I think my ASHP is performing poorly and I want to know what questions to ask when the installer visits

[miles (Miles Benson)]

Hi All,

I don’t think my heat pump is performing very well. It was installed by a Heat Geek (Elite, no less) and I have their engineer on site in a couple of weeks. So my goal is to get clued up and know what to ask about.

I have an LG Therma V 12KW unit, installed last spring. The house is reasonably insulated. The heating system is underfloor - in screed downstairs and polystyrene panels beneath floorboards upstairs. The underfloor was pre-existing, so the ASHP installation is as follows:

• The outdoor unit is about 12m from the tanks in the loft.
• There is a 50L 4 port buffer tank first, in the loft.
• I removed the zone valve/mixer/pump on each of the upstairs and downstairs manifolds and instead placed a single pump on the consumer side of the buffer. This helped.
• The underfloor system is zoned, with thermostats in each room, as it has been for the last 12 years. The goal is about 20C downstairs and 17 in the bedrooms.
• There is no useful room temp input to the ASHP. The control panel is not in a very representative location.

I’m not currently using LG’s “AI” setting (the quotes are sarcastic). It seems to just jack the flow temperature up to 45 degrees. Instead I’ve been monitoring and tweaking. Maybe it needs adjusting.

However, I am on an Octopus EV tariff, which means I get cheap electricity at night. So I run an overnight “set forward” where the downstairs is +3 degrees and the heatpump flow is set to 40. This banks some cheap heat in the thermal mass of the floor. Depending on the weather, this can mean that the ASHP doesn’t run for maybe 5-6 hours until mid morning. DHW heating and weekly disinfect cycles are also confined to the overnight period.

Data collection into emoncms is a Shelly clamp for the power and a Modbus adapter on the heat pump for everything else.

The data is here

Emoncms - app view

Don’t fret, this is not my live system. It’s a sandboxed copy which is overwritten from the live system every 10 minutes. Ignore the SCOP section - there’s no kWh data there for now. Oh, and the graphed Room Temp (black line) is showing as zero. That seems to be just a Modbus issue and will come back if I reboot the heat pump. Assume it’s “20” in reality.

I’d welcome any insights from experience.

Thanks
Miles

[Timbones (Tim)]

Some initial thoughts:

• are you able to collect a cumulative kWh feed for electric? That will give better daily chart.
• flow rate of 17 l/min seems reasonable for a 12 kW unit
• the behaviour of the heat pump looks okay overnight, with cycling during the day
• there is some spurious heat showing between cycles, which should not be possible
• overall, the system is performing way below simulated Carnot COP (will explain below)

Looking in more detail: this random section on Dec 31st

Here we see flow temperature pegged at 40° with the return temperature gradually rising and the heatpump throttling back. Carnot Heat calculated at 38.5% of the ideal formula; most systems sit at 40-60%. The end of the cycle is producing significantly below what is expected. Working backwards from Carnot, we should be seeing a dT of at least 2.9° here, not 1.7°.
I’m doubting the accuracy of the temperature probes here.

Return temperature is often seen to be higher than flow temp, which ought not be possible.

This section on Jan 1st is barely managing to beat a COP of 1.0 in conditions that ought to be performing between 3.0 and 4.0.

A good way to diagnose this would be with DHW cycle. An late evening reheat on Dec 11 looks to be heating the tank from 30° to 45°, and put just 2 kWh into the tank over 25 minutes. This would be about right for 120 litres, but would be way short if the tank was any bigger, like 3.5 kWh for 200l…

It would be helpful to see tank temperature at the same time, if you have it. Can stuff it into heatpump_roomT temporarily, or create a second instance of the app.

tl;dr I don’t trust the heating power measurements, and recommend calibrating the sensors.

[Timbones (Tim)]

Some followup questions:

• what is the floor area of your property, and what type?
• what was the estimated heat loss of the property?
• do you have historical energy use for the previous heating system?
  (most useful would be how much was needed on a cold day at -2°C)
• how big is the hot water tank?
• when are disinfect cycles scheduled for?

[miles (Miles Benson)]

Tim,

I’m glad you see the disappointing COP too.

Looking at my installation quotation, the floor area is 169 sq m. Ground floor is screed on top of insulation - I had the builders dig out the floor to the blindings and install 100mm of insulation before the underfloor pipes were laid. House is detached.

As for heat loss, I found a quote from someone I didn’t use in the end and it details the following

DHW is 200 litres, so not the good case you mention.

DHW heating happens from 04:30, not Monday.

Disinfect cycle is in a similar window only on a Monday.

I do have some cumulative kWh data, but not in emoncms. I should have the last few weeks at 30 minute intervals in an sqlite database that I could query - let me know what would be helpful. I did validate the power usage in the graph against the consumed kWh on the Shelly device and against an Octopus Mini accounting for general background consumption. The figures looked believable.

I’ll have a think what I could do with tank temperature. I might find a suitable time to reboot the pump this weekend and put the tank temp in place of room temp for now.

I don’t think I have useful historical gas usage, but I’ll have a root around. Don’t hold your breath. I’d have to find historical temperature data too, to make any sense of it.

[miles (Miles Benson)]

Some historical data.

Turns out my Ovo account is still accessible, and weather underground will show historical data for Manchester Airport (closest place, maybe 5-6 miles away). This is Feb 2024, scaled nicely to align

It’s not 100% comparable - I suspect I would have had most of the upstairs rooms shut and not heated.

More data is available if it would help.

[Timbones (Tim)]

@miles

I’m glad you see the disappointing COP too.

I believe the COP is disappointing because the heat calculation is too low, caused by the temperature sensors being inaccurate. Is there any way you can inspect them physically?

Without a calibrated heat meter, we won’t ever know the true COP of this heatpump. The measurements you have are way outside the Carnot formula, which suggests to me they are wrong.

The heat loss calculation says the property needs 9.5 kW of heat at design temperature (this is likely to be a considerable over-estimate, but let’s keep going), which scales¹ to about 8 kW at 0°C.
[¹ using Heat Loss Coefficient of 475 W/K]

Over this 24 hour period the outside temperature was 0.0°C heatpump averaged 2.7 kW of heat. This a looooong way below the heat loss calculation. A value of 4.8 kW from Carnot @ 40% would be more plausible.

Hard to see a strong correlation between gas consumption and temperature, but we’re looking at maybe 40 kWh on a mild day (10°), up to 70 kWh on a colder day (5°). There’s no data for a freezing day, but the day linked above only managed 65 kWh at 0°C; a little extrapolation² would take us up to 100 kWh, assuming the indoor temperature is the same. If your house is still as warm as last year, then it must have more heat than you’re measuring.
[² using Heat Loss Coefficient of 240 W/K]

[miles (Miles Benson)]

I do actually have some additional flow temperature measures. I’d forgotten about them because the networking is flaky up to the loft. Anyway, I’ve just managed to read them. Basically I have some Dallas temperature sensors tucked under the lagging against the pipes. I’ll verify tomorrow, but it looks strange.

Leaving the pump - 39.2, arriving at the buffer input - 35.8 (loss of 3 degrees in the pipe run)

Leaving the buffer output - 33.9, arriving at the pump 36.5 (umm, gain of 2.5 degrees in the pipe?)

That seems very wrong to me. I shall dig into that tomorrow. Check that the loft sensors are correctly placed etc. I don’t feel that I can open up the heat pump and fiddle with its sensors - that’s the installer’s job while it’s still under warranty.

If one or both of the pump sensors is out of spec, that not only means the calculated data is wrong, but may also mean the pump isn’t being controlled correctly.

I shall also graph some daily consumption for the last few weeks and see if I can find equivalent temperature days from last winter.

In that 24hr period, the house did actually lose heat and the heating system didn’t keep up with the loss. For much of today, the main downstairs rooms were nearer 17C than the target 19-20.

I’ve also realised that I could quickly set the DHW temp into the black line on the graph, so we should be able to see the tank temp against the flow and return temps for the heating at 04:30 tomorrow morning. It’s not showing on the public copy yet, but I’ll fix that tomorrow.

Thanks for your input, it’s much appreciated.

[Timbones (Tim)]

@miles

Basically I have some Dallas temperature sensors tucked under the lagging against the pipes. I’ll verify tomorrow, but it looks strange.

Some useful advice on this page, scroll down to “Liquids”: Temperature sensing with OpenEnergyMonitor – John Cantor Heat Pumps - basically, you want to get very good contact between the sensor and the pipe using thermal paste, aluminium foil or copper wire.

You can check them by putting them all into a cup of warm water (e.g. 40°) and see if they produce the same measurements. If one is slightly out, you can add an offset in software to correct it. If you’re able to get accurate measurements from your own sensors, you can probably use them for calculating heat instead of the ones inside the heatpump.

Remember: it’s only the difference between the probes that matters, not the actual temperature.

@miles

If one or both of the pump sensors is out of spec, that not only means the calculated data is wrong, but may also mean the pump isn’t being controlled correctly.

The heat pump controller doesn’t need accurate temps to perform properly, but calculating heat does. This is why manufacturers don’t care about them; I’ve seen this with other brands too.

[Timbones (Tim)]

@miles

The underfloor system is zoned, with thermostats in each room, as it has been for the last 12 years. The goal is about 20C downstairs and 17 in the bedrooms.

How many zones?
How do they call for heat?

---

Looking at the heatpump behaviour generally, there’s nice steady running when the flow target is set to 40°, and frequent cycling when set to 32°. While within operating limits of the compressor, this is not ideal and will be detrimental to performance.

link

There’s simply not enough (open) floor surface to emit the amount of heat being produced at that low temperature. Power output from the floor is a function of flow temperature - room temperature. It would run better at a higher temperature with all the thermostats open, if at all possible.

Can also see a change in behaviour at 03:30 where the flow temperature starts going up while flow target is still 40° and flow rate unchanged.

link

I suspect this is when one of the thermostats is satisfied and has turned off one of the floors, and the remaining emitter surface isn’t large enough to shed the heat being produced by the heat pump. So flow temp rises, exceeds the limit and then cycles.

---

The ideal configuration for a heatpump is to run continuously at a constant flow temperature, set by weather compensation, exactly matching the heat loss and maintaining a steady room temperature. However, when the heatpump is oversized, which I suspect yours is, this strategy doesn’t work so well.

As Glyn wrote in another thread: “The most cost-effective way to run an oversized heat pump is to treat it more like a gas boiler, running it at higher temperature for shorter periods. I would experiment, setting your minimum flow temperature to about [35C]. A higher flow temperature will allow your emitters to dissipate more heat which allows the heat pump to have longer run times, which counterintuitively should result in higher COP.”

I don’t know what options you have for syncing the zones or having a master thermostat, so something to look into. Maybe LG have a way to read the room temperature remotely?

---

@miles

In that 24hr period, the house did actually lose heat and the heating system didn’t keep up with the loss. For much of today, the main downstairs rooms were nearer 17C than the target 19-20.

This calls for higher flow temperature and less cycling to get heat into the house.

[miles (Miles Benson)]

Some very useful things to think about above. I’m learning a lot.

In other news, the overnight hot water cycle is now visible in the graph. There must be some cacheing going on - I just had to stop and start the docker container.

Emoncms - app view

Basically there’s a rise in temp of 200L of water from 37.8 to 45 using 1.137kWh. My kWh accumulator for the 30 minute period to 05:00 showed 1.106, so very similar (about 8p, which is acceptable in cost terms, but does it tell you anything about the pump performance in general?).

And then there’s the buffer. I currently have temperature sensors on all 4 pipes of the buffer and it’s interesting. I will read up on that link about thermal paste etc, but for now, for example:

pump → 37.1 → buffer → 35.3 → underfloor

pump ← 33.6 ← buffer ← 28.6 ← underfloor

I have read somewhere about 3 vs 4 port buffers - effectively biasing the buffer to only the input or the output. Is this something you are familiar with? And then there’s the question of whether the buffer is strictly necessary. I’m told it’s functioning to both add volume and to give hydraulic separation to the two pumps. I can understand not wanting two pumps to fight each other, but I don’t know about the “adding volume” that’s often mentioned. I’d be very surprised if the water pump in the LG unit could do the entire job on its own with the help of a second pump.

As for the zones, there is a Polypipe manifold and controller upstairs and another downstairs. Each room has a thermostat and that controls one or more solenoid valves on the manifold - on or off. In addition there are flow controls on each zone but they’re fairly crude and hard to adjust. The pump on the consumption side of the buffer is controlled by either of the manifold controllers having 1 or more zone valves open.

I could run the system wide open, but I suspect it would be a nightmare to balance the temperatures in different rooms. And I certainly want the bedrooms cooler than the living spaces downstairs.

Are there any examples of weather compensation curves that you know of? I’d like to see what they look like. I suspect wouldn’t take me more than maybe an hour to implement one to experiment with. I realise that the LG has a built in “AI” mode (sarcastic quotes again), but it’s a black box and I don’t like those, especially when they just ramp the flow temperature to 45. Even when you say “run it like a gas boiler”, 45 seems a stretch?

[miles (Miles Benson)]

Here’s an analysis of gas usage for heating for 2023 against average temperature outside. I’ve not often seen data that correlates quite so nicely, but I suppose you’d hope it would.

So from the fit line, on a day with an average temperature of 0 degrees, I would have expected to use 97kWh of gas.

For that comparison, we would need to look at the 2nd of January this year where the max was 3 the mean -0.1 and the min -5. From my accumulator data, the heatpump used 38.74kWh. Looking at the relative prices of gas and electricity, and the fact that I can use 6 hours of very cheap electricity at night (so net cost is about 21p/kWh), I would need a COP of about 3 to break even.

Does this data suggest anything useful to you?

[Timbones (Tim)]

I’m not personally familiar with buffers, so I recommend this article to get the background info: Low Loss Headers, the Complete Guide! - HeatGeek. Generally want to avoid adding a buffer, but sometimes it is necessary.

Other’s have had success removing a buffer, or converting it to a 3-port buffer, or to a volumiser:

• A Heat Pump Experiment. | Protons for Breakfast
• System Volume
• Fitting heat meter to 4 port buffer set up
• Just a thought on bigger buffer tanks

Alternatively, you can try to balance the flow on both sides of the buffer as zones turn on and off, which could be done by monitoring the temperatures at the 4 ports:

• Making the best of a Buffer
• Optimum CH Flow via 2-Loop Buffer Tank

@miles

The pump on the consumption side of the buffer is controlled by either of the manifold controllers having 1 or more zone valves open.

Does the heatpump turn off when the consumption pump is off? Or does it keep running to maintain the temperature in the buffer tank?

---

There is one other LG Therma V listed on HeatpumpMonitor.org, which may be similar to yours.
12 kW, 200 m², fully insulated, mix of rads and UFH, no buffer. Might be a useful comparison.

[Timbones (Tim)]

@miles

Are there any examples of weather compensation curves that you know of?

The heat required to keep a property warm is close to a linear function of outdoor temperature, for example 50° flow at -2° to 25° flow at 18°. There’s usually a minimum and maximum at either end too. This can be worked out if you know the heat loss of the building and the output power of the UFH, though trial and error can also work.

[Image stolen from https://renewableheatinghub.co.uk/forums/renewable-heating-air-source-heap-pumps-ashps/the-definitive-guide-to-weather-compensation-and-curves-for-air-source-heat-pumps]

[Timbones (Tim)]

@miles

Here’s an analysis of gas usage for heating for 2023 against average temperature outside.

That’s a nice plot, fits the data nicely. There will be some discrepencies due to cooking and DHW, though one can argue that both of those contribute to heating the house. The intersection at 17°C is also as expected, as heating isn’t generally needed above that temperature. Heat from other appliances and occupants make up the difference.

Divide by 24 hours to get average kW heating power required, so about 5kW at -3°C.
The trend line becomes -0.240x + 4, so you can work out the power required at any temperature. Heat Loss Coefficient comes out at 240 W/K.

@miles

For that comparison, we would need to look at the 2nd of January this year where the max was 3 the mean -0.1 and the min -5. From my accumulator data, the heatpump used 38.74kWh.

At a COP of 2.8, the heatpump would have produced 108 kWh of heat, which is not far off the previous gas usage. More than was estimated by the mass flow rate heat transfer formula.

Looking at the other LG Therma V linked above the minimum power draw is about 900W and the minimum (stable) heat output is 3kW. This means the heatpump will be forced to cycle above 5°C outside temperature. Commonly seen with oversized heatpumps.

The LG in Sheffield has similar numbers: https://heatpumpmonitor.org/heatloss?id=110

Oh, while we’re looking at this system, here’s what heating up 300l tank to 52°C looks like:

[miles (Miles Benson)]

Oh, while we’re looking at this system, here’s what heating up 300l tank to 52°C looks like:

I picked a different example, looking for a lower outside temperature. We have to assume this is water heating, but it seems reasonable

Emoncms - app view

COP of 1.73.

So looking at my most recent water heating,

Emoncms - app view - my water heating

1.6C outside, 200L from 38.3C to 45C used 1.128kWh and shows heat output to be 1.628kWh.

Using an online calculator, raising that quantity of water by that temperature requires 1.56kWh. So the heat output of my system looks similar to the theoretical calculation. I strongly suspect you’d need slightly more to account for inefficiencies, but we’ll run with what we’ve got.

The question is, given an outside temp of 1.6C at the time, is a COP of 1.44 reasonable? The other LG you pointed me at was operating at 0C and achieved a COP of 1.73. Unfortunately we can’t see what the starting and ending temperatures of the tank were to compare actual with theoretical. If that COP difference could simply be attributed to sensor problems, then that’s fine. The COP will improve when they improve. I’m just trying to find a way to judge whether anything else is going on. Maybe the truth is that 1.73 is reasonable and I’m achieving that (and hence more output energy, which is lost in transfer inefficiencies), but just not calculating it from the sensor data.

On another topic, the pump that I’m using for the consumer side of the system has various options. I’m currently using the “constant pressure curve” because it says that’s designed for underfloor systems. However, there are 2 variants. Is it better to use a higher or lower water speed through an underfloor system? How would that impact system efficiency?

[dMb (davidMbrooke)]

@miles

Is it better to use a higher or lower water speed through an underfloor system? How would that impact system efficiency?

Your best plan is to try to match the flowrate from the ‘primary’ circulation pump (although the flowrate will change a bit as the solenoid valves open and close). You’ll get best efficiency if your UFH is seeing all the flow from the heat pump, since then you might be able to dial-down the flow temperature to improve efficiency.

@miles

pump → 37.1 → buffer → 35.3 → underfloor

pump ← 33.6 ← buffer ← 28.6 ← underfloor

Those temperature readings suggest the ‘consumer’ circulation pump flowrate is higher than the primary circulation pump. The way I’m reading the numbers is that the 35.3 flow to the UFH is recirculating some of the 28.6 return from the UFH, hence ‘diluting’ the 37.1 flow from the heat pump. I reckon that means you’d be better off with a lower flowrate on the UFH pump.

Ideally you want the temperature readings to match on both sides.

[Timbones (Tim)]

@dMb

The way I’m reading the numbers is that the 35.3 flow to the UFH is recirculating some of the 28.6 return from the UFH, hence ‘diluting’ the 37.1 flow from the heat pump.

I read this differently as the return to the heat pump is higher than that from UFH, suggesting some of the incoming flow is going back out again, i.e. primary pump flow is higher than the consumer.

…though that’s assuming that these temperatures are in any way accurate. Might be useful to see these temperatures plotted over time.

[Timbones (Tim)]

@miles

Using an online calculator, raising that quantity of water by that temperature requires 1.56kWh.

A more accurate test would be to heat up from a cold tank, so the starting point is known, with 6 or so kWh put into it.