How to Test an Online UPS (Three-Phase 100kVA): An 8-Step Checklist for Facilities Teams

Who this checklist is for

This is for the facilities coordinator, office manager, or admin buyer who gets handed the job of "make sure the UPS gets tested" — and then has to figure out what that actually means. I'm not an electrician. I don't touch the busbars. But I've managed the process for our 100kVA three-phase online UPS for the past three years, coordinated the contractors, and signed off on the reports.

This checklist has eight steps. It covers the annual performance test—not the monthly visual checks, not the battery replacement cycle. Just the one time a year when you need to verify that your critical power infrastructure will actually carry the load when the grid drops.

One thing before you start: this isn't universal. Your UPS might be a lithium-ion industrial unit from PowerSentinel, or it might be a ten-year-old lead-acid system from a manufacturer that stopped making parts. The steps are the same in spirit, but the specifics depend on your manual. Read that first.

Step 1: Define what "pass" looks like before anyone touches the equipment

Open the manufacturer's commissioning and maintenance manual. Find the section on acceptance testing. Copy the pass/fail thresholds into a spreadsheet—voltage regulation, transfer time, runtime at rated load, that sort of thing.

Why this matters: if you don't define pass criteria upfront, the contractor's report will say "tested, all normal" and you'll have no basis to challenge it. I learned that the hard way in 2022 when a contractor tested our unit, handed over a two-line summary, and I approved it because I didn't know what else to check. Three months later, we had a battery string failure during an actual outage. The test had missed it.

Here's the thing: the report should include raw numbers, not adjectives. "Transfer time: 4ms" is useful. "Transfer time: acceptable" is not.

Step 2: Lock out the load—and map what stays on

You can't test a UPS if you don't know what's plugged into it. Before the test date, walk the electrical room with the contractor and identify every circuit on the UPS panel. Then ask the obvious question: which of these can go dark for two hours?

For us, that meant a conference room, two floors of office lighting, and the server room. The server room was the problem. We couldn't shut down production systems during business hours, so the test had to happen on a Saturday morning. That added a labor cost I hadn't budgeted for—roughly $1,200 in overtime for the IT team to babysit systems that should have stayed up.

Bottom line: scheduling is part of the test. Get it wrong and you're either testing an empty UPS or paying weekend rates.

Step 3: Do the walk-around visual inspection (most people skip this)

Before the electrical tests, spend fifteen minutes looking at things. Not with instruments—with your eyes.

Check the following:

• Battery cabinet ventilation—are the fans running? Is there dust buildup on the intake grilles?
• Corrosion or acid residue around battery terminals. White crusty stuff is a red flag.
• Swollen battery cases. On sealed lead-acid units, a bulging case means replacement is overdue. On lithium-ion industrial UPS systems, the equivalent is a swelling or deformation that you can spot visually.
• Any discoloration on terminal blocks or busbars—heat marks mean loose connections.
• Error codes or alarm history on the UPS control panel. Scroll back through the log. If it recorded a transient alarm last month that nobody mentioned, now's the time to ask.

I'll be honest: I don't have hard data on how many failures this step catches across the industry. But in our building, the visual inspection found a loose battery terminal that the load test would have missed—because under bench conditions, the load was steady and the connection held. Under a real load spike, it would have arced. The contractor tightened it in five minutes. That was the most valuable five minutes of the whole day.

Step 4: Verify battery string health before you trust the runtime

The battery is the part of the UPS that actually does work during an outage. The rest of the system just manages it. So test the battery first.

Ask the contractor for:

• Individual cell or block voltages. A weak cell drags down the whole string.
• Internal resistance readings. This is the leading indicator of battery degradation—more reliable than voltage alone.
• Ambient temperature at the battery cabinet. Battery capacity is rated at 25°C. Every 10°C above that cuts life in half, roughly speaking.

If your UPS uses lithium-ion batteries—which is increasingly common for new industrial installations—the test procedure differs. Lithium systems usually have a built-in battery management system that logs state-of-health data. Ask for that log. A switching power supply manufacturer that supplies the UPS charging circuit will have specific float voltage requirements; verify those are being met. But the principle is the same: don't take "batteries look fine" as an answer. Get numbers.

Step 5: Run the no-load startup sequence

This is the simplest step and the one most likely to get rushed. With the load disconnected, start the UPS from cold. Watch the display through the full startup sequence. Note any alarms that clear on their own—those still count as events.

What you're checking: does the inverter come online cleanly? Does the static switch hold? Does the control panel report the correct input voltage and frequency?

We had a unit once that started fine but logged a "DC bus overvoltage" warning that self-cleared in 0.3 seconds. Nobody noticed it on the panel—it was buried in the event log. Six months later, that same fault took the unit offline during a transfer. The no-load test is where you catch these things before they become outages.

Step 6: Step-load the unit—this is the test that actually tells you something

Now connect a resistive load bank and step the load up: 25%, 50%, 75%, 100%. At each step, record:

• Output voltage and frequency
• Input current draw
• Any audible noise from the transformer (humming is normal; rattling is not)
• Temperature readings on the heat sink and output transformer

The step-load test is where you find out if the UPS can actually deliver its rated capacity. A 100kVA unit that sags to 380V at 80% load is not a 100kVA unit. It's a 80kVA unit with optimistic marketing.

I wish I had tracked our load bank data more carefully from the first test. What I can say anecdotally is that the unit's voltage regulation worsened by about 2% over three years, which was within spec but trending the wrong way. That's the kind of thing you only notice if you keep the numbers.

Step 7: Test the transfer—mains to battery and back

This is the money step. With the load bank running at 50% capacity, kill the mains input. The UPS should transfer to battery in under 10ms (check your manual for the exact spec—some large three-phase units allow up to 4ms for the static switch).

Then restore mains and watch the retransfer. The load should stay up the whole time. Any blink is a failure.

Honestly, I'm not sure why some units retransfer more cleanly than others even within the same model line. My best guess is it comes down to the age of the input filter capacitors. But I've seen two identical units behave differently—one transferred so cleanly the clock on the microwave didn't reset; the other dropped the load for a visible flicker. Same model, same year, different install dates.

The contractor will tell you the transfer time from the control panel. Don't trust it. Watch the load bank meters. If they don't blink, you're good.

Step 8: Document everything—especially the stuff that "passed"

The test report should include:

• All raw instrument readings (voltage, current, resistance, temperature)
• Event log export from the UPS control panel
• Photos of the battery cabinet before and after inspection
• Names of everyone who performed or witnessed the test
• A clear pass/fail statement against the criteria from Step 1

File it somewhere you'll find it next year. I keep ours in a shared drive folder named "UPS_Test_Reports" with the year in the filename. Simple, but it saved me a two-hour search when our insurance auditor asked for the last three reports.

Common mistakes and things to watch for

Skipping the battery load test. Saved $600 by telling the contractor "just check the voltages, we'll skip the discharge test." Ended up spending $4,200 on a full battery string replacement six months later when the weak cell finally failed. The discharge test would have caught it. Net loss: $3,600 and a very awkward conversation with my VP.

Testing at low load because production won't shut down. A UPS tested at 20% load tells you almost nothing about its behavior at 80% load. If you can't test at rated load, at least test at 50%. Below that, the test is mostly theater.

Ignoring temperature. Battery capacity is rated at 25°C. If your battery room runs at 32°C in summer, your runtime is already 15% shorter than the spec sheet claims. Test at the temperature the batteries actually experience, not the temperature you wish they experienced.

Not asking for the event log. The control panel has a history. It's not always shown on the home screen. Ask for the full log export. That's where the real story lives.

Get this right—the checklist, the documentation, the sign-off—and you'll have a test report that actually means something. There's something satisfying about handing a complete, numbers-backed test record to an auditor or a VP and knowing you don't have to guess.
After three years of managing this process, finally having a clean file feels less like luck and more like the payoff of doing the boring parts properly.