A solar inverter failed at 2 New York sites 5 weeks apart and only one of the fires reached the grass, which is the difference between a 90 minute call and 2 burned acres

The interesting object is not the panels.

It is a grey steel cabinet about the size of a large refrigerator, bolted to a concrete pad at the end of a row, with louvres down one side and cable glands underneath.

Everything happens in that box.

Almost all of a solar site is panels and steel, none of which cares what is growing, while the one component able to hold a sustained arc sits down among the stems.

The array feeds it, the grid takes what comes out, and it is the only part of the installation where real heat is produced as a matter of routine.

It also sits in the weeds.

Panels stand up on racking with air moving underneath, while the cabinet sits at ground level in whatever has grown there since the last mowing.

Why direct current will not let go

Alternating current puts out its own fires.

The voltage crosses zero a hundred and twenty times a second, and every one of those crossings gives an arc a moment with nothing behind it, which is usually enough to extinguish it.

Direct current never crosses zero.

A fault on the panel side is fed by a steady, uninterrupted flow, so an arc that starts there keeps being supplied and will hold itself open until something upstream removes the source.

Duration is what turns a fault into a fire.

An arc that lasts seconds rather than milliseconds builds enough heat to ignite whatever is nearest, and the nearest thing to a cabinet at ground level in July is dry vegetation.

What is in the box and what is around it

An inverter is a conversion machine and conversion is never free.

A few percent of everything the array makes comes out of that cabinet as waste heat, which is why it has louvres, fans and a service clearance written into the drawings.

The panel side arrives at full string voltage.

Hundreds of volts of direct current come down the trench and into the glands underneath, which is where the isolators, connectors and terminals all live in one small space.

Then there is the ground itself.

Vegetation on a solar site is managed on a seasonal cycle rather than continuously, so the fuel around the cabinet is at its tallest and driest in exactly the months when the box works hardest.

The two failures

On a Sunday morning in early June, an array off a state route in Jefferson County lost an inverter.

The local chief put it down to an electrical short. Crews were on scene about 90 minutes and the fire never left the point of origin.

Five weeks later it went differently.

A converter at a community solar farm in Rensselaer County developed an internal fault, an arc followed, and the grass around the pad caught.

That one took two acres.

Crews reached it on all terrain vehicles with a water wagon, and no panel, rack or cable outside the failed unit was damaged.

The protection worked and it burned anyway

The breaker at the second site tripped by itself.

It disconnected the faulted equipment exactly as designed, the system shut down, and the ground fire carried on because the ignition had already happened.

That is the uncomfortable part.

Electrical protection exists to end an electrical event, and once the energy has moved into dry fuel outside the enclosure it is answering a question nobody is asking any more.

The pattern is documented at scale.

A British government review of 80 incidents found 58 caused directly by the photovoltaic system, with direct current isolators the probable cause in 26 of them, and those fires produced 13 injuries and three deaths.

Installation keeps turning up as the root.

A separate review of about fifty cases traced more than 36 percent to installation problems, alongside 18 tied to isolators, 10 to connectors and seven to inverters themselves.

The Rensselaer County fire has already been written up here, 2 acres and the letter the operator now owes before restarting.

What the fuel around the box is doing there

Nobody left that grass by accident.

Ground mounted sites are increasingly seeded with native meadow instead of gravel, and the ecological case for doing it is real and well evidenced.

The Arkansas research is the strongest version of that argument, with badgers, box turtles and monarchs recorded on land that used to be row crop.

A meadow is also a fuel load.

Almost all of a solar site is panels and steel, none of which cares what is growing, while the one component able to hold a sustained arc sits down among the stems.

Permits usually address this.

Clearance distances around electrical enclosures are written into the original approval, and investigators keep finding they were not maintained across the operating years that followed.

New York is about to have far more of these sites, with a million panels going up in a single project.

Neither of these fires hurt anybody.

Both were small, both were handled correctly, and both suggest the maintenance item deciding how a solar inverter failure ends is a mower rather than a relay.