UPS battery autonomy: why five catalogue minutes are not five minutes on site

The UPS battery has one job: to hold the load from the moment the grid drops until the generator picks up. That is an interval of tens of seconds. Even so, designs routinely specify five, ten or fifteen minutes — and almost always without any calculation behind the number.

The problem is not that five minutes is too little. The problem is that the catalogue five minutes are obtained under conditions that do not exist on site.

What actually happens, by the clock

The sequence on loss of mains has fixed stages, each with its own duration:

• 0 s — the UPS transfers to battery. There is no interruption; the load notices nothing.
• 0–3 s — the controls confirm the loss is sustained rather than a brief dip, and issue the start command.
• 5–15 s — the generator cranks, fires and reaches rated speed and voltage.
• +5–10 s — frequency and voltage stability are verified before the transfer switch takes the load.
• 20–40 s — the load is on the generator. Cooling pick-up runs separately and often starts in stages.

So the real requirement is under a minute. The rest is margin — justified, but it should be clear what is being bought: a failed first start and a second attempt, time for an operator to react, and a controlled shutdown if the generator does not start at all.

The three factors that eat the catalogue figure

Temperature. Lead-acid (VRLA) batteries are rated at 25 °C. Below that, capacity falls — roughly 1 % per degree under 25 °C on short discharges. Above it, capacity holds but life shortens sharply: every 8–10 °C above 25 °C roughly halves it. A battery kept permanently at 32 °C will not reach its stated years, however rarely it is discharged.

Ageing. End of battery life is defined by standard as a fall to 80 % of rated capacity. That means the design has to deliver the required autonomy at 80 %, not at 100 % — otherwise the installation drops below requirement in the final third of the battery’s life without anyone noticing.

Peukert’s law. A lead battery does not deliver the same capacity at different currents. The stated amp-hours are for a long discharge — 10 or 20 hours. On a discharge lasting minutes, the capacity actually delivered is appreciably lower. That is why autonomy is never calculated from Ah, but from the manufacturer’s constant-power tables in watts per cell for the specific durations.

How it is calculated properly

The order is the opposite of the intuitive one. You do not go from battery to time, but from time to battery:

• Set the required time — the measured generator pick-up time plus margin, not a round number from the brief.
• Calculate the power on the DC side: the active load power divided by the inverter efficiency.
• Divide that power by the number of cells to obtain watts per cell.
• From the manufacturer’s table for that duration, select the cell type and count, at an end voltage matched to the inverter’s lower limit.
• Multiply the result by an ageing factor (1.25 for end of life at 80 %) and by a temperature factor if the room is not held at 25 °C.

Lithium or lead

Lithium iron phosphate (LiFePO₄) systems have taken a large share of new installations, and with reason: they occupy about a third of the volume, withstand many times more cycles, tolerate higher temperatures and come with per-cell monitoring built in. Against that stand a higher initial price and stricter fire-safety requirements for the room.

The decision is not technological but operational. If the site has limited space, frequent short interruptions, or will also run in peak-shaving mode, lithium pays for itself. If the battery will simply sit there and discharge twice a year, a lead system remains the cheaper sensible choice — provided it is maintained and replaced on schedule.

One thing holds for both technologies: autonomy that has not been tested under load is an assumption. A load-bank test at commissioning and periodically afterwards is the only way to know what is actually in the room.

Related area: Data centres

All articles