Long Running Devices

The “Always On” piece of the device seems to be an interesting enigma, where we don’t really have any insight to the devices that create this category, but it clearly is not just a static value. We’ve seen swings of usage, and we’ve been told it’s an average (over some undetermined period of time).

My question is more about devices that have been identified, but then go into longer running process, that might be considered “always on”. Here would be a scenario:

The blower fan on my furnace, during the spring/summer/fall months is only kicking on periodically when my hot water tank needs to be heated. Lets say it gets identified, and is accurately reporting usage for a period of time (maybe multiple months). In the winter, when it needs to heat the house, it turns on far more frequently, and for far longer amounts of time. In some cases, multiple hours (lets say for this example, that it needed to run for 24 hours straight, even though that might be a long shot). So, some cold snap comes in, it’s brutal, fan kicks on for 24 hours or maybe 48 hours, at what point does that become an “always on” device? What would happen to the device usage from the previous multiple months?

Another example, IF my computers had been identified as a device (that is yet to happen). And for some reason, I leave it on for multiple days. At first, it shows up as “unknown”, and after some amount of time, it transitions that usage to “Always On”. IF it had been identified as “Computer”, and I leave it running for a couple days, even a week, how would that get converted in the “always on” metric? Would it show as both “computer” and “always on” (that seems like reporting double). Would it never go into the “always on” bucket (but it’s been running for maybe weeks, wouldn’t that classify as “always on”)?

Just some questions around that mysterious bubble. If I can dove-tail that into a similar question — I’ve had a device “Aquarium Heater” show up and report very accurately for a month. After a period of time it disappeared, and magically reappeared as “Unknown Heat 9”. I renamed that, but I was wondering where did the “energy usage” from the previous months device that vanished go? Is it all of the sudden bumping my “Unknown” up? Did the old device stay in the database, with just no way to see the information gathered? Where do the things that move around end up getting reported long term?



You must have been writing at the same time I was about my boiler pump :slight_smile:. Is Sense weather aware?

Could it be something as simple as if the kWh consumed this hour >= average W then allways_on? What does Sense do if it misses a change in state from on to off? Maybe that’s where your blower and my pump went?

1 Like

I just checked with our data science team. So the short answer is: yes, it could happen that a detected device which is on for a prolonged period of time (like 24/7) could become part of what Sense sees as “Always On”. That said, our team thinks this is pretty unlikely for most real devices, since even a furnace which is providing a lot of heat is unlikely to be just running at a constant rate all the time. It may instead be considered part of “Unknown” if the detected device is not showing as on but Sense still considers it not part of “Always On”.

Note that if it does happen that Sense thinks the device is part of “Always On” it won’t break the existing furnace model, it would just make Sense unsure what the source of the power use is for that period in time.

As you know, Sense relies on machine learning algorithms, and certainly we expect this type of detection and logic to improve over time!

1 Like

BTW @allen.barriere - regarding your Aquarium Heater: you can contact our support team (support@sense.com) so they can take a look at that specific device for you and why it went away and/or a new model for it appeared. If you can provide specific timeframes for when you think yore models appeared/changed (looking back at your email history for when you were notified of the new devices is one quick way to do this) that would be extra helpful.

Overall Sense does keep adapting models - so some devices can change over time. Although infrequent, its possible an adaptation can cause a specific model to get worse… obviously our models improve overall as they change but some specific ones may degrade based on overall improvements across the board. We obviously try to avoid this and don’t normally see models degrade.But again, if you contact support our team data science team can then investigate specific occurrences.

1 Like

I know that it has been a while since this was posted, but I have wondered some of the same things and have some additional data points to add (although, kind of in the opposite direction.) I installed Sense a couple of months ago at my vacation house and have experienced some of the typical growing pains reported in this forum.

I was at the house for the first few weeks after installing Sense and was able to heat the house almost exclusively with my wood stove. There is a fan in the wood stove and I also used a ceiling fan to help circulate the heated air. During this time my “Always On” reading was right at 290 watts (I turned each of the fans off several times, but they were usually on.) I have been away from the house now for three weeks and my “Always On” reading is 178 - 180 watts. I assume that this is a more accurate reading because there shouldn’t be anything running (other than detected devices such as water heater, well pump, fridge, and freezer.) However, I also have a pretty consistent 8 - 10 watt load that shows up as “Other.” I have no idea what this could be, but it sure seems like it should be part of “Always On.”

I think that it would be interesting to know if the electrical signature includes things like the duration of the load. I will also be interested to see, when I go there this weekend, whether my “Always On” load goes back up to 290 or the other things that I use while there will be included in “Other” and “Always On” settles in at 179.

A few thoughts based on my experiences so far:

  • “Always On” calculation is based on a 24 hour window of your usage, so when you visit you are likely activating something or a set of things that overlap to stay on for a full 24 hours prior to that reading.
  • The 8-10 Watt mystery load must cycle down occasionally during a 24 hour window or it would be included in “Always On”. Look at your daily usage traces to find the low point in the day to figure out when that load might be off.
  • The search for the “Always On” devices might be interesting - you obviously have an internet and WiFi “Always On” which probably burns up 20-30W. I was surprised to discover that each of my garage door openers continually burns around 10W continuously “listening” for the signal from the RF remote.

Kevin –

Thanks for this response. I wasn’t really thinking about the rolling 24 hour average, so I agree that the small load that I see is probably something more intermittent. It could even be something like the TiVo when it is recording something as opposed to just in standby.

BTW - Last weekend, when I was at the house the Always On load remained at 179 watts, but I was only there for about 36 hours. We are going there soon for another extended period, so I will see if the Always On load changes based on things that we may have on a lot while we are there (such as ceiling fans that remain on virtually all the time.)

I haven’t investigated to see specifically what my Always On amounts are for my different devices (I have a TV, TiVo, home automation controller, AT&T micro cell, modem, WiFi router, several digital clocks, and a garage door opener that I can think of right now.) Maybe I will look into that more when I am there.