UPS runtime: the misconceptions that turn a 30-minute spec into a 12-minute reality

UPS runtime is the spec that survives in the building's nameplate but rarely survives in the building's operation. The catalogue figure is a clean-room test result; the building's holdover at month 30 is what the operator actually has when the mains fails. The gap is engineered, not bridged after the fact.

## Battery chemistry decides the degradation curve

VRLA (sealed lead-acid) is the workhorse of mid-tier UPS deployments — cheap, well-understood, predictable. The degradation curve is also well-understood: 5-8% capacity loss per year under nominal conditions, accelerating to 12-15% per year above 30°C ambient. A 30-minute autonomy bank at year one is a 22-minute bank at year three and a 14-minute bank at year six. The AMC schedule treats VRLA banks as a 3-year refresh cycle for any holdover-critical load.

LFP (lithium iron phosphate) carries a much shallower degradation curve — 2-3% capacity loss per year under nominal conditions, 5,000-7,000 cycle service envelope at 80% DoD. A 30-minute LFP autonomy at year one is 27-28 minutes at year five and 24-25 minutes at year ten. The capex penalty is real (2.5-3× VRLA at parity capacity), but the lifecycle economics favour LFP for any deployment with 15+ year planning horizon.

## Temperature is the silent runtime killer

Battery cabinet temperature is the leading silent factor in runtime collapse. Every 10°C above 25°C halves VRLA service life; LFP is less sensitive but still de-rates above 35°C. UPS battery cabinets in tropical-climate buildings without dedicated cooling routinely operate at 35-40°C, which compresses the published 5-year VRLA life to 2-2.5 years. Battery-room HVAC is sized against the full-discharge thermal envelope, not the standing-charge load.

## Power-factor reality vs catalogue assumption

UPS nameplate kVA is published at unity power-factor (1.0). Real loads have lower power-factor: modern PoE switches and Wi-Fi 6 APs run at 0.9-0.95 PF, AV-over-IP encoders at 0.85-0.92 PF, server-class loads at 0.85-0.95 PF. A 10 kVA UPS at unity PF delivers 10 kW; the same UPS at 0.9 PF delivers 9 kW. The runtime calculation must use the actual load wattage at the actual PF, not the nameplate kVA.

## Peukert effect on high-discharge events

The Peukert exponent describes how battery capacity degrades non-linearly with discharge rate. A 100 Ah battery discharged at the 20-hour rate delivers 100 Ah; the same battery discharged at the 1-hour rate delivers 50-60 Ah depending on chemistry. High-discharge events (full-load UPS discharge during a sustained outage) consume more battery capacity than the nominal calculation suggests. The Peukert correction is built into proper UPS runtime calculators; the spec-sheet runtime is the nominal-rate figure, not the high-discharge corrected figure.

## Per-load UPS vs shared-bus UPS

Per-load UPS architecture puts every critical load class on its own dedicated online UPS — a fault on one load cannot cascade across the stack. Shared-bus UPS routes everything through one mainline UPS, cheaper capex and a single point of failure. The brief decides which is right: when uptime is the brief (perimeter security, command-and-control, medical-imaging), per-load is correct; when power-quality is the brief (data centre, broadcast), shared-bus with full redundancy is correct.

## AMC discipline is what preserves the runtime spec

VRLA battery autonomy tests at quarterly intervals — battery bank discharged under measured load, runtime measured against the spec, replacement triggered at 80% of spec. LFP autonomy tests at semi-annual intervals with the cell-level capacity baseline preserved. The test is the operational deliverable; without it, the runtime spec is theoretical and the first sustained outage is also the first discovery.

## Configuration baselines are part of the spec

UPS configuration (inverter mode, ECO mode, battery-test schedule, alarm escalation thresholds) is exported offline at handover and re-exported after every firmware refresh. A clean-slate UPS recovery within the same business day is the operational guarantee; without the configuration baseline, the recovery is hours of vendor-mode reconfiguration.

## Callout — what UPS procurement most miss

**Runtime spec is not runtime delivered.** The catalogue figure is the starting point; the AMC discipline, the battery-cabinet thermal envelope, the actual load power-factor and the Peukert correction together decide what the building actually has when the mains fails. Engineer against the delivered runtime, not the nameplate.

## References

1. IEC 62040-1 — uninterruptible power systems (UPS).

2. IEEE 1184 — guide for batteries for uninterruptible power supplies.

3. NFPA 110 — emergency and standby power systems (international cross-reference).

4. Peukert's Law — battery discharge-rate to capacity relationship.