Nominal voltage

What it is (precise)

A battery's voltage is not a constant; it is a curve. It falls as the cell discharges, sags under load, and moves with temperature and age. Nominal voltage is one agreed point on that curve, chosen as the label. For LFP the choice is natural — the chemistry's flat discharge plateau sits around 3.2-3.3 V per cell, so the cell really does spend much of a discharge near the value that names it — but nothing holds it there, and the window the BMS actually protects runs from about 2.5 V empty to about 3.65 V full.

Three different quantities get called 'the voltage' on a battery datasheet: the nominal (a chosen constant), the open-circuit voltage (a measurable curve against state of charge, taken at rest — it has its own entry), and the terminal voltage under load (open-circuit voltage minus the I×R sag of the moment). Only the first never changes, because it is the only one that is a name rather than a measurement.

Labels multiply exactly, which is what makes the convention useful. Series counts carry the nominal up the hierarchy unchanged: 104 cells at 3.2 V make the ~330 V module of current 5 MWh-class product, four such modules make a 416S string labelled ~1,330 V, and anyone with the cell datasheet and the series count reproduces the arithmetic to the volt.

But the window multiplies too, which is what makes the label insufficient: the same 416 cells taken to the cell's 3.65 V datasheet maximum would sit near 1,518 V — over the 1500 V system rating — which is exactly why deployed 416S platforms hold their operating charge ceiling below the cell maximum. The nominal label never sees that constraint; the design table does.

One label, a moving window: the DC ladder

Current 1500 VDC-class strings run on the order of 360 to 416 LFP cells in series — nominal roughly 1,150 to 1,330 V — while the voltage the PCS actually receives spans from a low-SOC floor near 900-1,040 V up to a full-charge top held just under the 1500 V rating. No binding check on the DC side is made at the nominal point. Insulation, creepage, contactor and fuse voltage ratings are specified at maximum system voltage.

PCS compatibility is checked at the corners of the window: the floor is set by cold, loaded sag — low temperature and full current together, worst at end of life — and the full-charge top is a hot-versus-cold question the design table answers. Nominal voltage is not one of the corners; it is the middle of the band, where nothing fails.

Current sizing follows the same rule. At fixed power, current scales inversely with voltage, so DC conductors, contactors and the PCS input current limit are sized at the bottom of the window, not at the label: delivering full power at the 900-1,040 V floor draws roughly 28% more current than the same power at the 1,150-1,330 V nominal — the ratio of the band's own endpoints.

A cable schedule or a protection coordination built on the nominal figure is undersized by exactly that factor on the days the plant is driven hardest. The label does earn its keep as a fingerprint, though: a ~330 V module nominal identifies current 104S 5 MWh-class product, while 150-170 V identifies the earlier 280 Ah generation's 48-52S modules — reading the nominal tells you which hardware generation you are looking at before the datasheet does.

AC nominals and the bands equipment rides

The AC side is nominal-labelled all the way up the ladder: PCS output in the 400-800 V range with 690 V the grid-scale workhorse, medium-voltage collection at 13.8, 20, 33 or 34.5 kV, and the interconnection at whatever the network names.

The nominal values themselves are standardized — IEC 60038 is the reference — and the per-unit system is built directly on them: 1.0 pu is by definition the nominal voltage, so load-flow studies, relay settings and grid-code envelopes are all written as multiples of the label. That is the convention working as intended — one agreed base, every document comparable.

What equipment actually rides is a band around the base. Continuous operation typically spans about 0.9 to 1.1 pu, with grid-code ride-through envelopes extending beyond it for fault seconds, and a rating quoted at nominal moves the moment the voltage does: a converter is a current-limited machine, so its kVA tracks its AC voltage — the voltage entry's 480-vs-690 V worked pair — and a depressed grid trims deliverable MVA exactly when it wants VARs most; the per-unit-system entry works that corner.

So 'the 690 V rating' is a rating at 690 V and nowhere else, while the connection agreement's obligations apply across the whole tolerance band. Transformers carry the convention as a pair: the ratio is stated nominal-to-nominal — 0.69/34.5 kV — with off-load taps at plus/minus 2 times 2.5 percent to move the real ratio around the labelled one, a reminder that even the ratio between two nominals is adjustable hardware, not arithmetic.

Nominal energy: arithmetic the meter never confirms

Energy is where the label does its heaviest commercial work. Cell datasheets rate capacity in ampere-hours; the kWh printed beside the Ah row is the Ah multiplied by the nominal voltage — 314 Ah × ~3.2 V ≈ 1.0 kWh — and the container headline is the same multiplication carried through the cell count: 416 × ~3.2 V × 314 Ah ≈ 418 kWh per rack, on the order of 5,000 cells behind a ~5 MWh label.

A real discharge integrates voltage times current along the actual trajectory — sagging with load, drooping with state of charge, shifting with temperature — so the meter never reads the label figure, and cannot: pinning the voltage at one agreed value is what makes the number reproducible across bidders, and the same pin is what guarantees it is not a measurement.

Which rung of the resulting ladder a price, a retention percentage or a capacity-warranty baseline references — and who owns the gap between the label and the commissioning test — is the nominal-energy entry's subject; the cause of the gap lives here.

Common pitfalls

The first family of errors treats the label as a setpoint.

Nothing regulates a battery to its nominal voltage — the string passes through it and moves on — so every check performed at the nominal point misfires in the same direction: dividing power by nominal voltage understates current at the low-SOC floor by roughly the 28% above, checking the PCS window at nominal misses both binding corners, and comparing cells across chemistries by their kWh rows compares their nominal conventions as much as their capacity — 314 Ah is ~1.0 kWh at LFP's 3.2 V and rather more at NMC's 3.6-3.7 V, before either cell has stored anything differently.

The second error conflates nominal voltage with open-circuit voltage. OCV is physics — a measurable curve against state of charge, the basis of SOC estimation, balancing decisions and the voltage-match checks before paralleling — while nominal is a bookkeeping constant, and LFP's flat plateau makes the two look interchangeable exactly where the distinction matters most: on the plateau, a small OCV offset conceals a large SOC difference, and no arithmetic done at 3.2 V flat will surface it.

The working rule: label arithmetic (energy ratings, series-count voltages, per-unit bases) runs on nominal; every operating decision — sizing, protection, window compatibility, state estimation — runs on the measured curves and the corners of the band. Keeping each calculation on its own side of that line is most of what this term is for.