Posted by: @robl

↑

@derek-m 

The sensors are tsic506f, and they connect up to an mbed(old micro type I’m used to) that I programmed with a free Arm account.  There’s a cheap flow sensor (2%), and an in line electrical meter (1%), that has an isolated output, so all these go straight into the mbed, and it can calculate the COP to 5% absolute accuracy.  It works well, I don’t understand why COP measurements aren’t front and centre on every heatpump.

@robl what flow rate sensor / meter and how do you read it with the microcontroller? 

@iancalderbank - you might be able to get a really quick and cheap solution up and running with a couple of $5 esp32's, some 1 wire sensors and toit.  https://toit.io/, at least for temperature.  Accurate flow rate is the challenging bit so I am interested in the sensor that @robl has used.

I put in a single grundfos magna 3 with the modbus CIM500 for a number of reasons, one of which was to get very quickly and easily to some basic monitoring.  So about 2 hours effort (install CIM500, configure home assistant to retrieve data).   This using second hand components on ebay, but even then it is not a cheap option.  But I do have something up and running with about 10% accuracy for minimal effort, and I needed a relatively beefy pump given the length of pipe I have anyway (about 40 meters of 28mm and quite a bit of 15/16mm) with quite a lot of bends.  My phase 2 is reading more data from the pump directly using modbus, then phase 3 is reading the output from the heat pumps flow sensor (sika) .... very carefully of course.  Then I will come back around and start adding/improving.  All data going into home assistant.

Posted by: @william1066

↑

@iancalderbank - you might be able to get a really quick and cheap solution up and running with a couple of $5 esp32's, some 1 wire sensors and toit.  https://toit.io/, at least for temperature.  Accurate flow rate is the challenging bit so I am interested in the sensor that @robl has used.

temps already done. ESP32's and DS18B20 sensors .  Dirt cheap. I have loads of them.

Likewise interested in a cheap microcontroller readable flow meter.

Posted by: @william1066

↑

I put in a single grundfos magna 3 with the modbus CIM500 for a number of reasons, one of which was to get very quickly and easily to some basic monitoring.  So about 2 hours effort (install CIM500, configure home assistant to retrieve data).   This using second hand components on ebay, but even then it is not a cheap option.  But I do have something up and running with about 10% accuracy for minimal effort, and I needed a relatively beefy pump given the length of pipe I have anyway (about 40 meters of 28mm and quite a bit of 15/16mm) with quite a lot of bends.  My phase 2 is reading more data from the pump directly using modbus, then phase 3 is reading the output from the heat pumps flow sensor (sika) .... very carefully of course.  Then I will come back around and start adding/improving.  All data going into home assistant.

 

ah so you had to have a big bad boy semi-commercial pump for distance / head reasons so adding the modbus to it was a minimal addition. I will need to look further at what can be done with more domestic size pumps.

@iancalderbank 

The flowmeters I use are cheapie £8 things that look like this from Amazon, which claims 2%.  I have two, one on the house circuit measuring waterflow, the other in the groundloop measuring 30% glycol flow, they have been fine, operating for just over a year.  I only calculate the heat power on the water circuit, as I don't know the glycol/water heat capacity, the glycol pipes are plastic so unsuitable for accurate temperature measurement, and we care most about heat delivered.  Linky:

https://www.amazon.co.uk/GREDIA-Sensor-Food-grade-Flowmeter-Counter/dp/B07MY7K249/ref=sr_1_2_sspa?crid=2Z0GF8PJH7QQG&keywords=flow%2Bmeter&qid=1673444069&sprefix=flow%2Bmeter%2Caps%2C93&sr=8-2-spons&sp_csd=d2lkZ2V0TmFtZT1zcF9hdGY&th=1

Here's another better made unit, costs £56 and is 3%:

https://www.hemomatik.se/PDF/VISION2000_h.pdf

Both of the items above have spinny things inside, and hall effect switches, producing pulses.  The code just needs to time a few pulses - I wait for a Hi/Lo transition, zero a timer, wait until the uP sees Hi then Lo then Hi then Lo (for good accuracy), then use the timer value to calculate flow.  There's also timeouts in it, so the code doesn't get stuck if there's no flow.

@robl - thanks for the details. £8 per sensor is affordable, but given more standard flow meters cost hundreds of pounds, I would want to be sure the £8 version really was capable of measuring to within 2%. As I said earlier, once you are up into the high single figures accuracy, and certainly by the time you have got to 10%, the whole system starts to fall apart. Rather curiously, the 3% £56 'Swiss made' (sales baloney methinks) doesn't list heating systems among its very long list of applications, though it does mention cooling systems.

The other concern I have is Observer/Hawthorne effects (the act of observing something changes it). An inline flow sensor must affect the flow rate, but how much? I also think the £8 sensor has a maximum (bore? nominal?) size of 1 inch, in which case it will become a physical obstruction on 28mm pipework, while the £56 one appears to be 3/8in (10mm) or thereabouts bore or nominal, which will be a major bottleneck on 28mm pipework. There is also a table in the spec sheet on page 2 showing pressure drop in bar at different flow rates...

I am rather surprised there isn't an ultrasound based way of doing this non-invasively, as is used for example in measuring blood flow in blood vessels. The reason it may not work heating systems is there are no (or at least should be no) particles circulating round the pipes for the US waves to bounce off, revealing a Doppler effect.

The ability to measure flow accurately is central to this whole project. If all else fails, the Midea units have primary flow rates available via modbus, and other units with modbus access (Samsung, possibly others) may also have readable flow rates. Larger circulating pumps (bigger than standard domestic ones) also have modbus eg the Grundfos Magna 3 mentioned above but they are significantly more expensive, and they are also physically larger, the Magna 3 for example could not fit in the space where my current secondary circulating pump is. Some of the more expensive domestic pumps eg the Grundfos Alpha2 do have a display that shows flow rate, but no obvious way of collecting the data short of getting out your camper chair and bedding down with pencil and paper in the airing cupboard for the duration. As ever, the Grundfos manuals get full marks for total impenetrability. I don't know if anyone has ever tried to hack a Grundfos pump to get data from it...

If only some bright spark would add a flashing LED to a circulation pump that indicated flow in the same way a flashing LED on a power meter can indicate kWh usage. Then all we would need to do is count the flashes, let's say one flash for every litre through the pump, then if there are ten flashes over a minute, that is 10L/min etc.             

For the record, I have just across this interesting page on flow rates, which also has a very useful flow rate calculator at the bottom for both water and water with 20% glycol:

https://heatpumps.co.uk/technical/pressure-drops-flow-rates/

According to the calculator, given my design conditions heat loss of 12.3kW, and a typical delta t of 5 degrees, using water plus glycol fluid, I need a flow rate of around 2.3 m³/hr. But my Midea controller reports the flow rate to be almost always 1.4 (something) m³/hr...    

In my experience, the temperature is much harder to monitor accurately enough than the flow rate.  We need two sensors well connected to the flow and return, each with good absolute accuracy.  I did try calibrating them initially, but gave up as I didn't trust the results - so I rely on the datasheet values.  

Assuming a DT of 5C, we can compare the accuracy of a temperature sensor to that of the flow meter:  If each temp sensor is 0.1C absolute accuracy (The sensors I use, which I was very happy at finding), then we hopefully measure DT to be approximately 5C +/-0.2C, giving an accuracy of +/- 100*0.2/5 = 4%.  The electricity meter is much better than this at generally 1%, and there are plenty of flow sensors with better accuracy than this - and they are simpler to install well than the temperature sensors.  I googled "flow meter" - dozens of suitable flow sensors popped up, the first one I picked cost £6, gave 3%.  Here's a slightly bigger G1 60lpm sensor:

https://wiki.seeedstudio.com/G1_Water_Flow_Sensor/#sensor-components

Perhaps you could get flow information out of a pump - it is generally suggested for good results that the sensor should be positioned after a straight section of pipe, so perhaps pump results would not be so good?

Thanks @robl, that's all very helpful. I have also been googling away, and it appears there are clamp on ultrasound flow meters, which solves the interfering with flow problem, but I am not sure yet how accurate they really are, although they usually claim they are in the sort of accuracy range we are looking for. Prices vary greatly, from under £100 for second hand ones on ebay, to thousands of pounds. I haven't found a domestic sized circulating pump that can output more than just a visual LED display of flow rate, and as you say, measuring at the pump may introduce it own problems, or perhaps its OK if it is measured by proxy say pump revolutions corrected for flow pressure/resistance. For the primary circuit, I do potentially have access via modbus to the Midea pump flow rate, which is given in m^3/h to two decimal places. But if I measure your blood pressure to be 132.22 over 84.39, that doesn't mean it is as accurate as it appears to be, and there is still the question of how and where the reading is taken.

I did wonder about using thermal paste, the kind used to attach heatsinks to CPUs, for the temperature sensors. Their purposes is after all to transfer heat. 

I started by cable tying temp sensors to pipes, and wrapping with loads of insulation, re-wrapping, getting a different answer...  There is no doubt in my mind that the pocket solution is significantly better.  I think you will struggle to get good accuracy without draining the system and soldering on pockets, also adding an inline flow sensor.  I understand your not wanting to do that though!  For ultrasonics, do you need to know the pipe inner cross-sectional area accurately - if so, can you just add one on existing pipes?  I did a bit of googling - found these units, all made up - (boo - where's the fun in that?).

https://www.bes.co.uk/vhu20-ultrasonic-heat-meter-kit-3-4-bsp-m-22290/?ref=gs&gclid=Cj0KCQiA_P6dBhD1ARIsAAGI7HA5_52ly5g8GbNi1MwDAGWu8on0Z2Kxb0Zo21j2M0ciMPTwcO4xbzAaAlVFEALw_wcB

Maybe you can recreate the bits from this kit?

https://shop.openenergymonitor.com/level-3-heat-pump-monitoring-bundle-emonhp/

@batalto - they are invasive meters, and very pricey. The electricity in (in kWh) I can do from the blinking LED on my ASHP dedicated meter, the flow/return temps we have various affordable solutions for, the unsolved bit is simple reliable affordable non-invasive flow rate metering. I can probably use the Midea operational parameters via modbus for the primary circuit flow rate, or even just set it in software at 1.42m^3/h, because that is what it always seems to be, but it would be extremely useful to have flow rate data for the secondary circuit. Semi-domestic/semi-industrial circulating pumps can do it, but at considerable cost, what I am hoping is that I can find a clamp on ultrasound based solution. I've had some near misses so far on ebay and elsewhere, but I haven't yet looked at all the possibilities. I continue to bear in mind this is a beginner's guide, and also any solutions proposed should be as widely applicable as possible.

Not getting very far with the clamp on (non-invasive) affordable flow meter. The best I have come up with is an ebay listing from China:

https://www.ebay.co.uk/itm/384834475776

I haven't been able to find any UK stockists, which makes me wonder why there aren't any...

I think this will do the job:

DN15-100mm (DN = nominal diameter), will fit on 28mm pipe ✔

Gives flow in m³/h ✔

Accuracy: better than ±1% ✔ (if true)

RS-485 output (same as the Midea controller uses) ✔

A picture from the listing showing the basic idea:

Trouble is, it is a Chinese seller, with limited feedback ratings (34), but what there is is good (100%). But at the same time, it is 'only' £160. Perhaps it is worth a punt?