Mothballed vs. idle: reading plant status in satellite imagery

A plant can sit dark for a lot of reasons, and from a single pass of imagery they often look the same. A unit down for a six-week planned outage reads almost identically to one that filed its retirement notice with the regional transmission organization fourteen months ago and never got the breaker closed again. The difference matters if you're modeling reserve margins, chasing a basis move, or just trying to explain to a client why a plant that used to clear the stack every winter hasn't shown up in dispatch data lately. EIA-860 retirement filings and utility integrated resource plans eventually tell you, but they lag the ground truth by a reporting cycle or two. Imagery doesn't have to wait on a filing.

The tells that separate a cold unit from a sleeping one

Start with the stack. A unit in a normal planned outage still has steam traced into it, insulation jacketed, and a crew on site running checks, so you'll often catch a faint thermal signature even with the turbine off-line, plus vehicle activity in the yard that a mothballed site simply doesn't have. A plant idled for cause, mechanical failure, a forced derate, an emissions consent decree, usually shows the opposite: cold stack, cold condenser water discharge, and a coal pile that hasn't moved in weeks because nobody's reclaiming off it.

Cooling water is the next tell, and it's a good one because thermal imagery doesn't lie the way a press release can. An operating steam unit pulls makeup water and discharges warm effluent, and on a cooling tower plant you'll see active plume even in shoulder-season weather when ambient conditions wouldn't otherwise produce one. A plant on standby, meaning staffed but not generating, holds its systems in a state where minimal flow continues for corrosion control. A truly mothballed unit, the kind that's been preserved but not formally retired, usually shows none of that: dry discharge channel, tower fans stopped, condenser loop isolated.

Then there's the coal pile, which is where a lot of retirement calls get made wrong by people reading a single image instead of a sequence. A pile that's shrinking week over week, with visible reclaim-tunnel activity and truck or rail traffic at the unloading trestle, belongs to a plant that's burning fuel and will likely show generation in next month's EIA-923. A pile that's static, maybe capped or showing vegetation starting to take hold at the edges, belongs to a unit that hasn't drawn down fuel stock in a long stretch. That's a stronger retirement signal than a single cold stack, because operators sometimes bank fuel ahead of a planned outage, but nobody stockpiles coal for a unit they've already decided to walk away from.

Reading the sequence, not the snapshot

None of these signatures means much as a one-off. A stack can look cold on a mild day for a plant running at low load, and a coal pile can look static for a week because of a rail delay that has nothing to do with retirement. What separates idle from mothballed from quietly retired is the pattern across several weeks: does the stack warm back up when temperatures drop and demand picks up, does the pile ever get touched again, does cooling discharge ever resume. A plant pulled offline for an economic reason, say, gas basis made it cheaper to run the neighboring combined-cycle unit, tends to come back the moment the spread flips. A plant that's gone dark for good just stays flat, outage after outage, season after season.

That's the gap between a snapshot and a nowcast, and it's the whole reason to track a named unit on a weekly cadence instead of pulling an image once and guessing. A weekly read on stack, cooling, and coal-pile activity for a specific plant turns those three signatures into a series you can chart against generation data, instead of a one-time call you have to re-verify every time someone asks if the unit's really done.

If you've got a plant on your watchlist where the retirement filing hasn't caught up to what the yard looks like, Power Plant Activity is built to answer exactly that question on a weekly cycle.