Yes it does and something I really like about it. I haven’t tried but would assume since it still shows in the app that export would be possible also.
I think @samheidie has done this mod. If I’m remembering the wrong person, apologies.
Whoever did it and however it was done I sure would appreciate a share. Just for temporary use for me, nothing permanent. I had an idea and I’m sure I could make it work but I’d rather use a design or configuration that’s already been proven.
I think he modded 2 HS110s to do a 240V plug for a car charger. Unfortunately my EVs and AC units use more than the 15A limit.
I’ll take the task of putting together a list of the smart plugs in my setup and posting online and link somewhere. Sounds like this could be useful and not that hard to do as I’m in the middle of doing a complete smart home redesign and inventory is part of that.
Understand about the smart plugs boiler plate code. just a dream as it seems like there’s a lot of people in the forums who have greater than average tech skills and would be willing to donate a bit of time and effort for a better experience.
Produce feature prioritization is always a hard task. Probably hardest part of the product managers role to honest ( it’s my day job ). Another reason why documenting an integration framework and open sourcing some boilet plate cold might not be a bad use of time and effort by the Sense team. Empower users to help themselves. Plus the fact that the value of the total solution grows as users contribute their integrations back to the community. :).
I’m running Alexa, GoogleHome, Homekit and others in the house, but I’m thinking the best integration would probably be to put it into something like HomeAssistant if I was going after a smart home. The ability to leverage the communties work and use the power state of Hass discovered devices could be a great hint to the ML Algo’s. (guessing here). Plus, if I REALLY needed it exposed to another ecosystem, HomeAssistant is pretty great at exposing interfaces to all the majors which, again, takes work away from the Sense team and puts it in the hands of the community.
Wanted to ask the original question another way.
Is there any benefit to sense if I buy a plug or two and then use the integration for a short period of time for device detection. I get the limitations around the current implementation, but I’m wondering if running the integration with a specific device for X period of time would help to train the Sense algorithms for future detection of my devices.
If the answer is “yes”, then follow up questions are
How long should I leave the plug in one place to gather enough data for this to be useful? 1 week? 1 month? 3 months?
I don’t have a supported plug currently so forgive me if this is obvious in the integration, but how do I put in the information of the device that’s plugged into the smart plug? Assuming that I can put device type, manufacturer, model number, etc… in there to help label the data.
There’s no way I’m going to buy new plugs for the entire house, but if I can buy a couple and help contribute data back for a better experience for everyone I’m willing to do that, I just need to understand the effort involved and how useful it will be to the Sense R&D team.
Two thoughts Chris, and thanks for your interest.
- There is an informal API into the Sense server-side data collection that a number of people have used, including me. Sense has done it informally since the headcount to document and support, plus the reduced flexibility to change a productized API, are a bit prohibitive to a small company that is still rapidly evolving their core product. But what you are suggesting, the ability add integrations, resides in mainly in the Sense monitor, a place where I would not want to introduce 3rd party code, even with a robust API. Based on some of the issues I have seen, the monitor is the trickiest part of the codebase to keep operating smoothly given all the realtime constraints, plus it’s centrality in the operation of everything.
Here’s the location for the API, and associated Python client on Github. You can also find interfaces built on top of this
- You can also find a user supported interface between Sense and HomeAssistant, but that is focused (I think) on bringing Sense data from the server side, into Home Assistant.
As far as your smartplug questions go, here’s what I know:
There is no known magic length of time specified today that would deem the collected dataset “useful”. Knowing a little about machine learning, the real answer is that you have to run the collected dataset through training with the right architecture of LTSM (a kind of neural network), until the predictive error in the training set flattens out and the error term for the test set gets close to that of the training set. But until Sense figures out the right LSTM, or other neural network for that device type, which probably requires a lot of automated experimentation, they can’t know how big dataset is needed to be useful. A long way of saying “it depends on a lot of factors” today. In the future, it might be that Sense will have an indicator that tells when they have seen sufficient data via a smartplug to consider it “learned”, but I don’t think you will ever see a stated time period, since it could vary immensely.
Sense offers a couple special options on the Device Manage card, plus a special card to allow you to configure info about your smartplug. You add the usual name, model, location info, but you also tell whether you want to control that smartplug from Sense, plus what type of device(s) it is connected to. In the case that you have it connected to a device that has already been detected they have an option for that as well.
@RyanAtSense Finally picked up a Wemo insight plug ( looking at TPLink in the future as well ).
Looking at the interface, why would I choose to connect a device I’ve already detected with Sense to the Wemo Insight plug (or TP Link assuming integration is the same-ish).
Is it purely for having on/off control capabilities within the Sense app or is there some other benefit to having a known device connected to a support Wemo/TP-Link plug?
Obviously there’s the benefit of providing Sense with more labeled data that makes me feel good about contributing back to a solution I’m using and helping make it better for everyone, but trying to Grok what else this is going to get me?
If your detection is already great without using a smart plug, then really the only reasons are what you’ve described: remote control capability and helping to feed us better data. Remote control capability is pretty fantastic though, for certain use cases. I use it pretty frequently to manage my Always On loads, but I also used it a lot in the winter to turn off my space heater remotely (both my wife and I seem to have a serious issue with forgetting about it…)
That’s a great use-case for sure! Just curious - What about cases where Sense seems to be a bit confused about devices?
I have a June Oven and a Microwave that Sense seems to believe is the same device. If I plug one of them into the Insight plug, will this help Sense to understand they are different physical devices?
That would be nice if Sense could learn like that, but in my case not useful. When my hot tub was recognized I was happy to be able to have stats collected for just the tub so I can work on ways to reduce costs.
Only thing is, when I turn on my oven it thinks the hot tub is on. There isn’t a way to teach the Sense that one is on one circuit and the other another. With the pumps and ionizer not being recognized either, it will probably be a while until Sense actually knows my tub.
I didn’t even see a category for hot tubs, just pool pumps. I can’t be the only one spending 1/3 of my electric bill on this luxury.
If the Sense algorithm doesn’t directly benefit from the data immediately it doesn’t matter because it will surely be able to use the data and improve as time goes by. In the meantime you will also immediately benefit from separating out the devices because you can then see and control them.
I imagine the hot tub (heater?) is a watt sucker and probably >1800W so a standard Smart Plug would not be possible. FYI: I have a SECOND Sense that is dedicated to my hot water tank (4.2kW). In the long run I think it will easily pay for itself because it gives me a very clear picture of my primary electrical consumption device … if nothing else it encourages me to take shorter showers. And anyway, in your case, how else do you really implement the Hot Tub Time Machine?