Sizing Backup Power: Why the Battery You Buy Is Never the Battery You Get

Sizing a backup system looks like arithmetic. Take the load in watts, take the autonomy in hours, divide by the battery's rated capacity, and order. Systems specified that way routinely fall short of their promised runtime — not because anyone made a mistake in the division, but because the rated capacity on a datasheet describes a laboratory condition that your installation does not reproduce. The useful question is not how much energy the battery stores, but how much of it your load can actually withdraw, at your discharge rate, at your ambient temperature, at the end of the pack's service life.

Start with the real load, not the nameplate

Nameplate ratings describe what equipment may draw, not what it does draw. Summing them produces a figure that is comfortably conservative for the steady state and dangerously optimistic about everything else. We prefer a measured load profile over an afternoon of normal operation, because two things emerge from it that a nameplate sum cannot show: the actual average demand, and the inrush. Motors, compressors and some switching supplies pull several times their running current at start-up, and if the backup source cannot deliver that surge, autonomy is irrelevant — the load never restarts.

It is also worth separating the loads that genuinely need backup from those that have been attached to the protected bus by habit. Autonomy is expensive per watt. Reducing the protected load is almost always cheaper than enlarging the battery to carry it.

Then subtract what you will never get back

Four effects stand between rated capacity and delivered energy, and they compound.

Discharge rate. Capacity is quoted at a specified rate, often a long one such as C/20. Draw the same pack harder and it yields less total energy — pronounced in lead-acid, where the Peukert relationship makes a short, heavy discharge markedly less efficient, and milder but still present in lithium chemistries. Size from the curve at your actual rate, not from the headline number.

Temperature. Capacity falls as temperature drops, and service life falls as it rises. A pack in an unconditioned outdoor cabinet in an Israeli summer and the same pack in a climate-controlled room are, for sizing purposes, different products. Decide which condition your design must satisfy and use that one.

Ageing. Batteries are commonly considered end-of-life at eighty percent of rated capacity, and many are replaced there. If the system must still meet its autonomy target on the last day before replacement, the sizing must be done against the aged capacity — not against a new pack. Designing to a new-battery figure guarantees the specification is met only at commissioning.

Usable depth of discharge. Very few packs should be run to zero. Lead-acid life shortens sharply with deep cycling, and lithium systems reserve a margin at both ends for protection and for state-of-charge accuracy. The battery management system's cut-off, not the theoretical capacity, defines the floor.

Two constraints people forget

The first is recharge time. A system sized for a single outage may be unable to recover before the next one. If the site sees repeated interruptions, the charger — not the battery — often sets the design, and the honest specification states both the autonomy and the time to return to full readiness.

The second is the end-voltage the load will tolerate. Autonomy ends when the bus falls below what the equipment accepts, which may be well above the battery's own cut-off. Where the equipment has a narrow input window, converter selection can add usable minutes that no amount of extra amp-hours would.

Established sizing methods exist for exactly this — the IEEE 485 and IEEE 1184 practices for stationary and UPS batteries are a sound starting framework — and manufacturers publish discharge tables for their cells. The discipline is in applying the derating honestly rather than treating each allowance as optional.

Then prove it

A calculated runtime is a hypothesis. A discharge test on the installed system, at the real load, is the evidence — and repeating it periodically is the only way to know whether the pack is still meeting the specification it was bought against. Monitoring helps enormously here, but a logged voltage is not a capacity measurement; only a controlled discharge is.

If you are specifying backup power for a critical load and want the sizing checked before it becomes a purchase order, we are glad to look at the load profile with you. Write to info@lmelectro.com — we respond within 24 hours.