Data Center UPS: The Essential Characteristics of an Optimised Solution

Overview

The paper organises UPS selection under three themes: continuity (availability, MTTR, N+1, hot-swap), total cost of ownership (efficiency, battery management, component life, serviceability) and adaptability to load, electrical infrastructure, room and user. It contrasts conventional parallel UPSs with modular and scalable-modular architectures.

On continuity, MTTR for modular UPSs can be about 30 minutes (power-module swap) versus roughly 1-4 h for conventional units; the vendor table lists Keor MOD/UPSaver 0.5 h, Keor HPE 1-4 h and Keor XPE 3-6 h. In the N+1 example, a 1000 kW N+1 modular system built from 333 kW modules needs 1333 kW installed, whereas conventionally it could be 6 x 200 kW, 5 x 250 kW or 3 x 500 kW units in parallel. Scalable systems reach 2.1 MW with 250/300 kW power units, and 2.67 MW (up to 21 MW in parallel) with 333 kW units.

On energy, the cited split is IT equipment about 55%, cooling about 30%, power infrastructure including UPS losses about 12% and lighting about 3%. UPS load is usually below 50%, and the cited efficiency curve peaks (up to 96.6%) in the 20-50% load band. For a 10 MW data centre a 1-point efficiency gain (95.5% to 96.5%) saves about USD 1.3 million of OpEx over 10 years; OpEx is typically 60-75% of UPS TCO.

The pay-as-you-grow examples show capacity added in steps with efficiency staying nearly flat, and the load-variation cases (a maintenance week from 75% to 50%, a cooling UPS varying up to 35%) change efficiency by only about 0.5%. This is a vendor publication, with figures drawn from Legrand product families.

Key points

1. MTTR is about 30 minutes for modular UPSs (module swap) versus roughly 1-4 h for conventional units.
2. A 1000 kW N+1 modular system from 333 kW modules needs 1333 kW installed; conventional alternatives are 6 x 200, 5 x 250 or 3 x 500 kW.
3. Energy split: IT about 55%, cooling about 30%, power infrastructure incl. UPS losses about 12%, lighting about 3%.
4. For a 10 MW data centre, raising UPS efficiency from 95.5% to 96.5% saves about USD 1.3 million OpEx over 10 years.
5. OpEx is typically 60-75% of UPS TCO; CapEx plus installation 25-40%.
6. UPS load is usually below 50%; the efficiency curve peaks (up to 96.6%) in the 20-50% load band.
7. Scalable systems: 250/300 kW units reach 2.1 MW per system; 333 kW units 2.67 MW, up to 21 MW in parallel.
8. Load-variation examples change efficiency by only about 0.5%.

Using it in design

• Give UPS records an architecture enum {conventional, modular, scalable_modular}, module rating, maximum parallel units, MTTR hours and a hot-swap flag; store efficiency at 25/50/75/100% load, not one number.
• Compute installed kW for N+1 as ceil(load/module)+1 modules and show the over-provisioning percentage.
• Warn when the expected operating point (2N halves the load) falls outside the vendor's peak-efficiency band.
• In the TCO tool default the OpEx share to 60-75% and run the 10-year loss cost per efficiency point.

Who should read it

Electrical designers and cost owners choosing UPS architecture, module granularity and efficiency curve.

Related standards

• IEC 62040-3

Publication details

Authors

Legrand

Publisher

Legrand

Year

2022

Document language

English

Source type

Manufacturer

Access

Free to read

Licence

Copyright retained; citation only — the publisher’s terms apply to the original document

Citation