JK bms fast(er) SOC learn

[Skypower]

I’m reasonably satisfied with my JK’s…now.
I thought I had a problem because my third battery commissioning went a little rougher than the first two JK’s. Something was definitely different with the third and newest bms. The first thing that popped up was the yellow flashing “Change password in time” notification and the layout of the data on the status screen was in different locations. So some firmware changes were probably made. The first two learned what full State Of Charge was after a few cycles but the third was flat out stubborn and got more stupid by the day, losing about 2% capacity readings per cycle. Tried to run that battery up to trip OVP at 3.650 volts and that seemed to work…. For a while. Figure I got a dud bms and ordered another. Install the forth bms and after 5 cycles it became obvious this wasn’t going any better as it was also getting worse by the day. So I decided that dang it, I’m gonna figure out what this JK wants to know to learn what a full state of is. So I went with a hunch. I charged to my regular absorb voltage of 55.2 (3.45 vpc) and knowing that my delta is about .020 when it just gets there, I temporarily lowered the OVP to 3.46 vpc figuring a high cell will trigger it which it did. Got the notification “Battery fully charged” and zing, the SOC went to 100%. Now I’m skeptical because I’ve been here before just to see the SOC fall every cycle. Then I went back in and raised the OVP to 3.55, a place that it would probably never go to.

Well, it’s been more than two weeks and all three batteries get to full state of charge (99%) at the same time and I’m not seeing any weirdness. So I’m thinking that the algorithm wants to see a cell over voltage protected triggered one or two times at your absorption voltage. It also apparently didn’t notice that I moved the OVP up a bit afterwards. I don’t want my last line of defense and the FETs working every cycle. As long as the batteries are all full at the same time, I’m not concerned about the different batteries trading SOC positions a bit during the cycle, they are after all, three different batteries bought at different times so they have different personalities.
I’m 99% ? sure it’s all good from now on.
And….. I got a spare bms.

 

[dRdoS7]

Hi,

Interesting. I tried this today as my 2 banks (1 old, 1 new) were both out of wack. I have 280Ah cells in each, but have the BMS each set to 250Ah. Over the last 3 weeks or so, one was only going to 247.5Ah, but it did go to 250Ah, only to drop back to again shortlty after. The other goes to 190Ah (had been ~215Ah for a while). The Absorb is set to 55.3, and the total current drops >4A, so must be full(ish). So, I dropped OVP to 2.45V on each, one at a time, the "full charged" message came up, and they both reset to 250Ah Remaining. I turned away for a while, when I returned they had gone down to 245.7V Remaining Capacity! I know that nothing was taken out because I have a shunt on the combined -ves. Weird. I raised both OVP settings to 2.65V (OVPR settings also changed each time on both)

I have seen the Battery Capacity value change by itself from 250Ah to 210Ah when a cell went low. Eventually I removed that cell (that was the 2nd one of 18 that I had bought originally). Down to 16 cells in the old bank. New bank is 16 cells (hope none go bad).

dRdoS7

 

[dRdoS7]

Hi,

Quick error correction & update:

The capacity they dropped to from 250Ah was 247.5Ah, not 245.7Ah. Seems that they increment in 2.5Ah steps now for some reason. I have V4.9.1 on one phone, and V4.15.3 on the other. Don't know when they changed, as they used to increment in 0.1Ah steps in earlier versions of the apk. The Cycle Capacity still does 0.1Ah steps.

They both reach 250Ah when full (55.3V & charge current drops > 4A). Ah readings (in & out) agree with the external shunt measurement which shows 500Ah.

dRdoS7

 

[Skypower]

To tell you the truth, I really haven’t paid much attention to the remaining capacity because the JK’s are communicating with Solar Assistant and I look at that 98% of the time and Assistant doesn’t report that feature. The range of the JK’s Bluetooth is very short and Assistant is everywhere. I don’t worry about state of charge between the different batteries below full. They are always trading places because they are from different sources and have different strengths and weaknesses at different SOC. But I do want them to report full at absorb. So far so good.

 

[dRdoS7]

Hi,

I have a laptop & phones (use scrcpy to view phones over USB or WiFi) near the battery banks, so I can remotely connect to the PC with Remote Desktop (or Teamviewer or UltraVNC).

dRdoS7

 

[Paulfrench35]

To tell you the truth, I really haven’t paid much attention to the remaining capacity because the JK’s are communicating with Solar Assistant and I look at that 98% of the time and Assistant doesn’t report that feature. The range of the JK’s Bluetooth is very short and Assistant is everywhere. I don’t worry about state of charge between the different batteries below full. They are always trading places because they are from different sources and have different strengths and weaknesses at different SOC. But I do want them to report full at absorb. So far so good.
View attachment 169840View attachment 169841

What did you use to connect your bms to SolarAssist?

 

[GXMnow]

I am using a JK BMS with a 14S 360 amp hour Li NMC battery bank. When I was only charging with my Schneider XW-Pro inverter/charger, it would "reset" to 99% SoC after the absorb time would stop. It worked great. But now I added a DC solar charge controller to help charge the battery bank during the day. I have the XW-Pro stop charging 0.5 volts lower now, and the DC charge controller keep charging and brings the battery up the rest of the way. It goes into a sort of float mode as the battery hits the Absorb voltage setting, and it sits at under 1 amps at this voltage for nearly an hour now as the sun goes down and the charge controller is essentially running the inverter. Then the voltage finally begins to fall. I didn't notice any issue at first, but about 2 months after adding the DC charge controller, I now have a seriously wrong state of charge reading. It now drops all the way to zero state of charge when the cells are still at 3.6 volts, which is about 45% for Li NMC cells. I am charging the battery back up to 4.1 volts per cell which should be a bit over 92%. The JK BMS does count up the amp hours flowing in and it reaches a state of charge reading of about 47% when the charging stops. So it counted up 169 amp hours into the battery, but starting from 0% SoC instead of the real 45% SoC it was really at.

I will try that trick tomorrow, and drop the cell OVP down to 4.05 volts and see if it trips it to go back to 100% again. I have a second battery cabinet in parallel, so having one BMS shut off will not shut my system down. The other battery is the same capacity, using the same cells, but with just pair of dumb Daly BMS units. Those cells are all holding within 6 mv of balance with just the 60 ma passive balance in the Dalys so I am not worried about them. The JK is working just fine as well, and I know by the cell voltages that I am not having any problem, but it is odd that the SoC on the JK app is just so far off now. It had been resetting to 99% basically every day it hit absorb for over 2 years. Then after just 2 months with the DC charge controller added, it drifted down to zero SoC any night the cells go below 3.6 volts. Zero SoC on Li NMC cells is about 2.8 volts per cell.

 

[Skypower]

I am using a JK BMS with a 14S 360 amp hour Li NMC battery bank. When I was only charging with my Schneider XW-Pro inverter/charger, it would "reset" to 99% SoC after the absorb time would stop. It worked great. But now I added a DC solar charge controller to help charge the battery bank during the day. I have the XW-Pro stop charging 0.5 volts lower now, and the DC charge controller keep charging and brings the battery up the rest of the way. It goes into a sort of float mode as the battery hits the Absorb voltage setting, and it sits at under 1 amps at this voltage for nearly an hour now as the sun goes down and the charge controller is essentially running the inverter. Then the voltage finally begins to fall. I didn't notice any issue at first, but about 2 months after adding the DC charge controller, I now have a seriously wrong state of charge reading. It now drops all the way to zero state of charge when the cells are still at 3.6 volts, which is about 45% for Li NMC cells. I am charging the battery back up to 4.1 volts per cell which should be a bit over 92%. The JK BMS does count up the amp hours flowing in and it reaches a state of charge reading of about 47% when the charging stops. So it counted up 169 amp hours into the battery, but starting from 0% SoC instead of the real 45% SoC it was really at.

I will try that trick tomorrow, and drop the cell OVP down to 4.05 volts and see if it trips it to go back to 100% again. I have a second battery cabinet in parallel, so having one BMS shut off will not shut my system down. The other battery is the same capacity, using the same cells, but with just pair of dumb Daly BMS units. Those cells are all holding within 6 mv of balance with just the 60 ma passive balance in the Dalys so I am not worried about them. The JK is working just fine as well, and I know by the cell voltages that I am not having any problem, but it is odd that the SoC on the JK app is just so far off now. It had been resetting to 99% basically every day it hit absorb for over 2 years. Then after just 2 months with the DC charge controller added, it drifted down to zero SoC any night the cells go below 3.6 volts. Zero SoC on Li NMC cells is about 2.8 volts per cell.

Im thinking that that should work.
You may have to do this method twice within a few days for it to be deeply ingrained in the JK’s memory.

 

[GXMnow]

Im thinking that that should work.
You may have to do this method twice within a few days for it to be deeply ingrained in the JK’s memory.

The sun is out and the solar panels are running my house and there is still 50 amps charging the battery bank. 25 amp are going into the battery bank with the JK BMS. I already went into the BMS and set the OVP to 4.1 volts per cell and the OVPR to 4.08 volts per cell. The voltage topped out at 57.4 the last few sunny days, so that should trigger it. If clouds move in or it does not quite get full, I can lower the voltage setting a little more. I'll report back what happens.

 

[GXMnow]

Of course I had to run out for a bit and I missed watching it happen, but it worked!

The BMS reset to full, not sure if it went to 99 or 100% as it was down to 95% when I was able to go check it. There is a little jump in the battery voltage trace in the Schneider inverter which I am guessing was the point the charge FETs opened.

I am happy that it worked to reset it to show a reasonable SoC, but it sure seems dumb to require it to hit a protection error mode for it to happen. I set the OVP back up to 4.2 volts and the OVPR to 4.15 volts.

Thank you for the tip

 

[Skypower]

Just to reiterate for everyone else watching, his voltage is NOT for Lifepo4 chemistry!

 

[Skypower]

Of course I had to run out for a bit and I missed watching it happen, but it worked!



I am happy that it worked to reset it to show a reasonable SoC, but it sure seems dumb to require it to hit a protection error mode for it to happen.

Thank you for the tip

Yeah, I’m not thrilled with that either. Would have been a lot nicer to just dial in a voltage in a settings window.

 

[GXMnow]

Yes, I have NMC cells which are 3.7 volts nominal, 4.2 volts max, and about 3.0 volts minimum.

I just noticed another odd thing. The cycle count changed.
Before I did the reset, the BMS showed that I had cycled 100671.9 amp hours in and back out of the battery. With a rating of 360 amp hours, it calculated that to be 279 cycles. 100679.1 / 360 = 279.66 so that number made some sense. It is now almost 7 hours later, the cycle capacity is now at 100693.7 so 14.5 more amp hours. That is far less than 1 more cycle. 100693.7 / 360 = 279.70 but the cycle count now shows 980 cycles. Hmmm

Did it store in memory and just count up how many times the voltage went above 4.09 volts?

It is still less than the number of days as it has been running well over 3 years. I typically cycle less than 40% capacity per day.

Here are the screen shots from earlier in the day and after the reset.

 

[Skypower]

Beats the heck out of me? I don’t know if I would have noticed the cycle count difference. I defer to your theory as good as anything. Mine doesn’t always count a day as a cycle. My depth of charge averages 35 percent. It may go 3 days without hitting 100% because I have it set so it avoids grid as much as possible. Rarely goes below 50%

 

[Skypower]

Note the days vs the cycle count. It’s typical on all my JKs

 

[GXMnow]

Before the reset, my ratio was close to yours, about 3 days to get a full cycle count. And if you divide the cycle capacity by the battery capacity, it does match your 98 cycles, like mine did before the reset. Very odd.

 

[Skypower]

Before the reset, my ratio was close to yours, about 3 days to get a full cycle count. And if you divide the cycle capacity by the battery capacity, it does match your 98 cycles, like mine did before the reset. Very odd.

Different use profiles and probably different algorithms in the BMS. Yours is one of the earliest. I know that I have at least two versions because of how the order of data is displayed on the app and different ways to enter the current and voltage adjustments. Yours are probably some of the best. More careful with assembly and pre-Covid parts shortage/substitutions. The latter has played havoc across the entire electrical industry. How many fords sat in lots just waiting for a ECM?

 

[GXMnow]

I think I figured out what the BMS did.

Yes, my BMS is almost 4 years old now. I have updated the phone app a few times, but the firmware in the BMS has never changed. I have never seen any way of pushing a new firmware into it. I took a screen shot of the "About" screen. The manufacture date is 2020-06-19 so it was shipped to me as it was built. The hardware version is V3.0 and the software is V4.1.4 From the first time I powered it up, it has been 3 years 107 days. But it was about 3 months before it went into daily service. So it has been cycling for right about 3 years now. At the bottom of the page, it also has a Version V171 but no idea if that is something inside the BMS or from the iPhone app.

Because it drifted so far off on the SoC, it had been hitting what it thought was 0%. Then when I forced it to see 100%, it saw that it only took 103 AH from when it was ot it's 0% to see 100%. Ah, the battery is now only 103 amp hours. If I calculate the cycles and SoC based on 103 Amp Hours, then it all agrees. Less than 1/3 the capacity means it made over 3x as many cycles, an the SoC drops 3x faster than it should for the AH removed.

It got as low as 13% SoC this morning before the sun started to charge it back up. But here is the one odd part. At 13% it showed 46.8 AH left. So it still calculated the AH remaining based on 360 AH total.

To try and figure this out, I will estimate a bit, so these numbers will not be perfect, but it seems to fit.
In the nearly 9 hours from midnight to before it started charging, the XW-Pro inverter shows that it has pulled 5.9 KWHs from both batteries. I will call that about 655 watts of average draw. I'll round it up a bit and say this one battery was supplying about 350 watts per hour. 9 hours this morning and another 7.5 hours last night from sunset to midnight. 16.5 hours total from sunset to sunrise x 350 watts of average power = 5,775 watt hours removed. The average battery voltage (just max + min / 2) over that time is about 54 volts. 5,775 / 54 = 107 amp hours. Very close to the 103 amp hours calculated capacity value that it seems to be working off of now.

These "smart" BMS units seems to be trying to second guess us. I wish they would just list the logic they use so we know what they are doing.

The SoC has gone back up 8% in less than an hour since it started charging today. The XW-Pro inverter is reporting that it has charges 943 watt hours to the battery bank and the DC Charge controller reports that it charged another 377 watts hours so far today. Add those up and 1,320 watt hours have gone to battery. But again, this JK BMS is only half of my battery capacity. The other bank is basically the same cells, but 2 years newer. The average voltage since it started charging is a bit lower. Looking at the battery voltage graph in the XW-Pro, I will call it 52.7 volts. 1,320 watt hours / 52.7 volts = 25 amp hours. My total capacity with both batteries is 720 amp hours. 25 / 720 = just 3.5%. But the JK BMS is already showing it has climbed more than 8% in SoC.

I am very curious to see how fast it keeps climbing the cycle count. The numbers do not mean anything to how my system works. I have a better idea of the system state than it is reporting. I have never found a cycle life spec for these exact cells, but since they came out of a Chevy Bolt EV, I know that Chevy is warrantying them for 10 years 100,000 miles in the car. My battery bank is a bit over 1/2 the battery in the car. I have a total of 168 cells, the car has 288 cells total. That gives the whole car 62 KWHs of capacity, and I have 36 KWHs. 58.33% of the car. 100,000 miles at 3.5 miles per KWH means a cycle life of just over 28,000 KWHs from 62 KWH of battery. 28,000 / 62 = just about 450 full cycles. So GM and LG were pretty confident about that life expectancy. But they still ended up having to replace 140,000 battery packs under the recall. Looking at other LG made Li NMC pouch cells, they list 2,000 to 2,500 full 100% cycles at 1C charge discharge rates. I am always under 0.2C rate and stay between 40% and 95% SoC. I am hoping these cells stay working for about 10 years in my setup. When they do start to show reduced capacity, I will shop around and see what cells I want to use to replace them. With all the EV's coming to market, there may be a flood of cells showing up. Or I may break down and switch over to LFP or something else even newer by the time these die.

 

[Skypower]

That made my brain hurt. I dub thee the new resident JK logic wizard LOL.
LFP prices are getting better. Probably because something else is about to come out. But I’ve learned never to be a beta tester or buy a first year model. I’ll stick with LFP for Long life(longer than me) proven and less volatile personality.

 

[GXMnow]

That made my brain hurt. I dub thee the new resident JK logic wizard LOL.

My JK BMS still has a few kinks to throw at me. I am not the wizard on this thing.

After 3 full days, I was curious to see how much the cycle count changed.
Did it take the 3 days to just count 1 like before?

Well, no. It went down.
It was 980 on Saturday, and stayed there until afternoon on Sunday.
But then when I checked it at3:30 pm on Sunday, it dropped to 891. That would be about when charging stopped with SoC showing 100%.
And then on Tuesday it had dropped again to 875, when I checked it at 5 PM. It had started discharging for a bit and SoC was showing 90%.

I certainly do not know what it is doing, but I have one guess. Maybe it is using the largest cycle size since the reset as the capacity?
The mystery continues.