JamesPa

No need to apologise, thanks to @IanR for pointing out the ambiguity and making it easy to resolve with a diagram! If water is completely mixed in the buffer and pipes well insulated then A-K = B-C = C-H and my conclusion still stands. However it would be nice to know whether it is completely mixed, which could be ascertained by measuring B-C as well as A-K or C-H. If the pump speed is constant then DeltaT across the emitters and HP will anyway reduce as the outside temperature increases, because the load on the radiators is less. My V2 spreadsheet models this (V1 did not). Note that this statement assumes that the HP targets a specific flow temperature not a specific deltaT (the latter would be a bit stupid).. The optimum strategy (ignoring the energy used by the pump) is to minimise deltaT at the emitters at all temperatures, because that in turn increases the average temp at the emitter for any given HP flow temp and this minimises the heat pump flow temp required to maintain a given output. Practically, however, the minimum deltaT for any given ambient and emitters is set by the maximum tolerable pump speed before the system gets too noisy. So basically the 'optimum' algorithm is to set the pump speed to the highest tolerable rate and leave it at that rate. The only way to better this would be to turn the pump speed up above the maximum tolerable rate only when its very cold, if one were prepared to tolerate it in these rare circumstances. probably not worth the effort though.

Honestly no idea. Everyone says that cycling, particularly short cycling, is bad, but I don't know of data to prove it and haven't seen any heat pump specs which allow the penalty to be inferred. What is certain, however, is that on/off at 55C is worse that on/off at a lower temperature (assuming the same average heat output). Operating at a constant 55C (ie without weather compensation) is very bad news

I agree. I thought (incorrectly it turns out) that @ReedRichardswas reporting B and C (or A & D). A & K alone don't really tell us much, as we don't know whether the temp drop is across the buffer, or across the emitters, or split between the two. We need A/B (either) and C/D (either) to understand what the buffer is doing. Don't most heat pumps have an input for a buffer temperature measurement? That figure would help a bit.

Please do.

Thats more good data thanks. 5C delta T across the emitters and 5C delta T across the buffer is consistent with more or less total mixing (ie no stratification) in the buffer. I have seen many state that this is likely, you have given us data which seems to show that it is what happens. Hopefully others will also post data. In general a smaller delta T across the radiators, achieved by cranking up the pump speed, will give higher efficiency. But of course pumps can only go so fast and pipes can only accept a finite flow. Also faster flow = more mixing in the buffer (if you have one = higher flow temp required at the HP). Having said that if mixing is already more or less complete the dominant factor will be first one

Im attaching version 2 of the model which addresses a factor previously omitted and mentioned earlier in this thread, namely that the delta T across the emitters will vary as the load varies, unless the pump speed adjusts to compensate. As I surmised this makes a small difference, but doesn't affect the general trends. The main findings in summary are this: WC makes about 25% difference at 55C, 20 at 50C, 15 at 45C, 10 at 40C and 5 at 35C (all flow temps) A linear approximation to the perfect WC curve degrades the performance by 2% or less The 'Lizzie' adaption to a linear WC curve (whereby the flow temp never falls below 37C) degrades the performance by between 1% (at 55C) and 6% (at 40C). Obviously this adaption makes no sense at 35C A 1C uplift to the WC curve degrades performance by 2-3% The degradations above can more or less be added together The simulation is based on radiators, it can be adjusted for UFH by changing the emitter coef. (to 1 I am told). Hope this is of interest WC Simulationv2.xls

That's excellent data thanks. In summary (I think) your are saying that the output of the buffer on the flow side is about 4-6C below the input to the buffer on the flow side. You don't say what the deltaT is across the emitters, but given you are running at 40+C Im guessing that they are radiators with a design delta T of (say) 7. That's roughly consistent with a tank where there is some stratification, but also quite a lot of mixing. If your delta T is less than this, then there must be more or less complete mixing in the tank. According to my weather comp modelling (which is based on Mitsubishi not LG), the penalty for the 4-6C increase in flow temp needed to compensate for temperature drop across the buffer is in the region 8-16%, depending on some other factors. Of course the buffer is doing some positive things which will tend to increase efficiency, particularly reducing short cycling and helping with defrost. I have not seen any data on how much short cycling really matters, so cant estimate these. Im attaching version 2 of the model which addresses a factor previously omitted and mentioned earlier in this thread, namely that the delta T across the emitters will vary as the load varies, unless the pump speed adjusts to compensate. As I surmised this makes a small difference, but doesn't affect the general trends. The main findings in summary are this: WC makes about 25% difference at 55C, 20 at 50C, 15 at 45C, 10 at 40C and 5 at 35C (all flow temps) A linear approximation to the perfect WC curve degrades the performance by 2% or less The 'Lizzie' adaption to a linear WC curve (whereby the flow temp never falls below 37C) degrades the performance by between 1% (at 55C) and 6% (at 40C). Obviously this adaption makes no sense at 35C A 1C uplift to the WC curve degrades performance by 2-3% The degradations above can more or less be added together The simulation is based on radiators, it can be adjusted for UFH by changing the emitter coef. (to 1 I am told). Hope this is of interest WC Simulationv2.xls

Unless of course it was possible, solely because of the removal of the buffer tank and thus any temp drop across it, to reduce the flow temp from the HP whilst maintaining the same flow temp at the emitters (which is the variable that needs to be fixed to compare like with like on a fair basis), but he doesn't say that. As @JohnMo says, it would be great if those with buffer tanks (2, 3, 4 port) post their own findings. Temp drop and buffer configuration would be good to know.

'As I stated above there is the additional factor to consider, namely the temperature loss in the flow across the buffer tank, which will reduce system efficiency. ' refers to a 3 or 4 port buffer where there is flow/return mixing. In this case the flow temp has got to be higher to compensate for the mixing in the buffer, which reduces the temp to the emitters from that exiting the HP. HPs have lower efficiency at higher flow temps, and this degradation is additional to the standing losses. On further consideration I think my post above underestimates the effect. In a 4 port buffer with little stratification (ie near-perfect mixing) the flow temp has to be raised by an amount equal to the delta T across the emitters to maintain the same emitter output, not half as much stated above. So if delta T across the emitters is 7C then the flow temp (with perfect mixing) will need to be increased by 7C to get the same output at the emitters. That's a 14% penalty on system efficiency according to my WC model. Not insignificant. There is no mixing in a 2 port buffer, and if its plumbed into the return any residual heat loss is less, so in this configuration the penalty due to increased flow temps vanishes.

As I stated above there is the additional factor to consider, namely the temperature loss in the flow across the buffer tank, which will reduce system efficiency. The penalty, based on my WC model, is about 2% per degree C. Assuming that the buffer is inside the thermal envelope and well insulated, with a delta T across the heat emitter of 7C and little stratification in the buffer, that would be something like a 7% penalty. If the delta T is smaller, or there is material stratification, then the impact is less, if the buffer is poorly insulated or outside the thermal envelope then more. Figures and model based on Mitsubishi 11.2kW. A well insulated 2 port buffer (sometimes called a volumiser) in the return would have negligible efficiency impact so far as I can see. That's surely(?) the way to go if the sole reason for a a buffer is to increase the system volume.

I know of two reasons why a buffer tank may reduce system efficiency if installed in a system which can be run without one. The first is heat loss from the tank, which can be mitigated by installing the tank within the thermal envelope. The second is mixing, which increases the required flow temperature. This cannot, to the best of my knowledge, be mitigated but may be more or less significant depending on the design and plumbing of the tank. If someone (or, better still, a few) with a buffer tank can measure the flow temperature either side (ie temperature drop across the buffer), it would be a simple matter to model the latter effect and get a first-cut feel for how significant it is.

Thinking about this a bit more, to first order it can be modelled by just knowing the temperature drop on the flow across the buffer tank, which can easily be measured if you have one (a buffer tank that is). This can then be used to offset the WC curve, and give an estimate of the effect. If someone can measure that in a scenario or two, an estimate of the 'cost' of the buffer tank can quickly be generated. Need some data!

Sorry, I wouldn't know where to start on modelling buffers and I have not seen much data on how much mixing occurs, which would be a basic necessity. I guess worst case is that flow and return mix completely, so they are effectively at the same temp at the buffer op. I think this might mean that the flow temp from the HP will need to increase by about half of the delta T across the emitters. That would be a big hit (which my spreadsheet could easily model once the increase in required flow temp is known).

Just thinking a bit more on the above my 'ideal' WC curve effectively assumes that the pump modulates to achieve a constant delta T across the radiators. Some I believe do, others don't. In the case of a pump which doesn't, the 'ideal' WC curve will be a bit closer to linear (I think). That won't affect the results much, but will (I think) mean that a linear WC curve, which is what most HPs provide for, is less of a compromise.

I don't and cant. I model the flow temperature needed to achieve sufficient heat output from the radiator, on the assumption that the flow temp (and thus the average radiator temperature) is correct at the system design temp. As I mention above the DT across the rad may change as the flow temp changes, and if it does the required flow temp wont change precisely according to the radiator power curve, but in most reasonable systems I would be pretty certain, unless someone can show otherwise, that is a second order effect, so wont materially affect the comparative WC results. Your pump seems to be set deliberately to mess around with the delta T, which obviously is a special case and no system will require a flow temp to vary over the range you quote. I suspect it does this because it is 'set up' for radiators at 70, and UFH at 35, not because the pump manufacturers anticipate that this variation in flow temp to occur in actual operation. A pump could equally well be set to achieve a constant delta T over the actual range of flow temperatures (I think some controllers do exactly that) in which case the 'second order' effect is completely nullified.

Definitely - its instructive to plug in various flow temps. At 35C 'ideal' WC makes about 5% difference. At 55C 'ideal' WC makes 25% difference. My initial calculations were at 45C where 'ideal' WC makes 15% difference.

To calculate the 'ideal' WC curve I assume that the emitter design and flow temperature is correct at the system design temperature and then, at other temperatures, use the radiator output curve to work out what the new flow temperature must be to give the correct output power. This relies on the fact that most radiators have an output power which goes as (Trad-Troom)^1.3. Of course there are various second order effects, eg as the flow temp decreases (assuming the pump speed remains constant) the delta t across the radiators may decrease and thus the average radiator temperature (which determines the output power) will not fall quite as quickly as a calculation based on flow temp would indicate. Also the HP CoP varies a bit with the HP total output which is also not accounted for. The Mitsubishi databook gives some figures for this so could be factored in, but its small so I doubt it will make much difference.

Yes, I had a quick play and it seems WC makes greater difference where the design flow temp is higher and less if its lower. Thats not an obvious result, but not implausible. Will explore it more over next days. Well thats a lot easier!

As I promised I have created a first-cut model of weather compensation, attached. Its based on the Mitsubishi PUZ-WM112 performance data, which is pretty comprehensive. I have modelled the CoP with 5 different W/C schemes and none, estimating the CoP as a function of flow temperature and ambient by linear interpolation. Weather data is average daily temperature for 2022 from the Met Office Central England database. The average daily temperatures are collected into bins 1 degree wide, the number of days in each bin counted, and then the load and consumption for the conditions calculated and multiplied up to get the total annual for that 'bin' according to the w/c scheme selected. The totals for all temperature 'bins' give the total consumption over the year. This is compared with the total demand calculated the same way to get an average CoP, and the total for any particular w/c scheme compared with scheme 'none' (ie no weather compensation) to estimate how much w/c saves. The 'ideal' W/C curve is based on adjusting flow temperature to precisely match demand at ambient temps other than the design temp, using the heat output curve of a typical radiator (which varies as (flow temp-room temp)^n where n=1.3). If someone can tell me how UFH output behaves as a function of flow temp I can model that. Other curves are variants of this, Simply put, it suggests that the various weather compensation schemes I modelled save between 11% and 15% over no compensation at all, which is, frankly, disappointingly small (so I am a little nervous a mistake has been made - but I cant find it). The model takes no account of inefficiencies due to cycling, only the improvement in CoP. This might be significant, if anyone has any figures it might be possible to add them in. I will write it up in more detail over the next few days and - health warning - there may still be errors so, until its been peer reviewed or checked against real results, treat with extreme caution. If anyone wants to critique, contribute suggestions, discussion or actual results of comparing the effects of weather comp, I suggest to do so here so as not to hijack other threads. WC Simulation.xls

Easier for the manufacturer is not necessarily better for the customer! I 100% agree that a central heating system which relies on the cloud for real time operation is not desirable (I would go so far as to say totally unacceptable). In 10, 20 years time will the 'service' still be provided?

Potentially that sounds pretty damn good provided that a. other forms of control (room stats and TRVs) are set sufficiently above the target temperature that they only kick in in extreme circumstances, and (of course) b. that in practice, it achieves the correct temperature. Please let us know if it does The Ecodan is popular on this forum and so understanding it helps. From a purely selfish point of view its also on my shortlist (of 2), and this feature, if it works well, would help swing it. It makes me question what the relationship in the Ecodan is between the WC curve and the autoadapt curve. After a while the former surely becomes irrelevant, or perhaps it always starts with the WC temp then adapts only if it perceives a difference. Autoadapt as described may even be better than a linear WC curve (which is what the Ecodan supports to the best of my knowledge). The theoretical WC curve (at least for radiators) is not quite linear, because output is not a linear function of the temperature diff between emitter and the room. If the autoadapt doesn't lead to wild temperature swings and generally hits the spot, then its very probably doing at least as good a job as WC or better. Im working on the spreadsheet I hinted at earlier to model the effect of WC on COP. Very preliminary results (which use the performance figures from the 11.2kW Ecodan and make no allowance for any loss of efficiency due to cycling) are tending to indicate that more or less any half reasonable WC curve makes about a 20% difference to total fuel consumption, and the differences between the various flavours of half reasonable WC curves is small. I need to refine the algorithm for calculating COP as a function of ambient and flow temperature however, before I can be confident enough to publish so please take this with a pinch of salt until I do.

In a low consumption house definitely, in a leaky house less so. MCS doesn't, to the best of my knowledge, allow for body heat etc, unless someone can correct me.

Yes No, neither the MCS calculation (upon which installers were relying to suggest a system design) nor mine (upon which I am more confident to rely, but modified significantly by the actual measurements) took into account internal heat sources, nor solar gain.

Not with ASHP (yet) but here are my load figures (heat demand at -2C) As calculated by and MCS registered company using the MCS 'method': 14kW As calculated using the same general methodology, but adjusting the U values where I know I have improved the insulation) and told the MCS assessor - but he ignored it):11kW As measured during the extended cold spell in December, when it was -2+/- a degree or so for several days : 7.5kW Says it all to me.

I had always assumed, possibly incorrectly, this was load compensation, adjusting for short term variations due eg to solar gain, rather than targeting a semi permanent change in the WC curve. One problem with control systems for heating is that there are several different time constants involved. I'm not clear whether it's 'good enough' to fix one but not the others. It would be good to understand clearly what these marketing terms mean, and how much they matter. I feel a spreadsheet coming on...