Hi,
I’m trying to generate a Machine Learning training set from the system information that’s publicly available.
I’m pulling down data electricity input, flow temp, flow rate and outside temp using this URL.
https://heatpumpmonitor.org/timeseries/data?id={id}&start=01-06-2023&end=01-06-2024&feeds=heatpump_elec,heatpump_outsideT,heatpump_flowrate,heatpump_flowT&interval=3600&average=1&timeformat=notime
When I format this into a CSV, I’ve noticed that the flow rate values are enormous
Here are a few examples.
540.6100546095106
538.4602754380968
70.95514704386393
21.969012451171874
56.996384684244795
537.2835748460558
537.8868751525879
548.0047322591146
I appreciate this value is m³/h, but it’s still insane.
What am I doing wrong? Could this be related to interval?
Hello @tomasmcguinness
The units for flow rate are not consistent for all systems, most are L/min, mine is m3/hr, I wonder if that one is L/hour?
You should be able to access the units via e.g: https://heatpumpmonitor.org/timeseries/available?id=2
Ah. So that endpoint will tell me the feeds (and units) available for each system if I change the ID?
I tried that and it’s making sense.
Some of the flowrate feeds don’t have a unit property available.
What is the default unit for flow rate? Is it litres/sec?
Litres/min is typical I believe, and should be in the range 5-30ish L/min. If you’re really stuck, perhaps as a sanity check:
heat_output = flowrate (L/s) x 4150 J/K.kg x (Flow temp - return temp)
flowrate = heat_output / (4150xDT)
We should automate a script on the server to apply these units consistently!
2 Likes
Had better success with that. You’re right about the units. The values come back where sometimes in l/h, m3/h or l/m. I’ve excluded any feeds without a unit value to save me from guessing.
I’ve also used the start_time and end_time to get better timeseries data.
The predicted value is better, but still off. I’m certainly no data scientist, but we all start somewhere 
I’ve written this up if you fancy reading more: Mass Flow Rate, not Room Temp – tweaking my Heat Pump Maching Learning Model – @tomasmcguinness
1 Like