I did made a heat meter - I used the tsic716 sensor (tho the one you point at is a lot cheaper, and similar accuracy, which is great).  I connected two of them directly to an ‘mbed’ running Arm C code, bitbashing the sensors.  The mbed is overkill - it doesn’t take a lot of processing power, although a low power lcd display would be nice.  The same part measured flow (a square wave o/p sensor, frequency proportional to flow), and also electrical power - from a mid certified meter with an isolated optical pulse output.  The mbed then calculates COP, displays it (add on 3 digit led display) and logs it (the mbed is a low power uP board, has an rtc, flash memory, both very handy for logging).  Proper software engineers would hate the code- no interrupts, just does one measurement after another - takes a second or two to do the lot, fine as the heatpump and pumps run at constant speed.

What will HA run on?  If it’s on a raspberry pi or similar, Googling ds18b20 and raspberry pi indicates there’s been a few projects connecting them directly - with a pull up resistor on the data line.  Am I looking at the right part - it’s a one wire digital o/p part?

The sensors really need to be well connected - I ended up soldering short lengths of thin copper pipe to my flow and return pipes, that had the sensors pushed in, and insulating the whole lot.  Just strapping the sensor on will almost certainly limit your accuracy.  If you don’t want to solder, can I suggest fixing the sensor to the pipe with several layers of copper tape, then covering with at least 1cm of insulation, making sure there are no air gaps cooling them down.

Posted by: @georgeedjones

↑

I’ve been hoping you’d do an update

I've been waiting for the heating season to start, once the heating is on I want to check and double check my monitoring system works as it should do, and then I will post the details. There are also some anomalies that need to be sorted out eg collecting the data and calculating the COP using different methods gives similar but not the same results, and I don't know which, if either, is right.

@robl - sounds interesting, but I have to say that one of my guiding principles has been simplicity - I want to have a system that anyone with enough interest and the ability to use a screwdriver without stabbing themselves can use. That effectively rules out Home Assistant (too complicated to set up, even in its 'green' incarnation) and raspberry Pis, arduinos and all similar hardware (again just too complicated for a novice who doesn't want to spend days leaning about such things). I don't want something that relies on proprietary or quasi-proprietary software (another reason for ruling out Home Assistant) because all too often the developers will introduce breaking changes and/or insist on arbitrary system requirements, and likely as not they will store the data in hard to access databases (I use csv files, very robust and accessible). The whole package also needs to be affordable - off the top of my head, I think I spent £50 or thereabouts in total.

In brief, I have a small fanless PC running linux (for historical reasons, it could also run Windows) with a wired modbus RTU connection to the Midea wired controller, USB connection at PC end, screwed wire connectors at Midea wired controller end. A simple, and I mean simple, python script uses the python minimalmodbus module to get data over modbus once a minute from the wired controller, and logs it to a csv file. Another simple python script then converts that minute data into hourly and daily data, and logs the hourly and daily data into their own respective csv files. For charting, I use a mixture of spreadsheets (another reason for using csv files, spreadsheets can chart then out of the box), highcharts.js and plotly in python. And that's it.

The obvious downside to this approach is that it is totally dependent on Midea data. If Midea cook the books (not an absurd notion at all, remember dieselgate), then my monitoring results will also be cooked. That said, I have taken some precautions to confirm that Midea aren't roasting the data before making it available, but I can't rule out a bit of gentle simmering.

Full details to follow...