At what point do you just give up on Device Detection?

This is after a little over a year.
Had such high hopes for all the data I’d be able to pull out of this system and use it to monitor and optimize our usage, but I’m stuck with two sad bubbles :frowning:

It would be interesting to see your power meter waveforms both at the daily resolution and for an hour or so when the waveforms are most active. The power meter waveform might also show whether you have a “noisy” device that is masking other detections.

ps: Your “Always On” is pretty large in relation to most people’s. That’s a place that you could optimize even without any detections.

pss: Probably worth checking in with support@sense.com on this topic as well.

3 Likes

Are you getting similar with both your Senses?

Is the big Always On/Other for the reptiles? i.e. a bunch of heat lamps? (the Solar Sense?)
That would explain a lot.

On another note: I kinda love the idea of having a purely solar powered reptile house … like having to forego skylights for PV and then using the PV power to make light.

There are those for whom Sense seems to work OK, then there are the rest of us. No clue what the ratio actually is. Apparently environment and kinds of devices vary widely.

Do you have many things plugged in to power strips or do you have each device plugged in to it’s own outlet?

I wonder if that makes any difference in detection of devices.

Power strips, unless they’re a TP-Link HS300, don’t make a difference for Sense detection.

2 Likes

What do your detections look like? Can you give more info regarding your energy usage and some of the larger electrical loads? Have you reached out to support? We’re missing a lot of information here so it’s hard to give you a good assessment.

1 Like

That’s certainly abnormal. I’d echo all of @senseinaz’s questions above. And agreed that that is a huge Always On. You should be able to make some optimizations there even without a better detection experience.

1 Like

I have the same situation which makes Sense not very useful at all.

After 2 years of training, this is what my normal display looks like:

Always on: 1656w
Other: 1264w

I would love to get some insight into how to solve the “other” and “always on” part of this.

@RyanAtSense did a great checklist on how to inventory your Always On sources here:

I would use a traveling HS110 smart plug to take aim at the Other. Try for a day or two on each suspect 120V device in your house. Sorry, but it won’t help with 240V dual phase devices or wired 120V.

2 Likes

I’m hopeful. I’ve only had sense for a week and a half and it’s discovered a few things but some of them not completely yet. Originally my mini split only showed a small steady usage and then it must have picked up another part of the unit and now it varies and shows up to 10 times as much use. It found my dryer the other day but it only registers as dryer in the bubble if I have it on high heat. Low heat still comes up as other which makes me think it’s only picking up the heating element on high and is probably not picking up the motor. Oddly though when it reads other in the bubble in my notifications underneath it will tell me that the dryer was turned on for 1 minute. It also discovered my stovetop but it only recognizes the one burner as stove, the other three and the oven come up as other.

2 Likes

It’s common for a detection on specific heat levels and not have them on others. With my dryer and stovetop burners, there are detections for high on some and low on others. Don’t be surprised if your dryer has more than one detection for heat only without the motor. Say, one at high and another on low. You can them merge them into one device or leave as is.

2 Likes

Apparently Sense discovered my heating element on low and merged it in with the high heat under dryer. Still doesn’t see the motor as part of the dryer. Also found something it called ‘motor 1’ that I think might be part of my heat pump water heater but not sure yet.

1 Like

I just deleted a device that Sense called “motor 1”. It clearly was not a device in my home…Sense device power meter showed it with a very distinctive power meter signature, even when the mains were showing relatively constant power draw. Looks like Sense would try to make sense of “Other” power by frequently calling part of it as “motor 1”.

I now have 9 smart plugs…mostly given up on reliable device detection. Just installed several this week on my refrigerators, ice maker, and wine chiller. Sense could detect these 66% of the time…I wanted better. Over the next week, I need to tell Sense which of the 12 fridge devices it thinks I have are the 4 devices I actually own. (To be fair, the 2 fridges are dual compressor Sub-Zero; so I really have 6 compressor devices in addition the the AC compressors.).

My experience with smart-plugs and Sense has been a mixed bag. My home Wireless LAN is comprised of a Plume wireless mesh, which has mostly been great…providing strong wifi throughout my home, even as I walk from one side of my property to the other side…but I think it is causing some of the UDP packets from the smart plugs to be dropped, so Sense sees some power samples as missing. I need to setup Wireshark on my LAN to see if I can determine why packets are sometimes dropped, as Sense doesn’t handle that well…wish it would.

2 Likes

My Sense has only been up for about two weeks. During that time it has identified the coffee maker on its own. It has managed to detect my Hue lights by reading the data on the Hue Hub and it has a list of things on a Kasa Smart power strip because I told the power strip what was plugged in and Sense read that data. So in reality the coffee maker is all that it has found on its own.

To me this is disappointing. While I know the developers are interested in refining the AI behind this; Personally I just want it to know my devices so I can use the equipment. I can look and the graphs and see my heat pump, hot water heater, micro wave and a few other devices that have very distinct graph signatures. In many cases these things kick on when the house is otherwise very quiet electrically such as at night when we have only a 66 watt vampire load. At those times the power curves for these devices are quite clean and distinct; yet the system does not identify them.

Not withstanding the desires of the development team; as a user I should be able to click on the flag for a large wattage item when it appears in the graph and tell the Sense AI what it is. As a consumer I find it unacceptable that I am unable to achieve my person usage goals because I am forced to be part of a development team. That is not how this product was advertised.

I’d recommend giving this a bit more time, @oasis. I know that’s not the first thing you want to hear, but Sense needs more than two weeks of data to identify the devices in your home.

Here is a good post from the Community taking a look at some of the reasons why we don’t “train” Sense, as you suggested. Our data science team is constantly working to improve our device detection algorithms based on the data we have been receiving from users like you. As time progresses and more Sense monitors come online, the device detection will continue to improve.

While you’re waiting for Sense to model your home energy usage and devices, there’s still a lot you can do through the Power Meter in the meantime to ensure you’re hitting your goals. That link provides a walk-through of how to use the Power Meter to reduce your Always On bubble and minimize energy vampires and large energy loads.

If you have any other questions about some things you can use Sense for while it’s actively building out your home energy models, please reach out. Our Community loves helping new users and we have some amazing content I can share with you in the meantime.

Thank you for your response; but it is the same as what I have read to user requests for a way to tell the system what a device is across your website and this forum. The point you make is always the same. You are working on your detection algorithms, and I should be content to wait however long it takes for the Sense to decide what devices I have. I have seen this comment so often that it is almost a boiler plate response.

Well that all fine and good; but Sense was advertised as a system that would identify and show usage for electrical systems in my home. There are reports on this very forum of people who have gone almost a year without any devices being detected. By the end of a two week install this device should be able to identify large wattage devices that are standards in any home, like water heaters, microwave ovens and heat pumps. These devices should not take months to detect. I can see them in the power meter in the app, and they present as very clean power curves that repeat identically each time they occur. Certainly by now your data base should have enough examples of standard water heaters and heat pumps to identify them if I can by just looking.

With the Solar option this device is almost $500 by the time it is installed. So far, my system has not even identified the power used by my inverter and after all it is a solar optioned device that should be able to detect a SolarEdge inverter and it power use curve.

This is supposed to detect networked devices. Well I have a lot of Nest equipment; all of it connected to my network. All of that is standard Nest equipment that is sending out network signals all the time and the Sense system has not found one single device so far.

Sense was not advertised as an opportunity for me to help you crowd source your AI development efforts. I am happy to help in that; but I would like to be met half way. Give me the option to tell my system about devices that I can clearly identify. Those identifications could be clearly marked as user identified and not placed into your device pool unless you determine that they could be helpful to you AI development efforts. In the meantime you would get specific information about specific devices in specific homes; and the users would get better use of equipment they paid a lot of money to install.

All I am saying is that the user community should be provided a way to identify devices that are so obvious that they can be found with ease in the graphs. If you are afraid such identifications would taint your data; don’t use them. But allow US to use them to make our devices more useful.

Never try to teach a pig to dance … It wastes your time, and it irritates the pig.

If you got one device reliably during your first two weeks, you are doing better than many of us did. As the Sense community keeps saying, “Patience!”.

So, Sense apparently isn’t what you want it to be, and it probably never will be…it fundamentally doesn’t work that way. I suggest that you give up and get your money back while you still can, or sit back, relax, and practice “patience”….lots and lots of it.

@oasis Two weeks is a short time for detection to really establish itself as @JustinAtSense points out. You may well have a more difficult home for Sense success. Many users, myself included, have seen significant advances in detection (disaggregation).

My house is non-standard in so far as in the early days I had to inform Sense, via support, that my hot water heater was not being found. It turns out that it was because my supply voltage is 208V and so the phases are at 120deg vs 180deg. I believe Sense has incorporated this type of low-level recognition into the setup sequence. There are many pitfall permutations (the basic physics stuff of many varied electrical networks) that Sense has ironed out, before you even get to the disaggregation layers of the machine learning (RNN).

I recently did an analysis of my hot water heater detection you may find informative and optimistic.