HeatpumpMonitor.org starts & starts per hour

I’ve added the recording of heat pump / compressor starts and starts per hour to the data points collected by HeatpumpMonitor.org. A couple of people have asked for this (Phil Jennings, Owen Burbidge). Phil is building a model to try and estimate compressor lifespan based on system volume & over sizing etc.

These stats columns are now selectable, see example for the last 90 days in the screenshot below (last two columns):

The data is recorded as starts per day in the database and so comparisons of e.g starts per day vs outside temperature will be possible (the graphing tools and API do not yet make that easy) but this is an example pulling out data from the database directly (x-axis: outside temp, y-axis: starts per hour, my system HeatpumpMonitor.org)


Il be doing a bit more work to make it easier to access this data next.


Also interesting to look at starts/hour for Vaillant systems, specifically, to see if we can spot ones with the bad firmware. Here’s data for January 2024:


Some quick thoughts/questions:

  1. Does this metric ignore idle time?
    For example, if a heat pump started 3 times but only had call for heat for 2 hours out of 24?
  2. Could be interesting to plot starts/hour vs. outdoor temperature to see when cycling starts.

The Vaillant keeps all-time stats of

  • Compressor Hours (which I’m assuming is running hours)
  • Compressor Starts

You can then do Compressor Starts / Compressor Hours = average Cycles per hour

In my case

  • Compressor Hours: 7194
  • Compressor Starts: 9292
  • 9292 / 7194 = 1.29 cycles per hour

Which I talked about my article about the firmware bug on earlier Arotherm models


Thanks @Zarch interesting that he HeatpumpMonitor.org figure for total starts is not too far off in the case of your system, at 8910 starts vs 9292 starts a 4% error. Perhaps we should also measure compressor hours…


From memory I didn’t get my remote OEM up and running properly until about a month into the heat pump’s life. So OEM is missing a whole October of compressor data.

Which might explain some of the discrepancy?


Yes, I think this would be useful.

a) to match up to the Vaillant numbers (for Arotherm owners)
b) to give context and a relationship to the number of starts.

1 Like

How are starts detected? My system is showing zero starts or starts/hour, but it’s definitely been running recently.

It’s detected by a threshold if more than 100W it’s running, if less than 100W it’s not, and then if it changes from one state to the other it’s registered as a start. Next step is to make that configurable and add a minimum run period for it to be a valid start.

Im also going to add run time per day.

1 Like

What is your system id? happy to take a look

Hmmmm, that 100W threshold should have caught it, I’d have thought. I’m system 147. Could the fact that it’s only doing DHW and not heating at the moment be at all relevant?

Ah interesting, it turns out processing PHPTimeSeries feeds takes way longer than the PHPFina feeds and this was causing the script to timeout. I’ve just recompiled your data but only loading 7 days at a time rather than 60 which seems to have worked. It’s now showing the starts data.

Thanks Trystan. Am I using the wrong feed type? I’m uploading to emoncms.org from from Home Assistant.

We dont specify that they should be PHPFina, but generally yes it’s best to use PHPFina rather than PHPTimeSeries for regular 10s data like this. Processing is much faster as the data is stored at a fixed interval and this fixed interval makes it very fast to find data points in the binary data file as there’s no time consuming search required.

I could probably convert these to PHPFina for you if you were happy for me to do that.

It doesn’t sound like there’s any downside. If you’re happy to convert, please go ahead. Many thanks.

My Daikin pulls 180W with the circulation pump running to check the flow water temp, without a compressor start. I know I can change the threshold figure when viewing the stats locally, I can’t see that on the HPM site? Otherwise it won’t be very accurate - the flow pump runs pretty often during an off cycle. It’s 111 on the heatpumpmonitor.org site.