Tesla monitoring incorrect power, usage, and misses end of charging

[dolfs]

I am reporting this mostly for the benefit of the sense team, assuming someone is reading this…

So my Tesla MY has been recognized now and I’ve done two charging sessions since. In both sessions during most of it, it reported a max of 9,618W, which corresponds, to a T, with 240V @ 40A. Thing is, I have a 48A charger and actual use is 48A @ 227.65V (avg) or 10,972 W. Somebody suggested sense does not make assumptions, but the 9,618 number is suspiciously like an assumption and it happened twice now.

The first charge session worked more or less correct in the sense of detecting ramp-up and end of session and, were it not for the incorrect power, reported a wrong (but consistent) overall power use.

The second session was different. It reported the same power, but while (in both cases) the first session did, the second session does not report an “off” notification, but looking at the power meter graph, I can see it thinks shut-off was at 03:25AM, which is incorrect. See the two tables below for comparison:

Charging table: actual (directly from car)

Start	End	Duration	kWh	Avg A	Avg V
00:17	04:53	4:36	50.66	47.99	228.66
00:01	04:25	4:23	48.26	48.00	227.75

Charging table: from sense

Start	End	Duration	kWh	Avg W
00:05	04:53	4:48	46.1	9,618
00:00	?	3:25	32.9	9,618

Further inspection of the overall power meter shows a “data gap” from 3:27:00 to 3:27:12 where there is no data whatsoever. In the car specific graph it stops at 3:25:51. While I am convinced that data gap is what caused the “false” end of charging detection, it is strange that the timelines do not line up.

Overall the power mismatch causes about 1 kW/h to be attributed to “other” rather than the car.

So the issues seem to be:

• Assumption about max power use that is incorrect, but identical in both sessions
• End of charging not notified in second session, in the sense that no end notification exists, but power meter shows when it thinks it happened and reported consumption is consistent with that time
• Consequently a very large contribution counted as “Other” in the second session
• Timelines for overall and device specific shut-off do not align (difference of about 2 minutes)
• Data gap is apparently interpreted as power consumption dropping to zero. I would argue that, at least for short drops, a better assumption is a linear interpolation between last before and first data point after the gap. This would then have avoided the incorrect detection of end of charging
• Mismatch between some part of sense thinking charging ended (because power graph goes to 0 and kWh build ends there), and on/off notification system where there was no “off” notification

[kevin1]

@dolfs,

Sense is in-between community managers, in the process of hiring a new one. I suspect you’ll hear more from Sense when they have a new community manager onboard. I will say that your experience in not unlike others when it comes to EV detection:

• I may have been too harsh in my rebuttal of “Sense assumptions” when it comes to EV charger detection level in an earlier post. There are certainly assumptions for the charging power level built into the Sense model for specific Tesla model detection, based on the shape of changing curves. Sense hits both my power level numbers more closely than you are seeing for 48A Model 3 charging and 80A Model S charging. And when I dial down my Model 3 to 32A, Sense changing power level commensurately. But they might not have full variability bail into the Model Y models, or there may not be a substantial charging curve ramp difference between 40A and 48A.
• Did you have an overall data dropout from your Sense monitor during the second charging session ? Just to be more specific, did your main Power Meter go empty during that period as well as data for the Tesla ? I ask because full Sense data outages are rare, and usually only happen coincident with a power outage or when something goes wrong in the monitor (a rarer and rarer event in my experience).
• One of the biggest modes of failure for me, with EV detection is the detection of a premature off, where the Sense power usage drops well before the actual charging ends, especially for the Model S.

[dolfs]

We’ll wait and see. I am not in dire straits, but wanted the sense team to be aware.

Specifically to your points:

• I understand that they may make certain assumptions. I guess what I am suggesting is that if power usage delta between pre ramp-up and post ramp-up should be a rough indication of power used. Just classifying those in “common” buckets for charging scenarios would improve this accuracy
• Yes, there was an overall dropout, although it registered at offset times in the overall power meter and the charger specific one, which is strange to say the least. There was not a power outage though.
• The second cycle I described seemed to be an example of premature off, although if it was indeed induced by the apparent fall to 0W usage during the drop-out I would classify it more as an anomaly. More concerning would be the case where the premature off happens without any apparent cause.