Any control theory bods here?

[Alan Ambrose]

Hi,

Am I right in thinking it should be pretty easy to figure out whole house U values and even some room level U values by blipping the heating with known heat inputs (e.g. on full power for a few hours) noting the outside temperatures and observing the inside temperature response curves?

 

Alan

[JohnMo]

Should be easy, until you add in ventilation heat loss, which is a big variable. Does the sun poke its head out and give some solar gain...

[ProDave]

But why need to blip it?

 

I had calculated the heat loss of our house using Jeremy's spreadsheet.  When the house was just a bare shell, I just put a small known power electric convector heater on in the middle of the downstairs for a week, and plotted inside temp vs outside temp over a few days (once it had reached equilibrium) and confirmed delta t between inside and outside was what was predicted for that amount of heat input. 

 

I can't see that working with a shot test as a modern well insulated house, the time constants are measured in hours if not days.

[joth]

 

31 minutes ago, Alan Ambrose said:

Hi,

Am I right in thinking it should be pretty easy to figure out whole house U values and even some room level U values by blipping the heating with known heat inputs (e.g. on full power for a few hours) noting the outside temperatures and observing the inside temperature response curves?

 

Alan

No, this is not possible. You have no idea about energy loss through air movement (drafts), energy gain from solar and other  gains (people, cooking, equipment, etc) , and depending on the definition of "blip", how much energy is going into warming up internal heat capacity (dealing with thermal mass) vs lost through the external envelope. Also, knowing the heat input during the blip is pretty tricky as even "boiler on full heat" does not account for efficiency of the boiler on that day or losses in pipework exposed directly to outside, etc 

 

You can get a good empirical model of the overall losses by monitoring long term steady state input needed to maintain a level temperature in the face of varying external temperatures, as @ProDave says above, as @TerryE had done and extensively documented on this site, and pretty much every AI-powered heating controls innovator is trying to automate.

 

 

 

Edited December 23, 2022 by joth

[SteamyTea]

Would be hard to disaggregate what is actually going though the walls/windows/floor/roof/doors and what is ventilation losses.

You would also need to know how much has been absorbed by the air and the building fabric.

 

There was a idea to do air leakage tests with a pulse of air, was that what you were thinking of?

[ADLIan]

Work has been done on this about 12 years ago. Search internet for ‘co heating test’. NHBC Foundation has a summary document which outlines the method and its complexities. Leeds Met University (now Leeds Becket) done testing to better understand the mechanism of heat loss from cavity party walls - as high as 0.7 W/m2K. This lead to party wall thermal bypass being included in Appr Doc L a couple of revisions ago and also the solution of fully filling the cavity with mineral wool to give zero U-value.

[Alan Ambrose]

Ah, interesting discussion as always on BH,

 

Thanks for the pointer to 'co-heating test', I see e.g.:

 

https://www.designingbuildings.co.uk/wiki/Co-heating_test

 

https://www.nhbcfoundation.org/wp-content/uploads/2016/05/NF54-Review-of-co-heating-test-methodologies.pdf

 

I guess I was thinking similar to ProDave - to verify that the actual build was close to the design calcs.

 

Also, to measure the actual numbers for buildings where you don't have the original design calcs (or buildings old enough that there were no/minimal calcs). I have a couple of those and I would be interested to see what the numbers look like.

 

The results could be used as an input to 'non-AI' heat controllers (a 'building model') for better temperature / energy optimisation.

 

Good points about air movement, solar gain, thermal mass etc. I'm thinking that you could actually measure wind and solar inputs and distill down the answers by regression and/or simple AI (actually I see that's what the NHBC researchers did in their analysis). Also a good point about taking out the thermal mass effect by aiming for steady-state.

 

The SAP and Passivehaus calcs are static ones and. in time (maybe) you could build up a dynamic model.

 

Curiously the NHBC paper suggests that in some circumstances they were able to get within 4% of the SAP calcs. I've never thought of SAP as particularly accurate, so that surprised me.

 

Alan

[TerryE]

20 hours ago, ProDave said:

But why need to blip it?

In principle you can work out the transfer function of any control system by putting a impulse or a step change in an otherwise constant input. In practice, doing this can prove almost impossible as the system can be non-linear. 

 

For example, most new builds use dot and dab fixed plasterboard on an inner leaf of blockwork. The blockwork usually leaks air like a sieve and you get convective circulation behind the PB which kills any nominal U-values. This convection is non-linear and also very dependent on external wind, etc.

One way of estimating spot U values is to use a single pixel FLIR meter to measure external wall temperatures. If the ∆t from internal room temp is x °C then you will be losing 7-10 W/m² into the wall depending on the amount of surface convective flow. The ∆t wall surface to external will allow you to estimate external U-values, and more importantly find any thermal breaches.

[ProDave]

18 minutes ago, TerryE said:

In principle you can work out the transfer function of any control system by putting a impulse or a step change in an otherwise constant input. In practice, doing this can prove almost impossible as the system can be non-linear. 

 

A lot of my earlier career was involved with servo motion control of machines, so you would do that to tune them, put a step change to the position demand and watch the resulting actual motion and tune it to get the characteristics you wanted, be that maximum speed, or no overshoot etc.

 

You could do that quite quickly with the position demand stepping backwards and forwards, and adjust it on the fly and pretty soon get the response you wanted.

 

But a house is much slower time constants, hours not milliseconds, so to do the same with the heating system in a house would take days.  More akin to tuning the heating system on a furnace to get to desired temperature as quick as possible without overshooting.

 

 

[Alan Ambrose]

>>> But a house is much slower time constants, hours not milliseconds, so to do the same with the heating system in a house would take days.  More akin to tuning the heating system on a furnace to get to desired temperature as quick as possible without overshooting.

 

OK I'll give it a go at some time and see whether I can figure any results. I have Heatmiser to measure the temperatures / Bright to measure the total energy input / and maybe I can pick up some solar and wind data from somewhere.

 

Alan

[Alan Ambrose]

I just noticed that there's a good way to measure actual overall U factor / energy/degree day described here:

 

https://forum.buildhub.org.uk/blogs/entry/965-update-on-energy-use-based-on-4-years-of-actuals/?tab=comments#comment-6148

 

See TerryE's post where he uses a regression of daily energy used vs. external temperature:

 

Returning to the data, the strongest trend is shown by the external temp vs daily total power use scatter plot, which fits to P = 60 - 2.45T, that is each drop of 1 °C in average daily outside temperature requires an extra 2.45 kWh heating.  I also checked this against my pre-build design calcs which predicted 1.92 kWh, i.e. the as-built house performs about 25-30% worse than as-designed.

[SteamyTea]

40 minutes ago, Alan Ambrose said:

I just noticed that there's a good way to measure actual overall U factor / energy/degree day described here:

It does rely on data collection, which everyone should be doing anyway.

It does amaze me that people will spend £100+ on smart light builds, but are unwilling to spend half that on energy monitoring.

[Alan Ambrose]

>>> It does rely on data collection, which everyone should be doing anyway.

 

True - but if you are electric and/or gas heated and have smart meter(s) this could be, say, bright app daily data power usage download vs. local weather station average daily temperature or heating degree days from https://www.degreedays.net/. 

[TerryE]

@Alan Ambrose, yup but you need to log out to be able to analyse it. 

[Radian]

38 minutes ago, Alan Ambrose said:

>>> It does rely on data collection, which everyone should be doing anyway.

 

True - but if you are electric and/or gas heated and have smart meter(s) this could be, say, bright app daily data power usage download vs. local weather station average daily temperature or heating degree days from https://www.degreedays.net/. 

 

The free use of the Bright app is highly commendable if you don't need real-time info. I guess they convert enough sales on the strength of people wanting to get real-time data. But the API still works even if you don't buy their hardware.

[TerryE]

As I said above I have the equivalent if the Bright app running as a Node RED script and logging to my MySQL nodered database.

[Alan Ambrose]

Ah, can I ask you guys then how you source your usage / heating data?

[SteamyTea]

4 hours ago, Alan Ambrose said:

can I ask you guys then how you source your usage / heating data?

I use an old CurrentCost energy monitor, connected to a Raspberry Pi, this creates a time stamped daily file every time I use a Wh, or an average over 6 seconds, for both temperature and energy usage.

I think use a local WeatherUnderground weather station for external condition, though have just added a basic external temperature logger (RPi and DS18B20) that is also a secure webserver for a bit of fun (found out I am colder than the local weather station, but I am higher up).

I also collect sea surface temperature and National Grid data.

All that is mashed into a spreadsheet and I then average it out hourly with some minimums and maximums.  That hourly data is searchable by date ranges.

 

I know that on the Friday the 6^th January, between Midnight and 1 AM, I used 0.01 kWh, Max Power was 0.1 kWh, 73% of that hour I used no power at all, the room temperature was 19.1°C, my back garden was 9.4°C, but the local weather station was 11.1°C and the sea surface temperature was 11.3°C.  I used 0.00000003% of the energy generated by the National Grid, which at that time was 82% low emissions and 14% high emissions.

 

I have no (expletive deleted)ing life, or friends.

Edited January 16, 2023 by SteamyTea

[Radian]

5 hours ago, Alan Ambrose said:

Ah, can I ask you guys then how you source your usage / heating data?

 

In case you don't already know, if you've got a smart meter, your energy usage (particularly Gas kWh consumed every 30 minutes) can be accessed by signing-up for the Bright App. Bright would love to sell you one of their Zigbee intercept dongles to let you access the traffic between smart meter and IHD but generously make the pooled data available to energy suppliers available to domestic users using the same app.

[SteamyTea]

CurrentCost used to make a logger for most gas meters as well as electric meters.

A million or so of these meters were given out free 15 years ago, most were never used.

 

It is a shame that there is not an easy way to locally log and store Smart meter readings.

Half hourly data is generally as good as you need, but knowing maximum power can be very useful.

 

There is always this though.