Battery Backup for a Commercial Building: Sizing a Four-Hour Guarantee

A commercial premises in Latin America loses power often enough that the owner has stopped keeping count. The brief was narrow: when the grid fails, the whole installation should transfer to battery backup without anyone walking to a switchboard, and it should carry at least four hours at high load.

Four hours is not an arbitrary number. It is the duration that shows up repeatedly in commercial and utility storage specifications, and it is usually chosen to cover a typical outage rather than a worst case.

We designed two versions of the answer. One 125 kW / about 261 kWh outdoor storage cabinet with an isolating step-down transformer and an automatic transfer switch, or two cabinets in parallel at roughly double the cost. The single-cabinet version does hold four hours — while the battery is new. At the end of its service life it does not. Most of this article is about that gap, because it is the part a proposal can quietly leave out.

This is a sample project and a preliminary proposal. The client has not signed a contract. The load figures are estimates rather than metered data, the power factor is assumed, and every runtime below is a calculated design value, not a measurement from a delivered system.

What the site told us, and what it could not

Item What we have Status
Scenario Commercial premises; frequent public-grid outages Reported
Estimated maximum load about 75 kVA Approximate
Simultaneity factor about 65% to 85% → storage carries about 49 to 64 kVA Approximate
Backup requirement at least 4 h at high load; longer with load shedding Reported
Electrical system three-phase 120/220 V, 4-wire, 60 Hz (120/208 also common) To be measured
Transfer fully automatic, no operator Reported
Power factor not supplied → design at 0.9, energy envelope at 1.0 Assumption
Load curve and outage statistics no 15-minute metering; no annual outage count Missing

Two lines matter more than the rest. The load is estimated, not measured, and it is a range rather than a number — 49 to 64 kVA depending on how much of the equipment actually runs at the same time. And the site voltage is unconfirmed. Three-phase 120/220 V is what the client reports, but in the region 120/208 and 120/240 are also normal, and the two cannot be told apart without a meter on the terminals.

Neither gap blocks a preliminary design. Both block a purchase order.

The answer, before the reasoning

At 56 to 60 kVA the single cabinet delivers 4.3 and 4.1 hours while the battery is new. That meets the requirement, with almost nothing to spare.

At end of life, assuming the usual 90% state-of-health floor, the same loads get 3.9 and 3.6 hours.

Bar chart of battery backup runtime at 56 to 75 kVA loads for one and two storage cabinets at beginning and end of life

Read the four pairs of bars rather than the two headline numbers. The dark blue bar is the same cabinet, same loads, same control logic — only the battery is ten years older. At 64 kVA it drops to 3.4 hours. At 75 kVA, the estimated maximum, it is 2.9.

So the honest statement of the single-cabinet option is not “four hours of backup”. It is “four hours for the first several years, then less”. If the requirement is four hours for the life of the asset, one cabinet does not meet it, and no amount of optimistic rounding changes that.

The two-cabinet option clears four hours at every load in the table, at end of life, with margin: 7.8 hours at 56 kVA, 5.8 hours at 75 kVA.

Two loads, not one

The 75 kVA figure and the 49 to 64 kVA figure are not two estimates of the same thing. They answer different questions and they size different equipment.

The estimated maximum sizes the conductors and the transfer switch. If every piece of equipment started at once, the site-side switchgear has to survive it. That is what the 75 kVA number — 67.5 kW at 0.9, 197 A per line at 220 V — is for.

The simultaneous load sizes the battery energy. What the site actually draws when it is running normally, with the simultaneity factor applied, is what the battery has to carry. That is the 49 to 64 kVA band.

The reason this distinction is worth writing down is that 125 kW of converter against a 60 kVA load looks, at first glance, like a system that has been over-specified by a factor of two. It is not. Power was never the constraint here. At 60 kVA the cabinet is running at roughly half its rated AC power, and the load could grow substantially before power became the issue.

The constraint is energy, and energy is what the next two sections are about.

Where the energy goes: the k factor

A battery’s nameplate capacity is not the energy it delivers to a load. Between the two sit four losses, and designing with the nameplate number is how a four-hour promise becomes a three-hour outage.

The calculation chain used here is:

Stage Efficiency
Inverter (DC to AC) 0.95
Depth of discharge 0.90
Auxiliary loads (cooling, controls) 0.98
Combined factor k 0.8379

So the nominal demand behind a given AC energy requirement is that energy divided by 0.8379 — about 19% more capacity than a naive calculation would call for. At 60 kVA and 4 hours, that is 216 kWh of AC energy, which becomes about 258 kWh of nominal battery capacity before any ageing allowance is applied.

Then the ageing allowance. A cabinet sold as 261 kWh delivers close to that when new and about 90% of it at end of life, which is where the 3.6-hour figure at 60 kVA comes from.

Chart comparing energy needed for 4 hours of backup under several load scenarios with installed battery capacity

The chart on the left is the same arithmetic across five load scenarios. The red line is the single cabinet’s 261 kWh; the green dashed line is the two-cabinet 522 kWh. Only the two lowest scenarios — 49 kVA and 56 kVA — stay under the single cabinet’s capacity at end of life. From 60 kVA upward, the nominal requirement at end of life exceeds what one cabinet holds, which is exactly what the runtime table shows.

This is the part of a storage quote where kW and kWh get conflated. A 125 kW cabinet sounds like it should handle a 60 kVA load for four hours, because 125 is bigger than 60. It will — for about two hours. The rating that decides runtime is the one measured in kWh, and it has to be read at end of life, not at commissioning.

A cabinet that does not speak the site’s voltage

The cabinet’s AC side is 380/400 V, three-phase, 4-wire. The site is roughly 120/220 V, and possibly 120/208. There is no setting that bridges that gap.

A Dyn isolating step-down transformer is therefore mandatory, not optional: 125 to 160 kVA class for the single cabinet, 250 to 300 kVA class for the pair. It does two jobs. It steps 380/400 V down to the site voltage, and it provides the neutral on the low-voltage side, which the site needs because single-phase, two-phase and three-phase equipment coexist on the same bus.

Conceptual topology of a commercial battery backup system with grid, ATS, Dyn step-down transformer, storage cabinet and EMS

The path is: grid → main incomer breaker and meter → normal terminal of the ATS → load bus, with the cabinet on 380/400 V feeding a step-down transformer that lands on the backup terminal of the same ATS. Charging can run from the grid through the same transformer path or through an independent circuit; that detail belongs in the formal design, not the preview.

Two things about the transformer have caught clients out before. First, it is a real line item — the CAPEX ranges below include it, and it is not a rounding error. Second, its rating follows the maximum load, not the battery. Sizing it against the 49 to 64 kVA operating band instead of the 75 kVA maximum is how a perfectly good backup system ends up tripping on the one day every load starts together.

The transformer also settles the three-phase unbalance question. With mixed single-phase and three-phase loads on the low-voltage side, the neutral is doing real work, and a transformer that provides one is doing more than voltage conversion.

Load shedding is a design lever, not a failure mode

Once the battery is sized, the runtime is not fixed. It is a function of how much load the site chooses to keep running, and the EMS can change that automatically as state of charge falls.

At the 60 kVA design basis, with the battery new:

Share of load retained Equivalent load Single cabinet Two cabinets
100% 60 kVA 4.1 h 8.1 h
70% 42 kVA 5.8 h 11.6 h
50% 30 kVA 8.1 h 16.2 h
35% 21 kVA 11.6 h 23.2 h
Bar chart showing how staged load shedding extends battery backup runtime for one and two cabinets

The useful way to read this table is not “we can get 8 hours from one cabinet”. It is that the client, not the equipment, decides where the four-hour line sits. Dropping to 70% of load buys 5.8 hours and turns a marginal design into a comfortable one, without a second cabinet.

That only works if someone has already decided which loads are expendable. The proposal puts three load levels on the diagram; the contents of those levels are marked to be confirmed by the client, because an engineer cannot know which circuits in a commercial building are allowed to go dark. Getting that list agreed before commissioning is worth more than a few percent of extra battery.

One caution on the numbers. The table is computed at beginning of life. At end of life, 70% retained load gives roughly the same runtime as 60–65% does when the battery is new. If the shedding plan is what makes the system acceptable, the plan should be checked against the end-of-life column too.

What “under 100 ms” actually covers

The cabinet transfers internally in under 100 ms. That number is accurate and it is also narrower than it sounds.

It describes the transfer inside the cabinet — the point where the battery picks up the load it is already powering. The site-side transfer is a separate event, and its duration is set by the ATS or STS that the client installs. A mechanical ATS transfers in hundreds of milliseconds, sometimes more, and during that window the load is briefly dead.

If the load cannot tolerate that — servers, medical equipment, continuous-process machinery — the answer is an STS, not a faster cabinet. The cabinet cannot compensate for switchgear downstream of it.

Everything else in the transfer chain is conventional and should be treated as such: grid-failure detection, ATS to backup terminal, return to grid after the supply is stable, protection coordination, earthing, cable sizing and the neutral arrangement all have to be confirmed by a responsible engineer after the site inspection. Black start is supported, which matters if the grid comes back weak or not at all.

When the second cabinet is the right answer

Three situations make the pair the correct choice rather than a upsell:

Metering shows a sustained load of 64 kVA or more. At that level the single cabinet misses four hours even with a new battery.

The four-hour requirement applies to the whole service life. If the specification says four hours, not “four hours when new”, the single cabinet does not meet it, and a proposal that says otherwise is describing year one as if it were year ten.

The client does not want to shed load. Load shedding is what makes the single cabinet workable. Refusing it removes the lever and leaves only capacity.

The cost of that choice is documented rather than estimated away. Conceptual CAPEX, including about 10% contingency, the transformer, ATS and distribution, EMS extension, ocean freight, customs clearance and commissioning:

Option Configuration Concept CAPEX (USD)
A, single cabinet 125 kW / about 261 kWh + 125–160 kVA transformer about 78–146 k
B, two cabinets 250 kW / about 522 kWh + 250–300 kVA transformer about 139–245 k

These are conceptual ranges for commercial alignment, not an offer. They exclude civil works, fire protection, long-term O&M contracts and tax. The ranges are wide because the equipment has not been frozen — the cabinet in this document is a reference specification and can be substituted at equivalent power, energy and voltage.

What still has to be verified

Seven items sit between this preview and a construction-ready design, and two of them can change the equipment count:

1. Fifteen-minute interval load metering, or better, plus a power-factor measurement. The simultaneity factor is the single largest assumption in this document.

2. Outage statistics — annual count, typical duration, longest continuous outage. Four hours is a design target; the outage distribution is what tells you whether it is the right one.

3. On-site voltage measurement, confirming 120/220 or 120/208 or 120/240, plus the neutral and earthing arrangement.

4. Per-phase load measurement and three-phase unbalance, including impact loads such as lift starting and air-conditioning compressors.

5. A signed list of critical and interruptible load levels, which is the foundation of any load-shedding plan.

6. Civil works: footprint, bearing capacity, ventilation and fire protection for both the cabinet and the transformer, plus the customs and installation window.

7. After contract signature: formal preliminary design, equipment freeze, commercial offer and lead time.

Until items 1 and 3 are closed, the runtime figures in this article should be read as the shape of the decision rather than the numbers on a purchase order.

FAQ

Q: Can one cabinet actually deliver four hours, or not?

A: Both, depending on when you ask. At 56 to 60 kVA it delivers 4.3 and 4.1 hours when new, and 3.9 and 3.6 hours at end of life. The requirement was four hours at high load; the honest answer is that one cabinet meets it for part of the service life and not all of it.

Q: Why not just size the battery for the 75 kVA maximum?

A: Because the peak is what the switchgear and transformer must survive, not what the battery normally carries. Sizing energy against the maximum that might occur once a year would add cost for a scenario the site does not run in. The correct sequence is: maximum for the conductors and ATS, simultaneous load for the energy, then confirm both with metering.

Q: Is the transformer really necessary?

A: Yes. The cabinet is 380/400 V three-phase and the site is roughly 120/220 V. There is no wiring arrangement that connects them safely without a step-down transformer, and the transformer is also what provides the neutral the site needs.

Q: Does load shedding mean the site goes dark in parts?

A: During an outage only, and only for the load levels the client designates as non-critical. The list of what may be dropped is decided by the client, not by us, and it is one of the items we insist on getting in writing before commissioning.

Q: How much of these numbers would change after real measurements?

A: The load figures could move the cabinet count in either direction. If metering shows a sustained 64 kVA or above, the pair becomes necessary. If it shows 49 kVA, the single cabinet holds closer to the four-hour line at end of life. The voltage measurement will not change the topology, only the transformer ratio.

About this proposal

This is a preliminary proposal preview prepared to show the analysis process, not a delivered project. The load data is estimated rather than metered, the power factor is assumed, and the equipment is a reference specification that can be substituted. All runtimes are calculated design estimates. The topology is conceptual and the transfer scheme, protection coordination and earthing are subject to a site inspection by a responsible engineer.

If you are being quoted a backup system by its runtime, the three questions worth asking are simple: what load is that runtime calculated at, is it the beginning or the end of the battery’s life, and what happens if you need to keep less load running. A supplier who can answer all three with numbers is worth talking to.

This proposal was first shared on Facebook and LinkedIn.

YURI POWER Engineering Team · November 5, 2026