How to Size a UPS

An uninterruptible power supply (UPS) carries the critical load through the gap between a utility outage and the standby generators taking over. Sizing one comes down to five decisions: the design load, the power rating in both kW and kVA, the loading limit, the redundancy scheme, and the battery runtime. They're made in that order, because each one feeds the next.

Step 1: Establish the design load

Start from what the protected equipment actually draws, not from its nameplate. A server's power supply rating is the most it could ever draw, and real IT load commonly sits well below the sum of nameplates. Measured load from a running facility, or the IT vendor's power calculator for a new one, is the better basis.

Then add growth. The UPS will be in service for years, and the load rarely shrinks. The growth allowance is a design decision, not a formula, but it should be written down, because it's the number most often argued about later.

Step 2: Check kW and kVA, both

A UPS has two output limits: real power (kW) and apparent power (kVA). The ratio between them is the UPS's rated output power factor. Older three-phase designs were commonly rated at 0.8, so a 500 kVA unit could deliver only 400 kW. Many current designs are rated at 0.9 or 1.0. Convert the load to kVA using its power factor (kVA = kW ÷ PF) and confirm it fits both ratings. The kW vs. kVA guide covers the conversion in detail.

Step 3: Don't plan to run it at 100%

Many designers keep steady-state loading at or below about 80% of the UPS rating. That leaves room for load steps, inrush from equipment restarting, and measurement error in the load estimate. This is design practice rather than a code requirement, and the right limit depends on the load profile and the manufacturer's guidance.

Step 4: Apply the redundancy scheme

Redundancy is about the whole path, not just the UPS. An N+1 UPS feeding a single distribution path is still a single point of failure downstream. N+1 vs. 2N Redundancy Explained covers what each scheme survives and what it costs.

Step 5: Size the battery for runtime

Runtime only has to cover the gap to the generators, with margin. Standby generators for emergency systems are commonly required to pick up load within 10 seconds (NFPA 110, Type 10). The battery has to cover that, a possible failed first start, and the time to transfer. Specifications commonly land between about 5 and 15 minutes at full load. Sizing for a long outage is usually the wrong instinct: that's the generator's job, and every extra minute is battery you replace on a cycle.

The energy the battery must deliver is the load multiplied by the runtime, divided by the efficiency of the conversion from battery to output. Then allow for the battery losing capacity with age: IEEE 485 applies an aging factor of 1.25 to lead-acid batteries, so the battery still meets the runtime when it has faded to 80% of rated capacity at replacement. For lithium-ion, the manufacturer's end-of-life capacity figure plays the same role.

Final selection is made from the battery manufacturer's constant-power discharge tables, not from kWh alone. Battery capacity depends on discharge rate. A lead-acid battery delivers noticeably less energy in 5 minutes than in an hour, which is also why halving the load more than doubles the runtime.

Worked example

A data hall with a measured IT load of 400 kW at 0.95 power factor, 20% growth allowance, N+1 redundancy using 250 kW / 250 kVA modules (unity output power factor), and 5 minutes of runtime.

• Design load: 400 kW × 1.20 = 480 kW, which is 480 ÷ 0.95 = 505 kVA
• Modules for N, by kW: 480 ÷ 250 = 1.92, so 2 modules
• Modules for N, by kVA: 505 ÷ 250 = 2.02, so 3 modules. The kVA check governs.
• N+1: 4 modules, 1,000 kVA installed. Normal loading is about 51%. With one module out, 505 ÷ 750 = about 67%.
• Battery energy: 480 kW × 5 ÷ 60 = 40 kWh at the output. At 95% conversion efficiency, 40 ÷ 0.95 = 42.1 kWh from the battery.
• With the lead-acid aging factor: 42.1 × 1.25 = 52.6 kWh, before any temperature correction or design margin

The 95% efficiency and 20% growth figures are assumptions for the example. Use the UPS datasheet's figure at the expected load, and your own project's growth plan.

Don't forget the input side

A UPS draws more from its input than it delivers at its output. It draws its own losses plus battery recharge current after an outage, when the batteries are depleted and the load is still running. Feeders, breakers and the generator upstream are sized on the UPS's rated input current from its datasheet, not on the output rating. The topology matters too: double conversion and eco mode draw from the input very differently. Generator compatibility deserves its own check. Older rectifier designs with high input harmonic distortion often forced a generator to be oversized relative to the UPS. Modern IGBT rectifiers with low input distortion reduce that, but the UPS manufacturer's generator guidance governs. How to Size a Standby Generator covers the generator side.

Common mistakes

• Checking only the kVA on the front of the datasheet. The kW rating is often the one that runs out first.
• Summing nameplates. This oversizes the UPS and leaves it running at low load for years, where efficiency is worst.
• Counting redundancy on kW alone. As the worked example shows, the kVA check can move N up by a module.
• Sizing the battery on new capacity. Without an aging allowance, the runtime is met on day one and missed a few years later.
• Treating runtime as linear. Battery capacity depends on the discharge rate, so use the manufacturer's constant-power tables.
• Sizing the upstream feeder on UPS output. The input carries losses and recharge current on top of the load.

Frequently asked questions