@robs   Regarding the November operation as attached, the DT is increased to 8C to get 9kW at the flow of 16 l/min with both compressors. Mitsubishi obviously set the flow at constant 16 l/min because that is rate for 6kW at 5C DT. Minimum output at 2kW shows a 3:1 ratio so this all stacks up. This blows my theory of manually turning down the flow out of the water.

The flow rate isnt controlled by Mitsubishi, the installers said 16-20 l/min was fine. Experimenting I've found 16 l/min to be a good balance, as you say 5C dT at 6kW and 8C at 8+kW.

The defrost cycles are clear with electrical energy used in the dips. 

For the previous January graph, the dips at the reduced flow level to the right show that there is no electrical energy used in the dip. This suggests that this is normal cycling, not defrost cycles?

The image below shows the two types of defrost side-by-side, two reverse cycle ones on the left and two hot gas ones on the right. The reverse cycle ones do use electricity, about 450w for 3-4 mins. The flow temp dips below there return temp in both (more on that below). The image also has the OAT measured by the heat pump, rather than the Met Office, and so shows the rise in localised OAT when doing a defrost. 

This discussion prompted me to zoom right in to a dual compressor defrost and it looks like it does the first part of a regular reverse cycle defrost before doing a hot gas defrost (I'm assuming that's what the 700-800w electrical use for 6-7mins is based on the compressor specs). So the house does lose some heat but noticeably less as the water pump runs for a shorter period during the defrost cycle. 

The only thing that puzzles me is that your reported room temperatures over the heating season vary in the range 16C-23C. This wouldn’t work for me. Is there any reason for this?

The IAT isn't the Mitsubishi thermostat and is an OEM emonTH. This has been moved between various rooms, the actual IAT is much more constant e.g. 21C plus or minus 0.5C in the lounge (except for solar gain when the temperature can get to 22C or 23C).

For the previous January graph, the dips at the reduced flow level to the right show that there is no electrical energy used in the dip. This suggests that this is normal cycling, not defrost cycles?

The only thing that puzzles me is that your reported room temperatures over the heating season vary in the range 16C-23C. This wouldn’t work for me. Is there any reason for this?

@robs  Thank you. Interesting! I presume that the switch from 20 l/min to 15 l/min was made automatically by the heat pump, and not manually, to effect the switch from dual compressor to single compressor.

The switches to the flow rate were manual and just experiments, the system should work fine at 15-16 l/min if not for the condensate issue that's now solved. The two compressors will run together at the lower flow rate (e.g. 28 Nov), but the second and third quiet modes seem to restrict the compressor frequency (and hence output) to the point that the 2kW compressor isn't used.

So, in summary, for those who want to cover the full heating temperature range efficiently, get one of these. It has all the advantages of an under-sized heat pump covering the top temperature end efficiently, but when the few cold weeks arrive to clobber you, the 2kW compressor kicks in to cover it. I initially thought it would have been the other way round with the 2kW covering the mild end, but this makes perfect sense. The hot gas defrost also covers the worst defrost period below freezing when house heat loss is critical. My old Ecodan will be proud.

A good summary!

I'm going to try to answer all your questions, just maybe not in the order you asked them - I think it'll make more sense this way.

First some additional info... the very cold weather in January highlighted that the handling of condensate could be better - the heat pump is on the northern side of the house (so no radiant heat from the sun) and it was rather humid, so it was damp and stayed damp. To cut a long story short our heat pump was running with a sheet of ice in the bottom of it for the better part of a week, so the performance was rather compromised but it still kept the house warm. Now resolved with more insulation and a trace heater. But back in January I tried a few different settings, one of which was to increase the flow rate and thus increase the mean emitter temperature without increasing the flow temperature. Our heat pump usually runs with a flow rate of 15-16 l/min, this seems to allow the heat pump to run continuously at mild OATs with good COP values while also working okay at lower OATs.

@robs  Thank you for this. I attach a graph from the change you made. It is very interesting that you have reduced the flow to broaden the DT as this is what I have been advocating and some of the cognoscenti poo-poo this. What do you conclude from this? What do you think would happen if you reduced flow further? What is the ideal setting? Do you believe that this is the way to tame down over-sized HP’s?

As above, the change to 15-16 l/min was just going back to the usual settings. The 15-16 l/min is the minimum the water pump can do (unless the strainer is partially blocked then it can go down to 11 l/min!). I think for our heat pump and emitters this is close to ideal, much faster and the dT will narrow too much in mild weather, and much slower and the dT will be too wide in cold weather to maintain mean emitter temperatures at reasonably efficient flow temperatures.

I think flow rate can be part of the mitigations to over sized heat pumps, the control strategy used by the best performing Daikin on heat pump monitor was initially because he had a 9kW (software restricted 16kW) heat pump that was way over sized. The other is software limiting of the compressor through quiet or noise reduction modes. Our heat pump is over sized and it is usually set to the first (of three) quiet mode levels, as a result it generally won't use the 2kW compressor and the ramp-up on start up is gentler.

The other issue is the residual cycling at low OAT temperatures. It broadened out when flow temperature reduced. This persists later in January. This suggests that the flow temperature may still be too high as the AVERAGE flow temperature is a few degrees lower than peak. Each start-up seems to cause a higher power surge before dropping which could affect COP. Have you tried lowering the flow temperature further to eliminate this cycling?

The cycling isn't, they are defrosts. The 6kW compressor has a regular reverse cycle defrost, the 2kW compressor has a hot gas defrost - so it has two defrost mechanisms! If it runs the 2kW compressor I think it does a hot gas defrost but if the 2kW compressor isn't being used then it does a reverse cycle defrost using the 6kW compressor. The defrosts of the left side of your image are hot gas and so the water pump is stopped, while the defrosts of the right side are reverse cycle and the water pump remains on. The hot gas defrosts take longer and use more electricity but don't take heat from the house, in very cold weather not taking heat from the house is a bonus.

This certainly looks like an interesting machine. Does it still have the external long NTC leads for measuring the flow and return temperatures? What happens at mild temperatures - presumably it turns one of the compressors off? Do you know at which point that happens?

Well done. Looks like you made a good choice.

Yeah, it's an interesting machine, quite complex inside with two compressors shoe horned into it! I guess not, there are only two cables - power supply and multi-core power/data to the FTC. 

It only uses the 2kW compressor if it needs it, most of the time it just uses the 6kW compressor as that seems to be able to do over 6kW even in freezing temperatures. 

The other issue is the residual cycling at low OAT temperatures. It broadened out when flow temperature reduced. This persists later in January. This suggests that the flow temperature may still be too high as the AVERAGE flow temperature is a few degrees lower than peak. Each start-up seems to cause a higher power surge before dropping which could affect COP. Have you tried lowering the flow temperature further to eliminate this cycling?

This certainly looks like an interesting machine. Does it still have the external long NTC leads for measuring the flow and return temperatures? What happens at mild temperatures - presumably it turns one of the compressors off? Do you know at which point that happens?

Well done. Looks like you made a good choice.

Without knowing what the modulation curve of any given heat pump/compressor looks like, about all we can do is assume the manufacturer has limited the modulation range so as not to allow efficiency to drop off a cliff. There's little point allowing a compressor to modulate down to 10% or 20% if the efficiency within that range is horrendous. It's probably going to be better to limit it to 25-30% and accept some cycling (substitute numbers as appropriate to your compressor efficiency).

If at mild OATs, maximum efficiency is in the 40% region, cycling may not be that bad, as cycling off and allowing the water to cool to say 25C and then reheating back to a flow temp of 30-35C may result in the heat pump working in exactly that region where it is most efficient with the extra energy used to reheat the water offset by the energy saved during the cycled off period, which may be more efficient than modulating down to 10-20% where efficiency is really poor and running continuously.

That's a good analysis.  Presumably heat pump manufacturers design on the assumption that the heat pump is right sized.  The typical oversized heat pump will probably therefore be working mostly at the bottom end of the modulation curve much of the time, potentially quite significantly compromising COP (notwithstanding the claims of some that oversizing is OK)

That's largely where I am, with a 12kW Samsung unit and on our coldest day last January with produced 96kWh of heat, so an implied heat loss of ~4kW.

I don't have any monitoring other than what's available from the Samsung controller (heat generated) and a MID electric meter recording electricity usage of the ASHP. We do spend a fair amount of time running at the lower end of the curve, and the observed COPs are very much inline with the published literature from Samsung, so we certainly seem to be in the right ballpark and I have no complaints. When it's cold we can achieve extended runs at low flow temps of 30-32C. When it's warmer, 32-33C is the lowest flow temp that will allow the radiators to emit the heat being produced, so we run for an hour or two as required, then switch off for a bit. We allow the room stats to control the on/off call for heat rather than let the heat pump cycle, but that's just how we prefer to manage it.

Our 12kW unit is actually a software-limited 16kW unit, so using the 25% rule of thumb, would expect it to modulate down to around 3-4kW. Minimum input seems to be around 800W, so with a COP of 4 we'd expect ~3.2kW output, depending on OAT and that aligns with the amount of heat being dissipated by our radiators at a minimum flow temp of 32-33C in milder weather.

TBH, because of the way we time shift usage on the Cosy tariff and use overnight switch offs (and commensurate reheats the next day), we probably do not spend that much time running on absolute minimum (25%), and are probably running a little higher up the curve where efficiency is better, although by no means do we ever stretch our unit (other than on the short daily DHW runs).

We are hoping to add a battery which will completely change the way we use the system next winter, allowing more constant low running as the TOU tariff is flattened by the battery storage. We will still set back overnight as the unit still puts out too much heat for comfortable sleeping, even on minimum, so will be working a little harder in the mornings, but nowhere near to the extent we do now, and we will no longer need to switch off during the 4-7pm peak. It will be interesting to see how this affects our monthly SCOPs compared to last winter

If at mild OATs, maximum efficiency is in the 40% region, cycling may not be that bad, as cycling off and allowing the water to cool to say 25C and then reheating back to a flow temp of 30-35C may result in the heat pump working in exactly that region where it is most efficient with the extra energy used to reheat the water offset by the energy saved during the cycled off period, which may be more efficient than modulating down to 10-20% where efficiency is really poor and running continuously.

That's a good analysis.  Presumably heat pump manufacturers design on the assumption that the heat pump is right sized.  The typical oversized heat pump will probably therefore be working mostly at the bottom end of the modulation curve much of the time, potentially quite significantly compromising COP (notwithstanding the claims of some that oversizing is OK)

@heatgeek 

Happy to answer questions about our 8kW Mitsubishi. Yes, it's able to run at ~2.0kW output for extend periods with a dT of ~2C and a COP of 5.5-6.5.

The 20 l/min in Jan was just me experimenting with different settings, basically 20 l/min with a dT of 6C or 15.5 l/min with a dT of 8C. 

Yes, it's inevitably oversized (MCS) but the two compressors have mitigated most of the consequences.

This is confounded by the likelihood that heat pump designers probably try to optimise for the things they are required by regulation to declare, which may or may not be what we experience, and also almost certainly design for the situation where the heat pump is 'right sized' whereas in reality that is probably rare.

My gut feel (which I hope is wrong) is that there is no silver bullet/right answer, because if there were surely most heat pump manufacturers would have drifted towards it.  The introduction by Mitsubishi of a two- compressor machine seems hugely significant to me.  It's not going to be cheap to add a compressor and develop the software to control it, so it feels like an admission that they don't have a way to extend modulation ratio purely by different control algorithms, and need to make a significant hardware change instead.  It will be interesting to see if others follow suit.

I would personally like to see more numerical data on the penalty of cycling.  Other than the examples on the John Cantor video I don't know of any!

Agree with the above. Without knowing what the modulation curve of any given heat pump/compressor looks like, about all we can do is assume the manufacturer has limited the modulation range so as not to allow efficiency to drop off a cliff. There's little point allowing a compressor to modulate down to 10% or 20% if the efficiency within that range is horrendous. It's probably going to be better to limit it to 25-30% and accept some cycling (substitute numbers as appropriate to your compressor efficiency).

If at mild OATs, maximum efficiency is in the 40% region, cycling may not be that bad, as cycling off and allowing the water to cool to say 25C and then reheating back to a flow temp of 30-35C may result in the heat pump working in exactly that region where it is most efficient with the extra energy used to reheat the water offset by the energy saved during the cycled off period, which may be more efficient than modulating down to 10-20% where efficiency is really poor and running continuously.