Current A

What it is (precise)

Current is charge per unit time — one ampere is one coulomb per second — and power is voltage times current: P = V × I on the DC side, P = √3 × V(L-L) × I(line) × cos φ on the three-phase AC side (the single-phase vs. three-phase entry derives the √3). In a storage plant the useful direction to read those equations is backwards. The contract fixes power; the platform fixes voltage; current is what falls out, and current is what the equipment schedule is written in.

A cable has an ampacity, a busbar a rated current, a contactor a making and breaking current, a fuse a rated current and a breaking capacity — none of them carries a MW rating, because the heat, the magnetic forces and the arc energy inside them all follow from amps. How many volts it costs to push a current through a resistance is Ohm's law, and what that current does quadratically to losses is the I²R entry; this page is about current as the sizing quantity itself.

A second precision matters more in BESS than anywhere else: a PCS is a constant-power device across its DC window, so its current is not fixed — it is P divided by whatever the battery voltage happens to be. A string's voltage is a window, not a number, which means the converter's DC current swings across the day and across the asset's life, and the maximum arrives at the minimum voltage: the cold, loaded, low-SOC corner. Every DC conductor is sized to that corner, not to nominal.

Ratings come in three distinct current families. Continuous current is a thermal-equilibrium figure at a stated ambient — the conductor or device can carry it indefinitely at that temperature, and the figure falls as the air gets hotter: a DC/DC converter stage rated 1,200 A at 30 C is a 1,040 A device at 50 C. Peak or overload current is a time-limited excursion above continuous, spent against thermal mass and always quoted with its duration.

Fault current is a different regime entirely — a prospective short-circuit magnitude the device must withstand or clear, quoted as kiloamps against seconds: medium-voltage ring switchgear rated 630 A continuous is simultaneously rated to withstand 25 kA for one second. Mixing the families up is the classic specification error, and the pitfalls section returns to it.

The current ladder of a plant

Start at the cell. A series string has one conduction path, so every cell, weld and module busbar carries the identical current (the series-connection entry owns why), and that current is set by the C-rate: a 314 Ah LFP cell at 0.5C carries 157 A, whether it stands alone or sits in a 416-cell string at 1,330 V. Grid-scale duty at 0.25-0.5C therefore puts the string in the ~80-160 A class — modest current at lethal voltage, which is why rack conductors are slimmer than intuition expects and why the insulation, not the copper, dominates module design.

Strings parallel onto the DC bus, and parallel currents add. A 5 MWh-class container is on the order of a dozen 418 kWh strings, so its 0.5C discharge lands near 1,900 A at nominal voltage — the same figure 2.5 MW ÷ 1,330 V gives directly. The DC inputs of current utility PCS product are rated to match: roughly 2,300 to 4,600 A of continuous DC current across mainstream frames, with the largest recent frames reaching ~5,500 A. This is the kA-class part of the plant, where busbars become laminated stacks, joints get torque-and-thermography programs, and the cost of a milliohm is real heat.

The AC side runs the same power at its own two voltage levels. At the 690 V low-voltage output that current grid-scale PCS product defaults to, the currents are the largest in the plant — 690 V frames span roughly 1,800 to 3,700 A of maximum AC output current at 40 C across the size range — which is exactly why the step-up transformer sits within metres of the PCS on a shared skid rather than at the end of a cable run.

One transformer later the picture inverts: 5 MVA at a 34.5 kV collection voltage is about 84 A, and a 630 A-class medium-voltage feeder loops several containers on a single ring. Same megawatts at every level; three orders of magnitude of current between the cell terminal and the collection bus.

Voltage buys copper

For fixed power, current is inversely proportional to voltage — and that single ratio is the economics of every voltage-class decision in storage. Moving a platform from 1000 V-class to 1500 V-class DC cuts the current for the same megawatts to two-thirds; conduction heat in an unchanged conductor scales with current squared, so it falls to four-ninths. Turn the argument around and the copper shrinks instead: holding the same heat, the 1500 V design needs less than half the conductor cross-section of the 1000 V design.

That saving repeats through every cable, busbar, contactor frame, fuse and terminal between the racks and the PCS, which is why the industry moved, and why 2,000 V architectures are the next step on the same line. The price is paid in a different budget — every component must now be insulated and certified for the higher voltage class — and the quadratic loss arithmetic itself belongs to the I²R entry.

The AC transformer is the same trade executed in one step. Stepping 690 V up to 34.5 kV divides the feeder current by fifty, which is what makes kilometres of collection cabling affordable at all; nobody runs 3,000 A across a site. Inside the fence the rule of thumb that falls out is simple: high current is tolerable only over short distances, so the plant layout clusters its kA-class connections — DC bus to PCS, PCS to transformer — into metres, and lets the long runs happen at low current and high voltage.

How it shows up in specs, studies and contracts

On a PCS datasheet, current appears as the honest version of the power rating: maximum AC output current at a stated temperature per frame and voltage variant, and maximum continuous DC current per input.

Cross-checking those rows is the reconciliation work — the AC current times √3 times the actual grid voltage should reproduce the kVA the sales sheet claims, at the temperature your site actually reaches, and the DC input rating must cover the container's current at the bottom of the string voltage window, not at nominal. The cable schedule then translates each current into a cross-section through ampacity tables, derated for grouping, burial or tray, and site ambient; the binding row is the hottest continuous case.

Studies split by current family. The load-flow and thermal work uses continuous currents; the DC short-circuit study sums the prospective fault contribution of every paralleled string to set what each fuse must clear and each contactor must survive — a duty set by the neighbours, not the faulted string itself, as the battery-string entry details.

Keep the withstand and contribution directions straight: a PCS datasheet's maximum DC short-circuit figure is what the equipment tolerates, not what your bus will deliver — the study computes the delivered number, and the comparison between the two is the actual check. Commissioning then measures current where design assumed it: current sharing across paralleled strings, phase currents balanced within tolerance, and thermographic scans of joints at full load, because a failing joint announces itself in amps and kelvin long before it announces itself in megawatts.

Common pitfalls

The expensive error is sizing DC conductors at nominal voltage. A constant-power converter draws more current as its source voltage falls, and the battery spends real operating hours below nominal: a 416S string nominal at ~1,330 V has a cold, loaded low-SOC floor near 1,040 V, and 2.5 MW at the floor is about 2,400 A against the 1,880 A the nominal arithmetic gives — roughly 28% more, before any ambient derate. The corner cases compound in the same direction: cold raises the sag, age raises resistance, and the thermal design meets its worst current on a day the nameplate never mentions.

The second family of errors is confusing the three current ratings. A fuse's rated current is not an operating allowance — continuous duty belongs to the BMS limits and the thermal design, and the fuse number is about what it can carry and clear. A peak or overload current is a stopwatch figure, not capability; quoting it against a continuous duty cycle is a claim the hardware cannot keep.

And a withstand rating is not a fault contribution — one is what equipment survives, the other is what the system throws at it, and only the short-circuit study says whether they are compatible. Each of these has sunk real equipment schedules, and each is caught by asking one question of every current on a datasheet: continuous at what temperature, peak for how long, or fault against what duration?