Seeking Normally Closed Latching Contactor and Q&A about balancing my Li Ion battery bank

karrak

If only it were so easy!

The problem, if you can call it that is that the charge/discharge curves for LFP batteries are too flat to use voltage as a reliable indicator of SOC. You can however very accurately measure the SOC by measuring the current going into and out of an LFP battery.

Due to the internal impedance of the battery depending on current, 3.45V while charging could mean  an SOC between 100% and 80% or even less, discharging at 3.26V the SOC could  be anything from 70%-30% and at 3.0V could be anything from 10% to 5%. Temperature and variability between the cells also has an impact.

If you want to accurately charge to a particular SOC you have to use an SOC meter to terminate the charge or shutdown the loads. The SOC meter will have to be calibrated on a regular basis by charging the battery to at least 3.45/cell with an end current ~C/50 and calling that 100%. You could use the equilise function on some charge controllers to do this.

If the system is just a backup, not using it full time with your loads only using around 20% SOC per day you could set the charging voltage to around ~3.35V/cell and drop to a float voltage of  ~3.30V/cell with no absorb time which would charge the battery to around 50%-70%SOC. You would have to fine tune the charge and float voltages.

Simon

jonr

I understand coulomb counting being great for knowing SOC - quite useful for a "fuel gauge".   Less clear to me is how useful it is for charging cells.  Is there some data/research showing how much it helps?    It must vary with charge/discharge profiles, but I could see it allowing faster charging and better life (easy to stop short of full charge).

LostinSpace

Yes. I sort of presumed that once out of the flat part of the curve (on either end) that volt readings would have more meaning.

I find it interesting that the cells stayed very balanced 60 hours into the discharge cycle. But I guess the drop off of some cells after that could be attributable to any of the three reasons you mentioned.

I need to order a new wizz bang. Will do so shortly.

In the meantime, there is an irresistible opportunity at this juncture to make a determination as to the true efficacy of the balancers. I am disposed to pull the array apart once again, charge all the cells in parallel with the LFP charger as before, and configure the array again... this time with the balancers connected. And then have a look at how well the cells are balanced after a discharge cycle.

That should be telling.

karrak

If it were me I would leave the battery as it is, connect the balancers up and do small daily cycles.

The reason for this is that currently charging to 3.4V/cell is not taking any cells out of their safe operating zone and that if you put the balancers on and they are working correctly they should over a period of time correct the imbalance.

As for getting a new wizz bang Jr. Have you thought about other options?

Simon

LostinSpace

karrak said:

As for getting a new wizz bang Jr. Have you thought about other options?

It seems a clean integrated solution. Do you have some other ideas? The Magnum MMP panel has a 50mv shunt.

karrak

jonr said:

I understand coulomb counting being great for knowing SOC - quite useful for a "fuel gauge".   Less clear to me is how useful it is for charging cells.  Is there some data/research showing how much it helps?    It must vary with charge/discharge profiles, but I could see it allowing faster charging and better life (easy to stop short of full charge).

I know from experience how accurate coulomb counting is for determining SOC. The reason it is so accurate is because the coulomb charge efficiency of LFP batteries is around 99.5% and doesn't vary much with differences charge/discharge rates. This is in stark contrast to lead acid batteries. From the logged data from two off grid systems that I have installed I have found that if the charge losses are taken into account the SOC readings  are accurate to within a few percent using coulomb counting, even after several weeks of the SOC meter not being reset by fully charging the battery.

As you say, stopping the charge below 100% will increase the lifespan of an LFP battery, especially if it is used for backup power and not used on a continuous basis and the only way to accurately do this and know how much charge you have in the battery is to do coulomb counting.

Simon

karrak

LostinSpace said:

karrak said:

As for getting a new wizz bang Jr. Have you thought about other options?

It seems a clean integrated solution. Do you have some other ideas? The Magnum MMP panel has a 50mv shunt.

For some reason I thought the Wizz bang Jr. was much more than just an interface to the shunt. I agree that it is a good clean integrated solution.

I was thinking of something more along the lines of my battery monitor that can be connected to the internet.

Simon

LostinSpace

karrak said:
For some reason I thought the Wizz bang Jr. was much more than just an interface to the shunt. I agree that it is a good clean integrated solution.

Oh it is. It is a physical device that attaches to a/the shunt and can track SOC (when reset properly) and help you discern the difference between generated power being used by the load vs going into the battery.

Raj174

LostinSpace said:

karrak said:
For some reason I thought the Wizz bang Jr. was much more than just an interface to the shunt. I agree that it is a good clean integrated solution.

Oh it is. It is a physical device that attaches to a/the shunt and can track SOC (when reset properly) and help you discern the difference between generated power being used by the load vs going into the battery.

I'd highly recommend using it also. It's well integrated with Classic and IMO is very accurate.

Rick

LostinSpace

UPDATE: The science project continues. I have been consulting with @karrak offline. Have been testing batteries further and some results are coming in.

Firstly, after dividing my qty 32 LFP cells into two banks (0 & 1) I charged each bank with the LFP charger to 3.6v (the aforementioned gold standard). I then connected each bank in turn to the inverter and ran my freezer for a load (note that this is a relatively light, but variable 0-250w load).

So all cells were equivalent voltage to start. For each bank, the cells were checked every few hours to see when the divergence in cell voltage occurred (using the celloggers). As it tuns out the cells stayed remarkably in balance through most of the discharge cycle for both banks. The attached "Cell divergence BANK x.pdf" files demonstrate that it is only when the cells dropped below 3.1x that the voltages began too sharply diverge.

In mapping the percent each cell dropped below the highest cell in the array (% of divergence BANK x.pdf) it turned out only a few cells were more than 3% below the highest cell. I counted 7 cells (out of 32) as being noticeable lower (within the sample). The worst cell was 8% off from the high cell (3.102v vs 2.853v).

So the question is, is this a problem? The evidence suggests that if the pack is never discharged beyond that point of divergence, say 3.150 (50.4v) then all is well. The answer it seems lies in what % of its rating has the pack delivered at this point. These are 72AH cells. Do I get 72 amp hours out of the pack when it has drained to 50.4v? I realized that I needed to determine whether I had a bunch of 72AH cells, and a few cells less than that rating ... or ... do I have a few 72AH cells and a bunch of cells rated over 72AH. Right?

I want to make that determination because in researching LFP cells my take was that discharging the pack to 2.7v (43.2 pack volts) was considered an 80% discharge, or 20% remaining. I got that figure from well known solar forums contributor PNjunction. So that was the discharge voltage I have been shooting for. But as karrack has been reminding me, it's really about watching individual cell voltages (whose charge and discharge rates vary over time), and not relaying on a set pack voltage.

So what I think got me into trouble is that I discharged the pack to what I thought was a safe level (and I wasn't even shooting for 43.2v so far. Rather, the inverter was set to to turn off at 48v, or 3.0v), and in doing so threw the pack way out of balance. Then, when charging an out of balance pack to 3.45v (55.2v) by the time the pack reached 55.2 volts some cells had gone high and were damaged. All this likely happened early on, before buying the celloggers. One obvious moral of the story is to never hook up an array of LFPs without cellogers or equivalent in place.

----------------------------
So what I am doing now is taking a full charged pack (to 3.6v) and tracking amp hours used during the discharge. The question to be answered is how many AH have been used when pack voltage reaches 50.4v? Which is the point the voltages started diverging last time. I will post the results for both banks when they are complete.

jim

(Simon, do you see any discrepancies in my take on this at this point?)

Raj174

@LostinSpace
I really don't see anything out of the ordinary with your cell balance. They are within ~10 mV at the top and are good down to about 20% SOC, where they begin a more apparent divergence. When they are recharged, they should come back up to the same relative balance. They should charge and discharge with very near this same relative balance for hundreds of cycles. My LFP bank is very similar, at some point below 20% the smallest cell will start radically diverging.  The bottom of a top balance is very much out of balance and it's really best not to go there. I rarely take my bank lower than 30%, but have taken it down to 20% a few times. For me, a floor of 30% is very safe and is better for the longevity of the cells.

BTW, I think 2.7V per cell is way more than an 80% discharge. I would think it's closer to 3.1V per cell.

I really appreciate you experimenting and posting your experience with your bank here, as we all learn and become more familiar with LFP chemistry in the process.

Rick       

jonr

IMO, if you don't have automatic balancing, the important thing to watch (after individual cell limits) is the rate of between-cell divergence in the top and bottom balance.    Knowing this, you can predict when to manually re-balance. 

LostinSpace

Raj174 said:

@LostinSpace
BTW, I think 2.7V per cell is way more than an 80% discharge. I would think it's closer to 3.1V per cell.

Rick       

The values for fully charged and fully discharged LFP cells has always seemed to be a moving target in the materials I've studied. I am just now starting to feel comfortable with what is considered 100% charged. Discharge limit still seems nebulous. The spec sheet for my brand of cell (New Energy) actually specifies 2.0v as the "DISCHARGE CUT-OFF VOLTAGE". WTF?

PNjunction has stated to cut off at 2.7 under load, but he also states the he normally cuts off at 3.0.

So this is why using a different metric to determine when the cells are 80% discharged may be of value; i.e. AH rating of the cell. The test I am running on my 16 cell pack started with a voltage of 3.57-3.58. Let's call that 100% charged. So how many amp hours would be expended to drain the cells to a 20% remaining charge? Well, we divide 72 into 5 parts (20%, 40%, etc). 72/5=14.4.

14.4 * 4 = 57.6. So from a 100% charge, there is 57.6AH of useful charge before the battery reaches 20% of charge. Right?

So what I am looking for is to see where cell voltages are when the pack has spent 57 or so amp hours. As I write this the pack has spent 51.6AH, and cell voltages are in the 3.2x range.



Cell specs: http://www.electriccarpartscompany.com/New-Energy-72Ah-Aluminum-Encased-Battery


(solar panels NOT connected. Discharging only)

karrak

LostinSpace said:So the question is, is this a problem? The evidence suggests that if the pack is never discharged beyond that point of divergence, say 3.150 (50.4v) then all is well. The answer it seems lies in what % of its rating has the pack delivered at this point. These are 72AH cells. Do I get 72 amp hours out of the pack when it has drained to 50.4v? I realized that I needed to determine whether I had a bunch of 72AH cells, and a few cells less than that rating ... or ... do I have a few 72AH cells and a bunch of cells rated over 72AH. Right?

Have you got your replacement Wizz Bang Jr yet?

I want to make that determination because in researching LFP cells my take was that discharging the pack to 2.7v (43.2 pack volts) was considered an 80% discharge, or 20% remaining.

Under light load 20% is around 3.2V/cell.

So what I think got me into trouble is that I discharged the pack to what I thought was a safe level (and I wasn't even shooting for 43.2v so far. Rather, the inverter was set to to turn off at 48v, or 3.0v), and in doing so threw the pack way out of balance. Then, when charging an out of balance pack to 3.45v (55.2v) by the time the pack reached 55.2 volts some cells had gone high and were damaged. All this likely happened early on, before buying the celloggers. One obvious moral of the story is to never hook up an array of LFPs without cellogers or equivalent in place.

It sounds plausible that this explains some of what has happened.

I have had to do a forced experiment of taking my battery down to zero due to three cloudy days with some of the worst solar input I have seen. About an hour ago the measured SOC was ~10% of the rated capacity, minimum cell voltage was 2.749V and maximum voltage was 3.059V, battery voltage was 23.7V. The inverter is supposed to shut down at 24V but had not shut down! Hopefully I will see what this has done to the top balance in the next few days when the battery is recharged.

So what I am doing now is taking a full charged pack (to 3.6v) and tracking amp hours used during the discharge. The question to be answered is how many AH have been used when pack voltage reaches 50.4v? Which is the point the voltages started diverging last time. I will post the results for both banks when they are complete.

(Simon, do you see any discrepancies in my take on this at this point?)

One important point is that you have to fix up the top balance before you do your capacity test.

Again, thanks for the information

Simon

LostinSpace

Simon, see above. Yes, should have mentioned that the new Whizbang is in place. And I did fix up the top balance following the "gold standard". I am discharging from a near 100% charge. :>

LostinSpace

karrak said: The inverter is supposed to shut down at 24V but had not shut down!

Simon

Exactly why I started this thread! With a multi-thousand dollar investment in these cells we everything has to be redundant. Murphy loves Solar systems!

karrak

Raj174 said:

When they are recharged, they should come back up to the same relative balance. They should charge and discharge with very near this same relative balance for hundreds of cycles.

This is the theory, Jim mentions this "Upon recharging to 54.4v (3.4v) cell voltages were 171mv apart (3.352 - 3.523)" in this post. It certainly has me puzzled. I can't see how discharging a cell down to a safe voltage of ~2.8V should have such a dramatic effect on the balance.

Simon

karrak

We must have been writing replies at the same time

LostinSpace said:The values for fully charged and fully discharged LFP cells has always seemed to be a moving target in the materials I've studied. I am just now starting to feel comfortable with what is considered 100% charged. Discharge limit still seems nebulous. The spec sheet for my brand of cell (New Energy) actually specifies 2.0v as the "DISCHARGE CUT-OFF VOLTAGE". WTF?

I think you will find that 2.0V would commonly be called the Absolute Minimum Discharge Voltage. There is no point going that far as there is only at most a couple of % of charge between 2.0V and 2.8V

So this is why using a different metric to determine when the cells are 80% discharged may be of value; i.e. AH rating of the cell.

I have set my SOC counter to read zero at 10% of rated capacity.

Simon

LostinSpace

karrak said:

I think you will find that 2.0V would commonly be called the Absolute Minimum Discharge Voltage. There is no point going that far as there is only at most a couple of % of charge between 2.0V and 2.8V

That was my assumption as well. But it seemed to me that with that low value of 2, and the stated high value of 3.65 you would have what the factory is specifying as 100% and 0% charge states, and could then calculate what 20% remaining charge is for that brand of battery. But that calculates to something like 2.33 volts, so this must be a dead-end approach given how LFP chemistry behaves. Oh well.

karrak said:

I have set my SOC counter to read zero at 10% of rated capacity.

Funny you should mention that. I just posted a thread on the Midnite forum asking how the Whizbang Jr, calculates 0% SOC. The only value you can set is AH rating of the battery. So they are calculating 0% internally, and I'm trying to find out if 0% equates to 0AH remaining, or perhaps 20% of AH remaining.

Of course I should see for myself once this test completes.

jonr

 > I'm trying to find out if 0% equates to 0AH remaining

Enter 100AH as the capacity, draw exactly 10AH and see what it reads.

LostinSpace

Thanx jonr. I also think I will find out when this test is complete. ;>

LostinSpace

Discharge test of this bank is complete. It appears I am getting the advertised amp hour rating from the batteries. The expectation was that the array could provide 57-58AH, or a discharge from 100% down to 20% remaining AH. At 57 spent amp hours pack voltage read 51.4 (3.2125). Cells were still in a tight voltage range.

So if I have this right, all cells are providing, at a minimum, the rated capacity. (see next comment though)

LostinSpace

Next step: How low can you go (with cells remaining in balance). This has already been established, but the highest voltage my Magnum inverter will turn off for LOW VOLTAGE CUTOFF is 48.8 volts. The setting will increment no higher than 48.8. So I wanted to see how the cell balance looked at that cutoff.

At 48.8v (3.05v) some divergence had occurred, but not to bad (see screenshots of the cellogers). In particular note cell #4 on the left bank. It is the only cell that has dropped into the upper 2 volts range. Is that important? Keep reading. This is where things get weird, and go right back to the issue I have had with a few cells going high and being destroyed when charging.

I turned on the onboard charger and started to charge the array. In general the cells were in balance as voltage rose. However at around 54.2 (or possibly sooner) one cell went off the reservation, jumping to 3.5x volts, with all other cells still in the 3.3x range. Curiously it is the same cell #4 that went lowest on discharge. This necessitated turning the charger off and leaving the pack very undercharged.

Is there any accounting for this behavior other than a bad cell?

karrak

I wouldn't use your 57 Ah figure to calculate the battery capacity. I would use the 66 Ah reading when the first cell reached 2.8V.

What current were you charging each 72Ah battery with?

Raj174

@LostinSpace
It looks like cell #4 has a diminished capacity. Right?

LostinSpace

Raj174 said:

@LostinSpace
It looks like cell #4 has a diminished capacity. Right?

Gosh. I guess so. I was thinking if that were true it would be lagging pack voltage. But yes, I see now. With diminished capacity it would reach full charge before the others. Thanx

LostinSpace

karrak said:

I wouldn't use your 57 Ah figure to calculate the battery capacity. I would use the 66 Ah reading when the first cell reached 2.8V.

What current were you charging each 72Ah battery with?

the charger was set for 30A max for the pack.

Raj174

LostinSpace said:

Raj174 said:

@LostinSpace
It looks like cell #4 has a diminished capacity. Right?

Gosh. I guess so. I was thinking if that were true it would be lagging pack voltage. But yes, I see now. With diminished capacity it would reach full charge before the others. Thanx

And as you noted, it reached the bottom first too. Again, indicating lower capacity.  

karrak

LostinSpace said:

Raj174 said:

@LostinSpace
It looks like cell #4 has a diminished capacity. Right?

Gosh. I guess so. I was thinking if that were true it would be lagging pack voltage. But yes, I see now. With diminished capacity it would reach full charge before the others. Thanx

This doesn't make sense if the battery is top balanced and remains top balanced.

Lets say we have two cells, one has a capacity of 90Ah and the second 100Ah. If we top balance them they will both be at the same voltage at 100%SOC. If we take 90Ah out of the two cells one will be at 0%SOC and the other one will be at 10%SOC. If we then add 90Ah to both cells they both return to 100%SOC and the same voltage if the two cells have remained top balanced.

Simon

LostinSpace

EDITED FOR ACCURACY:

Yes I was just thinking about this again after Raj's last post . Cell #4 was only about 30MV below avg when the pack was discharged. It didn't "crash" lower. Yet at the top it "blew out" so to speak by jumping about 200mv higher than the other cells.

By the way. I have not mentioned this yet to keep things on track, but on the other bank (the one NOT tested today), after the discharge test, which took the cells lower than today, I charged that pack to see how it would behave. It had a substantial imbalance; 124mv. I stopped charging before any cell could go too high, and attached the balancers. I was happy to see that overnight the balancers corrected the imbalance by closing the divergence to 7mv! I then completed the charging cycle, and got caught yet again upon discovering that one solitary cell had gone to 3.8v. No other cell exceed 3.4v. I hadn't been watching the pack minute to minute thinking the pack was balanced. But someone one cell took off like a rocket.

I don't know what to do except to repeat these top balance / discharge / recharge cycles until all the oddballs are culled out.