MVA headroom

Headroom is a distance, not a remainder

The apparent-power circle itself belongs to the MVA and P-Q capability entries; headroom is the question asked from inside it. Two measurements are in circulation and they are not the same number. Radial headroom is S − √(P² + Q²), the shortest distance from the present operating point to the boundary, and it is the one a converter's current limiter actually sees.

Reactive headroom is the vertical chord: hold P at its scheduled value and ask how far Q can travel before the boundary intercepts, which gives Q(max) = √(S² − P²). The chord is the number that matters commercially, because P is what the market schedules and Q is what the interconnection agreement demands on top of it.

What headroom is not is margin above the rating. Short-time overload — a figure such as 110% for 10 minutes on a PCS datasheet — is a separate, time-limited allowance, and it is not reactive headroom by another name.

Nor is it voltage headroom or energy: the circle governs the power path only, and a plant can be rich in megawatt-hours and completely out of reactive capability at the same instant. Headroom also inherits every condition attached to S. The rating is published as a matrix against AC voltage, ambient and altitude, so the chord computed at reference conditions is the best case, not the design case.

The sign of P does not enter the arithmetic. A converter charging at 0.95 of rating has the same √(S² − P²) available as one discharging at 0.95, because the circle is symmetric about the Q axis. What differs is what sits behind the terminals: DC-bus voltage at low state of charge and battery current limits clip the envelope asymmetrically, and charging is where designs most often come up short.

One further caveat from the MVA entry: vendors warrant different shapes. A semicircle gives you the square-root chord; a rectangle that caps Q regardless of P gives a flat allowance that neither grows at low P nor collapses at high P, and reading one shape's arithmetic off the other's chart is a standard way to promise capability that does not exist.

Why a plant at full megawatts has almost nothing left

Work the chord across the operating range and the shape of the problem appears. At P = 0.80 of rating, Q available is 0.60 of rating. At 0.90 it is 0.44. At 0.95 it is 0.31, at 0.99 it is 0.14, and at 1.00 it is zero. The circle is flat at its top, so the reactive capability does not fade gently as real power approaches the rating — it falls off a cliff in the last few percent. A plant specified so that nameplate MW equals installed MVA has, at full output, no reactive capability whatsoever, and no amount of control tuning will produce any.

The same geometry gives the exchange rate. Differentiating the chord, dQ/dP = −P/Q on the boundary, so each megawatt of real power surrendered near the top of the circle buys a disproportionate quantity of reactive power: about 3 MVAr per MW at 0.95 of rating, and about 7 MVAr per MW at 0.99. Take the canonical 4.4 MVA converter at 690 V.

Scheduled at 4.356 MW — 99% of rating — it has roughly 0.62 MVAr in hand, so backing off just 44 kW of real power has bought 620 kVAr. Back it to 4.18 MW and 1.37 MVAr is available, which is the 0.95 power-factor corner; back it to 3.96 MW and 1.92 MVAr is available, the 0.90 corner. Note that a point sitting on the circle at P = 0.95 S has power factor of exactly 0.95: the ratio and the corner are one statement seen from two directions.

The trade runs the other way at the bottom of the circle. On a semicircular envelope, at P = 0 the whole radius is available as reactive power, which is why STATCOM-mode operation costs nothing in converter hardware — on a circular envelope.

Vendors clip it: one 3,950-class frame publishes ±2,305 kvar at PF = 0 against a 3,620 kW corner at the same conditions, so an idle plant is a full-capability var source only where the published shape says so, and only while the DC bus stays energised and the controls support it. This asymmetry is the reason reactive obligations are cheap to meet overnight and expensive to meet during the evening peak, when the plant wants every megawatt it owns and the grid wants voltage support at the same hour.

How headroom is bought

The first route is to oversize the apparent-power rating. A 0.95 power-factor corner at full contracted MW requires about 1.05 MVA per MW at the point where the obligation is measured (1/0.95 = 1.053), and a 0.90 obligation raises that to about 1.11.

Developers commonly specify PCS blocks with kVA ratings 5 to 15 percent above their contracted MW share, which covers the power-factor corner and leaves something for derating. The cost is capital, and it lands across the whole power path: converters, MV skid transformers, switchgear, cable and the main power transformer are all priced against apparent power, not against megawatts.

The second route is to curtail real power. It costs nothing to install and a great deal to use, because the hours when the system operator most wants voltage support are frequently the hours when energy is worth the most.

Curtailment is also a controls question rather than a hardware one: when a var command and a power schedule cannot both be satisfied, something has to decide which yields, and that P-versus-Q priority lives in the plant controller and the converter firmware. It is set during commissioning and is worth confirming in writing, because a plant that defaults to holding P will breach its reactive obligation and one that defaults to holding Q will spill revenue.

The third route is to stop asking the converters. Shunt capacitor banks, reactors, or a dedicated STATCOM at the point of interconnection carry part of the reactive duty and leave the converter circle free for real power. The devices are not equivalent. A shunt capacitor's output goes as the square of applied voltage, so it delivers least exactly when a depressed grid needs it most; a current-limited converter's capability falls roughly in proportion to voltage, which is a shallower decline but a decline nonetheless.

There is also headroom to recover rather than buy: reactive consumption in the MV transformers and collection cables rises with the square of loading, so design choices upstream of the meter change how much of the converter circle survives the trip to the point of interconnection — the POI-referred accounting belongs to the poi-capability-envelope entry.

What erodes it before you can spend it

Voltage takes the first bite, and it takes it at the worst moment. A converter is current-limited, so at 0.90 pu grid voltage it delivers about 10% less apparent power. The 4.4 MVA unit becomes roughly a 3.96 MVA unit, which means a schedule of 4.18 MW — comfortably inside the circle at nominal voltage — now sits outside the envelope entirely.

Real power has to come down by about 5% before a single MVAr is available, and that happens precisely when the grid code is asking for maximum reactive injection to hold the voltage up. Any headroom calculation performed only at 1.0 pu is answering a question nobody will ask during an event.

Temperature and altitude take the second bite. Continuous MVA falls above roughly 25 to 50 °C ambient depending on the vendor, and above about 1,000 m of altitude, so the chord that clears the reactive corner on a mild day can close on a hot afternoon. Compute √(S² − P²) at the site's worst-case ambient and its actual AC voltage, not at the reference conditions on the first page of the datasheet. State of charge contributes as well: a low DC-bus voltage clips the achievable envelope from a third direction that the nameplate circle does not show at all.

Where the obligation is written decides how much headroom is enough, and the answer is jurisdictional. In the United States, FERC Order 827 requires newly interconnecting non-synchronous generation to provide reactive power across a ±0.95 power-factor range at the high side of the generator substation, available across the full real-power output range — the phrase "full real-power output range" is what turns a power-factor band into a headroom requirement.

IEEE 1547 Category B, the usual assignment for storage on North American distribution systems, requires injection and absorption of at least 44% of nameplate apparent power. IEEE 2800 sets comparable and stricter duties for transmission-connected inverter-based resources. In supply contracts, guarantee the chord rather than the circle: the continuous MVA at stated reference voltage, ambient and altitude, the reactive capability required at maximum charge as well as maximum discharge, and the reference point at which both are measured.