Fog. The thick kind, where the far end of the apron disappears completely. I was standing next to an airfield electrician at 1:40 in the morning, both of us staring at a row of runway edge lights, waiting for the transfer test. He held a stopwatch in one hand and a radio in the other. “Fifteen seconds,” he said. “Lights aren’t back in fifteen seconds, the tower starts diverting.” This was a CAT I runway. On a CAT II or III runway, the centerline, touchdown zone, threshold and stop-bar lights get one second. One. A diesel generator needs about ten seconds just to start.
That gap — between what the runway demands and what a generator can physically do — is where airport backup power actually lives. And honestly, most of the airport power plans I’ve read don’t survive contact with it. Below are the mistakes I keep seeing, on paper and on site.
The first mistake is treating “the airport” as one load. It isn’t. An airport is five or six power systems wearing a trench coat, and each one runs on a different clock.
The tower cab and the instrument landing system sit behind UPS units whose batteries carry them for ten, maybe fifteen minutes. The airfield lighting answers to its own rulebook — ICAO Annex 14 spells out switchover times down to the second: 15 seconds for a CAT I runway’s edge lights, but 1 second for the centerline, touchdown zone, threshold, runway end and all stop bars on a CAT II/III runway (ICAO Annex 14, Volume I, Table 8-1). The terminal? Its chillers and escalators can wait. Nobody diverts a flight because the food court went dark.
Yet the quotes I keep seeing propose one big standby set — say 2 MVA — feeding “the airport,” as if a single machine with a single transfer switch can serve loads that need one-second restoration and loads that can wait ten minutes. It can’t. The one-second loads need UPS-backed or DC-fed lighting circuits that ride straight through the generator’s start time; the generator covers the minutes and hours after. If your single-line diagram doesn’t draw that boundary explicitly, you don’t have a design. You have a wish.
Second mistake: sizing for the steady load and forgetting the first thirty seconds after transfer. An airport’s big motors — terminal chillers, baggage handling drives, fuel farm pumps — don’t start politely one at a time. After a transfer, half of them try to restart at once, and a motor pulling six times its running current doesn’t care about your spreadsheet.
I’ve watched a freshly commissioned 1.5 MVA set sag to the point where the ATS dropped it back off, re-transferred, and sagged again — a very expensive metronome — because nobody modeled the locked-rotor kVA of the chiller plant. The fix wasn’t a bigger generator. It was sequencing: staged restart, a soft starter on the worst offender, and a controller programmed to shed the car park ventilation until the critical bus stabilized. Motor starting is what really sizes a generator, not the nameplate total — I wrote about the math here — and airports, with their concentrated motor loads, punish anyone who skips it.
Third mistake, and it’s my favorite because it’s so avoidable: the UPS and the generator are bought from two different vendors, installed by two different contractors, and they meet for the first time during an actual outage.
Here’s how that first meeting usually goes. The grid dies. The UPS carries the tower and the nav aids on battery — ten to fifteen minutes of grace. The generator cranks, the ATS transfers, and the UPS looks at the generator’s output and says no. Frequency wandering, voltage dipping under block load, a waveform with enough harmonic distortion to make a double-conversion UPS stay firmly on battery until the batteries are gone. Then everything goes dark at once, which is exactly the outcome the whole system was built to prevent.
I’ve seen the alarm logs. “Utility fail.” “Generator on line.” Then, four minutes later, “UPS on battery — low.” The generator ran fine. The UPS ran fine. They just never agreed on what “fine” meant. UPS-generator coordination is a whole discipline: the generator needs to be sized generously for the UPS load (those rectifiers are nasty non-linear loads), the governor and AVR need to hold tight frequency and voltage during step loads, and — this is the part people skip — somebody has to run a real transfer test with the actual UPS online and watch what happens. On paper, at 2 a.m., with the runway closed. Not during a storm.
Fourth mistake: fuel autonomy calculated from the brochure. The sales sheet says 0.26 liters per kWh at 75% load, so someone multiplies by the critical load, gets a tidy number of hours, sizes the day tank, and moves on.
Reality is messier. During an outage the generator rarely runs at the tidy 75% — it runs wherever the critical load lands, often 40 to 50%, which is light enough to start wet-stacking the engine over a long event (light-load running is its own slow killer) and far from the brochure’s sweet spot. Meanwhile the fuel in the bulk tank has been sitting there for fourteen months because nobody polishes it, the day tank’s level gauge hasn’t been calibrated since installation, and the regional outage that killed the grid also has every fuel truck in the county stuck in the same traffic jam as everyone else.
Airports that get this right treat fuel as a logistics plan, not a tank size: contracted priority delivery, tested transfer pumps between bulk and day tanks, fuel sampling on a schedule, and an autonomy number with a margin that assumes the fuel truck is late. The airports that get it wrong discover the arithmetic at hour nine of an outage. There’s no good time to discover arithmetic.
Which brings me to the last mistake, the one that makes all the others invisible: the transfer test that never happens.
Monthly no-load exercise? Sure, everybody does that — the generator starts, runs unloaded for twenty minutes, logs a green check. It proves the engine turns over. It proves nothing about the ATS contacts that haven’t carried current in three years, the battery charger that’s been quietly dying, or whether the tower’s UPS will actually accept this particular generator’s output. A green exercise log is not readiness, and nowhere is that truer than at an airport, where a real load-transfer test means closing a runway or a taxiway in the small hours and getting operations, ATC, and the fire service to all agree.
An old airfield electrician once told me generators are the least important part of airport backup power. He has a point — the iron is the easy part. The hard part is the choreography: which load transfers when, what the UPS does during those ten seconds, who calls the fuel truck, and proving it all works before the fog rolls in. His airport runs a full transfer test twice a year, live, with the tower watching. Most airports I’ve visited have never done one. Guess which ones sleep well during storm season.
Next time your flight pushes back on schedule in the middle of a storm, look out past the wing at the runway lights burning steady through the rain. They aren’t on because the grid is reliable. They’re on because somebody sized the sets for the motor starts, made the UPS and the generator shake hands, kept the fuel fresh, and ran the test nobody wanted to run. Most passengers will never think about any of that. That’s the whole job.
Email：[email protected]
Work：Monday to Friday,9:30-18:30
