Tuning the Battery Aging Model

[Stoaty]

i downloaded the file & opened on the desktop w/ Excel. i don't get a version that has functions for calculations; it seems to be a flat data file. anyone have an idea of what i need on my Win7 computer to open the file w/ all the functionality?

Yes, you need the free OpenOffice 3.4.1, which is what the spreadsheet is created under:

http://www.openoffice.org/download/other-341.html" onclick="window.open(this.href);return false;

PS There is a bug in the End of Life Calculation that was introduced with the change to using corrected cycling loss. Sometimes you may find that the EOL calculation doesn't give a number (shows #N/A), it might show a wrong number in some circumstances. I know what is causing the problem, but haven't figured out a way to fix it yet with the relatively primitive tools available.

 

[DaveinOlyWA]

PS I am wondering what factors (besides heat) are causing some to match the model closely (e.g, DaveinOlyWa and Stoaty) and others to deviate far from it. Thoughts?

*speed. i rarely did as much as 65 mph (around here speed limit is 60 and traffic frequently limits that) so was 58-62 up until this Summer and its now more like 55.

*controlling temp bars. have only had a total of about 20-25 mins of TB7.

*and finally a weird work schedule. I am frequently on the road, driven my 20-40 miles in the morning and be parked for the day before 6 AM well before the heat of the day (and daylight 8 months of the year!)

I tried a few experiments to see how fast heat is built up and dissipated and found that there is a tipping point at roughly 60-62 mph and 75ish º F where heat buildup while driving is more than significant.

I was charging at night, finishing by 1 am or so and still seeing 6 TBs in the morning (high 70's, low 80's) which meant batt temps hitting the mid 80's if driving during the day. I was lucky in that the cool mornings in the 50's actually lowers my batt temps and typical drive is 32 miles, mostly freeway @ 55 mph.

Well today was day off with no real plans so I decided not to charge last night. Instead I got up first thing this morning (about 5:30 am) and opened garage door and plugged in and did an guesstimate charge for about 90% and ended up with 11 bars of charge and Batt temps were ranging between 72.7 and 71.2 so very cool temps. I took off driving, only did 25 miles mostly between 35-45 mph and temps did rise to 74-75º despite OAT being in low 60's.

On Wed i tried the opposite. finished charge less than an hour before I had to leave and temps were 77-79º. drove and it did its normal heat up 2-3º. On way home, drove 65 with A/C and OAT at 80ish (car said 74, phone said 81 and accuweather said 79 so take your pick...) temps started at 78-80 and shot up to 84-85 within 15 miles so I backed off... did not want to see 90

FYI; a few days ago I drove home at 55 (5 mile stretch at 10-30 mph) OAT was 90 and temps started at 82 ended at 84-85...

 

[surfingslovak]

I am not sure how you might incorporate any of this into your model.

All valid observations. Short of having the app log and protocol the temperature the battery sits at, we won't have a good measure of actual pack temperature. That said, and while your observations are correct, each LEAF will spend about 80% or 85% of its lifetime sitting parked somewhere. While it might take a while to dissipate operational heat, there will be a lot of time for that to occur. I had a look at radiant heat transfer when the problem with the Model S vampire load surfaced, and posted about it on TMC. It was clear already then that this energy loss was not caused by the TMS. Radiant energy and heat transfer even in a large flat pack with a metallic surface, as seen on the Model S, is not as significant as many would assume. Before we get lost in the minutiae of the argument, I believe that it would be more helpful to record the ambient temperature along with the pack temperature for a period of several days. Then analyze the data, and see if the effective battery temperature was different, and by how much. It would be also interesting to record the temperature in a garage and run a similar comparison to ambient air. The anecdotal temperature data Dave posted above is a good start.

 

[opencar]

that's a great suggestion. i can tell you that given that my car is a daily driver & averages 3 trips per day with nightly charging, i'd guess my battery temp will average a good 3-4 degrees above ambient for my location.

I am not sure how you might incorporate any of this into your model.

All valid observations. Short of having the app log and protocol the temperature the battery sits at, we won't have a good measure of actual pack temperature. That said, and while your observations are correct, each LEAF will spend about 80% or 85% of its lifetime sitting parked somewhere. While it might take a while to dissipate operational heat, there will be a lot of time for that to occur. I had a look at radiant heat transfer when the problem with the Model S vampire load surfaced, and posted about it on TMC. It was clear already then that this energy loss was not caused by the TMS. Radiant energy and heat transfer even in a large flat pack with a metallic surface, as seen on the Model S, is not as significant as many would assume. Before we get lost in the minutiae of the argument, I believe that it would be more helpful to record the ambient temperature along with the pack temperature for a period of several days. Then analyze the data, and see if the effective battery temperature was different, and by how much. It would be also interesting to record the temperature in a garage and run a similar comparison to ambient air. The anecdotal temperature data Dave posted above is a good start.

 

[Stoaty]

OK, here is version 0.94 of the Battery Aging Model:

https://dl.dropboxusercontent.com/u/48149991/Leaf%20Battery%20Degradation%20Model%20Version%20094.ods" onclick="window.open(this.href);return false;

Improvements:

--Calibration information (Capacity in AH) moved into Calibration Tab; Validation Tab now contains older readings using Gids
--Fixed bug that would not calculate EOL in both Prediction and Calibration tabs (still have to recalculate whole spreadsheet - Control-Shift-F9 - to get correct values)
--Tweaked aging factor - no change for aging factors less than that for L.A. civic center, but for factors above civic center used unscaled values rather than following Nissan's model; modest improvement in prediction for hotter cities by a couple of percent, but still work to be done
--Calibration tab now has semi-automated method for getting values using corrected cycling loss - enter row number (not the actual row number, but the number from the column marked "Row Number") of entry you wish to get more accurate results for, displayed in Fractional Year table and annual table in the same tab (then press Control-Shift-F9 to get correct result for EOL)

Known Issues:

--very short battery life (less than one year) may not give results for EOL in some cases (wouldn't be meaningful anyway if it did)
--still have to press control-shift-F9 to get correct value for EOL after making a change to any parameters

PS I probably won't make any more changes for a while once I recover from my cold

 

[Weatherman]

Before we get lost in the minutiae of the argument, I believe that it would be more helpful to record the ambient temperature along with the pack temperature for a period of several days. Then analyze the data, and see if the effective battery temperature was different, and by how much. It would be also interesting to record the temperature in a garage and run a similar comparison to ambient air. The anecdotal temperature data Dave posted above is a good start.

From what I've seen, unless the car is left idle and not charged for more than a day, the battery temp will be about 5F to 10F warmer than the air temp.

Since, down here, we go for months where the air temp falls to around 80 at night and rises the low 90s during the day, observations I've taken of battery temp are very consistent:

Early morning: starts out in the upper 80s (air temp around 80)
Following the drive in to work: around 90 (air temp in the upper 80s)
By late afternoon: lower to mid 90s (the afternoon commute doesn't affect the temperature, much, with air temps around 90)
After a 2 1/2 hour L2 charge is complete at midnight: mid to upper 90s (air temps in the mid 80s)
After 9 hours of overnight air temps in the lower 80s to near 80, the battery temp is back down in the upper 80s.

Even if it rains and air temps drop into the mid 70s, I haven't seen a battery temp below 80 since I started taking measurements in July. Average air temp during that time is about 84F. Average battery temp is around 92F.

 

[surfingslovak]

Even if it rains and air temps drop into the mid 70s, I haven't seen a battery temp below 80 since I started taking measurements in July. Average air temp during that time is about 84F. Average battery temp is around 92F.

Very helpful data! One more thing: how fast do you think you drive during your commute? It's been my experience with the ActiveE, which shows battery temperature in Fahrenheit on the dash, that spirited drives make the pack run considerably warmer.

 

[Weatherman]

Very helpful data! One more thing: how fast do you think you drive during your commute? It's been my experience with the ActiveE, which shows battery temperature in Fahrenheit on the dash, was that spirited drives make the pack run considerably warmer.

My commute is about 15 miles at 65 mph. It doesn't seem to warm the battery up by much from its already warm temperature.

 

[Stoaty]

OK, here is version 0.95 of the Battery Aging Model:

https://dl.dropboxusercontent.com/u/48149991/Leaf%20Battery%20Degradation%20Model%20Version%20095.ods" onclick="window.open(this.href);return false;

Improvements:

--found and fixed a bug introduced when I changed the Aging Factor to unscaled for climates hotter than LA Civic Center; the aging Factor inadvertantly got changed for cooler cities, which was not intended
--fixed the entries on the Calibration Tab that were wrong due to above bug
--general cleanup of some of the tabs, fixed some color backgrounds, reorganized some of the data to make it a bit easier to see

Known Problems:

--still have to press Control-Shift-F9 to get correct End of Life Values after changing any parameter that affects them

I have attached the calibration data of interest. You will note that the corrected total loss makes modest improvements (difference ranged from 0-1.5%), nothing revolutionary. However, in some cases the EOL decreased by a couple of years after the bug fixes.

 

[Stoaty]

OK, winding down on the changes to Battery Aging Model for now. Here is version 0.95b:

https://dl.dropboxusercontent.com/u/48149991/Leaf%20Battery%20Degradation%20Model%20Version%20095b.ods" onclick="window.open(this.href);return false;

Changes:

Mostly cosmetic - cleaned up the colors on the Degradation Tab, rearranged some items to be easier to make sense of.

Still to be done:

--add pie charts on Prediction tab to Calibration tab
--more thorough documentation

Comment: except for very hot areas (Texas and Ridgecrest, CA) and a Leaf with mixed heritage (14 months in California, then in Montreal), all of the "actual" capacity losses gathered so far are between 0-3.75% worse than predicted. However, note that Nissan states that after the P3227 update the capacity will read between 0 and 4% below actual capacity, so it is possible that the model is dead on. Time will tell of course. Anyone who hasn't submitted readings, please do so. We only have about 15 so far, need a lot more.

 

[KJD]

Comment: except for very hot areas (Texas and Ridgecrest, CA) and a Leaf with mixed heritage (14 months in California, then in Montreal), all of the "actual" capacity losses gathered so far are between 0-3.75% worse than predicted. However, note that Nissan states that after the P3227 update the capacity will read between 0 and 4% below actual capacity, so it is possible that the model is dead on. Time will tell of course. Anyone who hasn't submitted readings, please do so. We only have about 15 so far, need a lot more.

Nice work on the aging model Stoaty. My car has only recently had the P3227 update. I will send you my data after it has been 3 weeks since the update.

 

[edatoakrun]

...Nissan states that after the P3227 update the capacity will read between 0 and 4% below actual capacity...

Source?

 

[TomT]

We were told this by multiple Nissan sources at the meeting in Phoenix last month. The range was plus or minus 10 percent before the update and after it is plus 0 and minus 4 percent.

...Nissan states that after the P3227 update the capacity will read between 0 and 4% below actual capacity...

Source?

 

[vegastar]

We will see next winter the accuracy of P3227. Last winter I saw 13 gids@80% charge uptick in reported capacity since the summer low. If P3227 is -4% to 0% accurate I'm expecting a much lower uptick in capacity.

 

[WetEV]

Manufactured - 3/2012
Delivered - 5/5/2012
P3227 update - No
Location - In unincorporated area near Seattle WA
Miles/kWh - 4.3
Odometer - 15,600
Capacity (Ah) - 62.78
Date - 9/14/13
Parked in sun - Less than one day a week.

 

[edatoakrun]

...Nissan states that after the P3227 update the capacity will read between 0 and 4% below actual capacity...

Source?

="TomT" We were told this by multiple Nissan sources at the meeting in Phoenix last month. The range was plus or minus 10 percent before the update and after it is plus 0 and minus 4 percent.

The percentage range of what to what?

Capacity bar loss percentages of capacity, as compared to the percentages from the previously published chart?

So, "Nissan" said the first bar loss (for example) indicated an actual capacity loss range of between 25% and 5% before the update, and will indicate between 15% and 11% post update, correct?

Did they say if these percentages were of total battery capacity, or only of that (smaller percentage of) total capacity available between "100%" charge and shutdown?

In any case, Nissan (by saying the pre-update bar display might overstate capacity by 10 percent) seems to have reversed it's earlier statements, that the old bar display was "always pessimistic".

Which could explain why the update might have some (cool-climate?) LEAF owners disappointed...

 

[Stoaty]

OK, here is version 0.95c of the Battery Aging Model:

https://dl.dropboxusercontent.com/u/48149991/Leaf%20Battery%20Degradation%20Model%20Version%20095c.ods" onclick="window.open(this.href);return false;

Improvements:

--Calibration Tab - added the same pie charts which are on the Prediction tab
--Calibration Tab - changed color of cells which are looked up from the table using the row number from yellow to blue to indicate that you don't enter them directly

Bug Fixes: None

Example of calibration tab calculations attached (using the data from jmh614).

 

[Stoaty]

So, "Nissan" said the first bar loss (for example) indicated an actual capacity loss range of between 25% and 5% before the update, and will indicate between 15% and 11% post update, correct?

No, I think they are saying that when you have lost 15% of actual capacity, the capacity loss read by the software will be be between 15% and 19%. In other words, the software will be a little pessimistic, but never too optimistic. Tom, correct me if I am wrong.

 

[edatoakrun]

So, "Nissan" said the first bar loss (for example) indicated an actual capacity loss range of between 25% and 5% before the update, and will indicate between 15% and 11% post update, correct?

No, I think they are saying that when you have lost 15% of actual capacity, the capacity loss read by the software will be be between 15% and 19%. In other words, the software will be a little pessimistic, but never too optimistic. Tom, correct me if I am wrong.

However, since Nissan relies on the capacity bar display for warrantee purposes, and now claims that eight capacity bars indicates ~30% loss (of either the total or available capacity?) the accuracy of the capacity bar display, and whether/how much the update will improve it, is IMO, most significant to LEAF owners.

So, does Nissan now say the first bar loss now indicates between ~15% and ~11% (or another range of either the total or available capacity?) post update?

 

[DaveEV]

Comment: except for very hot areas (Texas and Ridgecrest, CA)

The Ridgecrest data vehicle does not have the P3227 update and thus, should not be included.

P3227 update - Not done

Edit: Keep in mind that while Stoaty will collect pre-P3227 update cars, the data won't be used to correct the model.