Start with watts, not VA
The most common UPS sizing mistake is reading only the big VA number on the box. A UPS is rated in two figures — VA (apparent power) and W (real power) — and the gap between them is the output power factor. Older units quoted 0.6-0.7, so a "3000 VA" unit might only deliver 2100 W. Modern IT loads draw near power factor 0.9-1.0, so you must match the UPS to both ratings, whichever you hit first.
Required VA ≈ total watts ÷ load power factor. A 4,000 W server load at PF 0.9 needs about 4,444 VA — so choose a 6 kVA/6 kW unit so neither rating sits above ~75%.
Never load a UPS past about 80% of rating: leave room for growth, for the inrush of switch-mode power supplies (crest factor), and for charging the batteries.
Runtime is not linear — Peukert's law
This is the surprise that catches everyone. Battery capacity is quoted at a gentle discharge; pull it hard and you get proportionally less out. This is Peukert's law — double the discharge current and you may keep far less than half the runtime. So a UPS that backs a 50% load for 20 minutes will not back a 100% load for 10; it will manage rather less.
A rough runtime estimate:
t ≈ (usable battery Wh × inverter efficiency) ÷ load watts
…where usable Wh = battery volts × amp-hours × usable depth of discharge. Then expect somewhat less at high load because of Peukert, and remember that heat shortens battery life dramatically (roughly halving it for every 8-10 degC above 20-25 degC). Site batteries in the coolest practical spot and load-test them — an untested UPS runtime is a guess.
When you need hours, you need a generator
If the autonomy you need runs to hours rather than minutes, the honest answer is a generator with a short-runtime UPS to bridge the start, not a battery room sized for the impossible. The UPS holds the load clean for the few seconds the generator takes to start and accept load; the generator then carries the outage. Designing the two together is the key to real resilience.
Choosing the topology
| Topology | Transfer time | Conditioning | Use for |
|---|---|---|---|
| Offline / standby | 2-10 ms | Minimal | Single PC, till point |
| Line-interactive | 2-6 ms | Voltage regulation | Networks, CCTV, POS |
| Online double-conversion | 0 ms | Full (voltage + frequency) | Servers, medical, PLC, lab |
Much of Kenya's grid is not just outage-prone but dirty — sags, spikes and frequency wander that age sensitive electronics. For anything critical, online double-conversion is the safe choice because the load is fully isolated from the incoming supply.
Redundancy for the things that cannot fail
A single UPS is a single point of failure. Critical sites add redundancy: N+1 (one spare module so any one can fail or be serviced with no loss) for clinics and SME data rooms, 2N (fully duplicated) for hospitals, banks and serious data centres. Match the redundancy to the cost of downtime, not to prestige.
The bottom line
Size on watts, respect the battery's non-linear runtime, choose the topology your load actually needs, and design the UPS and the generator as one system.
Tell us your critical load (kW and VA), the autonomy you need and your downtime cost, and we will return a UPS sizing, runtime calculation and redundancy recommendation. Call +254 768 860 665 or +254 782 914 717.