Balancing Unmatched Prismatic Cells

[Dave McCampbell]

We have a 4S2P bank, 540ahrs, and are charging with solar at 30-40a on a daily basis.  We can now track individual cell/parallel cell charge curves and most other parameters with Sherry's Node Red display.  Using this we can clearly see separate charge curves for our Lishen and RJ cells.  With these cells the problem using any balancer while charging seems to be that the voltage curves intertwine on the way up.   They did the same with just the RJ cells.  Our SBMS0  balancer will sometimes be taking milliamps from a cell that is high on the way up but low at charge termination.   So it may not be very effective at reducing the voltage delta at charge termination.  

We had been using default SBMS0 balancer settings, but have recently bumped the start charging to 3.3v and minimum delta to 15mv.  We are still seeing the same problem just a bit later on during charging.  Our delta now at charge termination is steady at about 20ma.

Does someone have an explanation for how a balancer can be effective with these types of seemingly unmatched cells?   Are we missing something here.?  Dave

[Dacian Todea]

Dave,

A photo of the SBMS0 main monitoring screen close to end of charge or a graph will be way more helpful.

If cell's stay below 10 to 15mV during charging and one of the cell increases above that around end of charge (above 98%) then likely the pack is balance or very close to balanced and will fully balance in weeks or months depending on how much you use the battery and how it is setup.

Why will you increase cell balancing minimum form 3.2V to 3.3V (assuming that is what you changed) ? As cell balancing is anyway only done during charging there will be no cell balancing during discharge to affect the balancing.

What sort of charge rate or charge current do you have ?

I'm sure you already seen this graph before as I post this quite often and what that shows is that in a relatively balanced pack voltage of cells will be very similar until above 98% where cell with lowest capacity will start to increase in voltage usually all cells will be around 3.4V by this point meaning all cells may be around 98 to 99% so there will be less than 1% delta between cell and even with less than 0.1% delta one of the cells just before end of charge will be the first to get to 3.55V. If the cell delta increase above 10 to 15mV happens in the last 2 to 3 minutes of charging then is likely just normal operation. Also when cell balancing is enabled there will be huge variation while cell balancing in ON and that is not real variation but is measurement error due to cell balancing wire voltage drop and SBMS knows to ignore that but it will display that for user.

  

[Dave McCampbell]

Dacian.  Thanks for the reply.  Will get a Node Red cell charge graph up soon.  Here is the issue:

How can the balancer be doing the right thing if a cell is HIGH in voltage and being balanced while at lower charge voltages, and then that cell ends up being consistently LOW near/at charge termination?  So the balancer is shunting current from a HIGH voltage cell early on during charging, when it should not be doing that.

We are just trying to understand how a balancer, such as is in SBMS0, can deal with cells that have slightly different charge curves.  This seems to be the case with all these Chinese prismatic aluminum case cells.  We can see how a balancer would work well if all the cell voltages stayed in their same relative positions throughout the charge curve until the end.  We probably would not have noticed this shifting of high and low voltage cells during charging had we not plotted all our cell charge curves during initial testing, and then later Sherry not developed the Node Red display that shows individual charge/discharge curves on a daily basis.  Unless we had been paying very close attention to cell voltages during charging, we would not have noticed this using the SBMS0 display. 

As we mentioned above, we are charging with 30-40a using solar daily here in the Philippines.  We did increase the cell balancing minimum from default 3.2vpc to 3.3vpc and the delta from default 11mv to 15mv to see it that would make any difference in what the balancer was doing.  It only keeps the balancer from turning on early in the charge cycle, but does not solve the issue.  Our delta is now pretty steady at about 20mv at charge termination which will work fine for now.  But I suspect it will not get better on its own with the balancer working against it.  We will let it run for a few more days and see.  We can, however, reduce the delta by using our bench power supply and adjustable load.   

We understand, and can clearly see, the large balancing deltas at the end of charge on the Node Red display.  Yes, I have seen the graph before, and it is helpful to understand the generic charge curves for LiFePO4 at different C ratings.  But that is not the problem we are trying to understand.  

[Dacian Todea]

Dave,

All cell's should increase in voltage equally when applying a charge current and cell balancing is not active (even if it is active but then you may wrongly interpret the data) so that low cells should never be balanced as it should always the the lowest voltage cell.  If that cell goes from lowest cell to highest voltage cell while charging that means one of two things (either cell has a bad connection or cell is damaged and has significantly higher internal resistance). Bad connection will be much more likely.

Without more details I can not be sure what is the issue. Even 100mV delta at end of charge 100% SOC will not be a problem and can be expected.

[Dave McCampbell]

Dacian,

Below see two Node Red graphs of the charge curves for our 4 paralleled sets of 271ahr cells, 4S2P, using a low charge termination for now.  We currently have our balancing set to start above 3.3vpc and at a delta above 20mv.  Earlier we had the balancing set at your defaults.

In the first graph it appears that the delta did not get above 20mv until the cells reached about 3.43vpc.  Had the balancer been working earlier it would have been pushing the green cell down even though that is not the high cell at cell termination.  You can also see that the orange cell starts out low when the cells are at around 3.36vpc, and then goes high near charge termination.  It is getting drawn lower by the balancer starting around 3.43vpc, while the green cell is high to begin with, then evens out with black and blue cells at charge termination.  We are not sure what the green cell indications at about 3.45vpc represent. 

In the second graph you can see more of the same as the orange cell goes from lowest voltage to highest.  Then later the orange and green cells start to swap high/low positions at around 3.42vpc and as they approach charge termination, but before the balancer cuts in. 

We can move the orange and green cells a bit closer to the black and blue cells manually by using our bench power supply and adjustable load.  But I am not sure that any balancer would do that properly with these cells.  Here is an explanation from another forum about this problem:

"When we reaching the charging shoulders, the cells with lower inner resistance are simply fuller and overtake the balanced cell. The balancers switch to them, but have very little time to fix the mistake they made in the first place leading to imbalances. The only mitigation is to start balancing as late as possible, when the cells enter the shoulders and leave them alone in the plateau phase".  He has set his balancer to start balancing at 3.42vpc.  

We are pretty sure that this is caused by the cells not being well matched for internal resistance and the change in internal resistance as voltage changes.  There are no loose connections and there has always been this mismatch of charge curves with our RJ and now RJ/Lishen mix of cells.  The same is true of the other 16 RJ cells that were bought with ours; the charge curves just do not exactly match.  We suspect that this is true of most of the lower priced Chinese prismatic cells, but most users can't see it.

We think that the best course of action is to get the cells as close as possible manually, then raise the start balancing to just above 3.4vpc or wherever the cell voltage positions stabilize, and then reduce the delta to the default at 11mv.  What do you think?

[Dacian Todea]

Dave,

I wish I had the graph for the battery current but it sor to looks the current significantly increases around 11

Cell 1 is the first to be 20mV above cell 3 so is the one that is balanced thus measured voltage is lower even if the cell voltage is of course not lower that is just the measurement error due to cell balancing current and same cell balancing current on cell 1 affects the measurement on cell 2 that shows the real voltage of cell 2 + the voltage drop due to cell balancing current on cell 1

All seems perfectly normal for a well balanced pack. Not sure why the charging seems to have stopped around 11:20 but maybe due to logging interval the averaging shows lower voltage or you modified the Overvoltage setting from default 3.55V and 6 seconds to something else.

The cell 1 seems to exceed cell 2 while battery charge current increases so it is very likely that is from a higher resistance contact either between cell 1 and 2 or on the cell 1 negative terminal that will also be the main battery negative.

It is also possible cell 1 has slightly higher internal resistance compared to the other cells but it is not super significant.

In any case all cells are fully charged at around 11:20 at least 99.9% SOC with no more than 0.1% SOC delta between lowest and highest cell.

There is just no advantage in changing default settings as with default settings maybe the cell 2 will have started balancing just a bit earlier than cell 1 but soon cell 1 will also have been balanced as both will have required that to get them closer to cell 3 and 4.