Posted by: @cathoderay

↑

[from yesterday] I am currently manually (via a script) upping the set LWT every hour (after the auto-adapt resets it) and it seems to be doing the trick. I will post a chart later of what has happened.

Here for completeness sake is the chart:

Another chart showing the Midea heat pump can't quite cope in cold weather. Blue upper chart is outside air temp (OAT), lower green chart is indoor air temp (IAT):

Once the OAT is in the 0-5 decrees C range, the heat pump can't keep up with the heat loss, and the IAT falls. Both cold spells triggered defrost cycles, which don't help. But notice also the slow IAT recovery once the OAT warms up. Recovery (time from end of cold spell to IAT back to where it should be) from the first cold spell took two days, recovery from the second cold spell took even longer, three days. This appears to be a common pattern when the rise in OAT is quite marked. You might imagine that would help the recovery, but paradoxically it appears to do the opposite, because the higher OAT means a much lower LWT and that in turn means less heat delivered to the building, when it actually needs a bit of extra heat, to raise the IAT back to where it should be.

Can you manually and temporarily raise the slope and or offset of your WC curve? On the Vaillant it would be by increasing the WC parameter but on the Midea you might have to lift the low end point or minimum LWT? Or the room temperature target?

it’s obvious that it will take longer to warm up for higher OAT but only once you know!!! Great find.

Posted by: @judith

↑

Can you manually and temporarily raise the slope and or offset of your WC curve?

Yes, though it is very tedious and boring doing it. I have a python script that does it manually over modbus (by manually, I mean I have to set the new right hand end point LWT of the weather compensation to a new value and then run the script). The new LWT then gets reset my current auto-adapt script every hour, and I have to reset the higher LWT again... I'm working on a way to automate all this, in effect building up my own auto-adaption logic.

Posted by: @cathoderay

↑

Posted by: @judith

↑

Can you manually and temporarily raise the slope and or offset of your WC curve?

Yes, though it is very tedious and boring doing it. I have a python script that does it manually over modbus (by manually, I mean I have to set the new right hand end point LWT of the weather compensation to a new value and then run the script). The new LWT then gets reset my current auto-adapt script every hour, and I have to reset the higher LWT again... I'm working on a way to automate all this, in effect building up my own auto-adaption logic.

I'm now a bit confused.  I thought we had more or less proved that the problem lies with the water pump and thus there is insufficient flow of heat from the heat pump.  Have you made any progress in getting the installer to engage with this?

Raising the ft may help because the rads will lose more so DT will increase and thus more energy will be transferred to the rads and house for any given pump speed. This isn't an optimum solution however because you are unnecessarily raising ft this reducing efficiency, better is to get the pump running at the necessary speed.

Or have I missed a development and the above no longer holds?

Posted by: @jamespa

↑

I'm now a bit confused.  I thought we had more or less proved that the problem lies with the water pump and thus there is insufficient flow of heat from the heat pump.  Have you made any progress in getting the installer to engage with this?

No you're not confused! There are actually two problems visible here:

(1) heat pump can't quite cope during periods of low OATs (IAT falls and stays low) - this is down to the capped flow rate, the heat pump can't output enough heat in these conditions, as you say.

(2) the poor recovery after a period of low OATs - here the heat pump can put out enough heat, but it doesn't, because the higher OAT and WCC = much lower LWT than is needed for the recovery (heat pump 'thinks' warm OAT, not a lot of heat needed right now). This can be over-ridden manually but as noted it is tedious and boring having to do it!

My installer was away last week, gets back today and will be in touch with Freedom before too long.

Posted by: @cathoderay

↑

No your not confused! There are actually two problems visible here:

(1) heat pump can't quite cope during periods of low OATs (IAT falls and stays low) - this is down to the capped flow rate, the heat pump can't output enough heat in these conditions, as you say.

(2) the poor recovery after a period of low OATs - here the heat pump can put out enough heat, but it doesn't, because the higher OAT and WCC = much lower LWT than is needed for the recovery (heat pump 'thinks' warm OAT, not a lot of heat needed right now). This can be over-ridden manually but as noted it is tedious and boring having to do it!

Phew! I thought I was going mad.

In terms of #2 rapidly warming OAT is an interesting phenomenon in terms of understanding the thermal dynamics of a house!  Ill give it some thought also.

@jamespa - I'll extract the November data, including the independent OAT sensor, probably later today, and send you a copy. November has been quite a useful month for weather variations, as we have had both very mild and rather cold weather, as well as the rapid change from cold to mild, which confuses the heat pump control logic (it assumes the house is warmer than it actually is).

Posted by: @jamespa

↑

In terms of #2 rapidly warming OAT is an interesting phenomenon in terms of understanding the thermal dynamics of a house!  Ill give it some thought also.

This is definitely an interesting scenario.  What I think happens is this.  The house fabric starts (when it's cold) with a thermal gradient continually falling from inside to out and the loss from the inside is determined by the temperature difference as expected.

This is then disturbed by a rapid rise in OAT which heats the outside initially but not the inside.  This will result in a temperature curve through the fabric which has a minimum within the fabric, so energy flows from the house to this point and from the outside to this point.  The rate at which energy is lost from the inside is initially the same as it was before oat changed, but then reduces as this point is heated up by the energy flowing to it in both directions. 

Eventually a new gradient with a continual fall from inside to outside is established, and during the process the fabric has absorbed energy partially from the outside air and partially from the inside (and thus from your heating system).

This results in a lag in the house response to change in oat and the transfer from the air in the house of more energy than the integral of the  iat oat difference would predict, the difference being 'stored' in the fabric.  When the oat drops this energy is released to the inside and outside so over a sufficiently long period the iat oat difference remains an accurate predictor of heat loss.

With the normal diurnal variation one 'cycle' happens sufficiently fast that the divergence in IAT is small, so we hardly notice.  However when both day and nighttime temperatures shift dramatically, the deficit/surplus becomes noticeable.

There are of course various ways the ashp could adapt.  The simplest WC control loop is proportional, where the heat pump adjusts the ft (and hence emitter output) in proportion to the oat difference from the target iat (or that raised to the power 1.3 for rads). I have read that some pumps gave an integral component as well.  The effect above could very likely be dealt with by adding in a differential component and maybe some heat pump controlkers do that.  We don't know because heat pump manufacturers dont publish their control algorithms.  In principle it would need to be tuned to each individual house (like the basic wc curve) but it may be that it's not so critical in practice to get the multiplier used spot on.

In my 'energy balance' simulations I have an internal and external heat capacity and a lag built in, which goes some way to modelling this effect.  Good luck with tweaking the algorithm 'externally', it might be  easier just to sonething manually and temporarily, as I frequently had to do with my now replaced gas boiler system.

@jamespa - a very interesting analysis. I think the key thing is the duration of the cold spell, if long enough (more than a day?), in effect the house stabilises at a new equilibrium, which is that given by the max heat loss that the heat pump can balance (which is why the capped flow rate matters so much), and then ends up at the IAT that corresponds to that heat loss. There are probably also variable weather effects eg if cold and windy the balance point might be a bit lower. The only way to fix this (using the heat pump, there are obviously other ways) is to get a higher heat pump output. 

The slow recovery I still think is a mismatch between the OAT and the actual lower than expected IAT. The heat pump logic says it is warm outside, so the house needs less energy (only enough to balance the heat loss at that OAT), but in fact the house needs more. My simple auto-adapt script does partially help things, but is deliberately rather gentle in effect (LWT only changes one degree for every degree the actual IAT deviates from the desired IAT, up to a maximum of LWT +/- three degrees). I think it may need to be a bit more aggressive at the right hand end of the WCC, ie when the OAT is higher; and if this is then coupled with a significant actual vs desired IAT differential, then the scenario is very likely the one shown in the chart I posted earlier, and an extra boost is needed. Something like 'if OAT > 12 and (desired IAT - actual IAT > 1) then add more to right hand end LWT'. I used 'more' rather than an actual number because the logic is the addition is made to a baseline, not the current LWT, otherwise it would just keep on going up! I will do some experiments, but will need to wait for another cold spell to see the results.

A thought, FWIW. Where is the OAT sensor located and does your system configuration allow you to set an offset? I've noticed that the OAT sensor (located on a north-facing wall overlooking a black tarmac parking area) for my recently installed Vaillant system is slower to report a temperature rise on a cold mornng than my weather station in the garden. This is beneficial as it results in a bit more heat going into the house on a cold morning. Conversely, in the summer, the OAT sensor was reading high most of the time but I didn't have the heating on so this didn't matter. If you can put an offset in the system configuration then you can get it to think that the weather is a bit colder than it actually is and therefore produce some more heat.

Posted by: @johnr

↑

Where is the OAT sensor located and does your system configuration allow you to set an offset?

This has been the focus of much interest. It is located somewhere in the outside box, but I have yet to discover exactly where. Nonetheless, it is not free-standing, and is definitely subject to local effects caused by the running of the heat pump itself. For these reasons, some have renamed it the AIT (air intake temperature) though I still tend to refer to it as the OAT as that is what people are used to. We also already have rather a lot of TLAs...

The answer to can you set an offset is yes and no. No in that you can't directly key in an offset, but yes in that you can add a 'soft' offset, for example when setting the weather compensation curve end points you can add or subtract an offset as you wish, along the lines you suggest.

In an attempt to calibrate the OAT (AIT) sensor I have used a RC-4 temperature data logger placed in an improvised Stevenson screen on the opposite side of the house. Insofar as the RC-4 data is accurate, the OAT (AIT) is broadly very similar. However, it should be noted that the RC-4 only records hourly values, and these miss the minute by minute variations revealed by the minute by minute values recorded from the OAT (AIT) sensor.  

Edit: you can see the minute by minute variations in the chart I posted at the top of this page