Charging in High Temperature environments 13A vs 32A vs 70A: cost of air conditioning

[markwj]

I submitted the following bug report to Tesla today:

I'd like to report a bug in the Charge Session History screen on my car. Not sure if something is broken, or if this is a general bug affecting everyone.

I have a separate meter on my charging station, so can see what goes into the car, as well as what the car displays.

I recently did three charges:

06/29/2011 18:28:02 GMT+8 @70amp ~30celcius - car showed 9kWh meter showed 8.5kWh
06/30/2011 22:36:48 GMT+8 @70amp ~30celcius - car showed 18kWh meter showed 17.0kWh
07/02/2011 14:03:20 GMT+8 @13amp (HPC dialed-down) ~36celcius - car showed 36kWh meter showed 16.2kW

As you can see from the screen picture, today the Charge Session History is showing 10kWh (0:54 duration), 9kWh (0:54 duration) and 36kWh (7:51 duration) respectively.

I notice two issues:

1] The first and second charges originally showed (ie; the day after) 9kWh and 18kWh respectively. Today, they show 10kWh and 9kWh (as on the attached screen picture). The information shown has changed. Note that the actual meter usage was 8.5kWh and 17.0kWh so the original values shown match much better than that shown in the history today.

2] The figure @13amp for July 2nd is way off. That was a 13amp charge with the meter showing 16.2kWh usage. The Charge History shows 36kWh. Even at the maximum of 13amp x 220volt x almost 8 hours that would be 22.8kWh (it should take 12.5hours at 13amp 220volt to consume 36kWh).

I have uploaded the logs to upload.teslamotors.com and I would appreciate it if you could forward this on as a bug report and let me know if you need anything further.

Thanks, Mark.

P.S. The figure shown for Jun 26th (35kWh) is for the 70amp charge after driving the 200km around Hong Kong. The previous charge was range mode, and that was a standard mode charge, but it cannot be that a 13amp charge for a 40-50km drive uses more power than a 70amp charge for a 200km drive.

 

[markwj]

Tom (thanks) did an analysis of the logs for me, and his summary is:

06/29/2011 18:28 (00:53:35) 73%->84% 70A 9.0 kWh 40 km 160 Wh/km 228 Wh/km
06/30/2011 22:36 (01:27:43) 62%->84% 70A 17.8 kWh 81 km 160 Wh/km 226 Wh/km
07/02/2011 14:03 (06:22:56) 69%->85% 13A 18.3 kWh 51 km 173 Wh/km 376 Wh/km

So I used about 160-173 Wh/km while driving. While charging at 70amp I put in about 228Wh/km driven, and while charging at 13amp I put in around 376Wh/km driven.

My analysis:

1] Toms log analysis pretty closely matches up with what the meter shows.
2] Given the estimate for the aircon at 2-to-3 amps @400volts, the aircon is presumably using around 1kW while running continuously. at 13amp 220volt we would expect that to extend charge duration by 50% or so.

Bugs in Charge Session History:

A] The Charge Session History figures may change when looked at on different days.
B] The Durations and "End Of Session" shown on the Charge Session History are inaccurate.
C] The "Energy Use" shown on the Charge Session History may be inaccurate at times.

I think this has shown that the Trip Counter is pretty accurate and clear, that charging in high temperature environments is going to have an impact on efficiency (you are always going to have to put in more charging than you took out driving, but high ambiant temperatures have a big impact on this), and that charging at low currents (eg; 13amp) vs high currents (eg; 70amp) in high ambiant temperatures appears to give a further 50% hit on efficiency.

 

[Doug_G]

That would suggest that higher power charging is more efficient because the air conditioning is running for less time. But that also implies that the battery temperature will be much higher during the higher current charge cycle. Depending on how much higher that may not be a good thing for pack life.

 

[markwj]

That would suggest that higher power charging is more efficient because the air conditioning is running for less time.

Also, because at 220volt 13amp, the aircon takes up about 1/3rd of the available power so further lengthening charge duration.

But that also implies that the battery temperature will be much higher during the higher current charge cycle. Depending on how much higher that may not be a good thing for pack life.

I don't have any hard data, and don't have a plot of pack temperature during such charges. From my understanding, pack temperature is much more of an issue during discharge (ie; driving) than during charging. If the aircon can keep it under control during driving, it should have no problem maintaining an adequate temperature during charging.

Maybe Scott or Tom have more info on this.

 

[markwj]

OK, I've got the wall meter in place, and now can get some good data points. All are around 30celcius (+/- 2celcius).

27 July 2011 6kWh net energy used by car, 70amp charge used 9.5kWh from the wall (car said 10kWh)
29 July 2011 21.95kWh net energy used by car, 70amp charge used 29.0kWh from the wall (car said 31kWh)
1 Aug 2011 7.15kWh net energy used by car, 70amp charge used 10.3kWh from the wall (car said 11kWh)
2 Aug 2011 5.64kWh net energy used by car, 13amp charge used 14.5kWh from the wall (car said 27kWh [bug!])

[ Remember: don't trust what the car says for charge history (especially if you are dialing-down the current on an HPC) ]

The three 70amp charges come out at +58%, +32% and 44% 'inefficiency' burden.
The single 13amp charge comes out at +157% 'inefficiency' burden.

Observations:
1] The 21.95kWh net energy used day was the most efficient charge. Seems to be the shorter charges are more inefficient.
2] The 13amp charge was dramatically worse than the 70amp charge. Seems to match Testsous's observations.

I suspect that the inefficiency is caused by the high temperature (having to run aircon), and is made up of a fixed component (per charge pre/post cooling/balancing the battery?) and a variable component (per minute of charge). I guess with more charge points, we could work out the two factors (but they probably depend on temperature).

I'll continue to gather data points (in particular at 32amp and 13amp). I'm also interested in seeing what happens if I refrain from charging for a few days, letting the battery drain down 6kWh per commute, and then do a big charge to see the resulting efficiency.

 

[tetsuos]

Hi Mark,

This is very helpful... it precisely matches my observations from 13A charging at 220V. Also, your note that larger percentage charge sessions are more efficient makes sense given some of the previous charge graphs I have seen, where towards the end of the charge cycle, the current usage drops quite a bit (the graph flattens towards the top).

 

[markwj]

3 Aug 2011 5.58kWh net energy used by car, 32amp charge used 7.2kWh from the wal (car said 35kWh [bug!])

That 32amp charge comes out at +29% 'inefficiency' burden.

One concern with these tests is that I'm assuming the end charge SOC is the same. However, that is not always the case. In particular, after a 13amp charge, the SOC seems higher than after a 70amp charge (makes sense, and expected). For long charges, this doesn't make much of an impact, but for short charges the error can be significant. The solution is to keep charging at the same rate for several days running, and ignore the first charge after a rate change (assuming the charges all start at approximately the same time of day).

Interestingly, the attached screenshots show the charge session screen bug nicely. The first one was taken on August 3rd morning and shows the previous afternoon's charge on August 2nd at 27kWh. The second one was taken on August 4th morning and shows the previous afternoon's charge on August 3rd at 35kWh. You can see the August 2nd charge has magically disappeared. I guess it got merged into the August 3rd one (as 35-27 = 8kWh which is approximately what the wall said).

 

[Doug_G]

I suppose I could add a feature to the Graphical Log Parser to calculate the integrated power over a charging session... would eliminate the VDS screen from the equation.

 

[tetsuos]

1 Aug 2011 7.15kWh net energy used by car, 70amp charge used 10.3kWh from the wall (car said 11kWh)
2 Aug 2011 5.64kWh net energy used by car, 13amp charge used 14.5kWh from the wall (car said 27kWh [bug!])

[ Remember: don't trust what the car says for charge history (especially if you are dialing-down the current on an HPC) ]

Hi Mark,

I suspect this might not be a bug. The charge history is on a per-day basis. It is likely that your Charge History on August 2 is an aggregate of both Aug 2 and Aug 1, if the session on Aug 1 extended past midnight. If there is a +10% variation between the reported charging usage and your wall meter, you would see the following:

Aug 2: 14.5 kWh * 1.1 = 15.95 (round up to 16 kWh)
Aug 1: 10.3 kWh * 1.1 = 11.33 (round down to 11 kWh)

16 kWh + 11 kWh = 27 kWh (i.e. your reported charge history on Aug 2)

Am I right?

 

[markwj]

I suspect this might not be a bug. The charge history is on a per-day basis. It is likely that your Charge History on August 2 is an aggregate of both Aug 2 and Aug 1, if the session on Aug 1 extended past midnight.

The August 1st charge was at 70amp. No way it extended past midnight.

I'm seeing two bugs: (1) the numbers are just plain wrong, (2) previously recorded dates and numbers change overnight.

There are two possible causes of the bugs:
1] Charge sessions go over midnight then they get treated as all the 'end' day rather than the 'start' day, or more likely the treatment changes with time. Perhaps time zone related?
2] Something to do with changing the power draw (dialing down the charge).

Tesla engineering have acknowledged the bug to me and say it involves calculation of charge duration (whatever that means).

Even if the charge went past midnight, and they book it to the 'end' day, it should have stayed that way. You can see it changed its mind overnight ;-) Firstly it is shown as Aug 2nd, then it is merged in with the Aug 3rd traffic. Perhaps if the charge took > 24 hours that could be the case, but definitely not that long.

 

[markwj]

I suppose I could add a feature to the Graphical Log Parser to calculate the integrated power over a charging session... would eliminate the VDS screen from the equation.

Doug,

This is something like what Tom has done for me a few times. It seems the most accurate way of calculating this. He calls it "Wh per ideal mile (or SOC %)". The figures he gave are earlier in this thread. But, at the time I couldn't match them up against times and current shown by the car without a meter on the wall (which I now have).

 

[Doug_G]

You're exactly right, markwj. In my experience the VDS charge history is, well, totally whacky. It loses things.

 

[Slackjaw]

I'm coming a little late to this thread but thought I'd mention that I did what may be a "worst case" test last weekend during a road course training day at Englishtown race track.

I talked a little about the charging over here in this thread but the summary is that I was drawing 24A @ 220V and getting about 10 ideal MPHC instead of the hoped-for 20. That seems to imply that 12A (==> 2.6 kW) was being used by the A/C. Or, possibly, that the high temperatures were making everything less efficient? The ambient temp was around 100F/38C and I was charging a black car in direct sunlight after coming off the track in Power Limit mode. My numbers may be a little off in terms of SOC so I'll try to get some better data using Doug's latest log parser.

 

[slcasner]

The August 1st charge was at 70amp. No way it extended past midnight.

I suspect that the combination of charges may not have anything to do with midnight, but rather just whether or not the car has been driven in between.

 

[markwj]

I finished my testing of charging the Roadster in hot climate at different charging rates (to answer Tetsous's question regarding 13amp charging).

I charged my car at around 31celcius every night for a couple of months. I had to cut the last test (32amp) short, as the weather suddenly got cooler.

Here are the results:

So, at 70amp, for every 1kWh we take out driving the car we have to put in about 1.5kWh when charging. This increases to 1.55kWh at 32amp and 1.74kWh at 13amp. These figures also include losses due to charging energy conversion, the car sitting idle when not charging/driving and the battery discharging itself, battery balancing, aircon and fan use while charging, etc etc.

If I drove the car a lot more (rather than normal commute, night out and the occasional airport run), we'd see the charge losses fall. Most efficient would be to drive the car whenever it is not charging (but I'm not a taxi driver).

End result is that I'd save around 15% on my charging bill by charging at 70amp vs 13amp. 32amp would be about the same as 70amp.

My advise to other roadster owners in Hong Kong would be go for a 32amp solution. Easy to install, fast enough for normal use, and only about 3% more on the electricity bill (more than saved by the reduced installation costs). 13amp is cheap and simple to install, but the charge durations are really too long and you are paying 15% more every time you charge.

This pretty closely matches what we've seen by looking at the logs from the car. It would be interesting to re-run this during the winter months, but it takes a lot of discipline to not charge anywhere else but on the meter at home ;-)

Hope you find this useful.

P.S. My numbers (particularly percentages) may be off, but the raw figures are right. Sorry, but its late here and it has been a helluva week.

 

[VolkerP]

great work, Mark! Will wire my garage for 32A (400V three phase, to be exact). Awaiting my Model S in early 2013.

 

[tomsax]

Nice work, Mark! Your numbers show an efficiency penalty for lower charging rates at higher ambient temperatures that is more pronounced than the results I found at more moderate temperatures. I should do a 240V/12A to add to my data set.

Did you record ideal miles before and after the charge? I believe ideal miles is a better measure for this because not all kilowatt-hours drawn from the battery pack are the same. Although ideal miles haven't been found in the log files, you can extract SOC %. It would be interesting to see how using those numbers affects the outcome.

What the battery really stores is electrons, measured in amp-hours. As the battery discharges, the voltage of the pack drops, so an Ah at the top of the battery produces more energy (kWh) than one at the bottom. The same thing happens when charging the battery: it takes more energy to put an Ah in at the top than it does at the bottom. Obviously, Tesla takes this into account when it computes the state of charge. If that was the whole story, kWh would be a perfect way to track driving and charging efficiency, but it's not.

When you take electrons out of the pack at a low rate, the pack voltage doesn't change much, and the above model is very accurate. However, as you increase the rate, the pack voltage drops so an Ah drawn out of the pack with the pedal to the floor produces less energy than one drawn more lightly. Said another way, drawing 1 kWh from the pack may reduce the remaining pack energy by more than 1 kWh, more so if you're really romping on it.

Here is a graph of my personal best 1/4-mile drag race run, 12.978 seconds ET at 103.40 mph:

The accelerator pedal is on the floor from go until I pass the 1/4-mile sensors, then I lift off the pedal and transition to maximum regen and hold there to the end of the graph. As the red current line rises, the green voltage line drops from 412.5V at 2A just before the graph starts to 336V at 637A (214 kW). After the run, looping back into line at 15 mph, the voltage holds very steady within two volts of 410 while drawing 5A to 15A. Voltage can be thought of as Wh per Ah, so my energy efficiency pulling power out of the pack drops to 82% at peak power. Said another way, every 1.0 kWh I pull out at maximum power removes as much charge as 1.22 kWh of driving at low power. When I charge the battery pack back up, I have to replace the 1.22 kWh, not just the 1.0 kWh I got out. So if I did a charge efficiency measurement at the track, and also under otherwise identical circumstances after a calm low-speed drive, my charging efficiency would look worse at the track only because of the difference in driving.

At a lower state of charge, the effect is larger. Here's another 1/4-mile run, this time at 41% Range Mode SOC (but driving in Performance mode):

In this (noticeably slower) run, the voltage drops from 384V at 1A to 300V at 639A (192 kW), or 78% efficiency or 1.28 kWh removed from the pack for every 1.0 kWh used.

By reading the SOC in ideal miles, you get Tesla's best estimate of how much energy is in the pack. By subtracting you can figure out how much charge you took out or put in, which is different than the usable portion of the energy measured coming out of the pack (some or all of the difference is lost as heat inside the pack). To get the most accurate reading, you should wait before reading the value after a drive or charge segment until the battery pack has had a chance to settle so the car can determine an accurate reading. This is typically about 10 minutes, so I usually wait 15 to 20 minutes when I want an accurate reading.

From the information here, the variation of charge/energy efficiency can be at least 28%, perhaps more at a lower SOC, but that extreme drop can only happen for a few seconds at a time. Still, a difference in day-to-day driving (a few extra stop/starts, pushing to get through a light, passing a slow car, etc.) can make different days have different charge/energy efficiencies. Drives that dip deeper into the pack will be more affected by differences in driving conditions.

 

[markwj]

Tom,

After charging at 13amp, the car shows about 302/303km ideal. At 32 and 70 amp, it shows about 298km. Both are 'morning after figures', and very consistent. For a single comparison charge, this would be an issue, and that is why I ran the tests over several days / weeks (so that the first day SOC difference at different current levels would have little impact on the overall result).

From our tests using the log file data, I was happy with understanding the effects that high temperatures and AirCon have on charging at 13/32/70 amp. But, as we know AirCon cooling is only one of the losses we have from using our car. With this test, I was interested in getting a real-world estimate of what those might be, and what percentage of those losses the AirCon is. A wall to wheel estimate.

The results I got are for me, my driving/charging patterns, in my environment and at the temperatures of a typical Hong Kong summer. I suspect efficiency is very tied in to how much you drive.

The losses we see are made up of many factors, including direct charging efficiency losses, AirCon energy use, and pack energy loss from the car sitting idle (both plugged in after charge, and unplugged at work). The AirCon is undoubtedly a fairly large hit, but less than the others. The surprising result was how close 32amp and 70amp were.

I definitely intend to repeat the test later in the year (when temperatures in Hong Kong drop to their winter norms around 15 celcius). It would also be interesting to see what would happen if I changed charging habits - such as a) leaving car unplugged at night and charging just before I drive to work, b) not charging every night, or c) unplug after charging at night.

But, winter is coming and temperatures are fluctuating too much now to get a reliable result.

For an individual car, real world, this makes so little difference. Over the month of the test, I used around 330kWh from the wall, and a 15% efficiency difference on that is somewhere around the price of a cup of coffee. The cars are so ridiculously cheap to power (about 1/4 the cost of gas for a Prius, here in Hong along), that it really makes little difference for 1 car. But, 15% over a fleet (or country) of cars - that is something worth investigating.

 

[hcsharp]

Tom,

This data is really useful and I want to thank you for your time sharing it! What I find interesting is that the maximum amp draw is the same regardless of SOC. I've been trying to figure out why the battery heats up faster at lower states of charge than at higher. Your charts make it clear. I can see that this will be even more of an issue with freeway driving or any time you are demanding the same performance from a lower voltage battery. At least at the track your car slowed down!

To stay on topic, I'm trying to extrapolate how this might affect charging efficiency. Sometimes I wait 2 - 3 days before charging if I don't drive much with the theory that storing the battery at a lower SOC (but more than 40%) is better for longevity. But now I wonder. If I drive exactly the same speed and acceleration every day then the days with lower SOC will heat the battery more which probably removes any advantage of a few hours storage at lower SOC.

As for charging, I think the best practices will be different in my climate where we go below 0 deg F (-20C) for a lot of nights every winter. But generally it looks like higher amp charging will be easier on the battery and more efficient when it's at a lower SOC than at a higher one.

 

[measton]

Mark, is this energy usage from start of charge to end of charge or over a fixed time

If I drive the car in 12 hours and it takes 12 hours to charge with the 110v 13amp and 3 hours with 240v set up are you calculating energy use over 12 hours for both or over 12 hours for the 110 volt plug and 3 hours for the 240 volt. This would make the efficiency of the 240 volt look much better as the car burns energy even when it is not charging. You need to remove energy used by the computer from the equation to figure the efficiency of charging.

I finished my testing of charging the Roadster in hot climate at different charging rates (to answer Tetsous's question regarding 13amp charging).

I charged my car at around 31celcius every night for a couple of months. I had to cut the last test (32amp) short, as the weather suddenly got cooler.

Here are the results:

So, at 70amp, for every 1kWh we take out driving the car we have to put in about 1.5kWh when charging. This increases to 1.55kWh at 32amp and 1.74kWh at 13amp. These figures also include losses due to charging energy conversion, the car sitting idle when not charging/driving and the battery discharging itself, battery balancing, aircon and fan use while charging, etc etc.

If I drove the car a lot more (rather than normal commute, night out and the occasional airport run), we'd see the charge losses fall. Most efficient would be to drive the car whenever it is not charging (but I'm not a taxi driver).

End result is that I'd save around 15% on my charging bill by charging at 70amp vs 13amp. 32amp would be about the same as 70amp.

My advise to other roadster owners in Hong Kong would be go for a 32amp solution. Easy to install, fast enough for normal use, and only about 3% more on the electricity bill (more than saved by the reduced installation costs). 13amp is cheap and simple to install, but the charge durations are really too long and you are paying 15% more every time you charge.

This pretty closely matches what we've seen by looking at the logs from the car. It would be interesting to re-run this during the winter months, but it takes a lot of discipline to not charge anywhere else but on the meter at home ;-)

Hope you find this useful.

P.S. My numbers (particularly percentages) may be off, but the raw figures are right. Sorry, but its late here and it has been a helluva week.

View attachment 2660