Detecting Electric Vehicle (EV) devices

I have Chevy Volt and Model 3 and am waiting too. I think asking when is an impossible question to answer. I think most of us are aware that the power signatures of EVs are difficult to detect by the Sense algorithms. I would however like a better description of what approach they are working on opposed to the generic response.

1 Like

Mark - I here ya… It drives me nuts how we can identify the “signature” of the device. In my case it appears to step up to 48amps like a staircase if ya will. Why cant I name that signature? It is constant…
I am not that concerned about lights or everyday stuff that is on. I am interested in learning about the consumption of larger devices in my house (EV Charger, Pumps, Stoves, Pool et cetera).
I am gravitating on switching toward the TED device and their use of spider technology where you can put a loop around devices you are closely interested in.
I am getting frustrated. Just my opinion.

Yup. I have a similar frustrating situation with the deep well constant pressure pump feeding my geothermal heat pump and home water supply. It’s VERY distinctive, but from the Sense point of view it’s “noise” that apparently prevents Sense from working very well with the rest of my home too.

Fortunately, I have a WELserver system that does tell me about my total use, solar production, and major consumers of power. Looking at TED, I might well have chosen that if it had been available back when I built my 100% electric home.

Andy - I switched to SENSE when I moved into a new house. I like the interface to SENSE, app et cetera.
I had The TED 5000 when I switched to SENSE. I realize that now that interface is worth 10% to me. Data, Device, Flexibility and accuracy are 90%.
I love DATA and the predictive analysis that can be done with it over time. I am afraid the data SENSE is gathering on what I consider important is falling under OTHER and useless. :frowning:

Sense - Here is my Tesla 48 amp Wall Charger ramping up to charge my Model 3. Why cant this be detected? It is always the same…

The short answer is that just because it looks constant in the Power Meter, does not mean it’s actually constant at the resolution Sense requires to identify devices. You can read a bit more about this here: Why can’t you train Sense?

Also, we went over the status on new EVs in our recent webinar. You can find excerpts from that conversation here:


I think maybe what should be done is train Sense to look at things at a much lower resolution to more easily detect these kind of “large” consumers to more accurately “mimic” human behavior. I understand that at 1,000,000 samples / second things look much different and can look very much alike to each other, and will be needed to detect those much harder to detect devices accurately. That said, for detection of these large and “more easily” detected devices maybe just look at 100 samples/second or even less so Sense can “see” more distinctive lower resolution “shapes” of wave forms, etc that any human can easily interpret as “that’s my EV charging”, or “that’s my heater turning on”, etc. Yes you probably won’t be able to tell at those lower resolutions “that’s my Tesla model S charging”, or “that’s my Honeywell model x heater turning on”, but “who cares” initially? If initially my heater is simply detected as “heater” and then a few weeks/months later it finally determines that it was a Honeywell model x heater that’s good enough for me. But at least for those weeks/months/etc I’ve already known what my heater has used in power.

Based on the responses I’ve seen in this and other detection threads I believe that Sense is currently working the “wrong way around” it’s trying to detect things using data sampled at 1,000,000 times a second, where that is probably not necessary to detect the “larger category” of the device. Yes it’s needed for detecting brands/models/etc, but not really necessary for “this is a heater”, or “this is an EV charger”, etc. It probably is necessary though for things like “this is an XBox One” and “this is a Blue Ray Player”, etc. but the difference between an XBox One and a heater should be pretty clear.

Just my $0.02.

1 Like

Absolutely - the current sub-second “recognition window” is too tiny for detecting things like EV charging ramps that are intentionally kept slow and steady. But Sense’s original strategy was to present the bubble immediately when a device turned on. As Ryan has said, the folks at Sense are now working on a framework for long on-event and off-event detections. The challenge there is that Sense might not flag the on-event until the device is 1-2 minutes into operation, though the time and energy consumed in arrears will also be known once detection has taken place. A second challenge, for those who know about machine learning, is that one has to use “memory” very, very carefully in learning networks. It’s hard to know what incoming data to save vs. what to get rid of when the Sense probe is supplying 4M samples per second (most EV chargers will use both legs of the mains, so 2 voltage and 2 current samples per microsecond), especially when Sense has to look at a minute or two of charging to identify it (240M pieces of complex data per minute to sift through).

1 Like

I hate that I missed the webinar but that was the information that I was looking for. Simply the approach they are looking into, nothing more. Thanks Kevin

1 Like

@RyanAtSense chargepoint has an api also:

I am not sure if that helps you at all or not.

I imagine the DS team knows about it, but I’ll pass it along, just in case.

FYI- I am not sure if it is helpful or not, but I have had a Model 3 Long Range since July 3rd that I charge using a Tesla Wall Connector on a 60a 240v circuit (48 continuous amps).

I throw it out there in case it is useful for Sense data engineers to use my car as another example device to study.

48a is the max charge rate for all currently shipping Model 3’s (eventually the short range one will only be capable of 32a).

1 Like

Nissan Leaf with Clipper creek charger. I have over a year of data on my sense now to help out :slight_smile:

I’m happy to report that Sense found my 2014 Chevy Volt! The detection is still a bit flaky, but it’s there and it’s recognized somewhat, sometimes. The current issues seem to be that the total usage is low by a few hundred watts, and it seems to incorrectly detect that it’s ending the charge cycle when it’s not actually doing so. It’s a start!


Glad to hear your Volt was detected, particularly because…we own 2 of them. A 2011 and 2012. I will admit that I’m a little concerned now moving forward with Sense that this could be a bigger issue for me…especially considering we routinely flatten both cars every day, so 22-25kwh of our daily consumption (the overwhelming majority) is our 2 Volts. I really want the consumption data to be categorized and separated, not lumped into the fray. Arguably this was one of the biggest reasons I bought the Sense, for that matter.

I did expect that given both cars are nearly functionally identical (even using an identical brand EVSE on both) that detecting 1 car vs the other could be difficult, and adding potential further confusion is that we typically charge both at the same time (with both fo them them programmed to start charging at precisely 7PM when our electricity rates go to off-peak) but I guess it all remains to be seen.

I just got my hardware installed last night and am still in the initial detection phase, but I expect to be a regular contributor to this thread moving forward.

Hopefully a welcome update for you Chevy users: Device Detection Major Update - 10/11/18


The good news – my Honda Clarity PHEV was detected on October 12. It accurately notifies when it is plugged in with 100% accuracy so far. It reports as turning off in the middle of its ramp-down at the end of a charging cycle.

Some oddities: (1) After full ramp up, the EV is drawing 7.2kW, but Sense consistently shows 6888W in the Electric Vehicle bubble, and the remaining 300 extra watts or so in Other. (2) When finished, on the “Devices” page, it shows the correct(ish) total kWh that was used (other than being 5% low due to issue #1), but also usually reports that it was only on for 1 second, rather than the hour or two that it was actually on. (3) If I’m charging the EV while also using the stove or oven, sometimes it is faked out when one of the burners cycles off, and erroneously reports the EV charger as turning off (and ends the recording of EV energy usage for that charge cycle).

Overall, despite these oddities, I’m ecstatic to have Sense more or less accurately reporting the energy usage of the EV, as it was a significant percentage of my overall usage.


My Clarity PHEV isn’t detected yet hope it is soon!

This is mostly a copy/paste from my response to the Chevy Volt detection thread, but it’s relevant here so I thought I’d post the content here as well.

So, now that the beta is over (I was part of the program), I can share some information for fellow Volt/EV owners specific to the freshly announced TP-Link HS110 / Belkin Wemo Insight integration. See the new thread specific to this if you’ve no idea what I’m talking about.

The HS110 is confirmed by TP-Link to be dual voltage (120V & 240V) capable, and yes, arguably the Wemo Insight has also been discovered to be dual voltage, albeit Belkin doesn’t (unlike TP-Link) confirm it. The HS110 is rated for 16A@240V, again, as confirmed by TP-Link. The decals indicate (120V / 15A) because they are, as TP-Link describes it, “Labelled to local standards”, but they are indeed 240V/16A capable.

So, I now have 2 HS110’s hooked up to our 240V 16A EVSE’s so I can now independently detect and log the consumption from both of our Chevy Volts. They are both charging right now, see attached image - no more massive “Other” bubble every night!

For anyone interested in more details, respond here and I’ll share as much as I can.

Please take note that this is ONLY valid for up to 16A EVSE’s, which coincidentally, is what the Volt max’s out at for a charging rate (our Gen1’s are actually 15A), but any larger EVSE’s such as the next step up (32A) won’t work as they exceed the amp carrying capability of the HS110/Insight.

For those currently charging on 32A L2’s but who are motivated to be able to break out detection of their EV’s with Sense, if you’re willing to step down to a 16A EVSE (I use a set of the popular <$200 Duosida EVSE’s commonly available on Amazon) there is now a solution at hand that will instantly break out your EV consumption directly in the Sense app for you. Yes, you’ll be charging at half the rate so it will take longer to charge your EV (in the case of those with the more common 32A L2 charge rates), but for those for whom 16A is adequate (IE, you’ll still get a full charge by morning, it’ll just charge later into the night) it’s an option.


Model 3 AWD Performance here; I’ve had it 8 weeks. Just heard (quick reply - thanks!) on my request about recognizing EVs and have now been reading this thread. I guess it takes some time, a lot of data, and an extended “startup window” to be able to identify EV charging. I am using a Tesla Wall Connector on a 14-30 dedicated circuit. Loving the car!

1 Like