Defeat "Always On" feature to obtain specificity of non-identified devices

device-detection
#1

Always on is based on an algorithm that averages 24 hour past usage. However, it is my understanding that the devices that are included in always on are then NOT individually identified or detected. This defeats one of the main purposes of Sense - to determine where there is usage that we may wish to identify and control.

I have been asking for a while to add the option to defeat the Always On feature so the devices that are using power continuously (or very frequently) can be identified

#2

Due to the way that energy disaggregation technology works, devices that are always on will never be identified by Sense. Sense needs to see an on/off signature in order to identify a device, so devices that are always on will, by nature, be unidentifiable.

#3

Also, using smartplugs is a workaround to measure the usage of always-on loads that you care about. I have one on my network rack, entertainment center, wine cooler, etc.

1 Like
#4

Thanks for the quick reply - my thinking is that since the number I’m seeing with Always On is always changing, devices included in that total are turning on and off and might be identified. The always on total varies from over 700 to low 400s, so if the always on feature was defeatable, then we might capture the identification of the devices as they turn on.

Also, regarding your other suggestion about the smart plugs - I don’t know where some of the devices might be that are being included, I was hoping that I could use the process of elimination to locate them.

#5

I would extend your explanation a bit to include components of a device. Sense (without a Smartplug) would never find the 7W component my furnace uses just to keep the control circuit board running and feed the thermostat power. Or the 8W of continuous power each of my garage door openers use, just to keep the radios “listening”.

2 Likes
#6

Well, I don’t think that sense would be prevented from identifying a particular device just because its usage is lumped in with the “always on” bubble. The monitor will still analyze any start/stop events it sees and attempt to match/develop a model. The odds of that happening for something without clear and repeated on/off patterns is very low, though.

#7

I understand that very low wattage devices won’t be identified under any circumstances, Always on or not. However, as I try to isolate “drains” that have eluded MY detection, having them listed as other and not part of a 24 hour averaging algorithm would permit more efficient “sleuthing.”

And in any case, isn’t this a bit like the Microsoft “clippy” thing where the software developer decided for the consumer what they wanted and then made it hard to defeat. (At least Clippy was defeatable - why can’t “always on” go bye-bye as well?)

#8

I usually use the real-time power meter, not the bubble display, when hunting for always-on loads. Does that not work for you?

#9

Well, my understanding is that if it is included in the always on total, it won’t be identified either separately or as other because then it would be counted twice.

#10

Many of us find the Always On capability and implementation very useful, especially when coupled with smartplugs. On the flip side, I’m not sure that an “Always On”-defeated waveform version of Other will give you the device isolation capability you are looking for. Sure, you would have a reasonable resolution waveform that represents Total Usage minus all the Sense identified “drains”, but I’m not sure it would be much more useful than looking directly at the Usage Power Meter, since that would only remove a portion of the waveforms that compose the aggregate Usage. I’m guessing the remain chatter in house, plus possible sampling and subtraction artifacts from the calculation, would easily hide a 5W uptick from your low wattage device being turned on.

#11

I really appreciate all the responses I’ve been receiving on this posting, and am taking them all to heart - however, somehow the discussion “evolved” into one where I was looking to isolate a low-wattage device. That’s not the case but I don’t know if that really has bearing on the bigger issue:

Ultimately, as my satisfaction with Sense as a “sleuthing” vehicle is very limited (as it has not even identified my 5 year old Kenmore refrigerator after 2 years,) I’m looking to use whatever tools the system has available to gather information about my usage. I find that there are variable delays in the power metering and am trying to holistically use contextual clues to see where my energy dollars are going. To that end, I don’t see why asking to make a feature defeatable seems to engender such an upswell of denial.

You have a feature you like, you can leave it on - you want to try it off, you like it that way, you leave it off.

Customer choice - that’s all… It can’t be that hard to do, is it?

#12

Sorry,
Just trying to understand what you are actually trying to do, so we can help suggest methods that you might not have thought of. @pswired suggested the smartplug approach but you said you didn’t know which devices were causing the variation. Now you reveal that you have a vexing Kenmore that’s not being detected, which would be an ideal usage model for a smartplug. But now it’s clear that the only thing that will really satisfy you is the “defeat” feature for the way you want to sleuth. Got it…

And I’m also sorry to stick up for Always On as a reasonable and useful feature, not a Clippy feature as you imply, but that’s customer choice - that’s all

#13

No need to apologize - and I have put a Kill-a-Watt receptacle on the refrigerator (not a smartplug, though) so I know what it draws. I was just using it as an example of how Sense hasn’t proven to be as effective a system as was originally advertised back when I bought it a couple of years ago. They haven’t identified nearly as many devices as I (and they) anticipated they would.

In fact, one of the reasons I chose Sense (instead of the competitors at the time) was because they specifically stated that the threshold for detection in their system was only about 20 watts, as opposed to 400 for a similar competitor.

So would there be any advantage to using a smart plug as opposed to the Kill-a-Watt meter? Is there a way for Sense to use the data from the smartplug to assist it in identifying the refrigerator?

#14

I’m with you. Sense isn’t identifying devices to the level I would like yet either - my house is only at about 50% identified and 10% Always On (40% Other) even with some judicious smartplugs in place. But I knew going in that this was a hard problem to solve, regardless of marketing pitches, so my expectations were fairly low. Plus I know another 25% in that mix is EV car charging for cars that haven’t been identified yet.

As far smart plug advantages vs. a Kill-A-Watt meter:

  • Integration - you get to see the readings from the smart plugs in the Sense app just like any other Sense device. The smart plug usage is part of the % calculation and $$ calculation, plus all the associated summaries and Device Power Meter
  • Smarter Always On - Sense pulls the always on data associated with the smart plug device(s) out of the Always On bubble/data and into that specific device. Instead, you’ll see an Idle power associated with always on power for each relevant smart plug device (my Furnace pulls 8W when it is off, goes up 130W when it is On).
  • Better picture of what whole devices are doing. Sense was able to identify my furnace fan automatically. But once I hooked the furnace to a smart plug, I was able to see 2-3 additional components that happen at the start of a heating cycle (igniter, ignition fan, etc.), that Sense wasn’t seeing (and aren’t there when AC or fan is on).
  • Control - you can turn the smart plug on and off from within Sense.
  • Future Learning - The best way for Sense to improve learning for devices is via “ground truth” data - that’s exactly what the smart plug provides. That’s not a now feature, but I know from my background, that Sense is will be most likely to improve on devices that it has “ground truth” data for.
2 Likes
#15

Tell me what do you think about my idea of clearing out Always On section. I was able to shake it up by resetting all breakers and running my Wi-Fi routers off UPS supply. I was able to almost kill Always On wattage from 500W to 10w, and transfer all those undetected devices to identifiable Other section for 1 day until it jumped to Always On again. I know most of my Always On loads, so now I’m thinking adding low wattage timer switches to cycle those on/off once a day to make Sense happy. Of course, timer switches will consume power itself but not as much. This way I could unload Always On, not that it brings ton of value, mostly for esthetics cause Always On thing is ugly thing to see everyday. One day those loads may become identified altgough it’s not clear if they will stay like that if you stop cy cling once identified, will they?

#16

Always On gets calculated over a 48-hour period, so yes, turning off those associated devices could remove them from the Always On calculation. However, you would not be able to entirely get rid of the bubble as, at the very least, Sense itself is still drawing about 4 watts. And, you very likely have Always On devices in your home that you don’t want to or can’t power down.

As for identification, we’re working on ways to better identify Always On devices but there’s a significant barrier there: automatic device detection requires Sense to see repeated on/off cycles of a device. If a device is running constantly, there are no on/off cycles to see.

Can I ask why you find Always On to be an “ugly thing”?

#17

Always On section is not ugly on itself, it is completely understandable why it is there. However, seeing it everyday and realizing that there are dozens of things that will never be discovered - this is ugly. It’s kinda boring section to watch on vs. “Other” section which could bring some new detections, it’s more exciting to have your loads there.

#18

There’s definitely a chance for future discovery. Integrations are key here and we’re working on ways to use network data to better ID those devices. Of course, not all Always On devices are networked In that case, smart plugs can help pull those from your Always On calculation. With a little tracking in the Power Meter while you cycle devices, you can also help to ID those devices on your end and decide if they’d be best left unplugged when not in use or left on a smart plug with remote control options. There’s a lot of opportunity for ROI in the Always On bubble and we’re working to get those opportunities more visible to users. Keep watching in 2019 :+1:

#19

Sounds like it would work, and a Sense engineer told me that I could reset the always on algorithm by disconnecting the power, etc., as you suggested… however, I DON"T know all of the always on devices - that’s the point.

But what the !!! - I’ll be darned if I’m going to go through all that just to accomplish what the Sense software team could (and should) do in response to a reasonable customer request.

#20

Unfortunately it’s just not as simple as removing the feature. Those Always On devices, whether or not they’re shown in a bubble or not, don’t have the repeating on/off cycles that are needed to differentiate them from the noise in your house. @kevin1 had some great points above on why smart plugs are a useful solution to these issues at present.

1 Like