It was 2:40 in the morning when the transfer switch threw. Rain hammering the canopy. The 400 kVA set I’d commissioned three months earlier rumbling to life behind the building. Ten seconds of beautiful, boring normal — then the building’s two chillers woke up at the same time. The lights sagged brown. The generator coughed like a smoker. And everything went dark again.
Not a generator failure. A load failure. The machine was fine. What killed us was a math problem nobody had solved: 400 kVA of generator, roughly 620 kVA of “everything.”
We got the lights back an hour later by doing it the caveman way — main breaker off, generator restarted, then breakers flipped on one at a time while I watched the ammeter like a hawk. It worked. It also took three people and a flashlight. That’s not a plan. That’s an anecdote you tell at conferences.
Every facility I’ve ever worked on has this same argument hiding inside its electrical room. Customers ask me three questions about it, always in the same order.
Honestly? You can. It’s just the most expensive answer on the menu.
A 400 kVA standby set versus an 800 kVA one isn’t twice the iron. But by the time you count the bigger breaker, the heavier cable, the larger day tank, the fatter concrete pad, the extra delivery crane — you’re flirting with double the installed cost. Rough numbers, your mileage will vary, but nobody’s ever told me I was wrong by much.
And here’s the part the quote never mentions: the monster set then spends 99% of its life at 15 to 20 percent load. That’s wet-stacking territory — black slobber weeping from the exhaust joints, cylinder liners glazed like a donut, injectors coking up from combustion temperatures too low to stay clean. Buying everything-size doesn’t just cost more up front. It quietly poisons the machine for the rest of its life.
The professional answer is boring and cheap: size the generator for what’s critical, leave yourself a margin, and shed the rest. If you haven’t done the sizing math yet, start here — it’s the same exercise, just with a red pen.
Partly the law. Partly judgment. Get them mixed up and somebody’s having a bad day.
The law draws the top of the list for you. A fire pump’s generator isn’t allowed to protect itself — NFPA 20 says the fire pump stays on, full stop, and you do not shed it to save the generator. Egress lighting and alarms fall under NEC Article 700 with its 10-second transfer rule; legally required standby loads get 60 seconds under Article 701; everything else lands in Article 702, optional standby, where you have real freedom. I keep a laminated copy of that hierarchy in my truck. Not kidding.
Below the code, it’s judgment — and this is where facilities quietly fail. In a hospital, the operating rooms stay and the parking deck lighting goes. In a data center, the white space stays and the office floors go. In a cold-storage warehouse I worked on, the priority list looked perfect on paper. Then I noticed the office was on the shed list. Fine, nobody needs desks during an outage — except the office held the only working phone line, and the fuel supplier’s number was taped to the wall in it. We moved the office to tier one. Cost: one breaker position. And check the weird ones: I once reviewed a plant that had proudly shed its basement sump pump as “non-essential.” The outage outlasted the rain. You can guess the rest — six inches of water in the electrical room, which is exactly where the transfer switch lives. Non-essential, huh.
That said, the worst priority list is the one that exists only in somebody’s head. I’ve walked into plants where the chief engineer could recite all three tiers from memory — beautiful — and then he’d retired six months earlier and taken the list with him. Write it down. Laminate it. Tape it inside the panel door. Future-you, at 2 a.m. in the rain, will be grateful.
Nobody runs anywhere. If a human has to make the call, you’ve already lost.
The usual setup: current transformers — CTs — clamped around the generator’s output feed the total load into a load-management relay or a module in the genset controller. You program staged thresholds. A typical scheme I install: shed stage one at about 85% of rated kVA sustained for 30 seconds; stage two at 95%, immediately, no waiting. Re-adding is where the craft lives. You don’t slam loads back on the moment current dips. You wait until the load sits below, say, 60% for several minutes — hysteresis, in the jargon — or the set starts hunting: shed, re-add, shed, re-add, like a nervous light switch. I’ve watched a generator hunt itself into a shutdown over exactly that.
Small sites don’t need the fancy module. A simple shed relay wired off the ATS’s generator-running signal drops the water heater and the decorative lighting the moment the set takes over — the same two-wire start logic that fires the generator can carry the shed command. Cheap. Reliable. Done.
Two wiring mistakes I’ve found more than once: first, shedding logic that fires on grid power too — a hot summer afternoon peaks the building load, the relay thinks “overload,” and dumps your air conditioning while utility power is perfectly fine. The shed command must only arm when the generator is actually carrying the load. Second, the shed circuit that was wired, labeled, and never once tested under real load. Monthly exercise runs the set unloaded, so the shed contactor never moves. Then the real outage arrives and the contactor that’s sat for six years is welded shut, or the CT wiring was never landed in the first place. Your weekly test proved nothing — the only test that counts is a load-bank run with somebody watching which breakers actually drop.
Look, next time you’re in your electrical room, ask the electrician one question: when the generator hits 90%, which loads drop first? If he has to think about it — you don’t have a load-shedding plan. You have a hope. And hope is a terrible transfer switch.
Email：[email protected]
Work：Monday to Friday,9:30-18:30
