Achievable accuracy of thermostatic radiator valves

[sophiecentaur]

TL;DR

How do TRVs manage to measure room temperature in the proximity of a hot central heating radiator.

I have recently acquired some remotely controlled radiator valves in an attempt to control our domestic heating better. The water enters at the bottom of the radiator and the valve opens proportionally to let just enough hot water through to bring the local room temperature to your chosen setting.

I have had some real struggles with the Zigbee wireless network - which is another problem - but now I mostly have some sort of command of all five valves and also the boiler thermostat. I am impressed how the temperature of the room seems to measure within a fraction of a degree (C) after settling down. How do they do it?

I have read that there are three sensors in each valve body so they have clearly gone to some trouble to achieve the performance. The valves are about £50 each so I don't intend to poke inside one. Have any PF members been here and found out any of the trade secrets - or even some basics of temperature control that doesn't use remote sensing?

Whilst I am at it, most of the network elements are battery operated (2XAA) so they will use minimal power. So my system is very sensitive to placement of units (rads are fixed, of course). The manufacturers clearly have 'issues' with weak signals and local WiFi interference but their recommendations are deliberately vague and there's little solid guidance. If there were an affordable scanner to look at the Zigbee environment, that could help but that sort of test gear would be expensive.

 

[.Scott]

For contrast, in the simplest kind of room thermometer:
There is a small resistor in the thermostat that mimics the heating of the room. So, as the room changes temperature, so does the bimetalic strip in the thermostat. Without this resistor, the thermostat would bring the room to its proper temperature, but wouldn't stop heating until the room warmed up the thermostat. By that time, the room would be too hot. Similarly, the heat would not turn back on until the cold room cooled the thermostat. The end result would be a room that goes through repeated hot/cold cycles. That resistor can be adjusted so that it matches the way the room heats up.

For an electronic thermostat, there is the "PID" loop. This is more sophisticated. It can work operate that variable flow control (not just on/off) and can be controlled by a microprocessor that will "learn" the room and tune the PID parameters accordingly. PID stand for proportional, integral, and derivative - the three factors that control the PID output (and thus the valve position). Each factor has a tuning parameter associated with it.

The "Proportional" factor is the difference between the set point (say 20C) and the actual measured temperature (say 20.1C). In this example, the difference would be -0.1C indicating that the room needs to cool down, so the valve position might be too open. The "P" value in this case would be a positive number - say 0.2. So 0.2*(-0.1) = -0.2 would be the "P" contribution to the valve position.

The "Integral" factor keeps score of how well the temperature has been tracking recently. In our example, the current error is -0.1C. If it the recent average has been -0.5C, then the "P" value is correcting so we may be okay. On the other hand, if the recent average has been +0.5C, the we have over corrected and we need to use "I" to close the valve even more.

The "Derivative" factor keeps track of how the temperature is changing and can be used to dampen the rate of temperature change.

 

[sophiecentaur]

Interesting and informative reply - thanks.

There is a small resistor in the thermostat

I seem to remember it was called an 'accelerating resistor'. The thermostat used to operate at higher frequency and gave a more steady temperature.

For an electronic thermostat, there is the "PID" loop. This is more sophisticated.

I can well appreciate that 'better' feedback will give a steadier temperature / heading / position etc. Many systems use a normal feedback loop which, with the right feedback parameter, will give a critically damped response. The more complicated feedback (PID) can improve things and produce a faster response and steadier output. That, I can understand and have managed to make a voltage regulator and a phase lock loop work.

However , those systems involve measurement of an 'output signal' and comparison with a standard. The Smart TRV can't see the room temperature because it's right next to the radiator's dominant effect.

The "Proportional" factor is the difference between the set point (say 20C) and the actual measured temperature (say 20.1C).

The TRV doesn't have access to the actual room temperature (the value of y(t) in the Wiki block diagram) so it has a harder job. I suspect that the mean 'measured' temperature in the TRV (near the bottom of the rad) would be the same as in the room when the water flow is just right. It won't be a simple as that because the water temperature will have to be 'warm'. People on the Facebook group say they work on a different aiming point for a wanted room temperature.

It impresses me that my posh little datalogging room temperature / humidity sensor seems to agree, after a short delay, with the 'actual' temperature on the TRV. They say there are three sensors in the TRV so it could know the local air temperature and the water temperature. Not trivial. They say you need good air circulation.

You may be able to help me to relate what Wiki says with my added problem.