Sign convention

Two conventions, and the one machine that is both

The two conventions are bookkeeping choices, not physics. The passive sign (load) convention points the reference current arrow into the device, so positive power is power absorbed — the natural frame for anything that consumes, and the frame battery vendors think in: a BMS typically counts charge current as positive, because charging is the direction its protection limits act on and its amp-hour ledger accumulates.

The generator convention points the arrow out, so positive power is power delivered — the natural frame for anything that sells energy, and the frame the PCS, plant SCADA and market systems face the grid in. Same copper, same electrons, opposite ledger; neither is wrong.

A gas turbine never has to choose, because its flow never reverses. A BESS reverses on schedule: a 100 MW plant is a 100 MW generator in one settlement interval and roughly a 100 MW load in the next, and both frames coexist inside one container — the BMS on the DC side logs a charging half-hour as positive current while the PCS on the AC side logs the identical half-hour as negative power.

Both are internally correct in their own frame. The failure mode is numbers crossing frames without translation: an EMS summing battery-side and meter-side registers, a historian trending mixed tags, an energy model fed one export figure with the wrong sign.

The convention this glossary fixes — the one the power entry carries as a key fact — is generator convention at the AC boundary: positive = discharge/export, negative = charge/import, agreed in writing before the first point list is built. Every derived quantity inherits it. State-of-charge accounting integrates signed power, round-trip efficiency divides a discharge sum by a charge sum, and a market bid to charge is a negative-MW offer in most dispatch systems; each of those calculations is one unstated sign away from nonsense.

The reactive sign: inject or absorb, over- or under-excited

Under generator convention, positive Q is injection: the plant runs over-excited, sourcing VARs like a capacitor bank and supporting local voltage. Negative Q is absorption: under-excited, pulling voltage down. The treacherous words are leading and lagging — a lagging generator injects reactive power while a lagging load absorbs it, so the same adjective points in opposite directions depending on which convention its author was raised in. A specification should state the direction as inject/absorb or over-excited/under-excited with an explicit sign, and never lean on lead/lag alone.

The BESS consequences are immediate. A volt-var curve assumes a Q polarity; flip it and a stabilizing droop becomes positive feedback — high voltage commands more injection. And because a four-quadrant PCS can exchange reactive power at zero real power, running as a STATCOM while the battery idles, the sign of Q is the only number that says whether the idle plant is holding the feeder up or dragging it down.

The quadrant map itself — which corner of the P-Q plane is which — belongs to four-quadrant operation; what this entry owns is the polarity of the axes, which has to be fixed before that map means anything.

The interconnection-vs-metering trap

The two document families of one project are routinely written in opposite frames. On the interconnection side, power-flow and dynamic studies model the plant as a generator — export positive — and the interconnection agreement caps injection as a positive MW figure, increasingly with a separate charging-load limit beside it.

On the metering side, revenue meters are organized around delivered and received registers, and a meter programmed from the utility's point of view can count energy delivered to the plant — charging — as its positive channel. Same plant, same megawatt-hours, opposite signs, and both documents internally consistent, which is exactly why the conflict survives design review.

Where it bites is settlement and testing. A flipped sign in the loss-compensation or telemetry chain can make a plant fail a capacity test on paper while performing correctly — the reconciliation duty the energy-capacity-test entry describes. Efficiency arithmetic at least fails loudly: divide a discharge sum by a wrong-signed charge sum and the answer is negative or impossibly high. The dangerous case is the double flip — two sign errors that cancel in the discharge-only test everyone witnesses, then reappear months later in charge-side settlement or a reactive-dispatch instruction.

The fix is a document, not a debugging session: a sign-convention sheet in the project dossier that lists every interface — BMS, PCS, revenue meter, SCADA, plant controller, market gateway — states its convention, and attaches a signed worked example to every point, in the form the setpoint entry insists on: a value of +50.0 at this register means 50 MW export at the POI.

Commissioning then closes the loop end to end: a known P and Q injection at the PCS terminals must reconcile with every layer up to the POI meter in magnitude and in sign before any guarantee, settlement, or control function is trusted.

Common pitfalls

The vocabulary traps come first. Lead/lag flips meaning between conventions, as above. Import and export are written from an unstated vantage point — the utility's export is the plant's import — so a bare "import limit" in a connection offer needs the same scrutiny as a bare sign. In prose, the battery-centric words are actually the safe ones: charge and discharge are unambiguous in a way that positive and negative never are, which is why good test procedures say "discharge at 50 MW" and reserve signed numbers for registers with a stated convention.

The control traps are worse than the accounting ones. Frequency response must answer a falling frequency with more export; reverse the sign and the plant charges into an under-frequency event — an anti-service that performance scoring, and possibly protection, will punish.

And a convention verified in one operating mode proves nothing about the other three: a telemetry mapping checked during discharge testing has said nothing about the charge sign, or either reactive direction, until each is exercised. A four-quadrant machine needs its sign reconciliation witnessed in all four corners, because that is where a cancelled double flip comes back.