Solving the conundrum of industrial heat

Start with how large and how varied the category is. O’Donnell puts heating and cooling at about half of world final energy use, heat at about 95% of that, and industrial heat alone at roughly a quarter of world energy use and a little more than a quarter of world carbon dioxide emissions, because more than about 80% of heat comes from burning coal, oil or gas. But it is not one thing. Pasteurizing milk wants about 80 degrees Celsius, refining about 200, and cement runs two thirds of its energy near 1,100 degrees and the rest near 1,800, which he calls the highest temperature in use anywhere. Roughly 80% of industrial heat sits below about 350 degrees. Under about 600 degrees the heat usually travels around a plant as steam from a boiler; above that it is often flame in direct contact with the product, as in a rotary cement kiln. And every substitute has to pass a cost test first, because these are commodity businesses where energy can be 40% to 60% of the cost of production.

That test sorts the options quickly. Biogas drops straight into existing equipment with no plant changes, but supply is very limited and it trades at roughly four times the price of natural gas. Hydrogen needs only modest burner and boiler modifications and reaches the highest temperatures, but it takes about two units of electricity to deliver one unit of heat, and O’Donnell is emphatic that this ratio is set by physics and will not improve even as electrolyzers get cheaper. Heat pumps invert that arithmetic, turning one unit of electricity into three units of heat, but only up to about 110 degrees, and they have to run all the time rather than only when power is cheap. Direct electric resistance heating and electric arc furnaces are mature and cheap to install, and operators already pair electric and gas boilers and switch between them on price. The limit there is the calendar: in the United States he sees at most about 30% of the hours in a year with clean power and perhaps 40% with low-cost power, so running on live electricity is interesting at the margin but will not get a plant a 90% or 95% emissions cut unless it sits somewhere like northern Quebec with abundant hydropower.

The reason that ceiling matters is the constraint every option has to satisfy. Large industrial processes run essentially continuously, roughly 8,700 hours a year with one annual shutdown for inspection, because restarting a smelter or a refinery can take weeks. Kann frames this as a purely economic problem; O’Donnell half agrees and half corrects him, saying it is economic and also somewhat technical, precisely because of that ceiling on cheap hours. So the move is to decouple buying electricity from using heat: charge a thermal store when power is cheap or curtailed, then deliver heat steadily. The class is above about 95% efficient from electricity in to heat out, and Kann draws the contrast that makes this surprising: very few batteries reach 95%, but here nothing converts heat back into electricity, which is where the losses normally live. Three costs decide which design wins: how fast it can charge, because the hours of near-zero prices are much shorter than the generating day, maybe four hours of curtailment inside an eight-hour solar day; what the core costs, where O’Donnell’s line is that the box costs a lot more than the rocks; and what it costs to convert stored heat into whatever the process wants. Rondo stores heat in brick, which blast furnaces have done since 1828, and claims a faster way of heating it; others use gravel or sand, molten nitrate salts, melting metals or superheated graphite.

Two conclusions follow. First, despite the long-duration storage framing, O’Donnell sees no business case beyond about 20 hours, because the fallback when renewables are absent is burning fuel in a boiler at about 90% efficiency rather than in a power station at about 50%. At 20 hours of storage, something else supplies roughly 10% to 12% of annual energy, fuel now and hydrogen later, and that is cheap enough to accept. Second, the binding constraint becomes access to cheap clean power at the scale of tens to hundreds of megawatts, which in 2022 means fighting permitting and multi-year interconnection queues, and he argues that some of the resulting price gap is rent extraction by projects already connected. Kann puts a thesis to him: falling renewable costs against rising delivered grid costs should eventually produce industrial grid defection and a migration of heavy industry toward the wind corridor and the solar South. O’Donnell takes half of it. He agrees about the migration, and Rondo builds off-grid projects today because a small company is impatient, but he does not accept defection. For the next 20 years the work is repowering plants where they already stand, and his expectation is that heat batteries become valuable to the grid rather than absent from it: a dispatchable load where the utility is asked not for 250 megawatts of firm service but for 8,000 megawatt-hours today, delivered on whatever schedule suits the system.