It's possible that something could be done around ToU tariffs, it's true. But due to the relative size of our 'shiftable' and 'unshiftable' loads, I think the benefits would be fairly marginal. It's probably something I'll try to simulate/model for a while before I consider doing it for real.

Things have settled down now, and I haven't had to make any substantial changes to the system in the last couple of months, so we've got some better numbers to work with:

A £300 saving on just the last four months, with February projected to come in with a saving slightly lower than December. We're pretty happy with that.

We can also compare the old and new system at specific OAT, looking specifically at January:

The average temperature for the month of January this year and last was remarkably similar: 3.6C

Last year did have a bit more variation, though, with the warmest day averaging 9.9C and the coldest day averaging -2.5C, compared to this year with a maximum daily average of 8.6C and a minimum average day of -1.8C. This is all measured with our own outdoor temperature sensor.

The similarity in the temperatures lets us do a rough comparison of how COP has changed from the old system (EAHP water heater + boiler) to the new system (monoblock heat pump + EAHP water heater). The new system had a monthly average COP of 4.28, and if we assume the old system had to do roughly the same amount of work last year, it would have achieved a COP of just 1.63.

What none of this captures, however, is just how much our experience has changed. The old system required my constant attention or else it would either get chilly, or very very expensive. It left us without ample hot water multiple times when the heating was needed overnight, and the temperature in the flat was frequently uncomfortably cool, with the bathroom often being pretty chilly in the morning (down to 17C or lower). We were also unable to control the system to the level of providing additional hot water for guests, or allowing the hot water to turn off when away on holiday (as the heating was reliant on there being hot water).

This winter, the flat has stuck solidly between 21.5C and 23C and I have not needed to tweak anything major at all. The bathroom is never cooler than 20C. The showers are always warm. This has resulted in more frequent and longer showers. I have also been able to connect everything to home assistant, and can easily top up hot water, quickly and cheaply, when we have guests, and can turn everything right down with the touch of a button when we go away for holiday, then have it come back on a day before we return to ensure maximum comfort.

The difference is night and day.

The outdoor unit has also stayed silent. The industrial aircon units on the roof of the Tesco across the road are far louder than our heat pump even when standing within a foot of the unit, and about the only indication of its operation are the occasional clouds of steam which rise from it when it does its defrost cycle.

The only thing missing, really, is an accurate and reliable flow- or heat-meter so that I can confirm the numbers coming from the unit, but given the improvements we can see in our bills, I am not sure it is necessary.

In other news, I've made big strides in improving our heating efficiency. I did this by adding an offset to the heating curve to bring it lower, then adding a minimum flow temperature of 28C. In order to stop the flat overheating at temperatures where the flow should be lower than 28C, I then added a fairly large room influence factor. This should be okay as the flat is well insulated and it takes a long time for temperatures to change significantly. Also, in the event that we do decide to leave all the windows and doors open, there's a maximum flow temperature of 40C configured.

This means that the heating will generally cycle on for a good few hours until the room influence grows enough (at around 1K over the target) that the unit turns off. It seems like if the target flow temperature (before the minimum flow limit is applied) is below 25 it just sets it to 0 and turns off until the room temperature falls enough that the room influence stops blocking the heating.

It remains to be seen how well this works across different parts of the performance curve, but for now we're comfortably around a COP of 4.5 while the outdoor temperature has been around 5.5C.

Sadly, it also seems the flow rate reported by the heat pump via the EMS isn't especially accurate. It looks like it could be off by as much as 25% when doing hot water, so I think any more detailed analysis will require a proper heat- or flow-meter.

Interesting about the Bosch unit being identical to the Buderus; is the Bosch a white-label version engineered by Buderus, or the other way round? The older Bosch units were indeed outsourced…

Have you ever heard of Warnock’s Dilemma? My guess is that it applies here because you’ve said all that needs to be said. A good job comprehensively chronicled.

Well done.

I'm not all that worried about the talking to myself thing, but I know some forums frown on bumping your own posts repeatedly. In this case, though, I'm hopefully documenting something useful for any future (english-speaking!) Bosch customers. 

I’ve been reading to find out about the Bosch and you’re learning fast. Just for info our Vaillant is giving a cop of 5-6 in this mild weather. Our design is for 42C at -2C, but I can’t see the present flow temp easily but cop for Nov is 6.1 (suspiciously high on just the Vaillant readings) 

Good luck with the investigation and improvements!

This is certainly the target. I've largely let it run without too much interference over the last few weeks to get a feel for how it's working. The upshot is that once we're below about 8C ambient temperature, the unit performs fairly well, but it can be a bit unruly and annoyingly inefficient when it falls below its minimum part load. Most of what I've been doing over the last few days has been aimed at curbing that unruliness. I think setting a minimum flow temperature has been a good start. If I put today side-by-side with a reasonably comparable day from earlier in the month, I see much less cycling, and a heating COP of close to 6, which is a big improvement over the cycle-per-hour of the comparison day, which had a heating COP of just 3.75.

The flat is still warmer than it should be, however, so there's still some work to do to find a way to turn it down further without impacting the newly-found performance.

Good luck with the investigation and improvements!

At the risk of looking like I'm talking to myself

…

Have you ever heard of Warnock’s Dilemma? My guess is that it applies here because you’ve said all that needs to be said. A good job comprehensively chronicled.

Well done.

Getting Good Data

Firstly, by purchasing an EMS Bus Wi-fi Gateway, I've been able to increase my logging capabilities by a huge amount. I'm no longer limited to the information Bosch have provided in the app as the EMS bus gives energy consumption and heat production figures to 2 decimal places, broken down into hot water, heating, and cooling, as well as totals. Based on chatter on the German language forum Haus Technik Dialog, where people have compared the numbers provided by the EMS against third-party metering devices, the accuracy seems fairly high. It even tracks the energy lost during a defrost by ticking the counter backwards.

I have been able to send this information into Home Assistant, which gives me the ability to track a large number of data points:

Using the Data

I am also able to send commands to the heatpump, which I have tried to minimise for now; however, I do have an automation which alters the silent mode setting for hot water vs heating. I have also set up an automation which minimises the use of the underfloor circulation pump. Now, rather than operating 100% of the time, it is off while the compressor is running (as the included pump is able to handle circulating the heating easily), but when the compressor goes off (and the included pump goes off), I now have the circulation pump pulse on for 4 minutes every 12 minutes to make sure the flow/return temperature readings are reasonable. This seems to work well.

Looking at the graphs also allowed me to identify some issues with my weather compensation curve. Our flat is very well insulated and has a heat loss of around 2.8-3.5kW at -4C (depending on how many window vents are open and wind direction, etc) which means that the unit will need to cycle at basically any temperature above about 8C:

I wanted to improve this cycling behaviour a bit, however and after watching the system for a while, it became clear that it can't sustain a flow temperature of less than 27 at all, and it is only able to maintain 27 for a while before it overshoots. The lowest flow temperature it seems to be able to manage for an extended period of time is 28C. With this in mind, I've implemented a minimum flow temperature of 28 in the weather curve, alongside a reasonably large room influence so that when it overshoots the target room temperature by a degree or more, the whole system will stop trying to heat until the flat's temperature has dropped sufficiently. In colder weather, it will also tail-off the heating power delivered above the target temperature. It's still early days on this control strategy, but I'm hopeful.

Investigation and reading on the German-language forums has also made a few other things clear.

Quirks and Features

• Compressor on/off is managed through flow-temperature hysteresis. The flow temperature is allowed to overshoot the target by up to 4K before the unit turns off. It will then wait until the measured flow temperature is 2K below the target flow temperature before kicking on again.
• The expected installation has a small 2-port buffer, and the installer should ensure that the heating system dT matches the heat pump's target dT or you'll end up with mixing and distortion. Some owners have used flow/return temperature sensors and Home Assistant to intelligently manage the system's dT, and have seen good results. The instructions do not really talk about this (they just give a flat 4.5K for underfloor and 7.5K for radiators).
• The unit has two (or three, kinda) different defrosting modes. 'Smart Soft Defrost' is used if the air temperature is over 5C and involves running the fan at a higher speed while still employing the compressor. This allows it to defrost while still providing heat. If the unit is in silent mode, it will instead turn off the compressor and keep the fan running, for a similar effect, but without the continuous heating. Once the ambient temperature drops below 5C the unit does a proper defrost by reversing the flow and drawing heat from the heating/hot water systems.
• There may be a bug in the ECO+ hot water mode which forces it to go to maximum power immediately, rather than ramping up through the heating cycle. This is a behaviour I've seen, but I can't be sure if this is because the bug isn't fixed, or if it's due to the peculiarities of our EAHP, which has a coil capable of sucking up a lot of heat.

We don't have OpenEnergyMonitor installed, and the resolution of the data provided by the Bosch app is limited, but I've been doing what I can with it. Here's a graph of our purely heating energy use for the day before I switched on the circulation pump:

A spike of higher-power use, then nothing for the rest of the day (admittedly, the flat barely needs to be heated at these temperatures, but as it's on purely weather compensation, that shouldn't really matter until I can dial in the curve). At first glance, the COP for the day was 4.21 for just heating (lower when hot water is included, but I'm optimising that separately), which isn't terrible, but it could be better. The pump didn't really modulate below the 50% upper bound imposed by the low-noise mode. Additionally, the data resolution is making things look better than they are, as it rounds the hourly idle use of the unit to zero when it's actually pulling an average of 25W constantly. Not much, but when not much heating is happening, that extra 0.5kWh of missing draw drags the true heating COP down to 3.75. For a mild 12°C day, that's not as good as I was hoping for.

After turning on the pump, yesterday was a very similar day in terms of temperatures:

Clearly, I need to adjust my heating curve down more, as the room temperature got up to 23°C, but the behaviour of the pump looks much better. It modulated down to its minimum compressor power of 18% and sat there for much of the day, achieving a true heating COP of 5.1 for the day (including the missing idle draw). The end result is we used 4.4kWh for heating on the 5th, and 4.5kWh for heating on the 8th, but got 6.3kWh more actual heating out of it.

It is heating we didn't really need, though. I suppose the next step is to keep turning the curve down until the temperature equalises. I may also want to turn on room influence, but I want to get the curve to roughly the right area before I do that.