Discuss data from the LEAF Battery app, and Comparisons

[lorenfb]

Just like with any battery type, a 'heavy' dynamic load test (delta V vs delta I) near max peak current output
(kinda like CCA with lead acid) needs to be performed to determine true output impedance.

 

[TickTock]

I've actually seen much more variation on impedance based on pack temperature than SOC..

Yup. The noise in the Resr plot does correlate somewhat to SOC (left out of the graph for clarity), but the pack temperature has the largest influence. Also, we had been wondering if maybe Hx was related to resistance, and from the data I've gathered over the last few months, it doesn't look like the case.

How is this "resistance" being measured and/or determined? And what are the units of measure?
Neither of these have been indicated in the post nor on the graph. And what device was used that
provided the data?

Sorry. Noone know what the units of Hx are - guessing to be % of *something*. Plotted as a straight scalar here (0.5 = 50%). Resr is in Ohms (so I am sitting around ~0.1 Ohm). Pack temperature is in Celcius. I have programmed CANary to continually compute Resr as the value of resistance that minimizes the short term (one second observation window) variation in the formula Voc=Vpack-Ipack*Resr under heavy driving conditions (I throw away all data sets with less than 40 amps change in current) and maintain a rolling average. At the end of every trip, I log the final Resr (along with many other parameters).

 

[mwalsh]

6/22: AHr=54.07 CAP=81.61%. Mileage: 30,455 miles. Lost bar 12.
6/24: AHr=54.01 CAP=81.52%
6/25: AHr=54.12 CAP=81.68%
7/11: AHr=53.16 CAP=80.23%
7/12: AHr=53.30 CAP=80.45% Hlth=76.57%
8/26: AHr=52.54 CAP=79.30% Hlth=75.16%
9/11: AHr=51.49 CAP=78.49% Hlth=73.22%
10/2: AHr=51.81 CAP=78.98% Hlth=73.82%. Mileage: 34,507
10/8: AHr=51.58 CAP=78.63% Hlth=73.39%. Mileage: 34,813. Lost bar 11.
10/23: AHr=52.12 CAP=79.51% Hlth=74.46%. Mileage: 35,220
11/5: AHr=52.68 CAP=80.31% Hlth=75.43%. Mileage: 35,723
11/12: AHr=52.47 CAP=79.98% Hlth=74.80%. Mileage: 36,071
11/22: AHr=52.74 CAP=80.40% Hlth=75.53%. Mileage: 36,445
11/25: AHr=52.89 CAP=80.63% Hlth=75.81%. Mileage: 36,509
12/22: AHr=52.74 CAP=80.40% Hlth=75.53%. Mileage: 37,596
1/21: AHr=51.38 CAP=78.32% Hlth=73.02%. Mileage: 38,471
2/25: AHr=51.37 CAP=78.31% Hlth=73%
3/12: AHr=50.75 CAP=77.37% Hlth=71.86%
3/24: AHr=51.93 CAP=79.16% Hlth=72.36%. Mileage: 40,228
4/17: AHr=50.81 CAP=77.46% Hlth=71.97%. Mileage: 41,232
5/5: AHr=48.64 CAP=74.15% Hlth=67.96%. Mileage: 41,754
5/21: AHr=48.10 CAP=73.33% Hlth=66.97%. Mileage: 42,485

6/24: AHr=48.50 CAP=73.93% Hlth=67.69%. Mileage: 43,277

Just noticed that's a year (almost to the day) of loss.

 

[lorenfb]

I've actually seen much more variation on impedance based on pack temperature than SOC..

Yup. The noise in the Resr plot does correlate somewhat to SOC (left out of the graph for clarity), but the pack temperature has the largest influence. Also, we had been wondering if maybe Hx was related to resistance, and from the data I've gathered over the last few months, it doesn't look like the case.

1. Is your Voc the loaded (I > 40 amps) battery voltage being measured by the VCM?
2. Your method of calculating the impedance requires three data points from the VCM,
Voc, Vpack, Ipack (load > 40 amps) with assumption that all points were measured contemporaneously.
3. You've masked any independent change of the output impedance over time by not monitoring
the impedance at a constant temperature. One would expect that the impedance is NOT constant
with temperature.
4. You have not indicated graphically what the AHr of your battery was over the time interval
as an actual AHr measurement. Your Hx appears to be an AHr reading which goes from 50 to 43
over the time period and is the fully charged AHr reading.
5. The impedance data from graph indicates about 100 milliohms. My data indicates about 63 milliohms
for a 59 AHr battery at about six months old.

Interesting statement from the LeafDD manual:

"The last line on Page 2 will be temporarily replaced with an impedance reading (mohms) as well as the pack SOC and temperature at the time of the test. This is basically an indication of how much the voltage sags under load (or rises on a DCQC or regen) and thus how much power you have. The Leaf battery has very low impedance with plenty of margin so this not so important as perhaps interesting. A new Leaf might be 50-60mohms whereas a degraded vehicle could be 2x that or more."

 

[TickTock]

I've actually seen much more variation on impedance based on pack temperature than SOC..

Yup. The noise in the Resr plot does correlate somewhat to SOC (left out of the graph for clarity), but the pack temperature has the largest influence. Also, we had been wondering if maybe Hx was related to resistance, and from the data I've gathered over the last few months, it doesn't look like the case.

1. Is your Voc the loaded (I > 40 amps) battery voltage being measured by the VCM?
2. Your method of calculating the impedance requires three data points from the VCM,
Voc, Vpack, Ipack (load > 40 amps) with assumption that all points were measured contemporaneously.
3. You've masked any independent change of the output impedance over time by not monitoring
the impedance at a constant temperature. One would expect that the impedance is NOT constant
with temperature.
4. You have not indicated graphically what the AHr of your battery was over the time interval
as an actual AHr measurement. Your Hx appears to be an AHr reading which goes from 50 to 43
over the time period and is the fully charged AHr reading.
5. The impedance data from graph indicates about 100 milliohms. My data indicates about 63 milliohms
for a 59 AHr battery at about six months old.

Interesting statement from the LeafDD manual:

"The last line on Page 2 will be temporarily replaced with an impedance reading (mohms) as well as the pack SOC and temperature at the time of the test. This is basically an indication of how much the voltage sags under load (or rises on a DCQC or regen) and thus how much power you have. The Leaf battery has very low impedance with plenty of margin so this not so important as perhaps interesting. A new Leaf might be 50-60mohms whereas a degraded vehicle could be 2x that or more."

The purpose of the graph was to corroborate GregH's comment about the primary impact of temperature. I have seen a degradation in ESR with SOH but I don't have this information readily available since I had not been auto-logging SOH in my trip log (I know at one point in I had been reading 70mO but need to go back to figure out the SOH and temperature at the time of that measurement). I have the information buried in my raw canbus logs but it will take some effort write the query to parse them (and computer time to execute - it's a LOT of data stored on a network drive with limited throughput).

I am confident the Leaf instrumentation knows how to read the voltage accurately but we know the current sensor is not particularly accurate. It is possible (if not probable) that the temperature variability in the resistance measurement is dominated by the temperature sensitivity of the Hall effect sensor but I don't know if it operates at ambient or at pack temp. Perhaps someone here knows it's location. If it operates at ambient (not thermally coupled to the battery box) we should be able to establish if it is sensor or pack since we do have both temperatures available for analysis).

Also, I originally said the data does not support a correlation of Hx with ESR, but on second look at the graph think it may - certainly still debatable. It does look like in April of 2014, the ESR is slightly higher that it was on October of 2013 although the pack temps are similar. Could very well be up 5% which is how much the Hx dropped over that same period. Hopefully, once more data is added to the set it will be more clear.

 

[lorenfb]

"The purpose of the graph was to corroborate GregH's comment about the primary impact of temperature. I have seen a degradation in ESR with SOH but I don't have this information readily available since I had not been auto-logging SOH in my trip log (I know at one point in I had been reading 70mO but need to go back to figure out the SOH and temperature at the time of that measurement). I have the information buried in my raw canbus logs but it will take some effort write the query to parse them (and computer time to execute - it's a LOT of data stored on a network drive with limited throughput).

I am confident the Leaf instrumentation knows how to read the voltage accurately but we know the current sensor is not particularly accurate. It is possible (if not probable) that the temperature variability in the resistance measurement is dominated by the temperature sensitivity of the Hall effect sensor but I don't know if it operates at ambient or at pack temp. Perhaps someone here knows it's location. If it operates at ambient (not thermally coupled to the battery box) we should be able to establish if it is sensor or pack since we do have both temperatures available for analysis).

Also, I originally said the data does not support a correlation of Hx with ESR, but on second look at the graph think it may - certainly still debatable. It does look like in April of 2014, the ESR is slightly higher that it was on October of 2013 although the pack temps are similar. Could very well be up 5% which is how much the Hx dropped over that same period. Hopefully, once more data is added to the set it will be more clear."

Thanks for the insight.

 

[DaveEV]

Also, I originally said the data does not support a correlation of Hx with ESR, but on second look at the graph think it may - certainly still debatable. It does look like in April of 2014, the ESR is slightly higher that it was on October of 2013 although the pack temps are similar. Could very well be up 5% which is how much the Hx dropped over that same period. Hopefully, once more data is added to the set it will be more clear.

It certainly would make sense if the Hx value is a representation of ESR...

That would also explain why Hx responds so quickly after a LBC reset as well.

 

[stjohnh]

Just like with any battery type, a 'heavy' dynamic load test (delta V vs delta I) near max peak current output
(kinda like CCA with lead acid) needs to be performed to determine true output impedance.

Could a quick and dirty test like this help judge battery degradation:

At a specified SOC (say 50%) and battery temperature, log the pack voltage, step on accellerator to max, accelerate for 5 sec, check pack voltage. Is the difference between the two voltage measurements a decent indication of pack degradation?

 

[lorenfb]

Just like with any battery type, a 'heavy' dynamic load test (delta V vs delta I) near max peak current output
(kinda like CCA with lead acid) needs to be performed to determine true output impedance.

Could a quick and dirty test like this help judge battery degradation:

At a specified SOC (say 50%) and battery temperature, log the pack voltage, step on accellerator to max, accelerate for 5 sec, check pack voltage. Is the difference between the two voltage measurements a decent indication of pack degradation?

That's what has been discussed. One wouldn't really need 5 sec, as the loaded voltage, i.e. the result
of the internal battery impedance (ESR), would stabilize in a few seconds (enough time to read it).
The ESR change over time should be indicative of battery aging.

 

[stjohnh]

Just like with any battery type, a 'heavy' dynamic load test (delta V vs delta I) near max peak current output
(kinda like CCA with lead acid) needs to be performed to determine true output impedance.

Could a quick and dirty test like this help judge battery degradation:

At a specified SOC (say 50%) and battery temperature, log the pack voltage, step on accellerator to max, accelerate for 5 sec, check pack voltage. Is the difference between the two voltage measurements a decent indication of pack degradation?

That's what has been discussed. One wouldn't really need 5 sec, as the loaded voltage, i.e. the result
of the internal battery impedance (ESR), would stabilize in a few seconds (enough time to read it).
The ESR change over time should be indicative of battery aging.

I am not aware of any data reported on these boards of the results of a test like this. Additionally, I was hoping for a way to gauge degradation more conveniently than doing a range test. LeafSpy reports average cell volts, and multiplied by 96 should give pack voltage, however the sampling rate seems a little slow, and it is hard to go more than 5 sec at max acceleration safely. I have a WattsLeft meter, which seems to show pack volt changes every 1/10 sec or so, but most Leaf owners will not have that. So, while I can do a test like I suggested, I have no way of knowing if say a 15 volt deltaE (volt change) corresponds to 0%, 10% or 20% degradation. The data reported by LeafSpy: "capacity" (Ahr), SOH, and Hx all have a general relationship to degradation, but, especially with the 2013 Leafs, have been programmed in a way that defies my understanding and have shifts in values that cannot be reliably attributed to battery degradation.

 

[JeremyW]

At a specified SOC (say 50%) and battery temperature, log the pack voltage, step on accellerator to max, accelerate for 5 sec, check pack voltage. Is the difference between the two voltage measurements a decent indication of pack degradation?

That's exactly what the LeafDD does. On the main screen one dot means low level voltage recorded and two dots for full throttle voltage. Hit the button once and the value is displayed on the bottom line. I have observed as high as 120 miliOhm and as low as 80 miliOhm. Temperature is the dominate factor by far. Higher temp, lower internal resistance. I run the CANary as well in tandem and it's findings are within a few miliOhms.

 

[lorenfb]

At a specified SOC (say 50%) and battery temperature, log the pack voltage, step on accellerator to max, accelerate for 5 sec, check pack voltage. Is the difference between the two voltage measurements a decent indication of pack degradation?

That's exactly what the LeafDD does. On the main screen one dot means low level voltage recorded and two dots for full throttle voltage. Hit the button once and the value is displayed on the bottom line. I have observed as high as 120 miliOhm and as low as 80 miliOhm. Temperature is the dominate factor by far. Higher temp, lower internal resistance. I run the CANary as well in tandem and it's findings are within a few miliOhms.

Right, so the test MUST be done at the same temp always, or the relative change in impedance over time
is of little to no value.

 

[lorenfb]

At a specified SOC (say 50%) and battery temperature, log the pack voltage, step on accellerator to max, accelerate for 5 sec, check pack voltage. Is the difference between the two voltage measurements a decent indication of pack degradation?

That's exactly what the LeafDD does. On the main screen one dot means low level voltage recorded and two dots for full throttle voltage. Hit the button once and the value is displayed on the bottom line. I have observed as high as 120 miliOhm and as low as 80 miliOhm. Temperature is the dominate factor by far. Higher temp, lower internal resistance. I run the CANary as well in tandem and it's findings are within a few miliOhms.

My version of LeafDD is 1.3.2 and as such appears to not have the feature of providing a calculation
of the impedance by accelerating at full throttle, i.e. if there is a FT switch as in an ICE.

 

[TickTock]

I decided to see if I could recreate the Hx data from other parameters. I started with the popular assumption that Hx is primarily inversely proportional to ESR (IOW, tracks ESC (Electrical Series Conductance)). I did try ESR at first but got better results with the ESC assumption. I then added terms to allow linear scaling for temperature and SOC to end up with the following formula:
Fn = C0(ESC-ESCo)(1-C1(T-To))(1-C2(SOC-SOCo)).

Then I used the Excel solver to solve for C0,C1,C2,ESCo,To,SOCo.

C0 -0.02108
ESCo 26.9221
C1 0.006568
To -9.4489
C2 -2.71244E-06
SOCo 141.1


The result is fn in the graph (Hx is the mystery value reported by the Leaf). Dissatisfied with this, I added another term (1-C3(AH-AHo)) and let excel solver go at it and got the following values (other values changed very little).

C3 -0.01727
AHo 44.4767

The result is fn2. Closer, but still not quite there. However, there are some observations to be made.
1) SOC has very little impact
2) Temperature, ESC, & SOH (AH) do seem to correlate

So it appears like it may be primarily comprised of a temperature-compensated ESC as well as including an SOH term. Perhaps some proprietary health indicator for in-house use?

I am open to suggestions for better model formulas to try to fit.

 

[TimLee]

Your fn2 looks pretty good. Excellent work
Seems to correlate pretty well, but Hx has a lot of smoothing applied.
I would suggest plotting some running averages of fn2 at various average durations of 2, 4, 6, and 8 weeks to see how those compare.
Or other more involved smoothing approaches.

 

[opencar]

lost my 2nd bar today @ Ahr=51.54, pretty close to my 2nd year anniversary @ 33K mi.

6/22: AHr=54.07 CAP=81.61%. Mileage: 30,455 miles. Lost bar 12.
6/24: AHr=54.01 CAP=81.52%
6/25: AHr=54.12 CAP=81.68%
7/11: AHr=53.16 CAP=80.23%
7/12: AHr=53.30 CAP=80.45% Hlth=76.57%
8/26: AHr=52.54 CAP=79.30% Hlth=75.16%
9/11: AHr=51.49 CAP=78.49% Hlth=73.22%
10/2: AHr=51.81 CAP=78.98% Hlth=73.82%. Mileage: 34,507
10/8: AHr=51.58 CAP=78.63% Hlth=73.39%. Mileage: 34,813. Lost bar 11.
10/23: AHr=52.12 CAP=79.51% Hlth=74.46%. Mileage: 35,220
11/5: AHr=52.68 CAP=80.31% Hlth=75.43%. Mileage: 35,723
11/12: AHr=52.47 CAP=79.98% Hlth=74.80%. Mileage: 36,071
11/22: AHr=52.74 CAP=80.40% Hlth=75.53%. Mileage: 36,445
11/25: AHr=52.89 CAP=80.63% Hlth=75.81%. Mileage: 36,509
12/22: AHr=52.74 CAP=80.40% Hlth=75.53%. Mileage: 37,596
1/21: AHr=51.38 CAP=78.32% Hlth=73.02%. Mileage: 38,471
2/25: AHr=51.37 CAP=78.31% Hlth=73%
3/12: AHr=50.75 CAP=77.37% Hlth=71.86%
3/24: AHr=51.93 CAP=79.16% Hlth=72.36%. Mileage: 40,228
4/17: AHr=50.81 CAP=77.46% Hlth=71.97%. Mileage: 41,232
5/5: AHr=48.64 CAP=74.15% Hlth=67.96%. Mileage: 41,754
5/21: AHr=48.10 CAP=73.33% Hlth=66.97%. Mileage: 42,485

6/24: AHr=48.50 CAP=73.93% Hlth=67.69%. Mileage: 43,277

Just noticed that's a year (almost to the day) of loss.

 

[TickTock]

Your fn2 looks pretty good. Excellent work
Seems to correlate pretty well, but Hx has a lot of smoothing applied.
I would suggest plotting some running averages of fn2 at various average durations of 2, 4, 6, and 8 weeks to see how those compare.
Or other more involved smoothing approaches.

You nailed it! I applied a rolling average on my function and had excel re-solve (LMS error) and it is pretty darn close. Closer than I hoped!

Here's the updated coeficients:
C2 -0.003955272
SOCo 140.7227557
C3 -0.037281634
Aho 42.77824966
C1 5.82199E-05
To -10.54527559
C0 -0.000487104
ESCo 29.23008492

Fn = C0(ESC-ESCo)(1-C1(T-To))(1-C2(SOC-SOCo))(1-C3(AH-AHo))

 

[TickTock]

I think we just put a stake in this. I zeroed out all the weak coefficients and now know Hx to be purely a function of Ah and nothing more. Simplifying the formula I get good results with:

Fn = 0.01645*(Ah-16)



Sorry folks - no ESR here.

 

[lorenfb]

I think we just put a stake in this. I zeroed out all the weak coefficients and now know Hx to be purely a function of Ah and nothing more. Simplifying the formula I get good results with:

Fn = 0.01645*(Ah-16)



Sorry folks - no ESR here.

REALLY?

1. You've overly complicated the analysis with this equation:

Fn = C0(ESC-ESCo)(1-C1(T-To))(1-C2(SOC-SOCo))

Simple testing of the battery at various levels of AHrs (SOC) basically indicates that
SOC does not affect the impedance. All that was needed was a simple linear regression
with one independent variable to develop a correlation coefficient.

2. As you previously stated:

"I have seen a degradation in ESR with SOH but I don't have this information readily available since I had not been auto-logging SOH in my trip log (I know at one point in I had been reading 70mO but need to go back to figure out the SOH and temperature at the time of that measurement). I have the information buried in my raw canbus logs but it will take some effort write the query to parse them (and computer time to execute"

So you indicate that there is/was a correlation between ESR & SOH, but now your "proof'
indicates otherwise. Scientific methods & analysis would question your results when they
are contrary to your other data acquired and a previous hypothesis.

4. Then you state:

"I am confident the Leaf instrumentation knows how to read the voltage accurately but we know the current sensor is not particularly accurate. It is possible (if not probable) that the temperature variability in the resistance measurement is dominated by the temperature sensitivity of the Hall effect sensor but I don't know if it operates at ambient or at pack temp. Perhaps someone here knows it's location. If it operates at ambient (not thermally coupled to the battery box) we should be able to establish if it is sensor or pack since we do have both temperatures available for analysis)."

Which then raises the issue of the reliability/accuracy of your data. Having not reviewed your
data retrieval process, posted data (little to none), and correlation, your data reliability becomes
questionable. Analysis of the data set is problematic because of the covariance between the impedance
and temperature and its reliable elimination.

5. And finally you state:

"Also, I originally said the data does not support a correlation of Hx with ESR, but on second look at the graph think it may - certainly still debatable. It does look like in April of 2014, the ESR is slightly higher that it was on October of 2013 although the pack temps are similar. Could very well be up 5% which is how much the Hx dropped over that same period. Hopefully, once more data is added to the set it will be more clear."

Here you indicate that the ESR has increased over time as AHrs have decreased which
indicates some correlation not indicated in your final equation. To a scientist, this would
'raise-a-flag'.

All that's really needed were a few data points of ESR and AHrs measured at a constant
temperature over time (24 months) and contemporaneously, while verifying the accuracy
of each data point, which it appears you haven't done. This analysis is hardly as complex
as you have presented. Also, you have failed to corroborate your data (CANary) with
another data measurement device, i.e. simple voltage & current measurements over time.

Only a long term analysis (2 years) with data gathered reliability, can battery impedance not
be assumed to be a proxy for battery health (Hx)/AHr capacity. Lithium batteries do have some
unique characteristics than other battery chemistry types, but the relationship of impedance
and capacity over time is very similar to other battery chemistries. As determined from very
simple data gathering & analysis, battery impedance doesn't change as significantly as AHrs
do, but it does change with AHrs. Furthermore, many research papers on this issue have
indicated contrary results to that posted here!

Bottom line: Reliable data needs to be gathered and analyzed.

 

[TickTock]

I think we just put a stake in this. I zeroed out all the weak coefficients and now know Hx to be purely a function of Ah and nothing more. Simplifying the formula I get good results with:

Fn = 0.01645*(Ah-16)



Sorry folks - no ESR here.

Grabbed a few datapoints from the forum to see how well it holds up on other's vehicles. It looks good on folks like Pipcecil who are at similar levels of capacity:
http://www.mynissanleaf.com/viewtopic.php?f=30&t=8802&hilit=hx+ahr&start=6874 *** 0.4495 ~= 0.4494 = 0.01645*(43.32-16)

..but not newer batteries like mwalsh': http://www.mynissanleaf.com/viewtopic.php?f=44&t=12789&hilit=hx+ahr&start=482 *** 0.6769 != 0.535 = 0.01645*(48.5-16)

So while strongly correlated to Ah for 4 bar losers we need more data, it appears, to cover the entire range.