How to Size a UPS for Your Site

1. Build a complete load schedule

Sizing starts with a list of every device the UPS will protect. Walk the room with a clipboard or use the connected-load report from your switchboard meter. For each device record the nameplate Watts (real power), and quantity. Watts is the steady-state real power the device draws — most modern equipment nameplates state Watts directly because that is what the user actually pays for in electricity.

You will sometimes see equipment rated in Volt-Amperes (VA) instead. VA is apparent power and is always greater than or equal to Watts. The ratio between them is the power factor (pf): Watts = VA × pf. UPS systems are rated in VA because their power-handling capacity is limited by current, not heat. To size a UPS you ultimately need a number in VA — but you start by collecting Watts because that is what equipment specifies.

Capture small loads too. A 50W ethernet switch and a 30W SFP module sound trivial, but ten of them across a comms rack adds up to 800W. If you have not protected them you will lose monitoring during an outage right when you need it most.

When in doubt, measure. A clamp-on AC ammeter on the input cable gives you actual draw in Amps; multiply by voltage (230V single-phase, 400V three-phase) to get VA, and apply 0.95 power factor for modern PSUs to get real Watts. A one-day metering exercise on the existing rack saves you from a year of UPS over-runs.

2. Apply headroom and power-factor conversion

Once you have total connected Watts, add 25% headroom. The headroom covers three things: PSU inrush at startup (PSUs draw 4-8x steady-state during the first 50ms after a transfer), planned growth across the UPS service life (typically 7-10 years), and the gap between nameplate and steady-state draw on equipment you have measured.

Mining and motor-heavy industrial loads need more — 40-50% headroom is normal because of motor inrush. Healthcare and steady IT loads can run as low as 15% if growth is well-controlled. The default of 25% is a defensible compromise.

After applying headroom, convert from Watts to VA at a conservative 0.9 power factor: required VA = (Watts × 1.25) / 0.9. This produces a UPS rating that has correct steady-state capacity and enough VA headroom to handle inrush and asymmetric loads.

Modern UPS units are increasingly sold with unity power factor (pf = 1.0). On those, your Watts and VA ratings converge — you can size at pf = 0.95-0.99 with confidence. Always check the manufacturer datasheet for the exact output rating.

3. Round up to a standard rating

There is no point specifying a custom 7.4 kVA UPS when a 10 kVA off-the-shelf unit is cheaper, has better lead times, attracts manufacturer warranty, and gives you 30% more growth headroom. UPS systems are sold in standardised steps: 1, 1.5, 2, 3, 5, 6, 10, 15, 20, 30, 40, 60, 80, 100, 150, 200, 300, 400, 500 kVA.

Round your calculated VA up to the nearest standard step. If your calculation lands at 7.4 kVA, you specify a 10 kVA. If it lands at 11 kVA, you specify a 15 kVA. The over-sizing is intentional — you are buying flexibility, not waste.

4. Select topology

UPS topologies fall into three classes. Standby UPS (sometimes called offline) is the cheapest — utility power feeds the load directly until an outage, then the UPS kicks in within ~10-15 ms. It is suitable for desktop PCs and basic comms but is too slow for modern PSUs and is rarely specified above 1 kVA.

Line-interactive UPS adds an autotransformer that buck-and-boosts utility voltage during sags and surges. Switchover to battery is faster (~5 ms) and the load sees a regulated voltage even on poor utility power. Line-interactive is appropriate for sub-1.5 kVA loads on clean utility supply.

Online double-conversion is the gold standard. Utility power passes through a rectifier (AC to DC), is stored on a DC bus, and then re-inverted to AC for the load. The output is always synthesised by the UPS, fully isolated from utility transients, and switches to battery with zero transfer time. Online double-conversion is the right choice for any mission-critical load above 1.5 kVA: server rooms, healthcare, data centres, telco.

5. Define redundancy class

Redundancy is how many UPS modules you specify above the load minimum. Notation: N is the base load capacity, N+1 is one spare, 2N is fully duplicated.

N (no redundancy) is fine for non-critical loads where a single UPS failure is recoverable. The cost is lowest but you have no protection if the UPS itself fails — a fact that is sometimes lost on small-business buyers who think the UPS is the redundancy.

N+1 means you have one spare module. If the UPS or one battery string fails, the load continues to be supported. This is the standard recommendation for any commercial server room, healthcare system, or revenue-generating IT environment.

2N is fully duplicated — two complete UPS systems, each capable of running the full load on its own. This is required for Tier IV data centres and concurrently maintainable Tier III deployments. Capital cost is roughly double, but it lets you do live UPS maintenance without ever putting the load at risk.

6. Battery runtime budget

Runtime is the second number you have to size for, after capacity. The question is: how long does the UPS need to hold up the load before the generator kicks in or the load is gracefully shut down?

For sites with backup generators: 5-10 minutes is normal. The UPS bridges the gap from utility loss until the generator is online and stable. Five minutes is enough for a well-maintained generator; 10 minutes gives you margin if the first start fails and you have to retry.

For sites without generators: 15-30 minutes is normal. This is enough time to gracefully shut down servers, save state, or transition to laptop power. Anything longer than 30 minutes is an unusual specification — you are essentially using the UPS as battery backup, which is the wrong tool for that job.

Battery sizing comes back to the runtime calculator: runtime ≈ (battery Wh × 0.8 × inverter efficiency) / load Watts × 60. Use our runtime calculator on a UPS rating + battery configuration to validate.

7. Battery chemistry: VRLA or lithium?

VRLA (Valve-Regulated Lead-Acid, also called sealed lead-acid) is the default. It is cheaper upfront, the supply chain is mature, every UPS manufacturer supports it. Service life is 4-5 years under typical Australian conditions, longer in climate-controlled rooms, shorter in 35°C+ environments.

Lithium-iron-phosphate (LFP) is the upgrade choice. Capex is roughly 2-3x VRLA but service life is 10-15 years and the BMS gives you per-cell state-of-health metrics that VRLA cannot match. Lithium also runs at full capacity above 30°C ambient where VRLA capacity drops sharply.

Total cost of ownership over a 15-year UPS life favours lithium for sites that have either: (a) >5 kVA continuous load, (b) >40°C ambient, or (c) growing runtime requirements. The upfront cost is recovered through avoided VRLA replacement cycles (typically 3-4 across a 15-year life).

8. Compliance — Australian Standards

Every UPS installation in Australia must meet AS/NZS 3000 (the wiring rules), AS IEC 62040.3 (UPS performance), and AS/NZS 5139 if the battery system exceeds the size threshold (typically 300 Ah/cell or stationary battery banks above ~10 kWh).

AS IEC 62040.3 classifies UPS performance using a three-letter code. The first letter is the input dependency (V = voltage and frequency dependent, S = voltage independent, I = independent). The second letter is the output waveform (S = sinusoidal, X = non-sinusoidal). The third is the dynamic performance class (1 = highest, 3 = lowest). A typical online double-conversion UPS classifies VFI-SS-111. A line-interactive unit might be VI-SS-211.

The most common compliance miss is the harmonic-emission requirement under AS/NZS 61000.3.2 / 61000.3.4. UPS rectifiers can emit current harmonics back onto the supply, distorting voltage waveforms for other equipment in the building. Modern transformerless UPS with active filtering meets the limits comfortably; older 6-pulse rectifier UPS may not.

9. Worked example: 50-rack data hall

Site profile: a 50-rack production data hall, average 6 kW per rack, with a backup generator on a 5-minute cold-start cycle. Site is in Brisbane, climate-controlled to 25°C.

• Connected load: 50 racks × 6 kW = 300 kW
• With 25% headroom: 300 × 1.25 = 375 kW
• At pf = 0.95 (modern PSUs): required VA = 375 / 0.95 = 395 kVA
• Round up to standard rating: 400 kVA
• Topology: online double-conversion, 3-phase, modular
• Redundancy: N+1 = 2 × 400 kVA modules in parallel, one as spare
• Battery: VRLA at 25°C, sized for 5 minutes at full load
• Runtime calculation: 400 kVA × 0.95 pf = 380 kW; battery bank ≈ 32,000 Wh
• Compliance: AS IEC 62040.3 VFI-SS-111, AS/NZS 5139 if battery >10 kWh

Specification: two parallel 400 kVA modules (Liebert APM2, Galaxy VS or Eaton 93PM), each with internal-cabinet VRLA at 32 kWh per module, 5-minute runtime at full load. Capex roughly A$280-340k for the UPS alone, excluding battery cabinet and switchgear.

10. Specification checklist

Use this checklist against any vendor proposal you receive.

1. kVA / kW rating documented and traceable to a load schedule
2. Headroom % stated explicitly (default 25%)
3. Power factor stated (output pf, not input)
4. Topology stated (online double-conversion VFI-SS-111 for mission-critical)
5. Redundancy class stated (N, N+1, 2N)
6. Battery runtime in minutes at full load and at 50% load
7. Battery chemistry, capacity in Wh, and service life
8. AS IEC 62040.3 performance class shown on the datasheet
9. AS/NZS 5139 compliance documentation if lithium
10. Manufacturer warranty terms and what voids them
11. Installation, commissioning, and handover scope
12. Maintenance contract pricing for years 1-5