Steam heat control theory: sensors and feedback

ariccio

I've been idly thinking about the theory of steam heat controls. Trying to understand all the ways they can and cannot work.

For many buildings, it seems the new hotness is installing LoRa sensors in some apartments to nudge the boiler in one direction or another. If you could put the sensors in every apartment, I can see how that would be even better… and that thought leads me to think about why the wireless thermometers are necessary in the first plate.

Here's what I'm imagining.

If a two pipe vacuum steam building installs correctly sized orifice plates and properly functioning thermostatic valves, I imagine the information provided by steam consumption, condensate return volume, and pressure in the mains, would be enough to estimate how closely the produced heat is matching the demanded load. I've read of systems like HAWK controls incorporating condensate return volume sensing somehow. What I'm guessing is that if all apartments are at a sufficient temperature, the TRVs shut, and radiators stop producing new condesate, which could be detected in the boiler room. If all radiators are shut, would that slow pressure decay in the mains enough to signal controls?

Then in the extreme case, how well would such a setup regulate temperature in apartments without any temperature feedback and only relying on maintenance of pressure in the mains and measured vs expected condensate return?

I really wish someone would write a basic simulator for steam systems, I learned a lot early on in my life by playing with the Falstad circuit simulator, and something like it would help me answer these questions myself!

Jamie Hall

If all the TRVs were shut — or even a substantial fraction of them — the pressure in the steam mains will rise, not decay, and will do so remarkably rapidly. Far more rapidly than the condensate return flow will show changes. Indeed, pressure in the steam mains (or header for the run of the mill single boiler system!) is the most accurate single indicator of the relationship between power generated by the boiler and power consumed by the radiation (or whatever).

This will be true, regardless of the type of steam heating system involved —- indeed, it is true whether the steam is being used for heat or power generation or whatever.

I have advocated for this approach to steam heating system control for a long time — and if one goes back to the dim dark ages of steam heat one will find that the more sophisticated coal fired systems used steam pressure to control the draught over the fire very sensitively (although with really horrible efficiency and emissions at part load…).

Now if all the apartments in an apartment building have TRVs, and you are dealing with a two pipe system, you have completely satisfactory control using only pressure at in the mains or header. In a one pipe system, you can also have decent, though not as responsive or tight, control, with TR VENTS rather than valves, provided you cycle the steam pressure to zero at reasonable intervals (say every 10 minutes or so) so the vents can exert control.

jumper

Two pipe TRV for local control. Feedback.

Outdoor temperature sensor (or predicted weather) controls steam pressure. FeedForward.

Sylvain

for simulation, one might use matlab

https://www.youtube.com/@MATLAB/search?query=simulink

The problem is to make the mathematical model and it is not simply steady state but also starting from cold, setback, cycling, reacting to outside temperature change, wind, sun, and perturbation like opening a window. And of course all this with thermal inertia, return delay and opening and shuting traps, vents and TRVs, EDR change with temperature (they are given for 70F if i am right). Changing temperature in the neighbor's apartment…

Then modelling the boiler and so on.

Of course one could start simple and refine the model progressively.

Jamie Hall

Ah yes. Mathematical computer models. I played with those, 40 years ago. Lots of fun. Problem is… you have too many unknowns, and too many of your supposed knowns have very wide uncertainties attached to them. At least most of the equations you would be using are linear. The models I was playing with (climate modelling) actually had smaller uncertainties in most of the parameters, but many of the equations were not linear (power or exponential relationships) which made life much more fun…

Or to put it briefly: either lots and lots of experimental measurements and historical data to calibrate with, or … garbage in, garbage out.

ChrisJ

@Jamie Hall

An awful lot has changed in the past 40 years in computers.

Sylvain

maybe online system identification while collecting data?

https://www.youtube.com/watch?v=uLbjeQrQJ3Q

or with AI?

At 72 I will not embark on this

reggi

https://forum.heatinghelp.com/discussion/comment/1800102#Comment_1800102

Still GIGO 👩‍💻🗑️

ChrisJ

https://forum.heatinghelp.com/discussion/comment/1800115#Comment_1800115

That's true.

But computers and software aren't even remotely similar to what they were in 1984 unless you're talking word processing or spreadsheets.

Jamie Hall

https://forum.heatinghelp.com/discussion/comment/1800118#Comment_1800118

This is true. The machines I was working with then — a CDC 6600 driving NOAA's two Crays and CSIRO's Cray were really feeble compared to what we have now. Which, however, doesn't solve the basic problem — solving matrices of non-linear equations (solving matrices of linear equations is a piece of cake. Lots of useful methods, depending on how well behaved they are. That's what Vector Fortran was developed to do). Nor does it solve the problem of relatively poorly defined constants, particularly for the exponents. That said, with the newer computers the models can use much finer grain data sets, and still solve them in a reasonable time frame.

jumper

How much precision do you need to control steam heating?

Analog TRV suffice. You're not going to obtain closer control. The feedforward does not have to be so accurate because the TRV s accomodate inaccuracy of the feedforward.

The pertinent question concerns the savings from resetting steam pressure.

Sylvain

"@Jumper Analog TRV suffice. You're not going to obtain closer control."

As far as temperature only is concerned, I agree.

"The pertinent question concerns the savings from resetting steam pressure."

Indeed, optimisation of steam production, distribution and condensate return.

And "@Jamiehall Now if all the apartments in an apartment building have TRVs, and you are dealing with a two pipe system, you have completely satisfactory control using only pressure at in the mains or header. In a one pipe system, you can also have decent, though not as responsive or tight, control, with TR VENTS rather than valves, provided you cycle the steam pressure to zero at reasonable intervals (say every 10 minutes or so) so the vents can exert control.."

Yes.

One could imagine cutting pressure by zones (not at the same time) in a one pipe system such that not the whole piping needs to be vented.

The quest of a Ariccio was first for a learning simulator and not so much for a regulation system.

KarlW

https://forum.heatinghelp.com/discussion/comment/1800132#Comment_1800132

In my single-family home, what I'd like is feedback from the TRVs driving the boiler cycle. I mostly have it using my European Netatmo system, but not fully. Right now Netatmo and all the other truly smart valves operate in one of two modes (albeit with different names for the mode):

1. "Eco Mode" - the Thermostat controls the boiler - this means that my thermostat room controls the boiler, much as the thermostat does in any house. The smart magic is that other thermostatic valves around the house will close when they reach their desired temperature.
2. "Comfort Mode"—Each thermostatic valve can control the boiler. This means that any [TRV-controlled] radiator can call for heat, and the other TRVs will open or shut accordingly based on their demand.

I so wish that "Comfort Mode" worked better. Unfortunately as implemented, it has the problem of short-cycling my boiler something fierce as my rooms heat and cool at different rates. The boiler will turn on for just one radiator, increasing my overall gas usage. I think the key to steam heating would be a better prediction of when the next call for heat will be, but I'll bet this group has better ideas than me.

Ultimately, Netatmo, Eve, and all the other valve makers don't care as their market is hydronic systems where the boiler cycles based on water temperature, and not (really) on room temperature, so the "Comfort Mode" makes total sense.

Jamie Hall

Steam heat, much as I love it and much as it is the simplest and generally most reliable system, does not lend itself well to individual space control in relatively small buildings — for exactly the reason you note: the various spaces call for heat at essentially random intervals, and the probability of having enough spaces calling at once in warmer conditions is near zero, making the boiler oversized and thus cycling. In much larger systems — say an apartment building — it is quite reasonable to find that at any given time a determinable fraction of the radiation will be calling and therefore the boiler cycles can be longer; in even larger buildings the probability of all the radiation calling at once becomes low, and the boiler can actually be significantly undersized for the installed radiation.

The related problem is that there are very few if any residential sized steam boilers which can modulate, other than by turning on and off.

jumper

In principle there are ways around problem Jamie describes. Resetting pressure. Undersizing steam system with supplementary electric heaters.