Auxiliary load

What it is (precise)

Auxiliary load is every kilowatt the station draws to keep itself running and safe. The dominant contributor in a modern LFP BESS is thermal management: the chillers, coolant pumps or HVAC units in each container / enclosure that hold cells in their target temperature window — often cited as roughly 15-35 °C — and suppress the gradients that accelerate degradation. Secondary contributors include BMS electronics, the energy management system and SCADA, PCS gate-drive and fan power, switchgear heaters, fire and gas detection, and site lighting.

Transformer no-load and load losses are usually booked separately in the efficiency chain, but some energy balances lump them in, so always confirm the boundary before comparing numbers. No-load loss is worth confirming rather than assuming, because it can be an ordering option: SMA lists MVPS transformer standby losses as industry standard, Eco design 1 or Eco design 2 — the tiering the European Union applies to transformer no-load losses under its Ecodesign rules.

Auxiliary power is normally served at low voltage — typically 400 V or 480 V three-phase — from a dedicated station-service or auxiliary transformer teed off the MV collection bus, a branch you should be able to trace on the project's one-line diagram. Station service is the name for that whole supply, and it is one of the balance-of-plant subsystems most often assumed rather than specified, because vendors ship it three different ways.

Huawei's JUPITER transformer station carries a dedicated dry-type auxiliary transformer, published as 5 kVA single-phase with a 50 kVA three-phase Dyn11 option, at 400/230/220/210 V. SMA integrates the supply into the converter station instead, with an 8.4 kVA transformer inside the SCS and an external auxiliary supply as the alternative.

EPC Power's M System makes it a selection — external single-phase 208-400 VAC (±20%), an internal AC-side supply, or an internal DC-side supply — so the same question has three legitimate answers depending on what was ordered.

Underneath all of them sits a DC control supply that is a separate system from the main battery: Ormazabal MV switchgear specifies 110 or 125 Vdc for breaker tripping, and Power Electronics' GEN3 commissioning sets a 120 Vdc auxiliary bus with 99 to 155 Vdc treated as acceptable. The quantity is expressed two ways, and both matter: instantaneous kW, which sizes the auxiliary transformer and cabling, and annual kWh as a percentage of discharged energy, which drives the efficiency guarantee.

The key distinction is load during operation versus load at standstill. Power Electronics publishes the shape of that curve for GEN3 rather than a single figure: on the HEMK/PCSK station, electronics only with ventilation off draws 0.4 kW, ventilation on at 0% load draws 3.1 kW, and the figure climbs through 4.7, 6.4 and 8.1 kW to 9.6 kW at full output on four modules, while the larger HEM/PCSM runs 10.6 kW at 0% load and 16.5 kW at 100%.

A factor of twenty-four separates the controls-only floor from full output on the same box, and the single largest step in the ladder is turning the ventilation on. When the plant cycles, auxiliaries scale with C-rate and ambient temperature. When it sits idle — which a merchant or capacity-market asset does most hours — the control hierarchy, its telemetry links and a baseline of thermal conditioning still run. That standby draw erodes the annual energy budget without appearing anywhere in a cycle-by-cycle model, and it is the part most often missed in a napkin estimate.

Why it matters in a real grid-scale project

Auxiliary load is a financial line item, not an engineering footnote. It is a main reason net RTE at the POI sits several points below DC cell-to-cell efficiency. If a tender guarantees, say, 85% net RTE, the integrator must budget conversion losses and auxiliaries together; under-counting parasitics is a classic way performance guarantees and liquidated-damages clauses get breached at the capacity test.

It also drives sizing and revenue. On a 100 MW / 400 MWh plant (a 4-hour, 0.25C asset) cycling once per day, a 2% auxiliary fraction is close to 3 GWh per year — energy either imported from the grid or subtracted from sales, and at typical wholesale prices worth a six-figure sum annually.

In hot climates, cooling load peaks exactly when the asset is most valuable, so designers oversize cooling and the battery to hold the contracted POI output. Auxiliaries are typically metered separately so that, in a tolling or PPA structure, it is contractually clear who pays for station service and idle-hour imports.

Gross, net, and the headroom auxiliaries spend

Gross output is what the converters produce at their AC terminals; net output is what the revenue meter counts at the POI after the auxiliary tap, the transformers and the collection cables have taken their share. Which of the two a number refers to comes down to one drawing detail — where the auxiliary transformer connects relative to the meter. Tee it off the MV collection bus on the plant side, the normal grid-scale arrangement, and station service is netted out of export automatically: the meter never sees those kilowatts as sales.

Feed it from a separate utility service outside the meter and export is untouched, but the same energy returns as a retail-tariff bill. Neither arrangement makes the load disappear; they decide which column it lands in. The sign is what makes it expensive. While the plant discharges, auxiliaries subtract from what is exported; while it charges, they add to what is imported. A round trip therefore pays for station service on both legs, which is a large part of why net AC-AC RTE sits below the DC figure.

Take the same 100 MW / 400 MWh plant. At roughly 5 MWh per 20-foot container that is on the order of eighty containers, so a few kW of idle draw each puts a few hundred kW on the site continuously, while a hot-weather cooling peak of several tens of kW each arrives as a couple of megawatts at once. The 2% annual case gives the average directly: 2.9 GWh spread over 8,760 hours is about 330 kW.

On those illustrative figures the coincident peak is roughly seven times the annual average, and the two numbers size different things — the auxiliary transformer, its feeders and its protection are specified against the coincident peak on the worst design day, while the energy model and the RTE guarantee run on the average.

Using either for the other's job is a design error in one direction and a revenue error in the other. Whether that peak can be grossed up is a separate question: if the converter fleet is rated at the POI export limit rather than above it, auxiliaries are not added on top of the schedule — they come out of it, and the plant delivers less than nameplate at the meter.

The megawatts go on to cost reactive capability, because the converter rating is a circle rather than a budget. Suppose the fleet is specified at 105 MVA against 100 MW contracted — a 5% margin, the bottom of the range the MVA-headroom entry describes.

Exporting 100 MW net while covering a 2.4 MW auxiliary peak puts the converters near 102.4 MW, and √(S² − P²) then leaves roughly 23 MVAr where 32 MVAr was available at 100 MW. On that arithmetic 2.4 MW of parasitic load has cost about 8.8 MVAr, a rate near 3.7 MVAr per MW, and the illustration isolates the auxiliary term while ignoring the transformer and cable losses that push the same way.

The auxiliary load's own reactive appetite is not the problem: Power Electronics' GEN3 consumption letter keeps |Q| at or below 0.3 kvar against a real draw of up to 9.6 kW, so station service on that platform sits close to unity power factor. It spends the circle through real power. How much of the loss survives to the measurement point depends on where the reactive obligation is written and on the envelope shape the vendor warrants, which is the POI capability envelope's question rather than this one.

Typical values and standards

Across recent utility-scale projects, total auxiliary consumption typically lands between 1% and 3% of annual discharged energy (up to ~5% in hot climates or at low utilization), with thermal management commonly the largest single share — often more than half of the auxiliary total. Liquid-cooled LFP designs generally run toward the low end, around 1-2.5%, while older air-conditioned HVAC containers in hot climates can sit at 3-5% or above, which is a major reason liquid cooling has become the default.

For a modern 20-foot container of roughly 5 MWh, idle draw is on the order of a few kW, while peak cooling during hot-weather cycling can reach several tens of kW. NMC systems demand similar or tighter thermal control; LFP's wider safe operating window eases the worst-case cooling case.

The denominator deserves the same scrutiny as the numerator: a percentage of annual discharged energy, a percentage of throughput counted in both directions, and a percentage of nameplate energy times cycles are three different quantities, and the first two differ by roughly a factor of two on a plant that charges what it discharges.

No safety standard sets an auxiliary-load number, but several dictate which loads exist and must stay energized. UL 9540 is the system-level safety certification for the ESS product; UL 9540A is the separate fire-propagation test method whose data feed NFPA 855, the North American installation standard; NFPA 68 and 69 cover deflagration venting and prevention.

Together these frame the detection, suppression and ventilation loads that can never lose power. On the performance side, the IEC 62933 series describes test methods for electrical energy storage systems, including how auxiliary consumption is accounted in efficiency measurement, and capacity-test protocols in supply contracts typically reference net-of-auxiliary values at the POI.

How it shows up in specs, studies and contracts

In vendor datasheets, look for auxiliary consumption stated per container: an idle or standby kW figure, an average operating figure, and a peak cooling figure, each tied to an ambient-temperature assumption — a value quoted at 25 °C can roughly double at a 45 °C site. Converter vendors publish the same three-figure shape.

SMA's SCS 2930 and 4400 UP-S-US list self-consumption below 370 W in standby, below 2.0 kW averaged across 5% to 100% of rated power at 25 °C, and below 8.1 kW at maximum; the MVPS 8000/8800-S4-US station, which packages two of those converters, roughly doubles all three to 740 W, 4.0 kW and 16.2 kW.

EPC Power's M System states 900 W maximum when operational and 3 kW while the inverters are starting or humidity protection is running — a figure its January 2026 datasheet raised from the 2 kW published in April 2025, which is reason enough to quote the edition alongside the number. None of those converter-station figures include the battery-container cooling that dominates the site total.

The system boundary shifts with architecture too: in an AC-coupled / DC-coupled comparison, check whether the quoted number covers only the DC battery block or also the PCS and MV skid. Ask the integrator explicitly which loads sit inside the efficiency guarantee and which are the owner's station service, and whether the design's 1500 VDC battery feeds any auxiliaries directly or imports them all from the grid.

In contracts and studies, verify three things. First, the RTE guarantee basis: net AC-AC at the POI including auxiliaries typically lands around 85-90%, whereas DC-DC figures of 92-95% exclude them, and mixing the two bases overstates performance by several points.

Second, metering: auxiliary supply is usually separately metered so a tolling or PPA structure can assign station-service cost and idle-hour imports, which a utility may bill at retail-like rates. Third, standby behaviour: interconnection studies model the plant as a small continuous load when idle — a recurring cost that belongs in the financial model, not a footnote.

Common pitfalls

The classic errors are boundary errors: quoting a DC-side efficiency against an AC-POI guarantee, forgetting transformer losses between the auxiliary meter and the guarantee point, or assuming the 1500 VDC battery feeds its own auxiliaries when the design actually imports them from the grid at 400/480 V.

That last one cuts both ways, because the architecture is a purchase-order line rather than a property of the technology — EPC Power's M System will take auxiliary power from an external single-phase feed, from the AC side internally, or from the DC side internally, and only the order confirms which arrived.

Each is a units-and-reference-point mistake, not a technology one, and each surfaces at the capacity test, when the plant delivers fewer MWh at the POI than the financial model promised. Confirming cells stay inside their VDC window and temperature window is a separate question from confirming who pays for the cooling that keeps them there.

The second trap is loss of auxiliary power itself. If station service fails, thermal management stops while the BMS and fire detection must keep running — which is why NFPA 855-driven designs back life-safety loads with UPS capacity, why the switchgear's 110/125 Vdc tripping battery is maintained as its own system rather than as part of the asset being traded, and why a prolonged auxiliary outage in extreme heat or cold can force derating or shutdown.

Ask every vendor how long the system rides through a station-service outage, what the cell temperature excursion looks like with the HVAC or chillers down, and how the plant restarts afterwards.