Thanks, that's excellent context.

Interesting comment! But wouldn't swapping in a nuclear reactor allow for the same cogeneration setup? So it would be more about the cost of the oil they're burning versus the cost of managing a nuclear reactor.

Combustion turbines operate at a much higher temperature (as much as 1300 C) than a nuclear reactor, especially a LWR (about 325 C), so the exhaust temperature can be much higher. Also, a turbine is much cheaper: a simple cycle combustion turbine is maybe $500/kW(e), vs. perhaps 20x that for a NPP.

The reason a combustion turbine can be so hot is that the heat is produced in the working fluid itself, by a chemical reaction. There is no need to conduct that heat from a solid material at higher temperature. All the solid parts that might be heated by the gas can be cooled by actively flowing a cooler fluid through them (air, steam).

But if the main thing they want is low temperature process heat, then the specific higher temperature for the method of generating electricity shouldn't actually matter? NPPs also produce electricity plus lower temperature 'waste' heat.

Unless the waste heat put out by the NPP is too low temperature to be used for process heat? Or the real advantage is that the refineries ramped up their electricity generation (turning waste fractions into electricity), with the process heat becoming a secondary concern?

Direct use of nuclear heat had nothing to do with PURPA. PURPA wouldn't have made it more likely to be considered, since it wasn't putting power onto the grid. If nuclear wasn't being used for industrial heat before that, it wouldn't suddenly have become reasonable afterwards. Instead, PURPA would have made that less likely, since it would have improved the economics of the competing technology of burning fuels for process heat (since a valuable and saleable additional product, power at very high marginal efficiency, could now be produced and sold to the grid.)

There is the question of why nuclear had not and has not penetrated the market for industrial heat. I suspect it has something to do with nuclear not scaling down well, not being suited to industrial applications that may require more flexibility, and concerns about nuclear requirements bleeding over into design and operation of otherwise non-nuclear industrial processes.

Where will process heat come from in the future? Possibly heat pumps. The thermal energy from these (or from resistive heaters, at high temperature) is much cheaper to store, per unit of energy, than electrical energy in batteries, so this would provide a large, cheap source of dispatchable demand to ease renewable integration.