I know, I know, patience.
But come on.
I have a fridge in the basement, it was found in week one.
The Samsung fridge upstairs, that gets the most use, hasn’t been found.
Add to that the dishwasher and not one single light bulb.
The ceiling fan in the bedroom is on 8 hours a night, not found.
Really other than the solar reporting it is a pretty big disappointment.
I have geothermal heating. It found the heat pump early on but not the two pumps that turn on and off to pump the loop water.
I feel like we are all paying Sense to be give them data to try and make their product work somedays.
I know, I know, patience.
Indeed patience is “king”! I’m confident that your detections will continue, during the subsequent months ahead. This was definitely my outcome.
No real need for disappointment at this stage. However, the expectation of quicker detection for higher use is totally understandable.
I think of it like, say, iCloud image storage and UI:
I want to be able to type in “ladder” and have Photos magically show me all my pictures with ladders in them. I have many … it doesn’t work … yet. Will it ever? How accurate will it be?
Meanwhile “beach” is pretty successful and “electric fan” is cool.
Not long ago doing people searches was pretty useless. Now it “just works”
I assume the old basement fridge (“beach”) has a non-DC compressor and the new Samsung fridge (“ladder”) has a DC inverter = very hard for Sense.
An interesting thing in both the Photos and Sense worlds is that at some point the detection becomes good enough that your “unidentified objects” constitute a definable set(1) that the algorithm (and/or algorithm building team) can focus on. The really hard stuff. We do this as humans very efficiently. This is probably what the Sense team is doing these days.
(1) A definable set of electric devices is certainly ramping-powered DC motors as found in many modern devices.
My strong recommendation is to put a modern fridge on a Kasa HS110 smart plug. The equivalent of doing some manual tagging in Photos.
Here’s a search for the set of “stairs” (via Apple Photos auto-magic) and "ladder " in my Photos (most of my pictures containing ladders, such as this one, I have manually tagged as containing a “ladder”). The result is pretty amusing …
But allegedly it gets smarter with all the other users. Surely there are a few Samsung refrigerators out there.
Samsung has around 20% market share in the US so undoubtedly the answer is yes.
You can see a few Samsung’s in the Community Device Library
Because all the DC inverters in fridges, freezers and HVAC (and elsewhere) present similarly difficult signatures for Sense to disaggregate, my guess is this actually at the top of the detection development list. Last year Sense made a concerted push to improve AC detection during the cooling season and aspects of fridge detection would be improved accordingly.
Is it truly save for someone that travels alot, or used too, to have a fridge on a smartplug?
No guarantees, but a number of folks on the forum have HS110s on the fridges and freezers. Some are converts from Wemos, because the HS110 powers up into ‘on’ mode, while Wemo, to ‘off’. I have personally had one HS300 have some plugs go off on me during a power outage, but never an HS110, of which I have 14.
The problem isn’t the number of units, it’s that Sense is primarily built around transition and inverter (and proportional speed) devices don’t work like that…it’s a fundamental difference in behavior. Perhaps they will eventually catch inverter devices…but probably never will get “always on” devices…no waveform edges to detect.
I’ve been doing this for a while, and as long as the smart plug recovers to “on” (for example after power line glitches), it seems pretty stable. Of course, the ability to keep an eye on it, to set alerts, and to control it remotely all add a safety backup.
This past Saturday, the HS110 on my freezer came back in the ‘off’ mode. luckily I found it Sunday morning. First time I’ve had this happen.
I just put one on the fridge. I set it to turn on three times a day. That should do it.
I have a Samsung RF28JBEDBSR/AA and it was detected pretty early in my first setup (which has since been reset). Help identifying a pulsing "Device 1"
I have since done a couple resets and it was detected again with more accurate wattage the second time but I’m on reset number 3 and it hasn’t been detected with almost a month in so I’m not holding my breath. Previously it had been detected within 3 days. A mini-fridge and stand-alone freezer with a more traditional “Fridge signature” have been detected already.
I think in some cases, the order of device detection can make a difference. An existing device detection that gets conflated with other devices can interfere with accurately detecting a new device. Just a theory.
Don’t forget! They power up into the same state they were in when power was lost. There is a slight delay to turn back on when power is restored if they were in the “on” state.
Agreed. Hope it comes together. It seems most things have multiple signatures so it’s hard to get 100% of a device or have confidence that you do.
I put an HS110 on that Samsung fridge and can see why it has trouble finding it. It basically never goes off. It pretty much always draws energy, from 5-180 watts and everywhere in between.
It is using 2.2 kWh per day.
We replaced our old refridge a couple of years ago with a Sears (actually Samsung), and Sense has never found it. The behavior is typical of modern appliances. The HS 110 does a great job, but of course doesn’t need Sense (or benefit from it), so if I had to do it all over I’d not invest.
I never thought to put HS110s on thinks like my fridge and was just waiting for sense to figure it out. I too have a Samsung so I’ll do this as well.
Why is there no way to manually identify devices and label them? It would seem this would drastically improve the development of a database to increase the onboarding and accuracy of new devices?
Check out this thread from one of the members of Sense.
You can also read this Blog entry from 2016 from Sense about how their training works.
@pbrsucks, Good question and one that many users ask. @ben gave you some good pointers for answering. Here’s one more. Hint: there are ways to help with training, but none of them directly involve you switching a device on and off.
There is a hand accumulated “database” of device waveforms here, that kind of shows some of the complicated aspects of on and off signatures described in the article.