Sizing BESS Transformers for the Actual Duty Cycle

A nameplate rating is a continuous load at a reference ambient, declared under IEC 60076-2. It says nothing about how often the unit will cycle, how deep the cycles will be, what the harmonic content of the current is, what the air temperature inside the enclosure will be, or how much reactive current the plant will be asked to deliver. All of those are part of a storage duty.

Sizing on the nameplate therefore fails in both directions. Oversizing buys copper and core that are never used, while no-load loss runs for 8,760 hours a year; over a transformer's service life, operating losses can reach several times the purchase price, a ratio the U4E programme under UNEP puts at up to four times. Undersizing shows up later, as hot-spot excursions on the hottest days of the year, which is also when a storage plant is most likely to be called on for full output.

The common convention is to size the transformer at 110 to 120 percent of the PCS AC rating. The margin is not arbitrary. The transformer carries RMS current, not active power, so harmonic current and reactive current both count towards its loading. If the PCS is quoted in kW at unity power factor, its AC current rises as soon as it delivers reactive power, and grid codes increasingly require storage to operate across a wide reactive range.

Two checks keep the convention honest. Size on the PCS maximum AC current at minimum operating voltage, not on its MW rating. And confirm whether the 110 to 120 percent figure was derived from the harmonic spectrum and the site ambient, or simply assumed. The lower end of the range suits a unit in an open yard with modest distortion; the upper end suits a containerised unit in high ambient with a wide reactive range.

The winding hot-spot follows the load with a time constant of minutes; the top-oil temperature follows it with a time constant of hours. A one-hour discharge is therefore long enough for the winding to reach its hot-spot, but short enough that the oil never reaches steady state. Over repeated daily cycles the oil temperature ratchets upward, and the hot-spot moves with it.

That matters because insulation ageing is exponential in hot-spot temperature, not linear. IEC 60076-7 and IEEE C57.91 both model the ageing rate as an exponential function, with a roughly 6 K doubling interval for thermally upgraded paper. A short daily excursion above the reference therefore costs more life than the same energy delivered at a lower steady temperature.

It follows that two transformers with identical nameplate capacity and identical total load loss can consume different amounts of life under the same dispatch, because their eddy-loss fractions and cooling designs differ. The comparison that matters is the calculated ageing rate over the profile, not the nameplate.

Where storage provides voltage support, the PCS can be delivering current at or near its limit with little or no active power. The transformer carries that current and generates the losses; the plant earns nothing from the energy. Sizing on MWh throughput alone misses this entirely, and it is one of the more common reasons a storage transformer runs hotter than its load profile appears to justify.

Standby is the other omission. With the PCS idle, the transformer still carries its no-load loss continuously. In a plant that completes one or two cycles a day, the annual no-load loss energy can approach or exceed the annual load-loss energy. Because no-load loss is a design property that cannot be reduced after the fact, it belongs in the sizing and purchasing decision rather than in a post-commissioning review.

Dyn11 is the usual vector group for storage step-up duty. The delta-connected high-voltage winding provides a closed path in which zero-sequence and triplen harmonic currents — 3rd, 9th, 15th — circulate instead of being transferred to the supply-side lines, and the connection gives the 30 degree displacement between the two sides that the surrounding distribution convention expects. The group must be declared, and units that will be paralleled must match it.

Impedance is a trade-off, and the usual band for these units is 4 to 6.5 percent. Lower impedance gives better voltage regulation and lower reactive consumption, but raises the short-circuit current that the windings and the downstream switchgear must withstand. Higher impedance limits fault current but costs voltage regulation, which matters most on weak grids where the plant is expected to hold voltage while exporting. Because a PCS is current-limited, the fault contribution from the converter itself is bounded; the dominant fault current usually comes from the grid. Short-circuit withstand is demonstrated to IEC 60076-5 by calculation or by a special test carried out on a sample, so the specification should call for the supporting report rather than assume the test is part of routine testing.

Two multipliers apply on top of the duty cycle. The first is harmonic derating: PCS switching produces a distorted current, and the eddy-loss weighting rises with the square of the harmonic order, so the calculation follows IEEE C57.110 and uses the declared loss split rather than a rule-of-thumb factor. The second is the thermal environment: a transformer inside a container sees air that is hotter than the site ambient, because the enclosure traps heat and adds solar gain. On a Gulf site with a 55 degree design ambient and a 75 degree solar-exposed surface, the container air temperature is the number that should be used, and it compounds with the harmonic derating.

Sizing therefore runs as a short procedure rather than a lookup. Assemble the dispatch profile: cycles per day, depth of discharge, charge and discharge duration, ramp rates, reactive duty and standby hours. Take the PCS maximum AC current and its harmonic spectrum at several load points. Fix the ambient basis, including the internal temperature if the unit is enclosed. Compute the harmonic loss factor and the resulting permissible current. Then run the thermal calculation over the profile to IEC 60076-7 and read out the hot-spot and the ageing rate. Choose the rating that holds the hot-spot inside its limit over the worst sustained period, not the average.