Direct answer: a UPS load bank test proves one thing no screen reading can: that this UPS, with this battery, carried a stated load for a stated number of minutes, at a recorded temperature, on a dated day. Ask for a discharge at your real site load (or a defined percentage of rating) to the runtime you need, with every battery block voltage logged and a signed report.
After a battery replacement: yes, run one once the new battery is fully charged. It confirms the installation and gives you the baseline every later test is compared with.
Written for facility, IT and compliance managers in Canada preparing an inspection, an insurance review or an audit. Sources checked 10 October 2026. GDF Technologies runs these tests on multi-brand UPS across Canada and replies within 4 business hours.
What a load bank test proves, and what it does not
A UPS spends almost all of its life doing nothing visible. The rectifier charges the battery, the inverter feeds the load, and the front panel reports an estimated runtime. That estimate is a calculation from the battery model and the present load. It is not a measurement. The only way to know how long the battery will really carry your equipment is to make it do so, under control, and to measure what happens.
A load bank is a controlled electrical load that can be set to a known kilowatt (kW) or kilovolt-ampere (kVA) value. Connected to the UPS output, it lets a technician discharge the battery at a defined power without risking the production load. Vertiv’s battery service data sheet puts the stakes plainly: UPS battery failure is the most common cause of unplanned outages [9].
A well-run test proves:
- the runtime the battery delivered at a stated load, temperature and end voltage;
- which battery blocks fell behind the rest, if every block voltage was logged;
- that the inverter held output voltage and frequency while the battery discharged;
- that connections and cables carried full current without overheating, when a thermal scan is taken under load.
It does not prove:
- how long the battery will keep that performance (it is a snapshot, so you trend it over years);
- that the generator, transfer switch and UPS work together, unless that sequence is tested separately;
- anything about blocks that were not measured. A total string voltage hides a weak block until the end of the test.
Scope, terms and assumptions
This guide covers single-phase and three-phase UPS with valve-regulated lead-acid (VRLA) batteries, the sealed blocks found in most UPS, in server rooms, data centres, hospitals and industrial sites. The system boundary runs from the battery string to the UPS output terminals. Lithium-ion UPS batteries are tested under their manufacturer’s procedure and battery management system, and are mentioned only where they differ. Generator testing is out of scope.
| Term | Meaning |
|---|---|
| Resistive load bank | A bank that converts electrical energy into heat at unity power factor. It loads the UPS in kW only [8]. |
| Resistive/reactive load bank | A bank that adds an inductive element to simulate a lagging power factor, typically 0.8, so the UPS is loaded in kVA as well as kW [8]. |
| Capacity test | A discharge at a published rate to a specified end voltage, compared with the battery manufacturer’s rating. It is the test IEEE 1188 uses to decide when VRLA batteries are replaced [2]. |
| Service test | A discharge at the site’s actual load profile for the runtime the site needs. IEEE 1188 treats an unsatisfactory service test as a possible reason for replacement [2]. |
| Runtime calibration | A manufacturer routine (APC Smart-UPS, for example) that runs the UPS on battery to correct its runtime estimate. It is not a capacity test [6]. |
| End voltage | The battery voltage at which the test stops, taken from the battery or UPS manufacturer’s data for the test rate. |
Assumptions. The battery is fully charged before the test, the critical load is protected during it (on a second source, on maintenance bypass with the risk accepted, or in a planned shutdown), and the person running the test is qualified for energized work under CSA Z462 [10].
Four tests that are often confused
Requests for quotation often say “load test” and mean four different things. Table 2 separates them, because each one answers a different question and costs a different amount of downtime.
| Test | Load | Stops when | What it answers |
|---|---|---|---|
| Capacity test | Constant current or power at a rate from the battery data sheet | Battery reaches the specified end voltage | What percentage of its rating does the battery still hold? [2] |
| Service (runtime) test | Your real site load, or a load bank set to it | The required runtime is reached, or end voltage, whichever comes first | Will the battery carry my load for the minutes I need? |
| Commissioning load bank test | Steps of 25, 50, 75 and 100% of rating, then a discharge | Each step is stable, then end of discharge | Does a new or repaired UPS meet its rating and transfer cleanly? |
| Runtime calibration | The connected load, at least 30% and steady within plus or minus 5% (APC) | The UPS shuts down or the routine ends | Is the runtime estimate on the panel realistic? [6] |
For an inspection or audit, the service test is usually what the reviewer wants to see, because it states the minutes your load received. A capacity test is what tells you how much margin is left and when to budget for a new battery. Many sites run both in one visit: a capacity test at the battery’s published rate when the schedule allows a long outage window, and a service test at site load when it does not.
How the test is set up
Figure 1 shows a typical arrangement. The load bank connects to the UPS output, downstream of the output breaker, while the critical load is fed another way. Measurement happens at three places: the UPS output (voltage, current, kW, kVA, power factor, frequency), the battery string (total voltage, discharge current and every block voltage) and the room (ambient and battery temperature).
Two practical points change the result. First, the cables from the UPS to the load bank must be sized for full current and kept short, and the bank needs somewhere to reject its heat: a 100 kW bank turns 100 kW into hot air. Second, block voltages must be logged during the discharge, not only before and after. The block that fails a test usually looks normal at rest.
Resistive or reactive: choosing the load bank
A UPS has two ratings on its nameplate: kVA (apparent power) and kW (real power). The ratio between them is its output power factor. A purely resistive bank works at unity power factor, so it tests kW only [8]. That is enough to discharge the battery at a known power, which is what a runtime or capacity test needs, because the battery supplies real power.
A resistive/reactive bank adds an inductive load and typically works at a 0.8 lagging power factor [8]. It is the better choice for a commissioning test, or when the question is whether the inverter holds voltage while carrying its full kVA rating, as it would with motor loads, transformers or older IT power supplies.
| Question | Bank | Set it to |
|---|---|---|
| How long will the battery carry my load? | Resistive | Your measured site kW (from the UPS log or a power meter) |
| Does the battery still hold 80% of its rating? | Resistive, or a DC bank on the battery | The constant power or current in the battery data sheet for the chosen test time |
| Does a new UPS meet its full rating? | Resistive/reactive | Steps up to 100% of rated kW, and kVA at the rated power factor |
| Will the UPS hold voltage with motor or transformer loads? | Resistive/reactive | The power factor of the real load, often near 0.8 |
Whatever the bank, check that its rated voltage and phase arrangement match the UPS output: 120/208 V, 480 V and 600 V systems all exist in Canadian buildings, and a bank rated for one does not load another correctly.
Method, step by step
Figure 2 shows the decision path most sites follow. The steps that follow it are what a provider should put in its method statement before the day.
- Collect the data. UPS model and rating, battery make, model, date code and number of blocks, the battery data sheet discharge table, the site load in kW, and the runtime the site requires.
- Check before discharging. Confirm the battery has been on float long enough to be fully charged, record each block’s float voltage and internal ohmic value, and take a thermal image of connections. A block already well above its neighbours on ohmic readings is a candidate to fail early in the test [4].
- Protect the load. Transfer the critical load to another source or agree a shutdown window in writing. Running a discharge with the production load still on a single UPS means the end of the test is also the end of the load’s power.
- Connect and set the bank. Set the load to the test value before opening the input, so the battery sees the intended power from the first second.
- Start and log. Open the UPS input (or simulate its loss as the manufacturer instructs) and start the clock. Log output and battery readings at fixed intervals, then more often as voltages approach the end value.
- Stop on the criterion. Stop at the required runtime (service test) or at the specified end voltage (capacity test). Stop at once if a block approaches reversal or heats abnormally.
- Restore and recharge. Restore the input, confirm the rectifier is charging, and leave the battery to recharge before the site counts on its runtime again.
Reading the result: the percentage that decides replacement
For a capacity test the result is expressed as a percentage of the manufacturer’s rating: the minutes the battery actually delivered to end voltage, divided by the minutes the data sheet promises at the same rate, times 100. Manufacturers publish their ratings at 25 °C [2], so the report must state the battery temperature and the temperature correction applied.
The threshold comes from IEEE 1188. Its 2005 edition, quoted in a paper Vertiv hosts from the Battcon conference, recommends replacing a VRLA cell, unit or battery whose capacity is below 80% of the manufacturer’s rating, and notes that at 80% the rate of deterioration is increasing even if the battery still meets the load [2]. DeMar quotes the clause to argue that capacity-recovery work should be tried before replacement [2]. IEEE 1188-2025 is now the active edition and supersedes the 2005 text [1]; check the current wording with your engineer before writing it into a specification.
| Item | Value |
|---|---|
| Data sheet runtime at the test power, 25 °C, to end voltage | 20 minutes |
| Measured runtime to the same end voltage, temperature-corrected | 15 minutes |
| Capacity | 15 ÷ 20 × 100 = 75% |
| Reading against the IEEE 1188-2005 criterion | Below 80%: plan replacement [2] |
| Service view | If the test power equals the site load and the site needs 10 minutes, it still has them today; the margin is shrinking |
The last row matters for audits. A battery can fail the 80% test and still carry the site for the minutes it needs, which is why a service test and a capacity test can give different but equally true answers. Report both numbers, and say which one the decision is based on.
How often, and when to test
No single Canadian rule sets one interval for every UPS. The references below are what most specifications draw on.
| Reference or trigger | What it says | Source |
|---|---|---|
| IEEE 1188-2005, VRLA capacity test | At intervals of 25% of expected service life or two years, whichever is less | Megger summary [3] |
| NFPA 70B, 2023, Table 9.2.2, UPS electrical testing | Every 12, 6 or 3 months for equipment in physical condition 1, 2 or 3; special procedures every 24 months, where manufacturer intervals are not available | [5] |
| APC runtime calibration | No more than once every 6 months, only when needed, because it wears the battery | [6] |
| After a battery replacement | Once fully charged: confirms the installation and sets the baseline for later tests | Good practice |
| Before an inspection or audit | Weeks ahead, so a failed result can be corrected before the visit | Good practice |
| Runtime alarm, sudden runtime drop, ohmic values rising | A targeted test to confirm or rule out a weak string | [4] |
Every discharge consumes some battery life. APC says so directly about its calibration routine [6]. Testing more often does not make a battery healthier; it gives you more data points. Ohmic readings between discharge tests are the cheap way to decide when the next discharge is worth it: in one refinery’s field study presented at Battcon 2002, cells whose internal resistance sat more than 23% above their own system’s baseline all failed a capacity test, while none below the baseline failed [4].
Canadian realities
- Temperature. VRLA ratings assume about 25 °C [2]. APC states that battery life halves for every 8 °C above that [7]. A test run in a cold room in January will deliver less runtime than the same battery in July, so the report must state the temperature and the correction applied. A summer test in a poorly cooled room can show a battery that is ageing faster than its date code suggests.
- Heat rejection in winter. A load bank exhausts its full test power as hot air. In winter that heat is welcome indoors only if the room can take it; an outdoor bank needs a cable route, weather protection and a clear exhaust path away from air intakes.
- Voltage. 600 V three-phase distribution is common in Canadian commercial and industrial buildings. Confirm the bank’s rating before the day.
- Language of the record. In Quebec, ask for the report in the language your inspector, insurer or public client reads. GDF Technologies issues reports in French or English.
Failure modes and safety limits
- Arc flash and shock. Connecting a bank to a live UPS output and measuring battery blocks under discharge is energized work. Follow CSA Z462 for risk assessment and protective equipment [10].
- Block reversal. When one block in a series string runs empty before the others, the rest of the string can drive it negative. Block voltage logging is how you see it coming; stop the test before it happens.
- Load drop. APC warns that a faulty or end-of-life battery can cause a load drop during calibration [6]. The same is true of any discharge, which is why the critical load must be protected.
- Interrupted test. A capacity test that is paused outside the procedure in the battery standard (for example, to bypass a failed cell) may not give a valid figure. APC states that its calibration must not be interrupted to give valid results [6].
- False reassurance. A short self-test or a panel runtime estimate is not a test. Neither is a total string voltage that looks fine at rest.
Checklist and records to keep
| Section | Content |
|---|---|
| Identification | Site, UPS make, model and serial number, battery make, model, quantity and date codes, date, technician name |
| Method | Test type (service, capacity, commissioning), bank type, set load in kW and kVA, end criterion, protection of the critical load |
| Before | Float voltage of each block, ohmic values, ambient and battery temperature, thermal image of connections |
| During | Time-stamped output V, A, kW, power factor and frequency; string voltage and current; every block voltage |
| Result | Runtime achieved, capacity percentage with the temperature correction shown, comparison with the previous test |
| Findings | Blocks that fell behind, connections that heated, alarms, and a recommendation with a date |
| Sign-off | Signature, and confirmation that the battery was recharging when the technician left |
Keep every report for the life of the battery and the UPS. The trend across several tests is worth more than any single result, and it is what an auditor reads to see whether the site manages its risk or reacts to it.
Quick answers
What does a UPS load bank test prove?
That the UPS and its battery delivered a stated load for a stated time, at a recorded temperature, on a dated day, with each battery block’s behaviour logged. It does not predict how long the battery will keep that performance.
Do I need a load bank test after replacing UPS batteries?
Yes. Once the new battery is fully charged, a discharge at site load confirms the installation (polarity, connections, charger settings) and records the baseline that later tests are compared with.
Is the UPS self-test or runtime calibration enough for an audit?
No. A calibration corrects the runtime estimate and needs a steady load of at least 30% [6]; it does not log block voltages or produce a capacity figure.
How long does the test take?
The discharge lasts as long as the runtime you are proving, often 5 to 30 minutes for a UPS, plus set-up, readings before and after, and the recharge, which takes hours.
Sources
- IEEE 1188-2025, Recommended Practice for Maintenance, Testing, and Replacement of Valve-Regulated Lead-Acid (VRLA) Batteries for Stationary Applications (active standard, supersedes 1188-2005). standards.ieee.org
- Peter J. DeMar, “Report on the Long Term Results of Battery Capacity Recovery Processes for VRLA Cells”, Battcon 2008 (quotes IEEE 1188-2005 clause 8, Battery Replacement Criteria). vertiv.com
- Megger, “Standard Battery Testing Requirements Summary” (IEEE 450-2010, IEEE 1188-2005, ANSI/NETA MTS-2015). megger.com
- William Bullis, ChevronTexaco, “Proactive Battery Maintenance”, Battcon 2002. vertiv.com
- NFPA 70B, Standard for Electrical Equipment Maintenance, 2023 edition, Table 9.2.2, as reproduced in the HVM maintenance interval chart. hvmcorp.com
- APC by Schneider Electric, FAQ FA284198, “How do I perform a Battery Calibration on my Smart-UPS?” se.com
- APC by Schneider Electric, FAQ FA158934, “What is the expected life of my APC UPS VRLA battery?” se.com
- Aggreko, “Resistive/reactive load banks”. aggreko.com
- Vertiv, “Battery Preventive Maintenance” data sheet. vertiv.com
- CSA Z462:24, Workplace electrical safety. webstore.ansi.org
Book a load bank test
GDF Technologies tests multi-brand UPS (APC, Eaton, Vertiv and Liebert, Tripp Lite, Mitsubishi, Delta and others) across Canada, with same-day service in Montreal, Quebec City, Toronto and Ottawa. We supply the load bank, protect your critical load, log every battery block and deliver a signed report you can hand to an inspector. Call 514-252-8324, write to support@gdftech.com or request a test; we reply within 4 business hours.
Related reading: our UPS preventive maintenance service, the preventive maintenance checklist to require from a provider, seven measurable signs of a battery at end of life, when to replace UPS batteries, the maintenance report format building inspectors expect, our UPS runtime calculator, three-phase UPS commissioning and UPS battery replacement. Replacement VRLA blocks and cartridges are available from UPSPLUSBATTERY’s replacement battery catalogue.
Written by Christian Barkley, Director, GDF Technologies (LinkedIn). Rebuilt 10 October 2026. The brand names cited belong to their owners.
