Tuning the Battery Aging Model

[opencar]

Stoaty,
thanks for your continued efforts to increase the fidelity of the model; i almost couldn't find the topic since it was buried down the "active topics" list. Given the current data set, any opinion about whether the slow down in degradation going into year 3 is a phenomena that the LEAF community is actually measuring? i'll be sending you my data at the end of the week since it will mark the 3 week passage since P3227.

Here is version 0.98 of the Battery Aging Model:

https://dl.dropboxusercontent.com/u/48149991/Leaf%20Battery%20Degradation%20Model%20Version%20098.ods" onclick="window.open(this.href);return false;

Changes:

--A bit more rearrangement of the entry fields on the Prediction and Calibration tabs to make it easier to follow
--Removed Documentation tab; there is now a separate PDF file (still being worked on) with documentation
--Worked around OpenOffice bug that caused a failure of Years to End of Life to update properly (thanks, thimel). You no longer need to press Control-Shift-F9 to make it show the correct value.

Newly identified bug:

If you make a change to one of the parameters on the Prediction tab and then press Control-Z to compare the before and after numbers, the EOL years do not update properly. Solutions?

 

[Stoaty]

Stoaty,
thanks for your continued efforts to increase the fidelity of the model; i almost couldn't find the topic since it was buried down the "active topics" list. Given the current data set, any opinion about whether the slow down in degradation going into year 3 is a phenomena that the LEAF community is actually measuring? i'll be sending you my data at the end of the week since it will mark the 3 week passage since P3227.

No opinion yet. It's only with the new data and the changes in parameters and correction for increased cycling loss due to higher depth of discharge that the model produces fairly accurate predictions of the AH capacity loss. At this point the model is fairly well calibrated, now we have to see if it can predict changes 6, 12, 18 or 24 months in the future. Time (plus more readings, especially repeat readings from the same Leafs) will tell.

 

[thimel]

Here is version 0.98 of the Battery Aging Model:



Newly identified bug:

If you make a change to one of the parameters on the Prediction tab and then press Control-Z to compare the before and after numbers, the EOL years do not update properly. Solutions?

Weird, but here is a solution:
Select cells A7:B7.
Right click on them and select "insert..."
Select "shift cells down" and click "OK"

That fixes it.

 

[Stoaty]

Here is version 0.98 of the Battery Aging Model:

Newly identified bug:

If you make a change to one of the parameters on the Prediction tab and then press Control-Z to compare the before and after numbers, the EOL years do not update properly. Solutions?

Weird, but here is a solution:
Select cells A7:B7.
Right click on them and select "insert..."
Select "shift cells down" and click "OK"

That fixes it.

Fix confirmed here, will be in next release of Battery Aging Model. How the heck did you ever find this solution? Seems like OpenOffice is rather buggy still. For the original bug, I tried downloading the latest release of LibreOffice to see if fixed the problem (it didn't).

 

[JeremyW]

Manufactured - 02/12
Delivered - 6/23/12
P3227 update - 7/13 (will try to find date and update post)
Location - Folsom, CA for 9 months, Pasadena, CA since March this year.
Miles/kwh - 3.9
Odometer - 18119
Capacity - 57.51
Date - 9/15/13
Parked in sun - 3 days/week (avg but varies substantially week to week)

Gids at 80% charge (this morning with the Brusa disabled)- 203. Last 100% charge two weeks ago showed 236.

Update:
Odo- 18573
Cap- 56.73
Date- 9/25/13
80% Gids- 200

 

[tivollix]

I lost 8% capacity since the P3227 update but it finaly stabilized at 86% and 57,25AH... It took many miles and more than 6 weeks for the BMS to find out the state of the battery...

Update, it is not stable yet. Will update in a few weeks. (removed chart taking too much space)
https://docs.google.com/spreadsheet/pub?key=0AlYaJGoW1QXYdC0xQnMxYTRCUkdZN1NYQ2FYaU9HZFE&output=html" onclick="window.open(this.href);return false;

 

[Weatherman]

Am I falling off a cliff?? I mean nearly an Ah in 400 miles!! :shock: :?

On July 11th, my LEAF's battery was at 57.45 Ahr. Today, it's at 55.08. How's that for a decline?

With the battery temp in the 90s, average drop of 0.032/day over the past 75 days. No slowdown in sight, since we'll still have summer weather (highs in the upper 80s to lower 90s and lows in the upper 70s) for at least the next four weeks.

 

[Stoaty]

Update:
Odo- 18573
Cap- 56.73
Date- 9/25/13
80% Gids- 200

Am I falling off a cliff?? I mean nearly an Ah in 400 miles!! :shock: :?

I don't know if you are falling off a cliff... but your Leaf is now tracking the Battery Aging Model a little closer:

Former Predicted Loss: 13.87%
Former Actual Loss: 13.19%
Actual Minus Predicted: -0.68%

Current Predicted Loss: 14.04%
Current Actual Loss: 14.37%%
Actual Minus Predicted: 0.33%

Scoreboard:

Battery Aging Model - 1
JeremyW - 0

 

[bradbissell]

Update on my 2013SV:

After a few 100% charges to make sure what I am seeing isn't an unbalanced pack....

64.090Ahr, 97%SOC, 98.270Health, 278 Gids, 7274 miles

 

[RegGuheert]

I lost 8% capacity since the P3227 update but it finaly stabilized at 86% and 57,25AH... It took many miles and more than 6 weeks for the BMS to find out the state of the battery...]

Since you live in Canada, perhaps most of that is simply due to the battery getting cooler. Let's hope so, anyway! It will be interesting to see how much comes back when it warms up next year.

 

[Berlino]

tivolix, the ºC to ºF conversion is wrong. 30ºC = 86ºF

 

[Stoaty]

Here is version 0.98b of the Battery Aging Model:

https://dl.dropboxusercontent.com/u/48149991/Leaf%20Battery%20Degradation%20Model%20Version%20098b.ods" onclick="window.open(this.href);return false;

Changes:

--incorporated fix so that Control-Z/Control-Y after a change in a parameter now correctly shows the EOL life years on both the Prediction and Correction tabs (thanks, thimel!!!)
--discovered and fixed subtle bug on Prediction tab that calculated calendar aging loss in the Annual table based on Years in Service, when it should be based on Battery Age. Note that in most cases the difference between Battery Age and Years in Service is only a couple of months, so the correction was relatively small for most entries. Also, the correction for Predicted Loss turned out to be very minor (e.g., a change of 0.01% for Stoaty's data) due to the fact that the way the calculations are done in the annual table has a very small effect on the calculation of Fractional Year loss. This problem was discovered when I realized that changing the date of manufacture had no effect on the calendar loss in the annual table, e.g., if the date of manufacture was moved 10 years earlier the calendar loss in the annual table didn't change (clearly, that made no sense). It now responds properly.
--discovered and fixed subtle bug on the End of Life tab that calculated calendar aging loss value for the Prediction tab incorrectly in the same way as described above. The EOL Years prediction now changes appropriately with date of manufacture on the Prediction tab. Correction is relatively minor in most cases, e.g., for Stoaty prediction changed from 6.90 years to 6.80 years.

Known Bugs/Issues:

--the bug discovered above in the Annual table also applies to the Calibration tab, and will take a bit more effort to fix there
--the bug discovered above in the EOL Years prediction also applies to the Calibration tab, and needs to be fixed
--EOL years on the Calibration tab is still (and probably will always be) hard coded, so once the fix to the bug on the Calibration tab is applied, these numbers will have to be updated manually (which is pretty easy, actually)

Request for more entries:

In the future:

--please limit new readings to 2011-2012 Leafs that have had the P3227 update at least 6 weeks ago (to be sure capacity has equilibrated).
--Do not submit readings on Leafs that need P3227 update but haven't had it
--Only resubmit a reading if it has been at least 3 months since your last submission, or if capacity has decreased by at least 2 AH from previous submission
--Only submit readings on 2013 Leafs if they have been in service for more than a year

Thanks!

 

[tivollix]

tivolix, the ºC to ºF conversion is wrong. 30ºC = 86ºF

Hello, it is exaclty that, the graph was showing only Celcius.

I have updated a google document for easy access for everyone:
In html:
https://docs.google.com/spreadsheet/pub?key=0AlYaJGoW1QXYdC0xQnMxYTRCUkdZN1NYQ2FYaU9HZFE&output=html
In docs:
https://docs.google.com/spreadsheet/ccc?key=0AlYaJGoW1QXYdC0xQnMxYTRCUkdZN1NYQ2FYaU9HZFE&usp=sharing

I'll continue to monitor but the battery capacity is dropping again... Temperature is mild and stable these days (around 60F).

 

[RegGuheert]

I have updated a google document for easy access for everyone:
In html:
https://docs.google.com/spreadsheet/pub?key=0AlYaJGoW1QXYdC0xQnMxYTRCUkdZN1NYQ2FYaU9HZFE&output=html
In docs:
https://docs.google.com/spreadsheet/ccc?key=0AlYaJGoW1QXYdC0xQnMxYTRCUkdZN1NYQ2FYaU9HZFE&usp=sharing

I'll continue to monitor but the battery capacity is dropping again... Temperature is mild and stable these days (around 60F).

Thanks for sharing your data! I find your results rather disturbing considering where you live and the fact that your odometer only reads 15,000 km.

 

[Berlino]

I find your results rather disturbing considering where you live and the fact that your odometer only reads 15,000 km.

Even though the data is only for the current location, could there be any lingering impact from the vehicle originally being in the USA for over a year?

 

[surfingslovak]

I find your results rather disturbing considering where you live and the fact that your odometer only reads 15,000 km.

Even though the data is only for the current location, could there be any lingering impact from the vehicle originally being in the USA for over a year?

I think that could explain a lot. I too was puzzled to see just how bad the data looked. Most other reports from Canada and Norway attest excellent battery longevity there.

 

[Berlino]

I think that could explain a lot. I too was puzzled to see just how bad the data looked. Most other reports from Canada and Norway attest excellent battery longevity there.

Is there a way to get precise location history from the VIN?

"San Francisco Area" is insufficient, given the huge climate differences when you travel 100km, or even just 50km inland in that region.

Of course, it doesn't help that the three months from manufacture to sale were summer ones where the vehicle may have been sitting fully charged under the sun.

 

[tivollix]

"San Francisco Area" is insufficient, given the huge climate differences when you travel 100km, or even just 50km inland in that region.

Of course, it doesn't help that the three months from manufacture to sale were summer ones where the vehicle may have been sitting fully charged under the sun.

I don't know for sure the location before May 2012. But starting spring 2012 it was at San Mateo or Palo Alto given the GPS history.

What is strange is that the capacity was really good until June this year, up to 95,25%. I was driving 65 Miles without hitting the reserve. It probably sat charged at 100% because the charging mode was set to 100% when I got it.

 

[Stoaty]

Here is version 0.98c of the Battery Aging Model:

https://dl.dropboxusercontent.com/u/48149991/Leaf%20Battery%20Degradation%20Model%20Version%20098c.ods" onclick="window.open(this.href);return false;

Changes:

--fixed subtle bug on Calibration tab that calculated calendar aging loss in the Annual table based on Years in Service, when it should be based on Battery Age.
--fixed subtle bug on the End of Life tab that calculated calendar aging loss value for the Calibration tab incorrectly
--EOL years on the Calibration tab numbers updated (manually)

Known bugs: None (currently)

Still to be done: Revise and finish documentation

Notes: No changes were needed to the calibration parameters, since the only significant changes were the EOL years for a few select Leafs. The Actual Loss divided by Predicted Loss has moved slightly closer to 100% (100.04%) and the standard deviation has gone down very slightly to 10.13%.

Please report any bugs on this thread.

 

[RegGuheert]

Stoaty, I know you prefer to have data post P3227 update, but I wanted to record my first reading somewhere as my Nexus One does not seem to allow a screenshot:

Manufactured - 06/11
Purchased - 3/16/12 (But it was a demo car put into service 9/28/2011. Charged to 100% until we purchased with 2011 miles on ODO.)
P3227 update - Not yet done.
Location - Winchester, VA
Miles/kwh - ~4.8
Odometer - 13,877
Capacity - 59.18 (90.21%)
Date - 9/28/13
Parked in sun - Almost never

Other data:
Health - 87.89%
SOC - 94.9%
Highest Cell - (Cell 1) - 4.106V
Lowest Cell - (Cell 77) - 4.089V (Next lowest is 4.094V)
Average Cell - 4.097V
Temps: 77.5F 77.7F 76.0F 73.9F
L1/L2 Count - 1214

I don't have your new spreadsheet installed, but the old one (0.83) is dead-on if I use the 9/28/2011 in-service date:
Total Loss = 9.64%
Calendar Loss = 8.29%
Cycling Loss - 1.35%