"Lean" Design

[jack]

14 minutes ago, Dreadnaught said:

 

@jack, with the benefit of hindsight, would you have changed anything about your TV room to correct for that propensity for summer overheating? Asking as its relevant for the design of my future house.

 

I had considered hiding heat-producing equipment like the games console in a cupboard that had an MVHR extract, but that didn't really work out for the space we had. In any event, it's only the TV and console (which we use a DVD player and streamer) that're on most of the time we're in there, so I'm not sure how much better this would have been.

 

Additional ventilation into that room would likely have helped alittle, as would comfort cooling via the MVHR (something I may well add in future).

 

Even something as simply as a ceiling mounted fan might have been useful - I've considered retrofitting those in the bedrooms, although I'm not sure there's enough head-height for safety upstairs.

 

For us, most likely we'll just get a nice fan to put in a corner for the worst days.

 

4 minutes ago, TerryE said:

In my new house, we rarely close doors because the rooms can overheat when containing a high body count, and start to feel a little stuffy. 

 

I'm constantly telling my kids to keep the door open if they're using this room, but it always ends up closed. Not sure why - they're the ones making all the noise! 

[TerryE]

10 minutes ago, jack said:

Even something as simply as a ceiling mounted fan might have been useful - I've considered retrofitting those in the bedrooms, although I'm not sure there's enough head-height for safety upstairs.

 

How about an in-wall 200mm vent between the snug and the hall?  You can get "silent" fans (e.g. ~25-30dB) which can shift maybe 2m³/min which will be enough to stop the room overheating.

Edited September 5, 2018 by TerryE

[ProDave]

10 minutes ago, jack said:

 

I had considered hiding heat-producing equipment like the games console in a cupboard that had an MVHR extract, but that didn't really work out for the space we had. In any event, it's only the TV and console (which we use a DVD player and streamer) that're on most of the time we're in there, so I'm not sure how much better this would have been.

 

I put satellite tv receivers / dvr's, the hifi, and the printer in the cupboard under the stairs. That is more about hiding "stuff" that you don't need to touch and all operates by IR remote control than anything else. Any "waste" heat from that will go up the stairwell to upstairs (that has no specific heating) so won't be a bad thing.

[jack]

11 minutes ago, TerryE said:

How about an in-wall 200mm vent between the snug and the hall?  You can get "silent" fans (e.g. ~25-30dB) which can shift maybe 2m³/min which will be enough to stop the room overheating.

 

Not a bad idea, thanks, as long as it can be concealed.

[TerryE]

Well maybe not conceal, but a fairly mute in wall plate comparable to an MHVR venturi.  How about something like http://www.justfans.co.uk/catalogues/silent.pdf. I am sure if you shop around, then you should be able to find something with a better acoustic spec.

[jack]

The main issue is that the wall it would go on is a very dark grey/green colour, so I'd need to paint anything like this to conceal it. There's also the issue of noise from the TV room escaping into the hall, although admittedly we don't have a home cinema or anything like that.

[Jeremy Harris]

Just wanted to say thanks, in general, for the idea of a heat transfer fan.

 

It's got me thinking that such a fan fitted to the water-cooled heat exchanger that I've already bought, and arranged to recirculate air and chill it around our first floor, may be far more effective in terms of providing first floor cooling, to augment the slab cooling we already have.  In effect, I could make a semi-hidden fan coil unit, with the fan and cooler fitted in our first floor services room (which is lined with acoustic foam to absorb noise anyway) and a duct running in the void over the second bathroom ceiling and coming out high up in the entrance hall atrium.

 

Recirculating air has to be more effective at cooling the house than drawing in warmer outside air and cooling it.

[pdf27]

3 hours ago, TerryE said:

@pdf27, Assuming that you have a conventional "double-back" layout for your zones and also have access to the Ansys heat flow modelling, then you can easily set up the heat flow model that I did, which is to approximate the slab as a concrete tube the length of a zone run and the radius set by the slab thickness and pipe spacing.  This radial symmetry makes give a 2 spacial and one time dimension model which is computationally solvable over 10s of hours.  OK, the radial symmetry assumption breaks down in reality because the UFH pipes are not in a cylindrical medium but set in a slab concrete with a insulating surface below and a radiant one above. Even so the model still gave an excellent prediction of the time response of the slab and the heat-off impulse response in the both the model and the actual slab shows that the radial component dominates the heat flow during heating: it's an extremely useful model.

Because they're cheap and it isn't my main job, I've only got a fairly slow laptop to run Ansys on so anything normally requiring 10 hours will run out of memory and crash long before it gets to a solution. The model I've run previously - basically a 150mm by 150mm section of floor using symmetry between pipes and a zero heat transfer condition at all edges except the top - would probably be quite amenable to looking at over time, and runs pretty fast. Because it's meshed anyway, there really isn't a lot of benefit to assuming radial symmetry - the software would just create a circular rather than rectangular mesh. One thing this might be useful for is working out what the minimum and maximum pipe spacing is - if you put too much pipe in, then depending on the flow rate of the circulating pump you might have short cycling issues, while with too little you won't be able to provide sufficient heat at a low flow temperature.

 

3 hours ago, TerryE said:

To me what this all underlines is that the slab itself is the biggest heat capacitor in the system, so there is little point in adding complexity of external smoothing using TMVs and buffer tanks.  So long as you are pumping enough heat per day into your slab then a passive class house with warm slab + UFH + cellulosic filler will stay comfortable. 

Absolutely. In any case, the buffer tank in most cases is there to smooth out oscillations/short cycling from the TMV, rather than the tank - and the TMV is there mostly because combustion systems can't regulate water temperatures down low enough not to get chronic temperature overshoot in most normal houses, let alone a properly insulated one. Fit an appropriately sized ASHP with the flow temperature turned all the way down, however, and all those problems go away. The amount of heat it can put in is limited so the reaction time will be quite slow, but because the slab itself has such a high inertia that will happen whatever you do. More importantly, you can't get the air in the house above the flow temperature without breaking the zeroth law - and the T⁴ relationship in radiative heat transfer will keep average slab temperature and perceived air temperature (itself heavily influenced by radiative heat transfer) very close together.

 

3 hours ago, TerryE said:

• No active heat management is required (roughly 6 months / year) because the intrinsic heat excess is enough to keep the house at a comfortable equilibrium and MVHR exchange / bypass gives adequate trim.
• One per day heat adjustment is sufficient (roughly 3-4 months / year). This is my overnight top-up / cool-down.  Again so long as this adjustment gives enough bulk heat-balance, the MVHR exchange / bypass gives adequate trim.
• One per day heat adjustment is insufficient (roughly 2-3 months / year).  Here for the mid winter months a single heating period starts to give a daily heat ripple that is noticeable, so you need multiple heating periods per day.  For the UK climate range IMO you will never need more than 3, but at 4 the ripple will be less than 0.1°C.

That's the one area I dislike your system - in the 3-4 month time section essentially it relies on guessing about right what the weather will be like tomorrow, over-warming the slab slightly and dumping any excess heat through the ventilation system. It clearly works well - unsurprising given the long time constant of the structure and slab plus the ability of the ventilation system to dump heat from the air - but it's fundamentally inelegant to my mind.

I would far rather use a standard air thermostat to call for heat, and the fact that an inverter-driven pump will have a maximum return temperature above which it can't maintain minimum output power to turn it off again. By the time it is allowed to run again, the room temperature should have risen and the system stabilised. The one weakness in this system is that there are two or three interlocking time constants - that of the slab warming up and warming the air, that of the return water warming up and tripping the heat pump, and the anti-short-cycling mode in the pump. Provided that the air warms faster than the water plus short-cycling requirement, it should work very nicely.

 

3 hours ago, TerryE said:

At the moment I use the day-to-day average temperature as a control feedback to compute the total amount of daily heat (and in future cooling) needed to apply.  Because I use a fixed input heater this maps directly to heating time.  If you are on E7 it does make sense to have an asymmetric cycle with a bulk heat overnight, but the tops can be spread through the day.  It sounds like you are going to adopt similar ASHP heating  scheme to the one that I plan to and @jack has done, which is to set the ASHP output temperature at a low setpoint (in my case around 27-28 °C).  You then need to control the mark/space ratio to maintain the overall daily thermal balance.   And the CoP at this set point is excellent.  (Though if you have kids and use lots of HW and want to use your ASHP to preheat this then this would greatly complicate this approach.)

• Kids are currently 2 & 4 plus hate bathtime, so at the moment hot water demand is very low. That'll change when they become teenagers, however - assuming we're still here. I'm not interested in using a buffer tank for preheating however - far simpler just to have a big hot water tank like @Stones used plus a shower heat exchanger.
• Mark/Space ratio would be driven by the thermostat, assuming the time constants match up which I think is probable but I'm not quite sure how to model. Essentially this is staying in your December/January/February mode all year round.
• Given how fast PV prices are dropping, we're almost certain to have a lot of it. That makes using the smart grid functionality in most heat pumps a no brainer - when PV is available, heat water first and when the tank is full turn the thermostat up by a degree. That's unlikely to be a comfort problem, and the long time constant means that no further (paid for) heating will be needed for quite a while afterwards.

 

3 hours ago, TerryE said:

There are a number of strategies here, eg. use a fixed cycle (say 6 hours) and then control the on time or use a fixed on-time and control the off-time to give a variable cycle, but IMO these are all based on macro thermal balance.  This will work well for my house, but if you have "acres of S facing glass" then day-to-day highly variable solar gain will become an issue that you need to factor in.

 

Jeremy and I started out at very different design conclusions from a very similar problem analysis and our solutions are conditioned by historic investment decisions.  Even so we have significantly converged in our approaches.  Jeremy uses a single internal datum, I daily average a couple of DS18B20s measuring room temperature.

My current thinking is to use whatever thermostat is packaged with the heat pump I end up with - I don't have a problem with temperature varying by a degree or so, so don't feel the need to go for super-accurate or low hysteresis temperature measurement.

 

3 hours ago, TerryE said:

To be honest, if mid-winter heating was my only concern then I'd stick with the largely E7 Willis approach.  Yes, the running cost is maybe  £300 p.a. more than using an ASHP,  but I have no complex mechanical systems to maintain and to replace every 10 years or so, so there is no cost benefit case here.  The real issue that makes me plan to introduce an ASHP is the summer cooling one: for about a month a year, I need to dump heat from the house actively to keep a comfortable internal environment, and I can't do this by an additive heat solution.

Agreed. The Passivhaus evangelicals talk about night venting being the solution to everything, but every year for a few days you'll have a period where it's 30°C inside and out for days on end. Treating it as a system an ASHP appears to me to be the most cost-effective way of providing this cooling since it can also provide heating and hot water very cheaply too.

 

3 hours ago, TerryE said:

One last comment.  I've mention the impact of solar gain which can throw a "big spanner in the works".  You can get very sunny days in December and if you have a large south-facing area of glass, this can be a big pulse of kWh into the house.  The other one that causes us fun is visitors.  When our kids+family or others come to visit,  then just the heat and activities around hosting these guests adds environmental control challenges.  For example, 6 people sitting in a room will cause it to start to warm up noticeably!

Yes. As you will have noticed I'm not at all a big fan of lots of glass (driven by the 15 kWh/m²/year requirement, largely). In our particular case the SE elevation at the front faces onto a busy road, while the NW elevation at the back faces onto a rather nice large garden. That means I want the house biased towards the garden, and don't want too much south facing glass. That supports my instinct that the 10 W/m² condition is a more appropriate one anyway in most cases - I get the feeling the use of lots of glass for heating in the shoulder seasons is a hangover from before heat pumps were readily available and people had to use gas or electric resistance for heat, in which case hitting the primary energy targets was almost impossible without a lot of glass for additional heating. It isn't needed any more, but when you start thinking in terms of a particular design solution it's very hard to shift out of it to another one.

[pdf27]

5 hours ago, JSHarris said:

We also have exactly the same issue with visitors.  A while ago we held an open weekend, showing groups of around half a dozen people around every couple of hours.  It got so I could pretty much predict when the MVHR would switch to cooling more (it's slightly audible on full boost daytime cooling) and that would be around 10 to 15 minutes after the visitors arrived.  People are surprisingly effective at warming up the air in the house.

A seated or sleeping person is worth about 100W. That many people is probably a third of the heating you would require on a design low temperature day!

[Jeremy Harris]

17 minutes ago, pdf27 said:

A seated or sleeping person is worth about 100W. That many people is probably a third of the heating you would require on a design low temperature day!

 

Either here, or perhaps on it's predecessor Ebuild, I'm pretty sure we did a bit of a joke calculation of how many dogs/cats were needed to keep a house warm...

 

Our open weekend was late spring, so well outside the heating season, which didn't help.

[Onoff]

4 minutes ago, JSHarris said:

 

Either here, or perhaps on it's predecessor Ebuild, I'm pretty sure we did a bit of a joke calculation of how many dogs/cats were needed to keep a house warm...

 

Our open weekend was late spring, so well outside the heatings season, which didn't help.

 

15 cats in a cage on the wall for the living room wasn't it?

[pdf27]

http://www.c60design.co.uk/many-cats-take-heat-passive-house/

 

"In summary then, cats are expensive and wasteful"

[Tennentslager]

Does a lady going through *the change* not produce X10 the heat of a typical person. Might be an important control matrix consideration, especially if her sister visits too

[vivienz]

4 minutes ago, Tennentslager said:

Does a lady going through *the change* not produce X10 the heat of a typical person. Might be an important control matrix consideration, especially if her sister visits too

 

I have my doubts about this.  Based on the law of conservation of energy, this much heat production should mean that x10 the number of calories are expended in the production of said heat.  Yeah, right.  I wish.  I have yet to see any other physical evidence of faster calorie burn.  Unless another part of me dramatically decreases in temperature at the same time.  Oh, of course, that will be my cold, stony heart if anyone dares suggest that I'm going through the change and seem a little grumpier than usual.

Edited September 5, 2018 by vivienz

[Onoff]

11 minutes ago, Tennentslager said:

Does a lady going through *the change* not produce X10 the heat of a typical person. 

 

Certainly feels like it here! 

[TerryE]

On 05/09/2018 at 21:00, pdf27 said:

Because it's meshed anyway, there really isn't a lot of benefit to assuming radial symmetry - the software would just create a circular rather than rectangular mesh.

 

Not true.  The whole point is that you reformulate the heatflow equation in radial form so that you work in (r,x,t) space instead of (x,y,z,t) space.  The package will still form a mess over (r,x), but because dropped one of the curved dimensions, the effect is to drop the computation load a couple of orders of magnitude.

 

On 05/09/2018 at 21:00, pdf27 said:

One thing this might be useful for is working out what the minimum and maximum pipe spacing is - if you put too much pipe in, then depending on the flow rate of the circulating pump you might have short cycling issues, while with too little you won't be able to provide sufficient heat at a low flow temperature.

 

Not really; just stick with the normal design guidelines which will be either 100mm or 150mm centres.  What you do need to get right is the number of loops (you don't control them separate so they aren't really zones).  But again here most people stick to just under 100mm since the UFH pipe comes on rolls which are a multiple of 100m and doing anything else creates wastage as you really don't want to have any joins in the concrete.  You do get a pretty strong temperature gradient along the pipe which is why IMO you should always do a loopback layout so that the hot and cooler runs alternate but still give an overall uniform heating effect.

 

On 05/09/2018 at 21:00, pdf27 said:

My current thinking is to use whatever thermostat is packaged with the heat pump ... The one weakness in this system is that there are two or three interlocking time constants

 

One for @JSHarris to comment on really, but this will vary from heat pump to heat pump.  AFAIK, what they control is the output water temperature which nothing to do with the environmental control.  Perhaps the main difference between Jeremy and me is that he has gone the off-the-shelf approach and I do all my control in nodeRED on an RPi, so he is constrained by different implementation strategies.   The time constants are on different timescales and the mechanisms sufficiently decoupled that the systems don't "fight".  It's all pretty stable.

 

[pdf27]

3 hours ago, TerryE said:

Not true.  The whole point is that you reformulate the heatflow equation in radial form so that you work in (r,x,t) space instead of (x,y,z,t) space.  The package will still form a mess over (r,x), but because dropped one of the curved dimensions, the effect is to drop the computation load a couple of orders of magnitude.

I'm not that good at fiddling with the mesh, so I'd be constrained to an (r, θ, x, t) space anyway. Absolutely it could be done (and certainly would be if you were doing it yourself), but I'm not sure how well supported it is in commercial packages where it wouldn't be something they would look at a lot.

 

3 hours ago, TerryE said:

Not really; just stick with the normal design guidelines which will be either 100mm or 150mm centres.  What you do need to get right is the number of loops (you don't control them separate so they aren't really zones).  But again here most people stick to just under 100mm since the UFH pipe comes on rolls which are a multiple of 100m and doing anything else creates wastage as you really don't want to have any joins in the concrete.  You do get a pretty strong temperature gradient along the pipe which is why IMO you should always do a loopback layout so that the hot and cooler runs alternate but still give an overall uniform heating effect.

At which point you really come back to not controlling anything - if you're fixed to 100m loops and a given floor area, then the only control you have is from picking the centres. Realistically that's going to be from a fairly limited menu anyway as you want to try to keep the loop length fairly close to 100m in order to ensure the time constant isn't too short.

 

3 hours ago, TerryE said:

One for @JSHarris to comment on really, but this will vary from heat pump to heat pump.  AFAIK, what they control is the output water temperature which nothing to do with the environmental control.  Perhaps the main difference between Jeremy and me is that he has gone the off-the-shelf approach and I do all my control in nodeRED on an RPi, so he is constrained by different implementation strategies.   The time constants are on different timescales and the mechanisms sufficiently decoupled that the systems don't "fight".  It's all pretty stable.

Instinctively the timescales should be fine and nothing should fight, but I'm always a lot happier when I have a realistic model, and only ever comfortable when I have proper, validated test data (not the same as experience or playing around with something for a few days to test it, unfortunately).

[TerryE]

2 minutes ago, pdf27 said:

At which point you really come back to not controlling anything - if you're fixed to 100m loops and a given floor area, then the only control you have is from picking the centres.

 

You are really losing me in this argument.  If you mean design freedom then perhaps I can understand more.   There is only a control (as in control system) issue if this type of arrangement is acts to constrain the control regime.  In our case at the dead of winter we only need to maintain the floor at an average ~2°C warmer than the target environment temperature to maintain overall heat balance.  We could easily heat the floor to 4 or even 6°C above room temp with our pretty standard UFH layout in terms of heat output (if we wanted to live in a dry sauna), so this design constrains nothing.  

 

All of the control parameters are still tunable: the heat input into each circuit, the flow speed, the min/max ontime, the on spacing.   The control algo i about 100 lines of javascript.  This is the easy bit. The main PITA in nodeRED is implementing a simple control panel.

 

And BTW, you don't need to have any weather forecast terms in the control.  We have a stone skin; this plus the cellulose-filled Larson strut frame give one averall decrement delay of well over a day, so we can just plug in actuals.

[Jeremy Harris]

Very good point about not bothering with weather compensation.  This was built in to the default settings for our ASHP (another thing I needed to correct in the installation settings).  My initial assumption, when I was working on using slab temperature control, was that I could derive the required slab temperature by measuring the outside air temperature, on the basis that the slab temperature controlled the heat input power to the house and the outside temperature determined the majority of the heat loss component.  This fell over in practice quite badly, but I did run through many, many iterations of control code before I decided to give up on it.

 

I will admit to be surprised at quite how well a simple (but sensitive, +/- 0.1 deg C hysteresis) room thermostat controls the whole house temperature.  The only slight problem with using a room stat is that I have to make sure that the flow temperature into the slab never exceeds about 25 deg C, ideally it needs to be around 24 deg C.  This was a problem when I was using a thermostatic mixing valve, as 25 deg C is right at the very bottom of their control range, and finding one that even goes as low as 25 deg C isn't easy.  If you do get one that can be physically set as low as this the chances are it won't control the temperature that well, I found.

 

The problem I've found with our house is that we get a higher than desirable temperature overshoot, from heat soak from the UFH pipes and deep in the slab, I think, that can give a 1 deg C or more higher than set temperature for a couple of hours or so after the UFH has turned off, due to the room stat being satisfied.  Keeping the flow temperature down significantly reduces this overshoot, simply because the core of the slab is a bit cooler. 

 

My heating strategy is time-shifted compared to @TerryE's though, in that I don't heat overnight, but heat during the day (usually only for an hour or two in the morning), really to take advantage of any PV generation.  This makes any temperature overshoot more noticeable, especially if there is a bit of additional solar gain later in the day.  My guess is that any overshoot in @TerryE's system is slightly beneficial, by peaking early in the morning, when a bit more heat may be welcome.

 

One reason I'm interested in getting a battery system (when the prices come down) is to use that to run the ASHP at night, or in the early hours, so that the house is slightly warmer from the heating first thing in the morning, and can then absorb a bit of solar and incidental heat gain during the day without getting a little too warm.  I'm not sure if this is me just being a bit fussy, though.  When you have lots of data, and can see small peaks in the room temperature, I have a feeling it tends to focus my attention more than it should!

[joe90]

I am all for simple controls. On another forum a chap with a passive house simply heated the slab to 1 or 2 degrees above his reqd room temp and found is self regulated. If we go E7 or E10 we may heat at night or when prices are low to take advantage of a heavy house. (We have no PV.). As Jeremy states above heat first thing in the morning would be welcome and solar gain may take over during the day (hopefully ?).

Edited September 7, 2018 by joe90

[pdf27]

Apologies, it's been a long week and I have a head full of cotton wool at the moment so I'm not expressing myself clearly.

• As I understand it, heat pump driven systems have constraints that your Willis Heater system does not - notably a minimum flow temperature (typically 25°C) below which they cannot add heat.
• For efficiency (Carnot) reasons, I want to operate any system at this minimum flow temperature for as much time as possible - ideally designing it such that the design 10 W/m² heat load can be provided by a unit running full time at 25°C.
• Most heat pumps have their own circulating pump built in - using this rather than an external pump, combined with the use of the heat pump for flow temperature control will give the simplest plumbing design available with a heat pump.
• As a matter of policy, I want to use absolutely standard OEM control systems only to provide heating and hot water - I appreciate that a homebrew system would probably be quite a bit better, but I simply don't have the time or energy to implement one nor do I want to deal with having to try to fix it remotely if there is a problem while I'm away from home. I'm happy to be more creative with Smart Grid controls, but with those the only impact would be a small increase in the electricity bill rather than any comfort effects, so no time constraints apply.
• Because I have a mechanical engineering background, I want to use parameters like pipe spacing and flow temperature to provide the majority of the control, with the residual being provided by a simple temperature switch (thermostat). When designing a mechanical system you typically aim to minimise the amount of control it requires, and to simplify the control system as far as possible - mostly because control systems are horrifically unreliable compared to purely mechanical systems. While this isn't a genuine problem in most domestic cases, I'm going to be seriously uncomfortable taking any other approach.
• I didn't realise your house time constant exceeds 24 hours - that makes using actual data a lot more feasible. You're still vulnerable to changes in internal heat generation however - your predictions are based on standard internal gains plus predicted external losses based on the temperature seen the day before. OK, it's overwhelmingly likely to be a "too hot" problem which the MVHR can handle, but it's too heavily driven by the requirement for E7 for me to be happy with.

[pdf27]

13 minutes ago, JSHarris said:

Very good point about not bothering with weather compensation.  This was built in to the default settings for our ASHP (another thing I needed to correct in the installation settings).  My initial assumption, when I was working on using slab temperature control, was that I could derive the required slab temperature by measuring the outside air temperature, on the basis that the slab temperature controlled the heat input power to the house and the outside temperature determined the majority of the heat loss component.  This fell over in practice quite badly, but I did run through many, many iterations of control code before I decided to give up on it.

Provided that the rate of change in slab temperature is slow enough, all the time constants are known and you can accurately measure average slab temperature, that should have worked fine. As TerryE notes, however, the building time constants are really long and the temperature you need isn't the current one but the average over the past 18-24 hours or so - if the flow temperatures are relatively high giving a fast responding slab coupled to a control system responding to current outside temperatures, you're going to end up with the thermal equivalent of a pilot-induced oscillation.

 

13 minutes ago, JSHarris said:

I will admit to be surprised at quite how well a simple (but sensitive, +/- 0.1 deg C hysteresis) room thermostat controls the whole house temperature.  The only slight problem with using a room stat is that I have to make sure that the flow temperature into the slab never exceeds about 25 deg C, ideally it needs to be around 24 deg C.  This was a problem when I was using a thermostatic mixing valve, as 25 deg C is right at the very bottom of their control range, and finding one that even goes as low as 25 deg C isn't easy.  If you do get one that can be physically set as low as this the chances are it won't control the temperature that well, I found.

 

The problem I've found with our house is that we get a higher than desirable temperature overshoot, from heat soak from the UFH pipes and deep in the slab, I think, that can give a 1 deg C or more higher than set temperature for a couple of hours or so after the UFH has turned off, due to the room stat being satisfied.  Keeping the flow temperature down significantly reduces this overshoot, simply because the core of the slab is a bit cooler. 

My belief is that the critical change was in reducing the flow temperature - a shift from say 35°C to 25°C flow should cut the overshoot for a thermostat set to 20°C by a factor of 3. Realistically going much lower than 24-25°C will be quite difficult in that the pipes might not be able to deliver enough heat on a very cold day. The next improvement would probably be in reducing pipe spacings a bit - this should help reduce the average difference between concrete and water temperature, and so mean that room temperature rises slightly earlier. That would potentially let the thermostat catch things before as much energy had gone into the slab, reducing overshoot (I think - that one needs a bit more thought).

 

4 minutes ago, joe90 said:

I am all for simple controls. On another forum a chap with a passive house simply heated the slab to 1 or 2 degrees above his reqd room temp and found is self regulated. If we go E7 or E10 we may heat at night or when prices are low to take advantage of a heavy house. (We have no PV.). As Jeremy states above heat first thing in the morning would be welcome and solar gain may take over during the day (hopefully ?).

It almost self-regulates: the heat moving from slab to room is a function of the difference in temperature between the two. Going by @TerryE's experience, at 1 degree temperature difference this is 7W/m². A 50% increase in losses would only cause a shift in temperature of 0.5°C. In a well-insulated house that's mostly acceptable - I'd be fine with it, others (including my wife) probably would not.

The one problem to be aware of (as @JSHarris found) is that measuring the slab temperature directly and then controlling it is quite hard. Because the slab temperature has such a strong influence on the air temperature in a well insulated house, however, you can get the same effect by just using a conventional thermostat. In both cases you're effectively controlling the slab temperature because that's what you're putting heat into - it's just that measuring air temperature is a lot easier and makes it less work to calibrate the comfort level you want as it can be set directly rather than via a couple of heat transfer coefficients.

[Jeremy Harris]

25 minutes ago, pdf27 said:

Provided that the rate of change in slab temperature is slow enough, all the time constants are known and you can accurately measure average slab temperature, that should have worked fine. As TerryE notes, however, the building time constants are really long and the temperature you need isn't the current one but the average over the past 18-24 hours or so - if the flow temperatures are relatively high giving a fast responding slab coupled to a control system responding to current outside temperatures, you're going to end up with the thermal equivalent of a pilot-induced oscillation,

 

The building time constant is what made this approach challenging, for sure.  I started out not appreciating just how long this was, which was a mistake.  Had I measured the building thermal time constant lots of times and determined a reasonably good mean, allowing for the inevitable variations caused by hour-by-hour fluctuations in heat loss and gain, over a week or two in order to try and get a reasonably sensible value, I could have programmed that in and had a half decent chance of getting the slab control system to work

 

25 minutes ago, pdf27 said:

My belief is that the critical change was in reducing the flow temperature - a shift from say 35°C to 25°C flow should cut the overshoot for a thermostat set to 20°C by a factor of 3. Realistically going much lower than 24-25°C will be quite difficult in that the pipes might not be able to deliver enough heat on a very cold day. The next improvement would probably be in reducing pipe spacings a bit - this should help reduce the average difference between concrete and water temperature, and so mean that room temperature rises slightly earlier. That would potentially let the thermostat catch things before as much energy had gone into the slab, reducing overshoot (I think - that one needs a bit more thought).

 

I didn't change the flow temperature for the UFH, as I started out knowing I would only need a low flow temperature, so used an externally sensing two port thermally actuated valve on the flow input to the manifold, with it's sensor deep in a pocket inside the flow manifold itself.  The lowest this valve would turn down to was about 25 deg C, but as mentioned, it wasn't that stable when down that low, and tended to oscillate about by a degree or two, as it was really operating outside, or right at the very edge, of its design parameters.

 

In practice I've found no problem at all with the pipes not being able to deliver enough heat, and our pipes are on a 200mm spacing, tied to the reinforcement fabric in the centre of the slab.  Concrete is a pretty good thermal conductor (dense concrete has about double the thermal conductivity of water) so in practice the heat flow rate to the surface seems to be more than good enough, even with a ∆t of just a couple of degrees.

[jack]

Assuming a single zone (as many of us have, made up of multiple loops), I've long thought that measuring the average return temperature at the manifold might be a useful approach. With an ASHP delivery temp of 25 C, you'd just set it up so that the call for heat is terminated at a particular return temperature. Set a reasonable hysteresis and minimum run/stop times for the ASHP and the temperature should be controlled pretty well.

 

I know a bit about control loops in theory, but not really enough to homebrew anything. However, my Loxone system has various controllers (including a self-learning PID controller, from memory) that should be plug and play. I plan to have a go at this approach this winter. 

[TerryE]

3 hours ago, joe90 said:

If we go E7 or E10 we may heat at night or when prices are low to take advantage of a heavy house. (We have no PV.)

 

I heat from 00:00 to 07:00 (or bring the start time back if this is too much).  This means that the slab is nice and warm under foot when we are pottering around at breakfast. The GFL room temperatures peak at around 11am and then fail by about ½°C by about 3pm, though we don't really notice this as we tend to be quite active in the house.  If we need a top-up we do this from around 2:30 to 4:30 typically so the temperatures again peaks around 7pm and the house is at a comfortable temperature whilst we are sitting around in the evening.   So this cycle is 21 kWh at cheap rate and 6kWh at peak rate, and the overall ripple is a little over ½°C.  

 

If the low point is uncomfortable then the easiest thing to do is to raise the overall temp another °C or so and keep the same profile.

 

3 hours ago, pdf27 said:

if the flow temperatures are relatively high giving a fast responding slab coupled to a control system responding to current outside temperatures

 

Sorry but this is wrong in my experience -- the main area of our slab is 10cm with the UFH at 5cm deep and pumping in 3kW on a top-hat profile still has quite a delay before enough heat reaches the slab surface to start raising a material amount -- probably about 3-4 hours, and that only starts the heating in the room which is why the peak room temp occurs some hours after the heating has stopped.   OK, if you have a 5cm screed with the UFH embedded then the slab will be more responsive, but MBC-style construction were the UFH is fixed to the rebar mesh before poor has long time constants, so you either need to adopt Jeremy's very low (but largely constant) temperature with a simple area stat or my daily computed total and a fixed heating plan.