bank test come in three forms, and reading the wrong one is how a test gets accepted that should not have. kW is real power, the work the engine does. kVA is apparent power, what the alternator and the conductors carry. kVAR is reactive power, the part a resistive bank never produces. They relate through power factor.
Power factor is the ratio of kW to kVA, and the rated 0.8 of most generators means the alternator carries about 25 percent more current than the kW alone implies. Frequency is in hertz, 60 Hz in North America and 50 Hz in much of the world, and the generator holds it by holding engine speed. Temperatures read in degrees F or C, and heat rise on distribution is read above ambient, not as an absolute.
- kW (real power)
- The actual power the engine produces and a resistive load bank dissipates
- kVA (apparent power)
- The total the alternator and conductors carry, kW and kVAR combined
- kVAR (reactive power)
- The reactive component a reactive or inductive load bank produces
- Power factor (PF)
- Ratio of kW to kVA; most gensets are rated at 0.8 lagging
- Block load
- A large load step applied at once to test voltage and frequency dip and recovery
- Wet stacking
- Unburned fuel and soot fouling a diesel's exhaust side from chronic light loading
- IST
- Integrated systems test, loading and failing the live power chain to prove ride-through
FAQ
How long is a generator load bank test?
A commissioning full-load hold is commonly 2 to 4 hours at 100 percent, after a stepped ramp through 25, 50, and 75 percent. The duration is set by the project specification and the engine manufacturer, not a fixed rule. It runs long enough to stabilize temperatures, prove the hold, and clear an engine of light-load deposits.
Why does my generator wet stack?
Wet stacking comes from running a diesel too lightly for too long. The cylinders never get hot enough to fully burn the fuel, so unburned fuel and soot foul the exhaust, valves, and turbo, and raw fuel dilutes the oil. The fix is a load bank run at roughly 75 to 100 percent of nameplate for several hours.
Resistive vs reactive load bank: which do I use?
Resistive load banks test kW at unity power factor and prove the engine. Resistive/reactive banks add kVAR to load the generator at its rated 0.8 power factor, working the alternator and voltage regulator to full current. Resistive-only leaves about 20 percent of the alternator untested, so use reactive when the spec calls for rated kVA.
What if the generator fails the dip and recovery test?
A deep, slow voltage dip points at the alternator and voltage regulator; a hanging frequency dip points at the engine and governor. First confirm the block step and power factor match the spec, since an unfair step fails a healthy machine. If the test is fair, the manufacturer's technician re-tunes, then you re-run and re-log the block test.
Is a resistive-only load bank enough for generator acceptance?
A resistive-only load bank is enough to clear wet stacking and prove engine power and cooling, but not enough for full generator acceptance at most data centers. It loads the alternator to only about 80 percent of rated current and never tests the regulator against reactive load. For rated kVA acceptance, use a resistive/reactive bank at 0.8 power factor.
How big a load bank do I need for a 2000 kW generator?
To load a 2000 kW generator rated at 0.8 power factor to full nameplate, you need 2000 kW of resistive load plus enough reactive load to reach about 2500 kVA at 0.8 power factor. A resistive-only 2000 kW bank loads the engine fully but drives the alternator to only roughly 80 percent current.
Does NFPA 110 require load bank testing?
NFPA 110 sets a monthly exercise and an annual load test for standby diesel systems, met by running at no less than 30 percent nameplate or to the manufacturer's exhaust gas temperature. That is ongoing testing, separate from commissioning acceptance. Confirm the current loading and durations against the adopted edition and the authority having jurisdiction.
What is tested in a UPS load bank test?
A UPS load bank test loads the output to rated kW and kVA, confirms the UPS holds voltage and frequency and transfers between normal, bypass, and battery without dropping the load, and discharges the battery to prove its runtime. A short runtime or a sagging cell is the finding you want before real IT load arrives.
What is an integrated systems test?
An integrated systems test, or IST, loads the live power and cooling with load banks, then forces failures like a utility loss or a UPS module trip to prove the building rides through. It catches the timing and coordination faults that a generator and UPS passing alone will miss, which cause most real data center outages.
People also ask
Codes cited in this guide
This guide is written and reviewed against the published standards below. Always confirm the current adopted edition with the authority having jurisdiction.