Voltage V

What it is (precise)

Voltage is defined between two points: the energy per unit of charge gained or lost in moving from one to the other, one joule per coulomb per volt. Multiply it by the current flowing and you get power — P = V × I on the DC side, P = √3 × V(L-L) × I(line) × cos φ on the balanced three-phase AC side — which is why voltage and current are the two axes every piece of BESS equipment is rated on.

The two-point nature is not pedantry; it is how measurements go wrong on site. Pole-to-pole across a DC string is one number; either pole to ground is a different one, set by the grounding design rather than by the battery; and a meter reading taken against the wrong reference proves nothing about whether a bus is dead. Isolation-verification procedures list every pair of points to be measured for exactly this reason.

A voltage figure also carries conditions. The number printed beside a cell, module or string is a nominal — a naming convention, covered under nominal voltage — while the terminals actually sit at an open-circuit voltage set by state of charge and move away from it under load as current works through internal resistance (the voltage-drop entry has the I × R mechanics).

So the plant never has 'a voltage'; it has windows — a range each source can present and a range each downstream input can accept — and the design work is making them overlap. The DC version of that exercise is the VDC window; the AC version is the tolerance band the grid code defines at the point of interconnection.

The reason to build voltage at all is current relief. For a fixed power, current falls as 1/V and conductor loss as 1/V² — the arithmetic that took DC architectures from 1000 to 1500 VDC (about a third less current and roughly 4/9 the resistive loss for the same megawatts) and that justifies every transformer between the PCS and the grid. Series connection covers how the DC side builds voltage and transformer turns ratio covers the AC side; this page is about the quantity they are both building.

The DC ladder: 3.2 V cell to the 1500 V class

The DC side is built by stacking. An LFP cell holds about 3.2 V nominal inside a protected window of roughly 2.5-3.65 V — an electrochemical fact no pack engineering changes. Current 5 MWh-class product wires about 104 cells in series for a ~330 V module (the earlier 280 Ah generation used shorter 48-52S modules at ~150-170 V), and four such modules in series make a 416S string near 1,330 V nominal; across vendors, 1500 VDC-class strings run on the order of 360 to 416 cells, giving nominals of roughly 1,150-1,330 V and low-SOC floors near 900-1,040 V.

The series count is chosen so the full-charge peak stays below the 1500 V system maximum — the top of the IEC low-voltage DC band, the ceiling every contactor, fuse, cable and PCS input in the chain is rated against. The 1500 VDC entry covers why the industry parked just under that line.

None of those numbers is what a voltmeter reads at an arbitrary moment. A string sweeps on the order of 480 V between empty and full — the ~1.15 V a single LFP cell traverses, multiplied across 416 of them — and load current, temperature and age push the terminals further.

Which corner of that excursion binds is a design-table question: the low corner is set by cold, loaded sag at low state of charge, the full-charge top by the hot-versus-cold comparison in the design table, and the whole swept range must sit inside the PCS DC input window at end of life, not just on commissioning day. Those mechanics belong to the VDC window and voltage-drop entries; the point here is that the DC voltage of a BESS is a moving operating point inside a designed envelope, and every DC rating in the plant names which part of the envelope it refers to.

The AC ladder: PCS terminals to the point of interconnection

On the AC side voltage is set by transformer windings rather than chemistry. The PCS produces low-voltage AC in the 400-800 V band — 690 V is the common grid-scale default — and AC figures follow their own conventions: RMS, line-to-line, at a stated frequency, unless the datasheet says otherwise (root mean square covers what the RMS convention means).

The voltage matters commercially before it matters electrically, because a converter is a current-limited machine: at a fixed current limit its kVA scales directly with AC voltage, and one shipping modular PCS delivers 374 kVA per module at 480 V but 537 kVA at 690 V — same hardware, same current, about 43% more power for naming a higher voltage. A vendor headline quoted on a voltage variant your project does not use is power your project will not get.

From the PCS terminals the plant steps up — typically once to a medium-voltage collection level such as 34.5 kV, and again at the substation for a high-voltage grid connection. Each step trades current for voltage so the collection cables and switchgear carry a fraction of the LV-side current.

The POI voltage is where the obligations attach: the interconnection agreement and grid code define the voltage band the plant must ride through and the reactive power it must deliver to support the grid's own voltage, all at the POI, and the studies express those duties in per-unit on the POI voltage base (the per-unit system entry covers the normalization). One quantity, three different jobs on the AC side: a rating reference at the PCS, a loss-reduction lever through the transformers, and a controlled variable at the grid interface.

How it shows up in specs, studies and contracts

Read any BESS datasheet as a table of voltage-qualified numbers. The battery container states a maximum system voltage, a nominal and an operating range; the PCS states an AC output voltage, a DC input window and a kVA figure that holds at exactly one AC voltage; cables and busbars carry an insulation rating and an ampacity, and the power they move depends on the voltage you run them at.

Reconciliation is the actual engineering: the battery's full excursion inside the PCS window at every design corner, the PCS AC voltage matching the transformer LV winding, and every DC component rated to the same system maximum with margin — not a mix of 1000 V-class and 1500 V-class parts sharing one bus.

In contracts, voltage appears as the compatibility clause obligating the battery's window to match the named PCS, as the reference conditions of the capacity test, and in augmentation language — racks added years later must land inside the original bus voltage window, or the shared bus pins old and new generations to a range neither prefers.

At commissioning, the series arithmetic doubles as a diagnostic: a string voltage that is not the cell count times a plausible per-cell voltage points at a wiring error or a failed module, and DC insulation-resistance tests are referenced to the actual system voltage rather than the nominal.

Insulation, touch and arc flash

Insulation is bought by system voltage, not by the job a conductor does. A sense wire tapping a cell deep in the string carries milliamps yet sits hundreds of volts from ground, so it needs the same insulation class as the power path; creepage and clearance distances inside modules, combiner panels and PCS cabinets are sized to the up-to-1500 V system voltage under IEC 60664; and the ~330 V module is deliberately the largest piece an installer handles, because the full string voltage only exists once the modules are linked inside the rack.

That is also why voltage-class changes are expensive: moving from 1000 to 1500 V, or toward emerging 2,000 V architectures, re-rates every insulated surface in the enclosure, not just the cables.

The reference point matters most in the safety analysis. Grid-scale DC buses typically run floating (IT) with continuous insulation monitoring, so on a healthy bus neither pole holds a firm voltage to ground — and after a first ground fault, the other pole sits a full system voltage from ground while no fuse blows and nothing trips, which is precisely the condition the insulation monitor exists to catch before a second fault or a technician completes the circuit.

Arc flash adds the energy dimension: DC has no natural current zero, so a fault arc at 1500 V must be stretched and cooled until it starves, and the DC arc-flash study — built on the impedances of the one-line diagram — sets the PPE boundaries inside the container. Live-dead-live verification across every pole pair and from each pole to ground is the procedural translation of the same physics: voltage exists between two points, so the absence of voltage has to be proven between all of them.