Smart Plugs - need to know

So I have a few smart plugs and detected by Sense. They are currently connected to a bedroom heater and the plug is also called “bedroom heater plug”. There are others. Such has been for +6 months now.

My point is this: Can I now take out the plug and connect the bedroom heater into a wall socket directly and Sense will continue to detect this bedroom heater? When mapping I had mentioned its a heater already and was NOT something already detected by Sense.

Then I can take this plug and plug-in a different device yet to be identified by Sense. Then = how long do I need keep this yet to be discovered device plugged in for Sense to capture data post which I can again successfully map it and move to the next unknown device?

Or that’s not how this works.

If only it worked that way… :frowning:

No, unfortunately, the energy use Sense has been detecting and reporting from the smart plug is only direct data from the smart plug… It doesn’t actually see, isolate and identify what is connected to the smart plug, it just reports what is coming from the smart plug…

So when you remove the device from the smart plug and plug it into a normal wall socket, the device has lost its “communication beacon” and now is indistinguishable from the rest of your usage (unless the on/off signature is picked up by Sense’s ML later).

In fact… Not only does the smart plug not speed up the ML process… But I have a theory that hiding the power signature behind the smart plug may actually remove anything plugged into it from the ML process… I have never had something on a smart plug be discovered/learned by Sense. I have no hard evidence, so I could completely off-base here… Has anyone here had something plugged into a smart plug detected by Sense’s ML?

Anyway… This is why I have 13 smart plugs littered throughout my house now… I never thought I would invest in smart plugs for device level detection because most of the value I get from Sense is in my total production and consumption… But I find it fun and entertaining to see device level consumption, including usage trending… I will never see an ROI/breakeven for the cost of all these smart plugs… But I am a data guy, and I enjoy seeing and observing the data over time…and I guess I am willing to pay for it…

I may buy a few more now that the KS125 is on sale at Best Buy!

It does not work like that yet. Sense is capable of still learning the device while its plugged into the smart plug. However it sounds like sense is currently only detecting the plug that you have labeled as bedroom heater plug. I have seen were other people are saying that sense learned each mode (Hi Med Lo) separately

So its basically useless. Frankly I don’t really care that my bedroom heater is now picked up by Sense. From the perspective of energy monitoring I can use Kasa app to do so. For automation, I can and have linked Kasa plug to SmartThings.

The way I see it is Sense team wants us to purchase these so that they can collect the data. They make no qualms about it in their help site (kudos for being transparent). But if that’s the case they should offer these at a reduced price point to consumers since they are the ones reaping the true benefits, not really the consumers.

That’s my take on it. As @MikeekiM said the cost incurred can never be capitalized.

1 Like

Interesting… Since Sense only detects power transitions between off and on, I wonder if Sense would EVER be able to detect a space heater that went from low to medium to high (and back again)… Since this would not be a power on/off transition, but a intra-device power draw variance…

I am not sure I would necessarily characterize the smart plug integration (or the smart plug itself) as useless… It is nice that there is a way to isolate a device (or set of devices) behind a smart plug and report on the usage/consumption… I have a power strip attached to my home office (which, of course, is a bunch of home office devices including computers, monitors, telephones, printers, etc…). And I love seeing what my “home office” draws during the work week, versus what it draws during the weekend, versus what it draws when I am on business travel…

Some things you buy for ROI and a net positive financial impact… Some things will never result in a net positive return…

But yeah…I think we have all had the conversation that the Sense marketing team oversells and over promises what it is actually capable of achieving… And I am not saying this is wrong…that’s their job… Same reason the Big Mac you get at the counter looks nothing like the photo posted on the posters littering their walls…

1 Like

I don’t see it that way.

  • More, better info from Sense - you only get limited power numbers out of the Kasa app, plus no real history. Sense with smartplug gives you an entire time history in the Device Power Meter - maybe not so important for an on/off heater, but much more useful for things like electronics. Sense also gives you the ability to pull out Always On data, plus set 3 states for the devices on smart plugs - On/Standby/Off.
  • Not every device is suited for detection - If you are a machine learning guy, @baivab.mitra , then you’ll understand that some devices (mostly electronics) have behaviors that are too random to be detected using waveform changes on the mains - things like TVs, servers, computers, when power is a function of what is being “played”. Smart plugs are a great way to track and account for these devices.
  • @MikeekiM is a little bit off - I have seen Sense detect devices, or components of devices that are on a smartplug, get detected.
  • Plus, as a machine learning guy, @baivab.mitra, you should understand how ground truth works for more generic models - Sense doesn’t do analysis and train on your smart plug data to create a home specific model. They have to generalize, which means leveraging data on lots of similar devices.
1 Like

Yeah…that doesn’t surprise me… As I said, I have no hard evidence other than my own personal experience that nothing on my smart plugs have ever been detected…

But I am open to the belief that it is possible… :slight_smile:

1 Like

Okay. So the topic of Sense being able to isolate based off the device the plug is connected to, is a moot point.

Coming back to what @kevin1 wrote i see his point. For example I didn’t know these advanced electronic devices behave the way he described. In which case having a smart plug capturing that device makes sense. Permanently.

Where I do differ though is let’s say my TV is thus connected. Then for a period of a month Sense is able to capture all the troughs and valleys and other innuendoes for the TV. I believe it ought to be enough to now identify the TV in isolation too, would it not? i.e. taken off the smart plug? Or that’s exactly why Sense recommends putting such plugs for their data scientists to come up with revisions to their identification logic? The final part also then clarifies i.e. with many such smart plugs in play connected to say similar type of device (TV) allows Sense team to generalize. That begs me then to ask and perhaps I am 99% certain Sense must be having an Azure/AWS cloud component where the processing happens. As such processes are pretty intensive with CPU/RAM. Did anyone ever tear down any unit to peak inside? Alternatively, the AI side is all done elsewhere and only dummy programs which are an outcome are loaded with configurations/captive DB into Sense and Sense works locally off of that.

Three answers:

  • First off, TV’s are highly unpredictable depending on the video content they are playing. Take a look at this little experiment and explain to me how Sense could detect the changes on the mains caused by my TV as a single device.
    Sony 77” 4K OLED TV
  • Sense does run a backend at AWS, but data volumes dictate that a lot of the processing of the microsecond samples go on within the monitor. If Sense uploaded all its incoming data to AWS at 4M datapoints per second (from tech spec), at 14 bits, that would be slightly more than 20TB per month ! Given the real data usage, Sense is probably sending something like 1/10th to 1/2 second data back to AWS, along with heavily processed data on the transitions of interest.
  • Someone has done a teardown on an early Sense monitor. The article gives a pretty good view of functionality - basically built for real-time data acquisition and heavy DSP processing.
    Confused Life - Reloaded: Sense Energy monitor teardown - sampling in MHz
1 Like

This ties in to my other post response w.r.t. battery backup. What was told is Sense power management unit has quite a bit of safeguards, etc. And the tear down does point to the fact the power supply is very custom.

The data being received/sent was bothering me. They can’t be sending too much for it’ll seriously blow-up their cost. As there’s no subscription service with Sense meaning the processing must be local. That’s really then a very complex piece of hardware capable for such almost real-time DSP processing and rendition.

1 Like