Battery string

What it is (precise)

The string is a circuit concept; the rack is the cabinet. In utility-scale product the two usually coincide — most racks house exactly one series string — and the datasheet pins it down in xPyS notation: 1P416S means one parallel group of 416 cells in series.

Modules are the assembly step in between: current 5 MWh-class product builds the string from modules of about 104 series cells at roughly 330 V each, four of them in series making the 416S string, where the earlier 280 Ah generation used shorter 48-52S modules and more of them. Vendor vocabulary drifts — the same object appears as string, stack or rack across datasheets — so read the configuration code, not the noun.

The voltage arithmetic is the string's defining property. Series connection multiplies voltage — that page covers why — so a 416S LFP string at 3.2 V nominal per cell sits near 1330 V, and the protected per-cell window of roughly 2.5-3.65 V stretches into a string-level span from a low-SOC floor near 1,040 V up to a full-charge ceiling the designer holds below the 1500 V system rating.

The series count is chosen exactly for those two ends: at the full-charge peak the string must stay under the maximum voltage of its contactors, fuses, cabling and PCS input, and at the bottom of the SOC window under maximum discharge it must stay above the voltage where the PCS can still make full power. Energy follows from the same structure — a 416S string of 314 Ah cells lands at about 418 kWh of DC nameplate, one cell's charge capacity at a thousand-volt-class voltage (derivation under series connection).

Because every cell carries the identical current, the string charges, discharges and stops as a unit — the first cell to reach a voltage limit ends the cycle for all 416. That rule, the weakest-cell limitation, has its own entry; here it needs naming because it is why the string, not the cell, is the level at which capacity, current limits and diagnostics are actually managed.

Per-string protection and control

At the top of each rack sits the high-voltage box — the hardware that makes a string an independently protected unit. Inside: DC contactors in both poles, a pre-charge resistor path that brings the bus capacitance up before the main contactor closes, a main DC fuse, and a shunt or Hall-effect current sensor.

The string-level BMS — the BCU tier of the three-layer hierarchy, sitting between the module measurement boards below and the system controller above — aggregates every cell voltage and temperature, sequences pre-charge and contactor closing, computes the string's state of charge and its charge and discharge current limits, and opens the contactors when a protective bound is crossed.

The contactor and the fuse are a coordinated pair, not alternatives. No single DC device covers the range from routine load switching to the maximum prospective short circuit, so the duty splits: the contactor makes and breaks operating current, and the fuse clears the faults above the contactor's breaking capacity. That is why a rack fuse is specified with a low minimum breaking capacity, around 2 to 3 times rated current — it has to take over exactly where the contactor gives up.

The fault duty itself is set by the neighbours, not the string: a fault inside one string is fed by every paralleled string on the bus, so the fuse is sized for the combined bus contribution rather than the string's own — the parallel-connection entry carries the figures. Some designs rate the contactor for load switching only and lean entirely on the fuse for fault clearing — legitimate, but something to know before writing a switching procedure.

Strings in parallel on the DC bus

One string is a few hundred kilowatt-hours; a plant is built by paralleling them. At about 418 kWh per 416S string, a 5 MWh-class container works out to on the order of a dozen strings on a common DC bus, and that bus feeds the PCS. Paralleling adds current capability and energy — the parallel-connection entry covers that arithmetic — but the property that matters at string level is what paralleling shares: the bus voltage is a single quantity, so every string on it is forced to one operating window, and the fleet can only sweep the band its weakest string tolerates.

That shared window drives real design decisions. Augmentation racks — fresh, high-voltage-for-their-SOC strings added years into the project — go on a separate DC bus or a dedicated PCS input rather than in parallel with aged ones, because a mixed-age bus restricts both generations.

The per-string contactor is also the plant's fault granularity: a tripped string costs its own share of container energy and power while the rest keep running, which is what availability guarantees and lockout/tagout procedures are built on. And since the paralleled bus floats with respect to ground, an insulation monitoring device supervises the whole rack-and-bus assembly for developing ground faults — one more function that only exists at string level and above.

How it shows up in specs, studies and contracts

On a battery datasheet the string is a table: configuration code (1P416S), nominal and minimum/maximum voltage, DC nameplate energy, continuous and peak current, and the UL 1973 listing. The reconciliation work is at the edges of that table. The string's voltage span must sit inside the PCS DC input window not just at 25 degrees C and beginning of life but at the full-charge top and the cold, loaded, end-of-life bottom — the corners move apart as resistance grows.

Continuous current must match the duty cycle the revenue model assumes, and the fuse and contactor interrupt ratings must be checked against the prospective DC short-circuit current of the fully paralleled bus, not one string's own contribution.

In studies and contracts the string is the unit of accountability. The DC short-circuit study sums paralleled string contributions to set what each fuse must clear. The witnessed capacity test ends when the first limiting string ends it — the test-day consequence the cell-imbalance entry unpacks — which is a good reason to contract for string-level state-of-health telemetry rather than fleet averages.

Replacement modules are matched and binned per string, mixing revisions within a rack voids fire-test basis and warranty, and the augmentation clause should state explicitly which bus later strings connect to and what voltage window they must fit.

Common pitfalls

The first trap is vocabulary. String, rack and stack are used interchangeably across vendors, and the xPyS code is read at a level: 1P at string level says one series chain per rack, while some products still parallel cell groups inside the module — so establish which assembly level a configuration code describes before multiplying anything. The related error is summing cell nameplates to a string energy: series strings deliver what the worst cell permits, and the weakest-cell limitation entry covers why the sum is a ceiling the string never reaches.

The second family of traps is window arithmetic done at one operating point. Quoting nominal string voltage where a study needs the full span understates both ends; checking the PCS match at 25 degrees C misses the binding corners — the full-charge peak at the top (hot vs cold per the design table) and cold loaded sag at the floor.

Two smaller ones: the fuse rating is a fault-clearing number, not an operating allowance — the continuous current limit belongs to the BMS and the thermal design — and pre-charge is not a formality, because closing a main contactor onto discharged bus capacitance is a switching event the contactor is not rated to survive routinely.