As I suspected, this UPS is a rather basic unit. You know the VA size of the unit (1500VA) so create a template sensor that is Power = (Load / 100) * 1500. Or you can use the Powercalc integration and it’ll do the same thing and provide you with a KWh sensor as well. This is what I do and while it’s not 100% accurate, it’s good and close enough.
1 Like
On my ups the load % only updates when running on battery power, when on line power it reports the value from the last run on battery.
I’ve found the Shelly Plug 2s do an okay job of power monitoring, but they’re WiFi. (That really shouldn’t matter unless you want to monitor power when the wifi is down but somehow your zwave mesh doesn’t go down?) You can also use Fibaro devices and wire them up manually to a plug to get the energy monitoring via z-wave. Fibaro isn’t cheap, but high quality.
(One quick comment: if you have multiple UPS devices and you use a single device connected to them all to monitoring them, you introduce a possible failure mode: if the single device gets a voltage overload it could take out your other UPS devices via those connections. E.g., each UPS device should be a surge protector for the devices behind it, but if you connect across UPS devices you’ve added a path that bypasses that protection. This is probably incredibly unlikely, though.)
Thanks, I’ve got the 1st PI Zero up and running with NUT (following these instructions) and it seems to be humming away nicely.
To infer wattage I’m multiplying load times nominal power. My load value does fluctuate a little over time (up or down 1%) — so I don’t think I’m in the same boat as @PeteRage . I haven’t done too much experimentation with adding loads and confirming that the reading really does change accordingly. (I will try to confirm that this weekend and I will also try to cross-check with another power meter to make sure that the nominal power (1500 VA) time load actually does realistically reflect power.) The one minor drawback is that the load entity seems to report in increments of 1% … which corresponds to a measurement granularity of 15 watts. But that’s not the end of the world.
1 Like
Also remember that VA is not watts. VA is the approximation of usage (apparent power) and Watts is “real power”. In a DC circuit where loads are purely resistive, they are the same which is why those are always represented in Watts. In AC though, there is inductance and the losses created from the sinusoidal wave, so power usage is represented in VA (literally volts x amps) and is only approximate. So being off 1% is really not that bad.
1 Like
Ah right thanks! I’m trying to remember years and years back to my E&M courses but do I need to include an extra 1/sqrt(2) factor (or something like that) to account for the fact that this is AC power (and I should therefore be looking at voltage times RMS current), or do we think the the “load” entity and/or nominal rating of 1500VA likely already take that into account?
Ok here’s an update: I’ve successfully installed NUT on 5 raspberry Pi Zero 2W’s monitoring 5 different UPSs. They are all humming along nicely, but I’m still not sure how to convert the “load” percentage to a wattage.
So I tried sticking an Athom smart plug between the UPS and the wall for a few days and recorded for a while. I found that the load did track closely with power (yay!) though it was much more coarse as noted previously since it only reports in integer percentages. However if I divided the measured power by the percentage load, I definitely did not get the 1500 that I expected. The ratio bounced around a bit (given that the smartplug measurement was fluctuating and the NUT load reading only moved by a full percentage point at a time), but the average over several thousand readings was 1073W. Interesting. (That’s kinda like 1500/sqrt(2)???)
So I tried with a different UPS, different brand, etc. And the experiment came out to 1072W.
So I’m tentatively concluding that “the way to measure power using NUT is to multiply the load percentage by the magic number of 1072W”
Does that strike anyone else as reasonable… or odd?
Thanks!
Seems reasonable to me, you’ve empirically derived a formula
y = mx
I doubt at 0% load it uses 0% power, so if you care about that, derive b by unplugging everything
y = mx + b
This assumes it is linear. It probably isn’t. But it also may not matter.
Have you been able to calculate the wasted power? Meaning the difference between what the devices are using and the UPS?
1 Like
Yes, definitely possible to do a linear regression between pairs of readings (though the usefulness of that presumes that the relationship between UPS’s “load” and Athom’s “power” is (a) linear and (b) consistent over time), but I’m still trying to make sense of what I’m seeing when I do.
Where does the seemingly magical number 1073 come from? It seems to be consistent over two different brands/models of UPS. Is it something specific to NUT? Is it a miscalibration factor from the athom smart plug (e.g., maybe it’s supposed to be 1000, but the smart plug is reading 7.3% high?)? Does it actually come from 1500 VA / SQRT(2)? etc. I would just love to understand what’s going on!