Five Things to Know About EV Detection

Sense EV owners, ever wonder what Sense does to detect, or possible miss your car charging ? Here are some user-discovered details, that may surprise you.

  • EV Detection is different - It uses different techniques from Sense’s main AI approach, because EV charging ramps are very different than the 1-2 second on/off transitions that Sense looks for and tries to classify. Charging ramps typically take multiple seconds to even minutes and often have unusual curves as seen in this blog.

Even though YOU can “see” the ramp-up and ramp-down curves in the Power Meter, Sense’s main AI built into the monitor CAN NOT. That, plus several other factors therefore make it hard for Sense to see the same thing you see, even though it’s a no-brainer for you.

  • EV Detection doesn’t use most of the million times per second data - Since the built-in detectors in the Sense monitor we’re designed to “see” EV ramps, or at best only see incomplete portions of them, Sense has to detect the ramp patterns using the half second power/energy usage data we see in the Power Meter, not the millions of times per second measurements the monitor is working with. That seems like it should still be easy to spot the on and off ramps EXCEPT:

  • Ramps are very different between EVs - Both shape and length of ramp-ups and ramp-down are very different between different EVs. In other words, both the pattern and more importantly, the length of the pattern Sense needs to look for, can vary widely for different cars, even between different models with the same make and even between the same model from different model year. This means that detailed detectors needs to be developed for each EV model, and possibly for each EV model year. That limits the number of custom EV detectors that Sense is likely to be able to support.

Here are two examples from my household.

Our Model 3 ramps up in 13 or so separate steps that take about 10 seconds, well beyond Sense AI’s viewing window.

Our Model S has an even longer ramp period to almost 12 kW, but had an even more difficult to detect rounded tail-off that lasts 15 minutes or so followed by a drop.

The Sense blog on EVs contains more of the EV charging variety.

Kudos to @JamesDrewAtSense for providing the currently supported list of EVs:

  • Tesla Model S, X, and 3
  • BMW i3
  • Chevy Bolt & Volt
  • Nissan Leaf
  • Chrysler Pacifica

Right now, Sense does not detect the following types:

  • Fords
  • Audis / VWs / Porsches
  • 120V charged EVs.
  • Ramps and Detection Depend on Charge Rates and Voltages - As that last bullet points out, there is NO detection for 120V charging. All EV detections are built to look for L2 / 240V charging ramps. Since charging ramps can also depend on charge rates, there’s a chance that setting a charge rate far different from the typical rate used for the Sense detection might also elude detection.

  • Ramps can vary based on software version - So the Sense detection models have to be regularly updated when this happens, or detections stops. As a Tesla owner, I’ve seen these several times, thankfully Sense has updated the specific detections as required, and backfilled the detections for a month or two. But regular changes likely limit Sense’s ability to scale EV detection across a wider variety of EVs.

in BOLD the EV’s that are currently NOT detected by Sense from the top 20 of 2023

Best Selling EVs 2023

  1. Tesla Model Y: 394,497 Units

  2. Tesla Model 3: 220,910 Units

  3. Chevrolet Bolt EV/EUV: 62,045 Units

4. Ford Mustang Mach-E: 40,771 Units

5. Volkswagen ID.4 37,789 Units

6. Hyundai Ioniq 5: 33,819 Units

7. Rivian R1S: 24,783 Units

8. Ford F-150 Lightning: 24,165 Units

  1. Tesla Model X: 23,015 Units

10. BMW i4: 22,583 Units

11. Kia EV6: 18,897 Units

12. Rivian R1T: 17,727 Units

13. BMW iX: 17,301 Units

  1. Tesla Model S: 16,466 Units

15. Mercedes-Benz EQS: 15,510 Units

16. Mercedes-Benz EQE: 14,895 Units

17. Nissan Aryia: 13,464 Units

18. Hyundai Ioniq 6: 12,999 Units

19. Polestar 2: 12,215 Units

20. Audi Q4 e-tron: 10,750 Units

In short: 5 of the top 20 are detected, and yes number wise they include the top selling.

1 Like

I’m guessing that the team has had to go back to the drawing board to come up with EV (and other slow ramp) detection that is more scalable.

1 Like