In practice, UPS hours usually means battery runtime – how long your UPS will provide power to your load during an outage. The words uptime hours, availability, and reliability hours (MTBF) may not be used correctly. The meanings are clarified in this guide, and then we dive deep on runtime: exact definitions, equations, correction factors (efficiency, Peukert/high rate effects, temperature, aging), power quality considerations (PF, crest factor, harmonics), and a complete sizing workflow with worked examples and validation methods.
Runtime, in its most basic definition, refers to the amount of usable DC energy from the battery which can actually be converted by the UPS into AC divided by the true (real) power that your load consumes after adjustments made for real-life losses and due to the fact that the battery delivers less usable energy at high discharge rates and at low temperatures.
2.1 Core equation (engineering approximation)
Runtime (h)≈PloadEbat×ηUPS×krate×kT×kaging
Where:
E_bat = V_bus × Ah (in Wh).
Summed across series/parallel to the pack level.
UPS discharge efficiency, noted as η_UPS, is typically 0.88–0.94 for double-conversion.
High-rate discharge correction (0.70–0.95; lower at higher C-rate, esp. VRLA).
The parameter k_T helps in correcting temperature for the VRLA type batteries.
k_aging is a state of health allowance like 0.8 to show that 20% of capacity fade has occurred.
P_load refers to the real power measured in watts, not VA. If only VA is known, use.
P ≈ VA × PF.
Quick mental check: 100 Wh ≈ 1 hour at 100 W before derating.
2.2 Peukert’s effect and C-rate (lead-acid)
Lead-acid capacity is specified at a low discharge rate (e.g., C/20). At higher C-rates (larger current), usable capacity drops. A rough Peukert form:
t=H(ICr)k
kkk typically 1.1–1.3 for VRLA.
Lithium-ion behaves far better at high rates (effective kkk closer to 1), another reason it sustains runtime under heavy loads.
2.3 Chemistry differences
VRLA/AGM: lower cost, predictable, but sensitive to heat/high-rate discharge; capacity falls at low temperatures; life 3–5 years typical.
Li-ion (NMC/LFP): higher energy per kg, better high-rate performance, wider temperature window, far better cycle life; higher upfront cost; may include active BMS limits on discharge.
UPS output varies according to power factor as rated in V and W. The PF of your load which is typically between 0.9 – 1.0 for modern IT loads determines the real power. Always size to watts.
Nonlinear IT loads have high peak current (typicalCF=3:1) drawing ability. Make sure the UPS inverter can handle peaks without cut off.
Harmonics or THD refers to heavily distorted current that increases the RMS current and therefore losses. Thus, the runtime shrinks compared to a purely resistive load.
Think of how like a lot of UPSs are very inefficient at lower outputs than 30% or higher outputs than 70%.
If we consider a load of 400 watts, we would see 685/400 equals 1.71 hours or roughly 1 hour and 43 minutes. If we consider a 800 watt load, we would get that this would approximately equal 51 minutes.
Connection to generator (start-up + stabilization, usually 1-10 min).
Extended runtime (external battery cabinets).
2. Audit the load (realistic watts, not nameplate).
Use a UPS or power meter to make measurements during a steady state and at peak events.
If the loads are non-linear, or if the UPS spec is tight, PF and crest factor must be determined.
Add 25-30% headroom to IT for growth for evolving sites.
3. Choose topology and output waveform.
The double-conversion or online UPS is capable of controlling the voltage and frequency of output supplies. It gives true sine output and is undoubtedly very useful for sensitive or PFC loads. It has high baseline losses on mains, but predictable on battery.
The line-interactive or standby types of UPS circuits give higher efficiency on mains but may have a stepped approximation on the battery. Therefore, confirm whether they are compatible with your active PFC PSUs in use.
4. Select chemistry and DoD strategy.
The VRLA is designed to minimize deep discharge for a target run time.
With vendor limits and BMS curves, Li-ion can support deeper DoD with lower wear.
5. Compute runtime using the correction chain.
Start with pack Wh; apply η_UPS, k_rate, k_T, k_aging.
Make sure to check against the load curve which the manufacturer made.
6. Check surge/crest and inverter current limits.
Make sure repeated peaks like a server cold start will not cause an early cutoff.
7. Plan management and shut-down policy.
Network card/USB agent to commence graceful OS shutdown at for example, 25% runtime remaining or N minutes.
Coordinate staggered or priority shutdown for clusters.
8. Document test and acceptance criteria
Load-bank test at representative load and ambient.
Record runtime, alarms, and remaining capacity estimates.
Inverters that use batteries have a high percentage of efficiency.
You might see a bit less efficiency for very light or heavy loads.
6.2 High-rate discharge(k_rate).
When using VRLA batteries at high C-rates, make sure to check with vendor tables. In the absence of these tables, make C-rate 0.75-0.90 depending on load.
Li-ion: flatter; kkk generally 0.9–0.98 at rated current.
6.3 Temperature(k_T).
VRLA capacity near 0 °C can fall 20–40% vs 25 °C.
At low temperatures, Lithium ion battery will stop discharging high current from BMS.
6.4 Aging(k_aging).
Design for end-of-life runtime. Assume that VRLA has around 20-30% capacity fade over the service life. For Li-ion, it is less, but still non-zero.
Yes—if the UPS supports larger/external packs and charger current is adequate. Follow vendor guidelines to avoid over-long recharge times and bus/current limits.
At a given nameplate Wh, not “longer,” but Li-ion typically sustains more of its nameplate at high load, in the cold, and after years of use—so effective runtime is often better.
Almost any UPS sized for your watts can deliver that—prioritize inverter quality and transfer behavior over massive battery capacity.
They don’t. Runtime is a battery energy question; uptime hours are a service availability metric across months/years.
