Detecting Electric Vehicle (EV) devices

device-detection
ev-chargers
#383

I just spoke to the DS team about the difficulties. These are the core challenges:

  1. We are doing real-time detection, so we need to figure out what it is by looking at the start of the charge cycle. With those slow ramp times, it’s hard to detect. So, yes, you could have an IF-THEN that says IF consumption equals, say, 2000W, THEN it’s an EV, but that consumption can take 30 minutes or longer to develop. This is something we’re working hard to refine for EVs specifically however.

  2. The start of the charge is much more variable than, say, refrigerators, which leads to the next point…

  3. We need to build a different model for each different type of EV. They all look different. And these differences can occur across EV models, which is why we don’t detect all Tesla S or X.

  4. EVs charge very infrequently (at best once per day, sometimes once per week, and often this doesn’t even occur at home) so we have much less data to work with.

#384

@kevin1’s model S is actually one of our “officially recognized” EVs. It’s the model 3 that is still in progress.

There’s a variety of reasons Sense may not be picking up on your Volt, but lack of data is the common culprit here. I’d suggest giving it more time. Have you mentioned this to Support at all? I’d be happy to open a ticket on your behalf. It’s possible something simple is blocking detection, though unlikely.

#385

Aha, sorry, thought it was the 3 that was recognized and the S that wasn’t. Had it mixed up.

And yes, I’d be interested in working with support, sure. I’m just surprised others with Volts above reported their cars being discovered literally within days of the roll out, but myself, not so much - despite having 2 of them, so yes, I’m more concerned that something is blocking detection vs anything else.

#386

I’ll shoot you a PM.

#387

For any in the same situation as me (2 identical EV’s) the followup to my ticket was that, unfortunately, this really complicates detection, so it may explain my experience.

Although the tech guys said they would look into things, long story short, it sounds like I shouldn’t hold my breath.

#388

This doesn’t just apply to EVs. I have 2 identical HVAC units and get the same issue. I don’t see any logical workaround at this time.

#389

I’ve been using my 240v EVSE to recharge my i3 for all of the 4 months I’ve been using Sense (and another 17 months before that). During those 4 months I’ve recharged roughly 75 times, but still no recognition. I guess I just need to be (more) patient.

#390

20 posts were merged into an existing topic: Device Detection Major Update: Chevy Electric Vehicles - 10/11/18

#391

Yes, I worked with Support many months ago, but there was not a lot they could do at that time. So I keep waiting, and looking at the app, and hoping. That’s where some of my frustration comes from.

I went ahead and reset the data on my device. I’m curious to see if that will help. I know longer recharges help as well, but that usually means skipping days, so then there are fewer recharging events, and I don’t know where the trade-offs are.

#392

@RyanAtSense this may have been mentioned but just in case it hasn’t I am a new Sense owner have had it in stalled for about 8 days, and it has found about 5 things so far, Nice! (downstairs HVAC, upstairs HVAC, Phillips Lights, Garage door opener, and Fridge).

Bottom Line Up Front : when talking about EV’s based off of what I have read on the community post thus far, it seems like your current sense tool is reading things at a microscopic level when sometimes a human eye level of detail will do, or a telescope level of detail will do. Example Look at a drop of water under a microscope and then look at it with your naked eye, very different views of the same thing

I, like others in this post have an EV, Tesla Model S, in fact I have two, got my wife a Model 3. So like all other Sense and EV owners out there having the sense device identifying the EV’s as quickly as possible, is a high priority. One could argue that the main reason that I got a Sense was because I have converted my cars to EV’s from gas and I need to know how much they are costing me much like we know how much gas is costing us every time we fill up. And your product/company seemed like a good way to see that outside of whats being reported in the car.

please share with your team this slightly different way of thinking about the data that you see, @ThomasBrock has commented (I think that it was Thomas) on how the EV’s are the elephant in the room and they are when you look at the energy draw going from 1000 watts or lest and shooting up to over 11k watts in a matter of seconds (thats what the teslas do) I have nothing else in my house that does that since my heat is gas, and cook top is gas, Hot water is gas.

I look forward to your thoughts on this different way of looking at the data.

Also again,
Tesla’s are designed to be connected to a persons home network and the names are pretty obvious why can’t you guys use that to also narrow down when a EV is charging.

image
image.png1269×1358 32.7 KB

#393

Thanks for the input, Rich. This has been discussed a bit more at length above. While part of the backend work involving EVs does necessitate a lower sampling rate, it’s ultimately not that simple. We wish it were! Many other devices operate within similar wattage and duration parameters, so there’s just not a simple one-size-fits-all approach to detecting EVs.

#394

Are there plans to add Prius Plug in or Prius Prime?

#395

I hope this is the case for I am using a European Myenergi zappi EVSE. FYI, it charges variable to the sun’s output. It is a 240V charger. I have a gen 1 Volt and Model 3. Don’t know if that is information the data scientist need. Neither EV is recognized at this time. Thanks for all the effort to make this critical step happen. I am in it for the long haul.

#396

I understand there are other things that look similar but i was suggesting that you look at that electrical signature and query the network to see if a Tesla_Model_3, or Tesla_Model_S are listed as devices on the network. I think that those 2 touch points could be used to more firmly identify that this particular signature is for a Tesla since it is also showing connected to the network.

In other words the sense signature would be the electrical usage and the fact that a Tesla is on the network.

#397

@richmoba,
Tesla enthusiast here too, with all three models. But given what I know about machine learning, a “presence” signal, like the car being on WiFi, is pretty much useless as an input to training unless the signal is highly time-correlated with both the on and off signatures. Plus, not all Tesla of even the same Model have the same charger type - different signatures.

At the same time, I’m not exactly sure how the humans that are working on new EV models or tuning models know exactly which homes/accounts to look at. So there might be a place for querying devices on the network. Unfortunately, I think that Tesla’s don’t give consistent information as to their origin since the mac addresses indicate other suppliers - 2 Parrots and 1 LG mac my cars.

#398

Sorry about that. I misunderstood. I’ll pass that along to the team to make sure it’s known about. However, it still wouldn’t fully solve the problem of identification. We might (and I emphasize might) be able to see that a Tesla is on the network and thus possibly accounts for that big load visible in Power Meter, but without very precise timestamps, this isn’t all that useful. It would really depend on what messages the Tesla is sending out to the network.

EDIT: Looks like @kevin1 beat me to it by a minute.

#399

You mention that the i3 is a detectable EV. Is that with a specific EVSE e.g. 30 amp vs 16?

I’ve been charging my i3 3-4 times a week using a 16 amp EVSE, for all 5 months since I’ve had Sense… and still no detection. It normally begins charging at midnight, when my other usage is minimal. I would think that would help with detection. Am I just going to have to be patient? Charging my car is about a third of my remaining “other” usage.

Device Detection Major Update: Nissan Leaf Electric Vehicles - 12/20/2018
#400

30 vs 16A EVSE should not make a difference. The signatures we’re looking for are from the car and not the EVSE, which tend to be somewhat generic.

Have you worked with Support on this at all? I’d suggest reaching out to them.

#401

Does your system/model work better with L1 or L2 charging?

I’ve setup our L2 charger for my wife’s Pacifica phev, but instead of sharing it for my Kia Optima phev, I opted to use a L1 charger for my Kia, and a tplink hs110, hoping that helps your data folks figure out a profile for it.

So, does L1/L2 matter for sense detecting our phev’s?

Device Detection Major Update: Nissan Leaf Electric Vehicles - 12/20/2018
#402

We’re actually only working with L2 charging. I believe that’s mentioned earlier in the thread. L1 charging is slated for work after we’re confident with our L2 detectors. But yes, L1s can work well with smart plugs, but just be sure to check the ratings