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

Following our recent Chevy Volt and Bolt EV announcement, we’re excited to announce a new device detector for Nissan Leaf electric vehicles, adding to our list of detectable EVs: Tesla Model S and Model X (with some exceptions due to manufacturing batch), BMW i3, Chevy Volt and Bolt, and now Nissan Leaf. We’re still working hard on a good detector for Tesla Model 3, and hope to roll that out soon.

Please note that not all Nissan Leaf EVs will be detected right away. Initial models have begun rolling out to homes, but further work is still being undertaken. As we gather more data, models will improve and more Leafs should begin to be detected.

The following was noted in the recent Chevy EV update and remains relevant for Nissan Leaf detection:

How often you use a device plays a significant role in its detection. For electric vehicles, if you don’t charge your EV at home every couple days, then Sense may not be able to extract enough examples to train a reliable EV model. For example, if you often charge your EV at work or other parking garage instead of at home, or if you often just “top off” at home rather than going through a full charging process, there likely won’t be enough examples for Sense identify detection.

Still, between the complexities of machine learning and the particularities of individual homes, even with time, Sense might not be able to detect a device. Based on specifics of other devices running in your home, or your unique usage of certain devices, certain devices can be tough for Sense to detect. EVs in particular present a unique set of challenges, due to their long-duration load profiles.

Thanks for your continued patience and feedback as we continue working on device detection.


Looking forward to mine being detected!

A post was merged into an existing topic: Device Detection Major Update: Chevy Electric Vehicles - 10/11/18

A post was merged into an existing topic: Detecting EV devices

A post was merged into an existing topic: Detecting EV devices

How much data does Sense need to recognize a Leaf? Using a Level 2 EVSE, I charge fully 5-7 times a week. Due to moving to a new home, I deleted my old home’s data, so I only have a few weeks’ data at the moment. I’m wondering how long it will take for Sense to detect the vehicle.

By “charge fully” I mean that the Leaf draws its full 6.6kW for at least some time (1/2 hour or more) before it starts its noisy 2-hour trickle-down, and ending with its 3, 5-minute-ish little pulses. I don’t often draw the battery below the car’s Low Battery Warning.

1 Like

My leaf charging at level 2 6.6kW was discovered after about 30 days of data. Charging around once daily. The detector gets the bulk of the charge, but when it starts ramping down, the sense detected device does it a tad too rapidly.

1 Like

I’m 6-7 weeks in, and still no detection of my Leaf by Sense. I charge at home virtually every day. The car is, I’m sure, a big reason that I still have 79% of my electrical usage in the “Other” category. My car’s a 2015 Leaf SL.

My Leaf showed up in Sense last night–woot! This should go a long way towards reducing my Unknown load from ~75% of my overall usage.

Now that my Leaf is detected I can see just what you mean. Sense recorded this usage for last night’s charge:

The recorded ramp-down was very quick–only about 10 minutes.

However, if you look at my full usage graph for the same time period, you can see that the actual usage ramped down much more slowly, and lasted for about 80 minutes longer than Sense recorded, not even including the three ~5 minute pulses that incorporate the end-of-charge for the Leaf. (Battery conditioning? Final top-up? I don’t know what the car is doing, but it’s always there.) I plugged the car in when my home’s heat had already set back for the night, and nothing else large was running; I’d say that less than 200W of this waveform over the entire captured range is non-Leaf usage.

Given how very noisy the Leaf’s end of charge cycle is, I wouldn’t expect Sense to capture it precisely. Could the device model be improved? Sure. But I’m happy that the bulk of the full-rate charge is captured correctly as device usage. This device recognition will go a long way towards reducing my ~75% usage that’s in the Unknown category.


Since discovering my Leaf several months ago, Sense no longer reliably detects charging. I’ve opened cases several times and have not seen an improvement. Is this expected behavior? Is it likely that Sense will begin reliably detecting charging cycles?

I charge the car a few times a week.

Three thoughts:

  • My experience is similar. Both my Teslas have slipped in and out of reliable detection over time. I also saw similar with my AC units back in 2018, though it looks like things have gotten better during the 2019 cooling season.
  • I think the issue is something called concept drift. As Sense adds more EVs to the dataset plus Tesla and others change their charging cycles, detections based on old datasets might get lost, or new data retrains the models leading to less reliable detections. I don’t know whether Sense tries to maintain home specific models or trie to eventually get to home-independent models…
  • I have deleted my Teslas over time and had Sense re-detect with greater accuracy with some reasonable backfill window (a few months).

Given the pivotal nature of EVs in the energy mix (household + global) …

  • relatively high (likely) energy use

  • solar interdependence for optimal charge timing

  • V2G potential (especially since the “G” here is mainly the local household supply)

… it seems like EV charge accuracy should be given higher priority than the (minority) Sense user base might otherwise warrant. Lack of “Ground truth” for EVs and batteries is a potentially significant drag on mass deployment akin to having solar but not knowing how much you are generating.

I also can’t help but imagine a dedicated repurposing of a secondary Sense as a “standard” install where the main or solar CTs could be used for the EV charging and the others could resolve what happens when you throw Powerwalls/batteries into the mix.

1 Like

I’m also curious about this. I’ve had sense for over 6 months now and have charged both my Leaf and Model 3 almost every day since then. It hasn’t detected either despite them being the biggest loads.

My Leaf was identified last summer about 3 weeks after first charge. Charging detection was intermittent for a while. Now, not at all since December. I charge only at home. Have sent several messages to Sense Support.

Hi @neketu. I welcome you to join the Model 3 thread (if you haven’t already) for updates on detection. We’re currently working on our M3 detection following the Tesla firmware update from a few days ago.

Are you using 120v or 240v chargers for your EV, @Spagnood?

I’m using a 240v JuiceBox from enel x.