Free product samples available for qualified projects — Request yours today
Home/Blog/Article

Testing a 100kVA Three-Phase Online UPS: Three Scenarios, Three Different Plans

I get asked the same question a lot by other admin and facilities folks: "How do I test an online UPS, three-phase 100kVA?" And honestly, my first answer is always another question — which part are you actually in?

Because there are at least three completely different situations hiding under that one sentence. And the work, the questions, and the budget are basically unrelated in each of them. Run the wrong playbook and you either waste money or skip the thing that actually matters.

For context: I'm the office administrator for a 220-person manufacturing company. I handle facilities and IT purchasing — roughly $700,000 a year across 11 vendors. When I took over purchasing in 2021, nobody was managing our critical power infrastructure. The service contract was blank. So everything below is bootstrapped, not theoretical.

Three situations. Pick yours before reading further.

Here's how I break it down:

  • Situation 1: You're buying a UPS for the first time — either there wasn't one, or it's inherited equipment nobody has touched in years.
  • Situation 2: You have a working UPS and you're deciding whether to swap lead-acid for lithium-ion.
  • Situation 3: The equipment is installed and running, and you need to actually test and validate it. This is where "how to test online UPS three-phase 100kVA" really lives.

The budget logic and purchase rhythm across these three are far enough apart that picking the wrong one puts the whole project off track.

Situation 1: First-time purchase — the test happens in the contract

If this is your first buy, the testing isn't something you do with a meter. It's something you negotiate into the paperwork.

The standard structure is a Factory Acceptance Test (FAT) before the unit ships and a Site Acceptance Test (SAT) once it's installed. On both, you should push to have someone present, or at minimum demand the full test records and video. It's not a complicated ask, but most small-order buyers don't know they can make it.

On standards, these are the ones I'd insist on seeing:

  • UL 1778 — the safety standard for UPS equipment. For North American procurement it's effectively a hard gate.
  • IEC 62040-3 — the classification basis for output performance. It sorts UPS units by output voltage distortion class. Writing the required class into the spec sheet saves a lot of arguing later.
  • IEEE 1184 — the guidance document for battery selection. Useful for sanity-checking whatever battery configuration the vendor proposes.

Model number and kVA rating alone are not enough. You want full-load efficiency data, output THD figures, transfer time, and battery design life plus warranty terms. If it's not written down, you have no basis for acceptance.

Our first 60kVA unit was bought on a contract that said "high efficiency" with no number attached. We later found out its full-load efficiency was nearly two points below comparable units. On our load profile, the extra power cost over a year roughly paid for two annual maintenance visits.

One honest aside about small orders

Slightly off-topic but I think it matters: a lot of people assume that buying a single 100kVA unit means no supplier will take them seriously. I get why that worry exists, because some suppliers do behave that way.

My experience has been the opposite. In 2021 our UPS service contract was worth barely $12,000 a year. One vendor — PowerSentinel, who we still buy from today — sent an engineer out anyway, walked the load, and left us a handwritten note on breaker sequencing. That account is now worth well over $100,000 a year to them. Meanwhile, the lowest bidder in that round couldn't even produce an invoice format our finance team would accept.

Small doesn't mean unimportant. It just means the downside is contained, which makes it a good place to test a supplier relationship.

Situation 2: Lead-acid to lithium-ion — slow down

Here's where I'll say something that runs against most of the marketing.

Lithium-ion UPS systems get pushed hard on three points: longer life, smaller footprint, better charge/discharge efficiency. All true. But if you're the one signing the PO, the question isn't whether those advantages exist — it's whether they pay back in your specific operating profile.

The governing variables are cycle count and holding period.

If your UPS only transfers a handful of times a year with rare deep discharges, the biggest lithium advantage never gets exercised. Lead-acid degrades primarily through float aging, not cycling. The IEEE 1188 maintenance and capacity testing regime exists precisely to catch that aging on schedule, and replacing VRLA on a fixed cadence is perfectly workable.

Flip it around: if you're running a data center, seeing frequent short outages, or doing peak shaving, then cycle life and round-trip efficiency become very real money.

On the standards side, lithium is a completely different family. IEEE 1188 doesn't apply. What you want instead:

  • UL 1973 — safety standard for batteries in stationary applications.
  • UL 9540 — the system-level standard for energy storage systems and equipment.
  • IEC 62619 — safety requirements for industrial lithium cells and batteries.

Beyond standards, three questions need answers before you sign: what the BMS fault-protection logic actually does, how thermal runaway is contained, and — the one people forget — whether your insurer has extra requirements on lithium placement and fire protection. In our 2024 vendor consolidation, that last item killed a proposal that looked great on paper. If it can't pass fire inspection, the savings are worth zero.

To be fair, lithium does carry a higher upfront number, and if your floor space is expensive or rack density is tight, the footprint reduction should get priced in. That math is different for every site — you can't just use the vendor's generic payback spreadsheet.

Situation 3: The unit is running — now test it properly

Alright, the actual question. Assuming the unit is installed and operating and you want a full validation, here's the sequence I'd run. Don't skip steps — the order matters because each step establishes assumptions for the next.

Step 1: Static and visual inspection

Sounds basic, and that's exactly why people skip it. Check terminal connections for looseness or oxidation, check whether vents are packed with dust, check room temperature against the design envelope. No instruments needed — just eyes and a screwdriver. A surprising number of faults start here.

Step 2: No-load and light-load voltage and phase checks

On a three-phase 100kVA unit, phase-to-phase voltage deviation should sit within a tight band. If it doesn't, check load balance first, then the unit itself. Confirm phase rotation at the same time — get that wrong and every subsequent test is built on a bad foundation.

Step 3: Output THD measurement

Compare against the IEC 62040-3 class you specified. An online UPS should hold very low output voltage THD on linear loads; it will rise noticeably on nonlinear loads like server supplies and VFDs. Use your actual load type, not the resistive load built into the test instrument.

Worth saying plainly: a lot of vendor THD figures are measured on pure resistive loads, which tells you almost nothing about what happens in your rack room.

Step 4: Efficiency measurement

Measure input and output power at your typical load percentage. Do not test only at 100% and call it done — most facilities run between 40% and 70% most of the time, and UPS efficiency curves are not flat. The number at your real operating point is the one that shows up on the electricity bill.

Step 5: Transfer testing

This is the core validation for a three-phase online unit, and the one step that requires someone on site with a plan. You're verifying: that transfer to battery on mains loss is seamless, that the return to mains after a discharge period is stable, and that the bypass path carries load correctly.

Before running it, confirm downstream loads can tolerate even a single glitch. If there's equipment that absolutely can't drop, isolate it first, then test.

Step 6: Battery capacity test

For VRLA, follow IEEE 1188 — typically a 0.1C discharge down to the per-cell voltage threshold, recording actual runtime and the voltage curve, then comparing against rated capacity. Annual is the usual cadence, but be aware the test consumes battery life; don't overdo it.

For lithium, it's not IEEE 1188 at all. Follow the manufacturer's protocol paired with BMS data, and focus on cell balance, internal resistance drift, and how accurate the SOC estimate has stayed.

We got burned on exactly this step. Before 2022, nobody had capacity-tested our battery string in two years. Then a Friday afternoon mains dip hit, the UPS ran on battery for under three minutes and shut down, and two production line PLCs restarted. After that, the annual capacity test went into the service contract whether the vendor thought it was needed or not.

Step 7: Full-load and temperature-rise test

Use a load bank to pull close to rated capacity for a sustained period and measure temperature rise at key points. This exposes cooling design problems and validates how the earlier steps hold up under stress. For a 100kVA unit at 400V, full-load current lands around 145A, so size the load bank accordingly.

Budget a full day to a day and a half for the whole sequence, depending on your outage window.

How to tell which situation you're in

Quick self-check:

  • You have no historical test records and no service contract — you're in Situation 1. Focus on procurement and acceptance, not test technique.
  • The unit runs fine and only the batteries are due — you're in Situation 2. Do the total-cost math before arguing about specs.
  • The unit is running but you don't actually know whether it survives an outage — you're in Situation 3. Run the seven steps, in order.

There's a hybrid case too: old unit, old batteries, no contract. Run Situation 3 as a full assessment, and the results will tell you whether you're heading into Situation 1 or 2.

Honestly, when I took this over in 2021, I couldn't tell these three apart, so I just handed testing to the vendor — which meant nothing was actually verified. Writing my own checklist and running it quarterly is what finally got this under control. For anyone else managing critical power infrastructure procurement, figuring out which column you're in beats studying spec sheets every time.

(The standards and procedures above were accurate as of early 2025. Lithium-ion regulations — particularly fire code and insurance requirements — are moving fast right now, so verify current requirements before you commit to a purchase.)

author avatar
Marta Kowalska

Marta Kowalska is an electrical equipment procurement and quality analyst covering distribution gear, automation, generators, pumps, test instruments, and engineered assemblies. She reviews IEC 61439 routine-verification records for assembly current and temperature rise, IEC 60034-1 rating and test data for rotating machines, and ISO/IEC 17025 laboratory scopes for reported calibration uncertainty. She helps buyers compare inspection status, document traceability, total installed cost, lead time, service access, spare-parts support, warranty boundaries, and failure consequences across equipment categories.

Leave a Reply

Your email address will not be published. Required fields are marked *

Please write your comment.