Phase unbalance

What it is (precise)

Start with the definitions, because the number on a motor datasheet and the number in a connection agreement are usually not computed the same way. The NEMA MG-1 definition is percent unbalance = 100 × (maximum deviation from the average) / (average), evaluated on the three line-to-line voltages; the same arithmetic applied to the three line currents gives percent current unbalance.

Line-to-line voltages are the right input for a subtle reason: each is the difference of two phase voltages, so any component common to all three phases cancels out of them — the zero-sequence entry covers that component, and the practical consequence is that the NEMA figure responds only to magnitude asymmetry, not to a shifted neutral.

The IEC family of standards instead defines the voltage unbalance factor as the ratio of the negative-sequence voltage to the positive-sequence voltage, expressed as a percentage. The decomposition of an unbalanced set into positive-, negative- and zero-sequence sets is the phase-sequence entry's territory; what matters here is that the two definitions are different calculations that agree only approximately at small unbalance, so a limit quoted without its definition attached cannot be tested against.

Voltage unbalance and current unbalance are different quantities with different owners. Voltage unbalance is a property of the supply at a bus; current unbalance is a property of the loading on it; each drives the other through impedance. The direction that surprises people is voltage-to-current: a load whose negative-sequence impedance is low — an induction motor's is close to its locked-rotor impedance, the same low impedance that draws several times rated current at starting — converts each percent of voltage unbalance into several percent of current unbalance.

The reverse coupling is gentler: unbalanced load current produces voltage unbalance only through the source impedance, so a stiff bus tolerates lopsided loading that a weak one will not. Measurement follows the IEC chain in practice: power-quality instruments built to IEC 61000-4-30 evaluate unbalance from sequence components and aggregate it over ten-minute windows, which is why grid limits are written as weekly statistics rather than as a value you can read off a multimeter during a site walkdown.

Where it comes from in a BESS plant

The main power train is balanced by construction. The PCS switches its three phase legs symmetrically, and the transformers, cables and switchgear behind it are three-phase equipment carrying three-phase current. The unbalance lives in the auxiliaries: control cabinets, lighting, receptacles, trace heating, UPS supplies, fire-alarm and communications panels are single-phase, and every one of them lands on one phase of the auxiliary transformer.

Individually they are trivial; collectively, left to whichever way each panel happened to be wired, they can leave a standing unbalance on the auxiliary supply — the single-phase vs three-phase entry makes the same point from the other direction, and its advice stands here: ask the electrical contractor for the auxiliary schedule with phase assignments on it, not just a total in kVA.

The plant is also a victim, not only a source. Voltage unbalance arrives at the point of interconnection from the upstream network — untransposed overhead lines have asymmetric per-phase impedances, and single-phase distribution loads elsewhere on the feeder pull the phases apart — and everything inside the fence then rides on that unbalanced supply.

Unequal loading among LV feeders inside the plant works the same way at smaller scale: a distribution board serving mixed single-phase loads presents its unbalance to everything sharing the bus. The two directions matter contractually because they are policed separately: the network operator limits what the plant may emit, and the plant's equipment has to tolerate what the network delivers, and neither obligation cancels the other.

What it costs

The plant's rotating machines pay first, and in a BESS the rotating machines are the thermal-management fleet — HVAC compressors, coolant pumps, fans. Negative-sequence voltage drives current through an induction motor limited only by that near-locked-rotor impedance, and the resulting field rotates against the rotor, inducing rotor currents near twice line frequency where skin effect raises the effective resistance.

Heating scales with current squared, so a small voltage number becomes a large thermal one: NEMA MG-1 tells motor users to derate above 1% voltage unbalance and does not recommend operation above 5% at all. Transformers pay differently — unbalanced loading works one winding harder than the others, so the hottest phase reaches its thermal limit while the three-phase kVA total still reads as margin. A cooling system specified against balanced-load arithmetic loses life on its most loaded phase in exactly the summer weeks the containers need it most.

Three more bills arrive downstream. In a four-wire LV system the unbalanced portion of the load current returns through the neutral — the neutral-current entry owns that consequence and its conductor-sizing arithmetic. Protection is set on the assumption of a balanced supply: phase-loss and unbalance relays on motor starters, negative-sequence elements, and phase-failure monitors all lose margin to a standing unbalance, and what the operations log then records is nuisance trips of healthy thermal-management equipment — availability events with no failed component behind them.

And the PCS itself has less to give than its nameplate suggests: a converter's real limit is per-phase current, so under unbalanced terminal voltage the most loaded phase saturates first and the deliverable three-phase kVA drops below rating, while the unbalanced set puts a double-line-frequency power component onto the DC link that a balanced three-phase design never has to buffer — the single-phase vs three-phase entry derives where that pulsation comes from.

How it shows up in specs, studies and contracts

The reference limit on the supply side in Europe is EN 50160: the negative-sequence component of the supply voltage within 0 to 2% of the positive-sequence component for 95% of the ten-minute mean RMS values in each week, with up to about 3% occurring in some areas where partly single-phase-connected installations dominate.

Connection agreements then pass a share of the headroom down to the plant as an emission limit at the point of interconnection. Notice which definition that chain uses — the sequence ratio — while a motor datasheet quoting NEMA derating uses deviation-from-average; reconciling the two is the reviewer's job, not the vendor's, and the reconciliation starts by writing the definition next to every unbalance number in the specification.

In studies, unbalance is what the default model cannot see. Load-flow and short-circuit models are balanced positive-sequence representations unless someone asks otherwise, and everything single-phase inside the fence is a source of negative- and zero-sequence current that such a model sets to zero. Hand the study consultant the auxiliary schedule with phase assignments, so the standing unbalance is a modelled quantity rather than a commissioning surprise.

At commissioning, the cheap fix window is open exactly once: record per-phase currents at the auxiliary transformer under representative load — thermal management running, not the empty-site condition of a quiet acceptance day — and rebalance panels before commercial operation, because moving a single-phase breaker to another phase costs minutes before handover and an outage request after it.

Common pitfalls

The first trap is treating 2% as 2%. Under the NEMA calculation and the IEC sequence ratio the same physical condition returns different numbers, and a voltage percentage is not a current percentage: the multiplication through motor negative-sequence impedance means the heating consequence of a 2% voltage unbalance is far larger than the figure suggests, which is why the derating obligation starts at 1%.

The second trap is the snapshot: the limits are weekly statistics of ten-minute aggregates, so one balanced reading during a walkdown proves nothing about the week, and one bad reading during a switching operation breaches nothing. Trend the logged data from the power-quality meter the connection agreement made you install.

The last trap is treating standing unbalance as a fault to be silenced. If phase-loss and unbalance protection keeps operating on the thermal-management fleet, the temptation is to widen the relay settings until the tripping stops — which removes the protection while leaving the heating. Standing unbalance is a design condition to be budgeted like harmonics: measure it, trace it to the phase-assignment schedule or the upstream network, and fix it by moving load between phases. The relay was doing its job; the auxiliary panel schedule was not.