Nuclear

Gas venting from a hydrogen tank outside the fence arced to the lines at a 500 kilovolt switchyard in Pennsylvania, and the reactor beside it tripped off 100 percent power the same minute

By Hugo Rojas · September 6, 2026 · 10:50 AM · 5 min read
Hydrogen tank vent stack beside a high voltage switchyard

Hydrogen tank vent stack

There is a lot of hydrogen at a nuclear station, and none of it is nuclear.

It cools the generator, because a gas one fourteenth the density of air lets a huge rotor spin without fighting its own wind.

It also goes into the water, in small amounts, to suppress corrosion in the steel of the cooling circuit.

Gas leaving the vent stack reached the energized lines and produced an arc flash, bright enough that a firefighter driving toward the plant caught it on video.

So tanks get refilled, and refilling pushes gas out through a stack.

That step is written down and done routinely.

So what makes a routine vent dangerous?

The gas with almost no ignition threshold

Hydrogen burns across an enormous range of mixtures, from roughly 4 percent in air up to about 75 percent.

Almost every other fuel has a narrow window. This one is a wide open door.

The energy needed to light it is smaller still, an order of magnitude below what most fuels need, which is well under what an ordinary spark carries.

Then there is where it goes. Hydrogen is the lightest gas there is, so a vented plume rises fast and drifts with whatever wind is present rather than pooling where it was released.

Put high voltage conductors downwind of that and the geometry stops being academic.

Energized lines maintain an electric field strong enough to ionize air at their surface, and a disturbed or conductive path across a gap is exactly what a flashover looks for.

The flame that results is almost invisible in daylight and the event is over before anyone can react to it.

The yard where the plant meets the grid

The station in question sits in Luzerne County and runs two boiling water reactors for a combined output near 2,400 megawatts.

Everything those machines produce leaves through a switchyard operating at 500 kilovolts.

That yard is not a safety system. It is the commercial connection: transformers, breakers and buswork that tie the generator to the transmission network.

The hydrogen storage sits outside the protected area, which is the fenced security boundary around the reactors themselves.

Two different worlds, a short distance apart.

One is regulated as a nuclear facility. The other is regulated as an industrial gas installation next to a substation.

Nine minutes on the log

At 7:04 in the evening, in late October of 2025, a tank outside the fence was being filled.

Gas leaving the vent stack reached the energized lines and produced an arc flash, bright enough that a firefighter driving toward the plant caught it on video.

The flash struck conductors on the bus feeding Unit 2. The turbine lost its path to the grid, the protection system registered the loss, and control rods went into the core.

Salem Township crews arrived at 7:13. There was no fire to fight: the arc had burned itself out in the yard.

No radiation was released, nobody was hurt, and no safety systems actuated inside the plant.

The unit settled in hot standby with decay heat going to the main condenser through the steam system.

The operator’s line was that the plant responded as designed and the unit went offline without incident.

The sentence the regulator actually wrote

The federal regulator did not close the question. Its wording was that whether the reported fire and the reactor scram are related is being investigated.

That is standard practice, and it is also the honest position, because the causal chain here runs outward from a non nuclear system into the grid connection and back.

The operator has 60 days to investigate and report findings, with any corrective steps.

Context helps and does not excuse: this station has had eight scrams since 2020, against 29 recorded across the country in that year.

The question worth watching is not the reactor’s response, which worked, but the siting: the geometry between a vent stack and a live 500 kilovolt yard, and whether the design ever accounted for wind carrying a plume toward the conductors.

That is a supply chain and layout question of the same family as the component supply and new build programs now being written.

Why one unit tripping matters more than it used to

A 1,200 acre campus next to this plant was sold to a cloud operator for about $650 million.

The supply agreement attached to it runs up toward 1,900 megawatts over its life, which is most of one reactor and then some.

A trip that takes a unit offline for days used to be a wholesale market event spread across millions of customers.

Tying that much contracted load to a two unit site changes what an unplanned outage means for the buyer, and nothing about the arc flash was inside the fence, as state coverage of the shutdown recorded.

The reactor did its job in seconds.

The hydrogen tank was never the reactor’s problem, as local reporting traced it.

It was next door.

Hugo Rojas
Hugo Rojas

Hugo is an engineer with strong technical expertise. Multilingual from an early age, his writing combines technical clarity with a strong interest in science and energy.

Hugo_writer
Hugo Rojas

Hugo is an engineer with strong technical expertise. Multilingual from an early age, his writing combines technical clarity with a strong interest in science and energy.