I don’t have an EV (yet). I’m thinking about it though, certainly since we are getting solar soon.
We did solar and EV in unison as well. I can honestly say I can’t imagine purchasing an ICE car again.
EV’s appeal to geeks, early adopters, the environmentally conscious, and the financially savvy…with disposable income.
I strongly suspect that Sense appeals to many of the same groups.
I don’t find the correlation surprising, honestly.
On another topic, I have not yet seen either of or Volts get detected yet. Patiently waiting.
How about a conversion, VoltsRabbit conversion done in 1994, with Zivan K2 120V charger? I’ll bet Sense users don’t have many conversions, but that was the main thing going on 15-30 years ago. However, mine is only charging to top off the batteries once a month until I get around to fixing the balky main contactor. Another charger I have been known to use rarely is an auto-transformer (variac) with bridge rectifier. I’ll bet Sense will never identify these chargers
Another vote for model 3. I believe this should be an area of stronger focus for Sense, because as others have pointed out, there are many sense users who also have an EV and the EV is going to be one of the primary loads at the house and thus is key to realizing full utility of the Sense product. In my case 78% of my usage is unknown and the great majority of that unknown is charging the model 3.
One more thing, The link shows EV sales for 2018, it supports my desire for Model 3, but I think also likely helps create the overall priority list for EV detection when you look back in time at EV adoption.
Maybe Sense has already done this research or better…
I think we all want our EVs to be discovered, as I’m sure Sense is aware. One month of sales does not make a sample though especially if production just kicked up. All time Volt (and bolt) was the best selling all time it makes sense to focus on that.
Personally I hope they are looking at the all time and going after number 2…
- Leaf owner
EVs are definitely a strong focus for us, but they’re incredibly challenging devices for a few reasons (one of those reasons happen to be a lack of data, despite their growth in popularity). On top of that, they all look different so each one needs a custom detection model. This is a lot of work and is only compounded when we don’t have buckets of data like we do for refrigerators or ovens. While the model 3 may be surging in popularity, we still don’t have many online in our userbase, and without data from wide-ranging contexts, we can’t crowdsource detection models quickly. It’s also not a matter of getting a few model 3s in the office and testing them (though, can’t say I’d mind that…). We need to see them in real-life contexts, with real-life patterns of usage, and we need to see these patterns many times over. This is mostly true for all devices, but especially so for EVs.
And like @oshawapilot noted, Teslas in particular are complicated. We’re pretty good at detecting S and X, but but not all S and X due to subtle changes in the load profile over the model’s history. It’s a tough nut to crack, but we’re working on it.
Totally agree that it’s not just 2018 sales to look at!
How do you guys know how many Model 3’s you have in your database, given the difficulty in detecting them? As another person waiting for this detection, I would love to have my account flagged to test any Model 3 models against.
I’ve seen you mention this before in previous responses, as well as mentioning that the long duration/heavy draw signature also complicates detection.
Can you explain why they’re so hard to detect?
From my perspective, simply watching the live display it’s blatantly obvious when someone plugs in one of our Volts - the wattage climbs quickly but is noticeably stepped as it ramps up to the ~3.3kw our gen Volts charge at. The entire duration of the charge seems very steady. Towards the end of the charge a notable ramp-down can be easily seen on the live display, slowly but consistently ramping down to somewhere in the 1kw range, and then it suddenly kicks off.
This “signature” on both the ramp up, duration, and ramp down seem like it would lend itself towards easy detection, not difficult detection.
What am I missing?
At least on our Volts every charging cycle seems to follow that exact scenario in both cars. The only exception is if the car is charging but incomplete and someone unplugs the car to drive somewhere, in which case the consumption goes from ~3.3KW to zero in a split second.
Excited for additional EV device detection. I charge our Ford CMax Energi over 110V daily and would love for the car to be discovered.
I am using a Seimens 240V charger to charge a Volvo XC90 T8. I did some basic math by watching Sense see the “other” category go up. I was kind of shocked to see how much power it uses but its stilll nice to know its something that may be able to be detected at some point.
My Tesla Model S 2016 is not detected. I really wish it was.
If it’s useful for reference, my 2018 Chrysler Pacifica Hybrid (on a Siemens lvl2 charger) was detected a few weeks ago, seemingly around the time this update came out.
I have been patiently waiting almost 2 years for Sense to recognize my Tesla Model S and it still doesn’t
Have you worked with Support on this at all? There are so many factors that go into device detection that it’s tough to say from afar what might be blocking it. I do know that we don’t work with all Tesla Model S as the signature differs depending on release date and certain release dates are tougher to track than others.
Yes I have and I just emailed them again.
+1 for detecting Tesla Model 3.
I would also like to know why it’s so difficult to detect an EV. In my case, I use the Tesla charger @240V 40A. It always starts charging after midnight. Is there anything else that generates a 10KW load for hours at a time at midnight? Seems like the complex machine learning algorithm could use a simple if statement.
[quote=“petersen.mc, post:46, topic:3643, full:true”]I would also like to know why it’s so difficult to detect an EV.
Seems like the complex machine learning algorithm could use a simple if statement.[/quote]
Funny, I was thinking the exact same thing last night.
If consumption = (typical make/model of EV L2 charger draw) then device=EV, else=Other
The on and off signatures from our Volts are also consistent, and for my human eye at least, reasonably easy to pick out.
Here’s one of our 2 Volts starting a charge cycle.
And here’s the end of the charge cycle, also quite an identifiable signature.
Between the 2 events, it’s a steady 3.14KW draw, and on the downward taper it kicks off at 1.15W.
I too am stumped why this is hard for the sense ML to pickup on. I’m left with both of our Volts undetected still despite both consuming ~>10KW per day each.