NFPA 110 Load Bank Testing Requirements: Monthly, Annual, and Triennial

Most facilities we inspect are doing a monthly generator test. Far fewer can tell us whether that test actually satisfies NFPA 110 — because the monthly run is only the first of three separate requirements, and the other two are triggered by conditions most facility managers have never been told to check for.

This guide covers all three: the monthly exercise, the annual supplemental load bank test, and the 36-month Level 1 run. It gives the load percentages, the durations, and the section references so you can check your own program against the standard.

What Does NFPA 110 Require for Load Bank Testing?#

NFPA 110 requires a diesel emergency power supply system to be exercised monthly for at least 30 continuous minutes, loaded to either 30% of standby nameplate kW or the manufacturer's minimum exhaust gas temperature. Facilities that cannot reach that threshold owe an annual supplemental load bank test of 50% for 30 minutes followed by 75% for one hour. Level 1 systems additionally run every 36 months for their Class duration or 4 hours, whichever is less, at no less than 30% of nameplate kW.

The single most misunderstood point: the annual load bank test is conditional, not universal. If your generator genuinely carries 30% of its nameplate rating during the monthly exercise, you do not owe an annual supplemental load bank test at all. If it does not — and on an oversized standby set it usually does not — you owe it every year.

Which edition are these numbers from?#

This guide is written to the 2025 edition. That matters more than usual here, because the 2025 edition changed the annual test: it eliminated the 25% first step that earlier editions required.

Two practical consequences:

• Many jurisdictions have not adopted the 2025 edition yet. If yours is on an older edition, you owe the three-step, two-hour sequence. Confirm which edition your AHJ enforces before you rewrite a procedure around the shorter test.
• If you see a "30% → 50% → 75%" sequence quoted anywhere for the annual test, it is wrong in every edition. The old first step was 25%, not 30%. The 30% figure belongs to the monthly exercise, and the two get conflated constantly — including, until recently, in some of our own guides.

Section numbers below reference Chapter 8 (Routine Maintenance and Operational Testing) of the 2025 edition. Section numbering is not stable across editions — §8.4.2.3 is the automatic-load-replacement clause in recent editions but was the annual load test itself in 2010 and 2019 — so pair any section you cite with the edition year. Your AHJ can also enforce a stricter interval than the standard, and many hospital accreditors do.

The Monthly Exercise#

NFPA 110 requires the emergency power supply system to be exercised under load every 30 days for 30 continuous minutes, timed from the moment the EPS accepts load — not from the moment it starts.

This section applies to diesel sets#

Before the numbers: the 30% threshold, and the annual supplemental test it triggers, are diesel requirements. NFPA 110 separates the monthly load criteria by engine type, and the committee's reasoning is explicit — spark-ignited engines are not subject to wet stacking, so there is nothing for the 30% floor to protect against. A natural gas or propane set is exercised under the EPSS load that is available.

The two clauses read differently. A diesel set is exercised on loading that maintains the manufacturer's minimum exhaust gas temperature, or at normal operating temperature at not less than 30% of standby nameplate kW. A spark-ignited (gaseous) set is exercised monthly on the available EPSS load, for 30 minutes or until water temperature and oil pressure have stabilized — no percentage floor.

The practical consequence for gaseous sets: you are not subject to the supplemental load bank requirement, and a vendor quoting you an annual load bank test on a natural gas or propane set should be asked which clause they are citing. See diesel vs. natural gas generators for how the two differ in service terms more broadly.

The load must satisfy one of two criteria (§8.4.2):

1. The 30% method. The set runs at normal operating temperature at no less than 30% of its standby nameplate kW rating. Note which rating: standby, not prime or continuous. On a set with more than one published rating this is a real distinction and it changes the target.
2. The exhaust gas temperature method. The set reaches the minimum exhaust gas temperature recommended by the manufacturer.

If you cannot hit either one, you do not skip the month. The set is still exercised monthly on whatever building load is available — the requirement to run does not lapse just because the load target cannot be met. What chronic light loading changes is that you now also owe the annual supplemental test below.

The second option exists because the point of the requirement is combustion quality, not the number itself. A diesel engine running lightly loaded does not reach the cylinder temperatures needed to burn fuel completely. Unburned fuel and soot accumulate in the exhaust stream — the condition commonly called wet stacking. The 30% figure is a proxy for "hot enough"; the manufacturer's exhaust temperature is the direct measurement.

Two details that catch facilities out:

• Cold starts are required (§8.4.4). The test must include a complete cold start, with the EPS accepting load within its Type window — 10 seconds for a Type 10 installation. What this rules out is idling the set up first and then applying load: the run has to begin with the engine off and be started by the test-initiating signal. It does not mean the engine must be cold. NFPA 110 separately requires jacket-water heaters on Level 1 sets — they are how a Type 10 installation makes its 10-second window at all — so a compliant Level 1 generator is always thermally preconditioned, and that is not a defect in the test.
• Budget 40–45 minutes on site, not 30. A 5-minute unloaded cool-down before shutdown is standard practice and protects the turbo. Note where it comes from, though: NFPA 110 puts a 5-minute cool-down in Chapter 7 (acceptance testing) and a 5-minute minimum shutdown time delay in §8.4.5(4), which is a controller setting rather than a step in the monthly test procedure. Chapter 8 does not impose a cool-down step on the monthly, annual, or 36-month test. Do it anyway; just don't cite it as the reason.

Why the 30% threshold is the crux of the whole program#

Emergency standby generators are sized for a worst-case simultaneous demand that, in practice, almost never occurs. A 500 kW set installed to cover a peak that materializes once a decade may see 60–90 kW of real building load on a Tuesday morning. That is 12–18% of nameplate — comfortably under the threshold.

This is not a defect in the installation. It is the normal consequence of sizing for life safety. But it does mean the monthly building-load test fails the 30% criterion, which is precisely what triggers the annual requirement below.

The Annual Supplemental Load Bank Test#

When the monthly tests do not reach 30% of nameplate kW and exhaust temperatures do not demonstrate adequate combustion, NFPA 110 requires an annual supplemental load test (§8.4.2.4). The current protocol is two steps:

• 50% of nameplate kW for 30 continuous minutes
• 75% of nameplate kW for 1 continuous hour

That is 1.5 continuous hours of loaded operation. "Continuous" is doing real work in that sentence: the sequence is not satisfied by two separate visits, and a shutdown mid-test restarts the clock.

Because the building cannot supply that load, it comes from a portable resistive load bank connected at the generator or at a designated connection point.

The requirement inside the requirement#

§8.4.2.3 requires that during the test, the supplemental load must be automatically replaced with essential building loads if normal power fails while the test is running.

This one is worth dwelling on, because it is the requirement most likely to be quietly unmet. If the utility drops while your generator is sitting on a load bank, the building's life-safety loads must get the generator automatically. A crew that has to notice the outage, walk outside, and throw a switch does not meet the standard — and, more to the point, does not protect the building. Ask any vendor quoting you an annual load bank test how they satisfy §8.4.2.3. The answer tells you a great deal about whether they have read the standard or are simply selling load bank hours.

The Triennial (36-Month) Test#

Level 1 EPSS installations carry a third obligation (§8.4.9): every 36 months, the system must run continuously for its assigned Class duration, or 4 continuous hours, whichever is less, at not less than 30% of nameplate kW.

For a diesel EPS the loading requirement is the same pair as the monthly test — not less than 30% of nameplate kW, or whatever maintains the manufacturer's minimum exhaust gas temperature — and a supplemental load bank is expressly permitted to meet or exceed the 30%. Most facilities need one, for the same reason they need one for the annual test.

Note what this test is and is not. It is an endurance and fuel-system test — can this system actually sustain the run time it was classified for? It is not a stepped-load performance test. And because it is bounded by the lesser of Class duration and 4 hours, a Class 48 system does not owe a 48-hour run; it owes 4 hours.

The test is initiated by operating at least one transfer switch test function and then activating the remaining ATSs, or by opening all normal power supply breakers — in other words, by making the building genuinely believe the utility is gone.

You may combine the annual and the 36-month test#

In the year both fall due, they can be satisfied by a single 4-hour event: the first 3 hours at ≥30% of nameplate, and the final hour at ≥75%. One mobilization, one load bank, one shutdown window — a meaningful saving on a large set, and the reason "4-hour load bank test" is the phrase many facilities use for this work.

Two cautions. Combining is an option, not the default: in any year where only the annual test is due, you still owe the two-step 50% / 75% sequence, and the 36-month clock is not reset by an annual test. And a standalone triennial run is not a stepped-load performance test — performed on its own it does not require the 30/50/75-style sequence and does not have to reach 75% at all.

Level, Type, and Class — The Three Classifications#

These three get conflated constantly, and you cannot determine what you owe without all three.

Only Level 1 systems owe the 36-month test. Level determines which requirements apply; Class determines how long the triennial run lasts; Type governs the transfer window your monthly cold start has to meet.

If you do not know your system's Level, Type, and Class, that is the first thing to establish — it is normally recorded on the EPSS documentation from commissioning, and it drives every interval above.

Documentation#

A test that is not documented did not happen, as far as an inspector is concerned.

NFPA 110 requires a written schedule for routine maintenance and operational testing (§8.3.3), and permanent records of all EPSS inspections, tests, repairs, and modifications (§8.5).

For each monthly test, capture:

• Date and time of test, and engine hours
• Nameplate data and ATS identification
• Transfer time and retransfer data
• Voltage and amperage per phase, and frequency
• Oil pressure, coolant temperature, exhaust temperature
• Cool-down readings and switch position verification

Maintenance records additionally need the date of the report, who performed the work, any unsatisfactory conditions found, and the corrective action taken — including verification that repairs were themselves tested.

NFPA 110 does not itself set a retention period. Your AHJ may, and healthcare accreditors generally expect three years of records available on site.

What a Load Bank Test Actually Involves#

If you have never watched one, the annual supplemental test is less dramatic than it sounds and takes most of a working day.

A resistive load bank — a bank of heating elements with forced-air cooling — is trailered to site and connected either at the generator's load terminals or at a permanently installed load bank connection point. Cable sizing is set by the test amperage, not the building's normal draw, so a 750 kW set needs substantially more copper than the crew brought for last month's inspection if nobody checked first.

The generator is started, brought up to voltage and frequency, and load is applied in steps. The technician holds 50% for the full 30 minutes, steps to 75%, and holds for the hour — recording voltage and current per phase, frequency, oil pressure, coolant temperature, and exhaust gas temperature at intervals throughout. What the test is looking for is not a pass/fail number at the end but drift: coolant temperature that keeps climbing rather than stabilizing, a phase current imbalance that widens under load, a frequency that sags at each step change.

Plan on 4–6 hours on site for a single set including setup, the 1.5-hour loaded sequence, cool-down, and teardown. Two things routinely extend that:

• The load bank connection point. Facilities with a permanently installed connection cabinet are far quicker to test than those requiring a direct connection at the breaker.
• Findings. A test that surfaces a cooling or fuel problem turns into diagnostic work, which is the entire point — you would rather find it on a load bank than during an outage.

Resistive vs. reactive load#

Most NFPA 110 supplemental testing is done with a resistive load bank, which loads the engine at unity power factor. That tests the engine, cooling system, fuel delivery, and exhaust — the things wet stacking damages.

A reactive load bank adds an inductive component so the set is loaded at a realistic lagging power factor (typically 0.8), which additionally exercises the alternator and voltage regulator. It is not what the annual supplemental test requires, but for a facility whose real load is heavily motor-driven or full of switch-mode power supplies, resistive-only testing leaves part of the system unexercised. Worth a conversation when you specify the work rather than an automatic upgrade.

How the Three Tests Fit Together Over a Year#

For a typical Level 1, Type 10, Class 4 healthcare installation whose building load sits below 30% of nameplate:

• Twelve monthly exercises, 30 minutes each under building load, with a cold start and documented transfer times.
• One annual supplemental load bank test — 30 minutes at 50%, 1 hour at 75% — because those monthly runs miss the 30% threshold.
• One 36-month run of 4 continuous hours, initiated through the transfer switches.

In the year the triennial falls due, you are performing all three. The triennial does not replace the annual supplemental test, and neither replaces the monthly exercise. Scheduling them as though one substitutes for another is the single most common program-level error we correct.

What Usually Goes Wrong#

From our own inspections, the recurring failures are not exotic:

• Running the monthly test unloaded. An unloaded exercise is not an EPSS test. It proves the engine starts; it proves nothing about the transfer switches or the system's ability to carry load.
• Assuming the annual load bank test is universal — or assuming it is never required. Both errors are common, and both come from not measuring the actual load during the monthly run. Measure it once and you know which regime you are in.
• Treating the 36-month test as a load bank test. It is a Class duration run initiated through the transfer switches, not a stepped-load performance test.
• No documentation of the transfer times. The generator starting is half the system. The ATS is the other half, and transfer time is what a Type 10 classification actually promises.
• Deferring the annual test after a light year. The requirement is not indexed to how much the generator ran.

Related Reading#

• NFPA 110 Required Equipment — what the standard requires you to have installed, not just tested
• Hospital Backup Power Compliance Checklist — the Level 1 / Type 10 healthcare case end to end
• NFPA 110 Compliance in California — state-specific overlay and AHJ expectations

We perform NFPA 110 load bank testing across California and Nevada, including the automatic-load-replacement provision in §8.4.2.3 and full documentation packages sized for accreditation review. See generator load bank testing and NFPA 110 compliance services, or get in touch to have your current program reviewed against the intervals above.