HeatpumpMonitor.org flagging monitoring errors

[TrystanLea]

It’s been on my list for a while to implement a way to flag and hide systems with monitoring errors on HeatpumpMonitor. Some have raised concerns about data quality seeing systems listed with issues such as heat output when the compressor is off due to temperature sensor offset errors or defrost energy not being taken into account. It’s clearly important that the data on HeatpumpMonitor can be relied upon to accurately reflect the real system performance.

This hasnt been that much of an issue yet as all of the systems listed in the top spots do not have these issues but it would become more of an issue if a system with significant measurement errors did occupy a top spot and so it’s good to get ahead of this issue. It can be more of an issue when making comparisons on particular brands e.g Daikin where there are more systems using heatpump integration as the data source.

At the same time we do not want to discourage contributions that use heat pump integration and even systems with measurement errors still provide useful data on a whole range of system attributes such as system control/cycling/modulation.

Those of us with experience in looking through the data can identify common measurement issues and can take this into account when comparing systems but it’s less straightforward for newcomers to the site.

I noted that we wanted to look at this issue in the recent HeatpumpMonitor.org next steps post:

@TrystanLea

Add an option for site admin’s to flag systems with data issues and add a comment as to why the flag has been added. This would show up as an icon, e.g red circle with exclamation mark. It would be good to have the option to hide/show systems with data issues.

and have since implemented a first pass at this feature, systems with flagged metering errors are now hidden but can be enabled from the ‘show systems with’ box:

Selecting only the systems flagged with metering errors, we can see a little yellow exclamation mark, hover over the mark to see a note about what errors were identified.

it’s also possible to select the node field and show it in the table for quick reference:

Not including defrost energy does not have that much of an effect on long term COP figures, it usually results in a 0.1 COP difference e.g it might list 4.1 rather than 4.0.

Temperature sensor offset error is more significant. Here’s a good example, the small DT error here is resulting in a suggested heat output of 385W when the compressor has clearly been off for some time.
The COP in window of 4.79 drops to 4.43 without this heat contribution:

I’m not 100% sure that I’ve got this feature quite right yet. Should we flag systems that do not include defrost energy differently to those with temperature sensor offset errors perhaps? Perhaps a yellow flag for defrosts and orange for temperature sensor offset errors? Is it fair to hide systems on the main list which do not include defrosts or is a flag icon fine?

[Timbones (Tim)]

@TrystanLea

Should we flag systems that do not include defrost energy differently to those with temperature sensor offset errors perhaps?

I feel defrost energy should be captured next to the other options for what the metering includes, as it is specific to how the metering is set up rather than an error in the the measured heat like for sensors:

[KevinF (Kevin Firth)]

Thanks for prompting the discussions. Its a tricky challenge and depends on what the long term aim of heat pump monitor is.

If it’s direct comparison between systems and focused on COP the data quality and variation becomes more important.

If it’s to enable users to ask better questions and understand opportunities for learning / adapting then understanding why the data is different is more important.

In reality both data quality and context are important.

Personally I would favour including data but with flags and perhaps a brief description of the possible impact of the error somewhere. There is such a wide variation in systems that I think making the variations visible is really helpful. Then individual users can ask better questions and make their own judgment with advice from the forum that’s available.

[TrystanLea]

@KevinF

In reality both data quality and context are important.

Agreed with both points, quality and learning both important, learning perhaps slightly more so.

Il try and implement two flag settings, yellow for not including defrosts and orange for temperature offset errors. We can try having these systems visible by default with an easy option to hide them with the system selection tool and review the result?

[KevinF (Kevin Firth)]

Thanks @TrystanLea - any advice on dealing with my own system error welcome! It’s important to me to understand the limitations of data and keep trying to improve it!

[alibloke (Ali)]

I’m confused on what defrost energy is, are we talking backup heaters or something else?

[Timbones (Tim)]

This is when the heat pump runs in reverse, cooling the house a little to warm up the outside unit to melt the ice that has accumulated on the coils. So, negative heat is produced during the defrost. Some of this heat is recovered when the system resumes heating again.

[alibloke (Ali)]

Thanks Tim. For quite a few of the 13 systems with metering errors I can see that negative energy is included when checking the box for “Show cooling / defrosts”, for example Ken Bones’ system in Basingstoke:

Link:
https://emoncms.org/app/view?name=MyHeatpump&readkey=8556f35f36e52bd7e3772816e6b9cb41

Am I missing something?

[Timbones (Tim)]

@alibloke

for example Ken Bones’ system in Basingstoke

This one is flagged for a different reason, not defrost energy:

[alibloke (Ali)]

Oh of course. How about the Speak to the Geek system in Port Talbot?
https://unpraying-ape-8145.dataplicity.io/app/view?name=ASHP&readkey=82af8f22985d47049461f3fbbc16632a

[Timbones (Tim)]

Same issue. See system details page: https://heatpumpmonitor.org/system/view?id=117

[alibloke (Ali)]

I could have sworn it mentioned lacking defrost data. Thanks for the clarification Tim.

[MikeEvans]

My system (147) has been flagged “Sensor offset issue”. What does this mean, and what can I do about it?

[MikeEvans]

On April 17th 2024 I added a 0.1° offset to my return temperature, to counteract an offset between the reported flow/return readings. Wondering if this was the cause of the reported “sensor offset issues”, I thought it best to investigate again.

My Daikin can be set to do an “air purge”, where it runs the pump for a while, but does not start the compressor. I started such a sequence at 16:58, and within a few minutes could see a negative ∆T of 0.1°. Around 17:05, I removed the 0.1 degree offset, but once the system stabilised I started seeing the “free energy” positive ∆T of 0.1° at around 17:15 - this was the reason I’d put the offset in in the first place.

I reinstated the offset at 17:21, and aside from a few noisy samples, the overall picture to the end of the purge run, and throughout the next, was of equilibrium or 0.1° loss, which seems reasonable to me given that it’s around 1.5° colder outside than in. It seems to me that the current arrangement is about the best that can be achieved, given the precision of the sensors involved. The heat reported is on the order of -20W across an hour or so, which seems very much “in the noise”

Could the data prior to April 17th be what’s getting flagged as subject to an offset error? Or is it a current issue? Is my approach anywhere near valid in terms of verifying/addressing sensor offset? Is there anything else I can do to resolve this?

[TrystanLea]

Thanks Mike, I will take a look at this shortly

[meatballs (Ben)]

Is it possible to suppress the errors? Mine has air ones where immersion input energy was reported but not included in the heat energy…

[TrystanLea]

Not yet, but how we deal with errors is an evolving aspect of the project, something that we will need to get much better at as the project grows.

[Timbones (Tim)]

I’ve synced my data to emoncms org, and now I’m seeing data errors that I don’t see on my original instance: e.g. when looking at the past year.

Each error is only 2 minutes long, but when you click on one of those days, there’s no apparent error. e.g. March 2nd.

Could this be due to only have 1-2 minute resolution in the data feeds?

---

I also notice that the ALL view is showing data before the start date I had set (25 Feb 2022)?

[dMb (davidMbrooke)]

I’m seeing a very similar 2-minute error being flagged for yesterday, but when clicking through there’s no detail of any error in the daily view and I don’t think the data is actually wrong.

Here it is shown in the weekly view:

Apart from the COP being rubbish (23hrs 40mins of ‘standby’) it all looks OK to me.

All my feeds are on 2-minute resolution because of how I pull data back from NIBE Uplink.

[TrystanLea]

I’ve increased the number of page setting url parameters supported on the system list page. Here’s a view of all public systems with flagged data errors:

https://heatpumpmonitor.org/?period=all&minDays=0&errors=1&mid=0

I’m working on automating the air error flag to only show if the impact of the air error is > -0.2 COP in any given period, perhaps -0.1 COP on the annual view, based on removing the kWh electric consumed during air error periods.