Electrical design
Sizing the UPS, and sizing the battery behind it
Two separate calculations that get run as one. The UPS is sized by the load it carries and the configuration it carries it in; the battery is sized by what has to happen in the seconds before the generator takes over.
Start with real power, not apparent power
A UPS is sold in kVA and limited in kW, and modern IT load runs at a power factor close to unity. A 500 kVA unit rated 0.9 pf delivers 450 kW; a unity-rated unit delivers 500 kW. Size on the kW the load actually draws, then confirm the kVA figure clears it. Sizing the other way round is how a plant ends up kW-limited with kVA headroom printed on the nameplate.
| Input | Where it comes from | Common error |
|---|---|---|
| IT load, kW | Design density × cabinet count, or the client’s own figure | Using nameplate ratings of servers, which overstates draw substantially |
| Mechanical load on the UPS, kW | Room units, chilled-water pumps and controls, if continuous cooling is claimed | Leaving the pumps out, then claiming continuous cooling anyway |
| Growth allowance | The brief — stated as a phase, not as a vague percentage | A growth figure that quietly becomes permanent loading |
| Power factor | The equipment, not the UPS datasheet | Mixing the UPS output pf with the load pf |
Then apply the configuration, not just the number
The module count follows from the redundancy, and the redundancy decides the loading each module actually runs at. This is where a “500 kW UPS” stops being one number.
| Configuration | 400 kW load | Loading in normal operation | Loading with one module out |
|---|---|---|---|
| N | 1 × 500 kW | 80 % | Load lost |
| N+1 | 2 × 500 kW | 40 % each | 80 % |
| 2N | 2 × 500 kW, one per cord | 40 % each | 80 % on the survivor |
| 2(N+1) | 4 × 500 kW | 20 % each | 27 % each |
Two things follow. First, the single-fault case is what sizes the module, not the normal case: in 2N each UPS must carry the whole load, not half of it. Second, a heavily redundant plant runs its modules lightly, which is where the efficiency conversation belongs — a module at 20 % load is not at its best point, and that cost is paid every hour of every year.
What the battery is actually for
In a facility with generators, the battery exists to carry the load from the moment the supply is lost to the moment the generator is on load — typically ten to fifteen seconds — plus a margin for a start that does not succeed first time. Five or ten minutes of autonomy is not there to ride out an outage; it is there so that a failed start is not an immediate shutdown, and so an operator has time to act.
In a facility without generators, the battery is the whole story, and autonomy is a business decision about how long an orderly shutdown takes.
The four margins that decide the string
- Discharge rate, not capacity. A battery is rated for a duty — so many kW per cell for so many minutes to a defined end voltage. A 10-minute duty draws far more than a tenth of the 100-minute figure. Size from the manufacturer’s discharge tables at the duty you need.
- End voltage. The string has to stay above the UPS’s minimum DC bus voltage for the whole discharge. The end voltage per cell sets the cell count, and it is the first thing to check when a string “tests short”.
- Temperature. Capacity figures are quoted at 25 °C. A battery room that runs warmer gives shorter life; one that runs cooler gives less capacity. Both directions matter, and only one of them is usually allowed for.
- Age. End of useful life is conventionally 80 % of rated capacity. If the string is sized at exactly the duty on day one, it is undersized for most of its service life. Size it with the ageing margin included.
The load the battery is not told about
Two loads are routinely left out of the sizing and then appear on site:
- The recharge load. After a discharge the UPS draws its load plus the recharge current. The generator and the upstream feeder have to carry both, at the moment the plant is least comfortable.
- The mechanical load on the UPS. If the pumps and controls are on uninterruptible power — and for continuous cooling they have to be — they are part of the battery duty, not a footnote to it.
A worked set of figures
| IT load | 400 kW |
| Mechanical load on UPS (pumps, controls) | 60 kW |
| Design load | 460 kW |
| Configuration | 2N, one UPS per cord |
| Module sized for | 460 kW alone → 500 kW module at unity pf |
| Loading, normal | 46 % per module |
| Loading, one cord lost | 92 % on the survivor |
| Autonomy required | 5 min to the UPS end voltage, at 460 kW |
| Margins applied | Ageing to 80 % capacity, temperature at the room’s design figure |
The row worth arguing about is the second from the bottom but one: 92 % on the survivor. It is inside the rating, and it leaves nothing for the growth the brief asked for. That is a decision to take at design stage, in writing — not one to discover during the first failure.
Sizing that reads the drawing
Data Center Simulator sizes the UPS, the battery, the generator and the cooling from the design you drew — the loads are the loads on the diagram, so the calculation and the single line cannot disagree. Run the failure case and every source reports what it carries, in kW and as a percentage of its rating, with the methodology printed alongside the result.
Download the free trial See what the report contains