Look Inside Kona Electric Battery And See Real Capacity

Yes, my error.

I'm sure I've posted this before, cumulative net discharge (blue line) plotted against SoC(displayed) down to 30%.
The blue line is the cumulative gross battery output to the Kona (ΔCED) and the green is the cumulative regen input (ΔCEC). The blue line is just the net , red minus green.

I've allowed 96% efficiency for regen energy as it cycles through the battery to be reused.

Note the linear trendline slope is 64 kWh.

 

I re-plot my data the same way as yours. We get very similar graphic. If you look closely and compare the actual data trace with the linear trace, it is not perfectly linear. I can see the same thing on your graphic. On my previous graphic, using the same data, I calculated the amount of kWh used for each %SoC and plot it against "SoC Display". Presenting the data differently shows the non-linearity of the Energy capacity vs SoC. There is much less energy available in the last 50% SoC; a bit more than 30 kWh vs a bit less than 34 kWh from 100% to 50%.

 

I'm getting a lot of scatter in those values undoubtedly due to the clash between the resolutions of the three data sources, the very hilly drive the data was drawn from and perhaps the methods I've used in Excel. Nonetheless, assuming I've done this correctly here's the net energy out per 1% loss of SoC(disp) at each 0.5% increment. I used the same 0.96 efficiency on regen energy to accommodate battery cycle losses. Added for reference is a linear trendline and a 5% SoC moving average.
Outliers are likely due to SoC increases along some parts of the drive.

 

Hi Kiwime! You are suggesting a 0.96 efficiency on regen. I record my data using an esp32 microcontroller and calculate my battery degradation with the Net kWh results. From this, I have estimated my current battery energy capacity to 62.3kWh. My average regen per trip is about 15%. So I guest, I should apply a 0.96 correction to this and my battery degradation will be a little more.

 

Hi KiwiME! You are suggesting a 0.96 efficiency on regen. I record my data using an esp32 microcontroller and calculate my battery degradation with the Net kWh results. From this, I have estimated my current battery energy capacity to 62.3kWh. My average regen per trip is about 15%. So I guest, I should apply a 0.96 correction to this and my battery degradation will be a little more.

 

In order to measure net discharge from the battery against SoC by driving and using CED you need to take into account regen via CEC. To normalise regen energy with the battery output you’ll need to apply battery cycle efficiency.
I’ve measured that between .96 and .98.
I’m unclear how that relates to SoH.

 

I managed to define the net energy decharge per 1% loss profile. The only unknown is the last 4% SoC since I never discharged under that level. By extrapolating the last 4% using the curve equation, I got a total energiy of 62.3 kWh. I'm assuming with a new battery I would obtain something closer to 64kWh. I use CED and CEC to get the net energy discharged but I don't apply a battery cycle efficiency. Should I apply a 0.96 correction factor to the net energy or to the CEC values?
The equation I used, corrected to get 64.08 kWh when I integrate under the curve from 100 to 0% SoC is:
2E-7 * x3 + -2.4E-5 * x2 + 0.002194 * x + 0.562

 

You would apply that 0.96 factor to CEC before subtracting that regen energy from CED.
So, all I've done at each 0.5% drop in SoC is calculate:

(ΔCED - (ΔCEC x 0.96)) / 0.5

... which is the net kWh discharge per 1 percent loss in SoC. I'm not aware how to fit a curve (other than the trendlines) to the data so suffer with the clash of resolutions which causes the steps in my plot. Nonetheless I'll agree that it's clearly not a constant change over the SoC range. I'll have to try the same with the cumulative current registers ... which do not need any correction for resistive battery losses.

 

Are we discussing Display or BMS SoC?
You guys are over my head on the maths, but I note from Car Scanner that BMS SoC and Display SoC are about 3% points apart at 100% and have been ever since I got the car, except when I need regen. Then BMS SoC will go a couple of points higher. I suppose I'm saying that there's some manufacturer programmed deterioration

 

Ok KiwiME! Makes sense. This means I should obtain a slightly higher Net kWh at the end. Thanks!

 

Just to add to herode10's comments from a few months back, I've collected data from a repeat of a 300 km drive going from 90% down to 15%. This time I've logged the Ah (coulomb) registers as well so I could plot Ah as well as energy against SoC. It's pretty clear now that SoC follows Ah more closely than kWh and that meshes with my experience on this round trip where I noticed a greater rate of loss in SoC as I approached 15%. The plots are not about driving efficiency, rather they are more about if the SoC reading correlates linearly to energy remaining. For what it matters battery temp ranges from 11 to 22°C, noting that it's a sunny day but still the middle of winter.

The upper two lines are cumulative net Ah out of the battery plotted against SoC displayed and BMS respectively. The lower lines are cumulative net energy out against the two SoCs, with a correction of returned regen energy to accommodate normal battery loses. All four lines have linear trendlines overlaid with the slopes noted. It's hard to see in the image this forum renders but the linear trendlines fit the upper lines somewhat better than the lower.

As we've noted before the slopes of the upper Ah line (1.805 x 100) and upper energy line (0.639 x 100) closely match the 180 Ah and 64 kWh battery pack ratings, respectively. This should allay speculation that the usable capacity is less or greater than advertised.



The second plot is how I determined the cumulative battery cycle efficiency of 98%. Once every charge cycle the CDC and CCC (colulomb) counters pass each other. Coulombs don't disappear as losses like energy does so they will match every time a certain SoC is passed, which I would assume matches the battery state when the counters were zeroed. Here I have all four odometers plotted over 90% down to 15%. As I pass through 55% SoC(disp) the charge and discharge coulombs match and at that point the ratio of the energy counters will reveal the cycle efficiency achieved since these counters were zeroed. That would be at the 196 update I had in Dec 2020.



This plot is the cell voltage while driving compared with LG's definition of "open cell voltage" vs SoC for a new cell.

There are two caveats here: One is that, while driving the voltage is all over the place and is certainly not "open circuit". It sags on power and rises on regen. The second is that we don't know if the SoC(BMS) matches LG's published definition of SoC. It appears that it may match but even though that's probable, it's not certain. The matching of the black 15% moving average curve to the blue LG spec does somewhat support this conclusion.

In this plot of 2590 data points you'll note that the peaks (during regen) are narrower than the sags (while powering up hills), because they are plotted against SoC rather than time. The SoC doesn't drop much on downhills so the data stays horizontally compressed.



One last plot, the red dots show 12V system voltage every 5 seconds over the same trip. The point here is that the 12V battery does not have a good opportunity to charge while driving, however it is not allowed to discharge. I'm pretty sure this was a change since Campaign 196 because it used to charge for the first 30 minutes of the first drive of the day. The normal charging voltage is 14.65 V, yet it dithers around 13.5.

This plot is against time with a 15 minute moving average in black. You can see a point where I stopped for about 4 minutes and despite the car being "off", because I was still drawing OBD data it pulled down the aux battery quite a bit.

 

I pulled data from 4 different dates

The SoC value where the CCC and CDC counters passe each other has drift with time for me. I gather some data close to the date I got the latest BMS update and some others each months after that. I observed at least 2 time a SoC correction without any CCC, CDC, CED or CEC changes. I chose a point on March 10th and one on March 15th, since on March 15th I observed a 3% SoC correction. I can't explane this correction but it explanes some SoC shift at CCC and CDC crossing.

 

I'd expect that SoC is derived primarily from the CDC and CCC coulomb counters with corrections made by assessing open circuit pack voltage when the voltage has settled and the main contactor is still open, possibly during the first few seconds of start-up. I think this is why they align so well while driving where no correction can take place. But I don't know if the LG-published SoC vs OCV is stable with cell age or not. When did you observe the corrections in SoC? I think I've seen mine move on startup now and again.

I'm unsure as to why you're seeing a wide scatter of SoC values at CDC=CCC. There certainly will be small cumulative errors from the shunt resistor calibration and perhaps with time the SoC at intersection will drift away from being within the range of 0-100%. It would be interesting to plot those SoCs values over a longer period to see if there is some pattern like a gaussian distribution. You're certainly doing a lot more driving than I am and logging more data!

 

One time I left the car On to monitor the diagnostic cycles. The SoC was around 50% and the power pulled from the battery was around 0.2 kW max. After 15-20 mins in that condition, I observe the S0C shifting by 2.5 % within 5 secs with no change in CCC or CDC. I repeated the same test (around 50% SoC) a couple of months later and this time I saw the SoC shifting by 4 %. After reading your post, I checked the instances where CCC=CDC and what was my SoC to make this table. Around july 4th, I had another big shift of my SoC. I don't have data when this happen. It's not a gradual shift in time, it's always a sudden correction. For the past 3 weeks, the number has not moved (around 3000Km).

 

The SoC at CCC=CDC seems to vary little bit up and down but the trending seems to be more a drift going down from the last bms update.

 

The shift happen on July 2nd. I was camping and put the car in "utility" mode to be able to power a small cooler. So it was not drawing much power from the battery. There has been a SoC shift of 7% at CCC=CDC on that day. As I said, it hasn't moved since then.

 

I’ve seen the display jump now and again when I select the relevant screen but it’s indeterminate if it’s an actual correction of SoC or just a display artefact.

 

The following plot shows what I observed as SoC correction. CDC and CCC stay constant but both SoC drop suddenly:

 

That must be off OBD data so excludes issues with the dash display updating. Are you also logging cell or pack voltage?

 

Yes, I record the min and max cell voltages and the pack voltage as well. During that SoC shift, the voltages did not change.