Parallel Cell Capacity Balancing (PCCB) Procedure.

Short_Shot

Photon Sorcerer
Joined
Jul 13, 2021
Messages
1,802
Posplayr,

Just seeing this thread today. The engineer (mechanical) in me has been thinking of using this type of cell matching approach when my cells finally arrive. Your results have convinced me to pursue it.

Thanks!

100 Proof
Yeah. It's effectively mandatory when building a battery.
 

Posplayr

Solar Enthusiast
Joined
May 27, 2021
Messages
140
At your point of view it must be truly appalling. i was simply looking at it from a GS12 perspective on the management side.
Probably worse than appalling at the end of one's career to get slapped in the face with the ultimate recognition that it is NOT important WHAT you know but rather WHO you know. While this is true in both the commercial and government sectors. In government despite all the pretenses of competitive bidding and fair procurement laws, who you know predominates as the factor to success.
 

Posplayr

Solar Enthusiast
Joined
May 27, 2021
Messages
140
Yeah. It's effectively mandatory when building a battery.
I'm also building a 32 cell 16S battery. I just placed an order with AMY for the latest LF280K matched 280 amp-hr EVE cells.

This PCCB matching process can easily be completed during the normal capacity testing that is recommended when you receive cells. However, it just occurred to me that with the detailed test reports coming from the manufacturer (EVE) that the matching can probably be done from the manufacturer's capacity testing results.

I am planning on using an active balancer and so the top balance is probably not even that important as it will quickly become irrelevant under the operation of the balancer.

So for matched cells, I will probably put the 32 cells in parallel and charge them to a common voltage at say 3.2V (but not a 3.65V top balance) and group them in pairs based on MIN-MAX of the EVE capacity measurements. I then do not have to worry about compression for top balancing and can just go to my final configuration for a compression frame.

I can monitor the matching with the OverKill BMS which will display each S cell pair individually. It would expect this to track within a few millivolts and the only sign of imbalance will be in measuring equalization currents across pairs.
 
Last edited:

ken morgan

Solar Addict
Joined
May 11, 2021
Messages
370
Probably worse than appalling at the end of one's career to get slapped in the face with the ultimate recognition that it is NOT important WHAT you know but rather WHO you know. While this is true in both the commercial and government sectors. In government despite all the pretenses of competitive bidding and fair procurement laws, who you know predominates as the factor to success.
yep seen plenty of that as well. thats why I do my own thing these days work wise.
 

Posplayr

Solar Enthusiast
Joined
May 27, 2021
Messages
140
I'm cross-posting some info on my 32 cell order here to show the anticipated results using a Factor Matched set of battery data from EVE.


For my location AZ 85704, I have an 8/21/2021 order into Amy for 32 LF280K for $126/cell plus shipping $4743 grade A matched cells (additional $140 credit card fees for AMX). They were to ship (to me 50-60 calendar days) last week but EVE apparently had a delayed shipment now expected to arrive this coming Monday. This was Friday morning status.

That is $148.21 per cell or the equivalent of $592.87 for 4.
I got a notification that the 32 cell purchase has shipped (it took about 20 days to actually ship due to a delay from EVE). Amy provided the capacity data as well as pictures that you can find in the attached spreadsheet.

Here is the raw data

View attachment 64207

The raw unordered 32 cells have the following statistics:
View attachment 64206

I then did a cell data sort from Min to Max , and paired the cells starting with the lowest and highest to create pairs and then repeated 16 times. the resulting in this final capacity estimate for the now 16 pairs.



View attachment 64208

The statistics of the 16 MIN-MAX pairs is now
View attachment 64212

The driver for loss of capacity due to cell mismatch is the lowest capacity cell which in the group of 32 was 0.64% below the average but now after MIN-MAX matching is only 0.03% corresponding to 0.186 Amp-hr! . The factory matching of the original 32 is within 1.878 Amp-hr so this process using 32 cell for 16S resulted in a 10:1 improvement in capacity matching. This might arguably be the best case where factory matching reduces dispersion resulting in a uniform distribution rather than Gaussian. This is the distribution of the 32 cells.

View attachment 64214
Here is what the MIN-MAX pairing distribution looks like. TWT how stable this pairing remains. Regardless it shows exceptional promise, making concern about cell matching a nonissue for large arrays.

Using the factory capacity data from EVE, this balancing procedure is a piece of cake where I can jump right to step #3 after top balancing Step #1 . This will reduce concerns about charging and discharging with uncompressed cells. With an active balancer, you can even just bring the 16S stack up to a 3.6V equivalent voltage (perhaps even 3.55 of lower) and let the active balancer bring the cells into SOC/Voltage agreement


View attachment 64221
View attachment 64215
View attachment 64216
 

lolailando

Solar Enthusiast
Joined
Mar 9, 2021
Messages
130
Yes, there are many things that can be done, but the problem is the return on investment. It is hard to justify the time/expense involved in this type of development for a 1 off. Although for a large house battery system the safety issues are very compelling and worth putting in the time. I have been evaluating various Arduino compatible ESP32 devices. I do mostly C++ development for embedded systems and these devices make it easy to develop a product where the CPU is a plugin. Of course, if you can find an open source project that does what you want there will be much less effort. I have noticed that Overkill Solar , Victron and Renogy/Rich have Arduino libraries to talk to their respective devices so there are several integration options available.

I am well into designing an alternator temperature-based charging controller for kind of put that on the back burner when I figured out I can buy a 300 amp alternator for $220!.


I was looking at this as a platform for a more sophisticated touch screen UI.


These are all BT, WiFi capable so the sky is kind of the limit, but I'm not falling all over myself to push data to the cloud to access analytics from my beach chair in the Bahamas LOL.

My first priority is to find a topology to isolate a single point cell failure to that cell so that it can be isolated (disconnected electrically) and insulated thermally in the pack. I don't know if that is better done in parallel or serial because I'm concerned with detection (as described before). The instrumented busbar is most useful for parallel cells although there are benefits that a sting approach could benefit from as well (i.e. thermal sense/shutdown at the cell/string level).
There's already a open source project that put out an alternator charge regulator like you are describing, thought i mentioned it. You can DM since this is offtopic, the project went commercial but it was open source for the first 3 versions.
 

wholybee

Solar Addict
Joined
Feb 11, 2021
Messages
364
tldr;
You could simply do a capacity test of each cell individually, then pair them so the sum capacity of each pair matched the other pairs. I think that is what you are trying to achieve, but with a longer procedure and more spreadsheets.
 

Posplayr

Solar Enthusiast
Joined
May 27, 2021
Messages
140
tldr;
You could simply do a capacity test of each cell individually, then pair them so the sum capacity of each pair matched the other pairs. I think that is what you are trying to achieve, but with a longer procedure and more spreadsheets.
Basically, that is correct. The point is to demonstrate what one would analytically predict. Specifically, that with MIN-MAX matching within a population's capacity spread (say 32 cells) generates a substantially reduced capacity spread in the match pairs (now only 16). This reduction in the spread is shown in this example to be 10:1 reduction for matched cells.

This means you can get a set of unmatched cells to match better than factory-matched cells if you have a large enough array. The expected reduction is a function of the total cells in the array.

The procedure is somewhat complex as it shows an actual process to measure the capacities which allow you to order them. What I am doing now is simply using the manufacturer's data of the capacities to do the ordering which should achieve the same effect.

The spreadsheets and procedure are simply to document in detail what is required to measure the various capacities (in order to rank accurately without special equipment) and then what the effects are after you do that in charge /discharge cycles. Now that I have done this you can safely ignore all the details and simply apply the principle of MIN-MAX using manufactures data for matched cells or for unmatched cells if available.
 
Last edited:

Posplayr

Solar Enthusiast
Joined
May 27, 2021
Messages
140
There's already a open source project that put out an alternator charge regulator like you are describing, thought i mentioned it. You can DM since this is offtopic, the project went commercial but it was open source for the first 3 versions.
Thanks, if you have a link I can take a look. As mentioned I have moved on because I purchased the higher performance aftermarket alternator. However, the original project I was working on was to sense the alternator temp but control the LiFePO DC-DC charger (from Renogy).

I was adding in some other features for displaying charging currents at the alternator, and load which provides a variety of warnings and automatic controls. This kind of evolved into a total power system control center, but with the chips shortages I became less interested.
 

wholybee

Solar Addict
Joined
Feb 11, 2021
Messages
364
Basically, that is correct. The point is to demonstrate what one would analytically predict. Specifically, that with MIN-MAX matching within a population's capacity spread (say 32 cells) generates a substantially reduced capacity spread in the match pairs (now only 16). This reduction in the spread is shown in this example to be 10:1 reduction for matched cells.

This means you can get a set of unmatched cells to match better than factory-matched cells if you have a large enough array.

The procedure is somewhat complex as it shows an actual process to measure the capacities which allow you to order them. What I am doing now is simply using the manufacturer's data of the capacities to do the ordering which should achieve the same effect.

The spreadsheets and procedure are simply to document in detail what is required to measure the various capacities (in order to rank accurately without special equipment) and then what the effects are after you do that in charge /discharge cycles. Now that I have done this you can safely ignore all the details and simply apply the principle of MIN-MAX using manufactures data for matched cells or for unmatched cells if available.
Factory Matched cells are not only matched for capacity, but for internal resistance. You could have 2 cells matched for capacity perfectly, but due to IR difference when paralleled together one will deliver more current than the other and be depleted faster.

Given a largish batch of unmatched cells, simply paralleling them in random ways will "improve" the match. You have optimized that by paralleling them in a non-random way. To be perfect, you would determine the IR of the cells and include that in your determining of parallel groups.
 

Posplayr

Solar Enthusiast
Joined
May 27, 2021
Messages
140
Given a largish batch of unmatched cells, simply paralleling them in random ways will "improve" the match.
It is a fundamental principle of statistics that any population of random numbers will have an increased variance with the increasing additive count. In this case, by adding two capacities together, there should be no expectation that the combined capacity STD is smaller. So this first statement of yours is incorrect. The population standard deviation increases by sqrt (2) and despite the fact that we're now only have 16 pairs vs the initial 32 cells the lowest cell capacity is what limits the overall array capacity.
You have optimized that by paralleling them in a non-random way.
Yes as stated MIn-MAX for parallel cells is the combination that minimizes dispersion. Not only it is non-random, but it is also optimal.
To be perfect, you would determine the IR of the cells and include that in your determining of parallel groups.
With the various mysteries of the battery chemistry characteristics, there is no such thing as perfection and we are not trying to achieve that. If you want IR-matched cells they would need to be matched at the factory, however, in theory, your could do MIN-MAX cells matching for IR as well but it has so little to do with capacity that matching capacities is far more productive.

The whole point of the data exercise is to show actual capacity test results to see if after matching there is reduced dispersion corresponding to improved array capacity. The results show clearly that the cell capacity spreads can be minimized using MIN-MAX pairing to the direct benefit of increased array capacity.
 

wholybee

Solar Addict
Joined
Feb 11, 2021
Messages
364
It is a fundamental principle of statistics that any population of random numbers will have an increased variance with the increasing additive count. In this case, by adding two capacities together, there should be no expectation that the combined capacity STD is smaller. So this first statement of yours is incorrect. The population standard deviation increases by sqrt (2) and despite the fact that we're now only have 16 pairs vs the initial 32 cells the lowest cell capacity is what limits the overall array capacity.

Yes as stated MIn-MAX for parallel cells is the combination that minimizes dispersion. Not only it is non-random, but it is also optimal.

With the various mysteries of the battery chemistry characteristics, there is no such thing as perfection and we are not trying to achieve that. If you want IR-matched cells they would need to be matched at the factory, however, in theory, your could do MIN-MAX cells matching for IR as well but it has so little to do with capacity that matching capacities is far more productive.

The whole point of the data exercise is to show actual capacity test results to see if after matching there is reduced dispersion corresponding to improved array capacity. The results show clearly that the cell capacity spreads can be minimized using MIN-MAX pairing to the direct benefit of increased array capacity.
Admittedly, I am not an expert in statistics. However, given 4 cells with random capacities between 70Ah and 100Ah. Now, randomly pair those with 4 more cells between 70Ah and 100Ah. Maybe this is my mistake, but isn't it unlikely that 2 70Ah cells would be paired and also that 2 100Ah cells would be paired? Anything other than would result in a better match than the original set. It seems more likely (3 out of 4 chance) that the 70Ah cell would pair with a cell of higher capacity, and that the 100Ah cell would pair with lower capacity. So the pairs of cells would be better matched than the original 4. This would only get better as more cells were added.

Re-reading your first paragraph. I am not talking about going from 32 cells to 16 pairs of cells. Agree, that will create more variation. I am talking about going from 4 cells, to 8 cells that are paired to act as 4 cells. Every time you add 4 more cells paralleled with the original 4, the match becomes better.
 

lolailando

Solar Enthusiast
Joined
Mar 9, 2021
Messages
130
It is a fundamental principle of statistics that any population of random numbers will have an increased variance with the increasing additive count. In this case, by adding two capacities together, there should be no expectation that the combined capacity STD is smaller. So this first statement of yours is incorrect. The population standard deviation increases by sqrt (2) and despite the fact that we're now only have 16 pairs vs the initial 32 cells the lowest cell capacity is what limits the overall array capacity.

Yes as stated MIn-MAX for parallel cells is the combination that minimizes dispersion. Not only it is non-random, but it is also optimal.

With the various mysteries of the battery chemistry characteristics, there is no such thing as perfection and we are not trying to achieve that. If you want IR-matched cells they would need to be matched at the factory, however, in theory, your could do MIN-MAX cells matching for IR as well but it has so little to do with capacity that matching capacities is far more productive.

The whole point of the data exercise is to show actual capacity test results to see if after matching there is reduced dispersion corresponding to improved array capacity. The results show clearly that the cell capacity spreads can be minimized using MIN-MAX pairing to the direct benefit of increased array capacity.
Hi PosplayR and thank you for sharing so much. Since you seem so knowledgable on this would you like to comment on how important would it be to get matched cells if I'm buying lets say 200 cells to do banks of 16, 32 and 48 cells at 48v the way you've done them for solar house storage? Couldnt we just organize them by capacity in groups of similar capacity and also min-max to further optimize? I understand some of my banks will be higher and other's lower capacity but are we talking about a 10% max in capacity? some prices for matched cells seem to cost much more than that. Trying to see the big picture here. I sell solar panels and I have many inqueries for battery banks and I would like to proceed in an efficient manner; my clients wouldn't care about a 10% capacity difference when switching from lead, and I can just charge according to the real capacity, there are always bargain hunters who will take the lower bank at a discount. I would appreciate your advice, thanks in advance.
 

Posplayr

Solar Enthusiast
Joined
May 27, 2021
Messages
140
Admittedly, I am not an expert in statistics. However, given 4 cells with random capacities between 70Ah and 100Ah. Now, randomly pair those with 4 more cells between 70Ah and 100Ah. Maybe this is my mistake, but isn't it unlikely that 2 70Ah cells would be paired and also that 2 100Ah cells would be paired?
MIn Max pairing would mean you pair a 70Ah with a 100 Ahr resulting in the first 170 amp hr pair. The remaining cells would be paired say 80 Amp-hr and 90 A-hr resulting in 170 Ahr again. Obviously, this is a contrived case to get perfect matching between the two pairs. So if you have a complementary cell that is as high above the means as another cell that is lower than the mean (i.e. same absolute distance from the mean) then they cancel to the mean.

More realistically the worst case for the 70 A-hr cells is to be 5 higher to 75 A-hr and the 100 A-hr to also be 5 high to 105 A-hr so the sum is 180 A-hr only 5 A-hr from the cell mean. The mean of the original 4 is (70+70+100+100)/4 = 340/4=85 and the lowest cell is 70 or 15 A-hr low.

If you study the numbers in the spreadsheet you have exactly what you are talking about 8 cells ordered by capacity and MIN-MAX matched to yield a 4S battery of 2P cells.

How well this works is going to be affected by the actual distribution of the cells. A uniform distribution is best, and an asymmetric distribution is worst. From the plots, you can see that the matched cells seem to be pretty uniform although not perfect.
Re-reading your first paragraph. I am not talking about going from 32 cells to 16 pairs of cells. Agree, that will create more variation. I am talking about going from 4 cells, to 8 cells that are paired to act as 4 cells. Every time you add 4 more cells paralleled with the original 4, the match becomes better.
You responded after I posted the 32 cell matching data although the original was 8 cells were matched into 4 pairs. The principles are the same, you just have more opportunities to match the more cells there are.
 

Posplayr

Solar Enthusiast
Joined
May 27, 2021
Messages
140
Hi PosplayR and thank you for sharing so much. Since you seem so knowledgable on this would you like to comment on how important would it be to get matched cells if I'm buying lets say 200 cells to do banks of 16, 32 and 48 cells at 48v the way you've done them for solar house storage? Couldnt we just organize them by capacity in groups of similar capacity and also min-max to further optimize? I understand some of my banks will be higher and other's lower capacity but are we talking about a 10% max in capacity? some prices for matched cells seem to cost much more than that. Trying to see the big picture here. I sell solar panels and I have many inqueries for battery banks and I would like to proceed in an efficient manner; my clients wouldn't care about a 10% capacity difference when switching from lead, and I can just charge according to the real capacity, there are always bargain hunters who will take the lower bank at a discount. I would appreciate your advice, thanks in advance.
Lolailando: I'm not an expert on LiFePO batteries but I am an EE with 40 years of experience much of it in systems engineering for complex systems. There are many issues that come to bear in what you are describing which amounts to ultimately a business plan for selling batteries.

If the only thing that, mattered was capacity matching, then as I stated originally as a goal, you could combine cells with MIN-MAX matching and essentially take B grade cell matching and turn them into A-grade matching. Some people will mention IR mismatching as a factor which I tend to think nothing of already it is already baked into the cake of capacity matching. The bigger issue in my mind is getting swollen Grade B cells and the likely expansion of those cells over their lifetime. Mine seem t have full capacity but what does the future hold is unknown.

The implication of this is the complexity of a compression fixture that you plan on incorporating into the battery pack design. You do not want to go through daily bulging cycles with cells pushing against their respective posts. I'm moving away from busbars entirely unless they are oriented orthogonally to the lateral bulging (i.e. end to end bus bars only). My small set of cells is bulged 2-3 mm and if these were clamp together with rigid busbars that would be putting a lot of pressure on the posts. The danger is damaging the posts, causing a short and then a thermal runaway as the shorted cell dumps all of its energy. Now we are getting into system-level safety which is beyond the scope of this post.

But to not leave you hanging I plan on a 16S battery of 2P cells (280 A-hr cells). I believe it will be safer and cheaper than two parallel 16S batteries. These are matched cells (as reported above) because they are not going into the house but into a detached shop but they still need to be safe and so paying the premium for the top quality and highest expected longevity is worth it for hybrid grid-tie battery backup.

On the other hand, my much smaller 4S2P 200 amp-Hr 12V battery is going to be abused in a moving vehicle exposed to the Tucson summer heat (usually above 110 degF outside). I'm not expecting them to last as long mainly because of the heat. In this case "more disposable" Grade B (swollen) cells matched MIN-MAX is the way I ended up but also feel comfortable with this situation.

As far as selling batteries, I think I would stick with commercial packs. where 2 100 amp-hr batteries are approaching $1000 especially if this is a sideline. For large battery packs (e.g. 32x 280 cells), this is prohibitive and much more endginning should go into the fire protection and reliability of the pack.

So to conclude I want to emphasize that I see the need to go far beyond what you see in almost all other commercials (integrated battery system) offerings to use some type of compression plate system and some type of thermal control(hot and cold). So depending on how well you are going to do this and what the life expectancy you want are going to drive the choice as to whether you want to buy Grade B vs paying the 30% premium for Grade A. MIN-MAX matching will improve either. Grade B will really need it and Grade A it is "gilding the lily" but in any event, it will always be better than what you start with.
 
Last edited:

wholybee

Solar Addict
Joined
Feb 11, 2021
Messages
364
Hi PosplayR and thank you for sharing so much. Since you seem so knowledgable on this would you like to comment on how important would it be to get matched cells if I'm buying lets say 200 cells to do banks of 16, 32 and 48 cells at 48v the way you've done them for solar house storage? Couldnt we just organize them by capacity in groups of similar capacity and also min-max to further optimize? I understand some of my banks will be higher and other's lower capacity but are we talking about a 10% max in capacity? some prices for matched cells seem to cost much more than that. Trying to see the big picture here. I sell solar panels and I have many inqueries for battery banks and I would like to proceed in an efficient manner; my clients wouldn't care about a 10% capacity difference when switching from lead, and I can just charge according to the real capacity, there are always bargain hunters who will take the lower bank at a discount. I would appreciate your advice, thanks in advance.
I very highly recommend that you look at larger capacity cells, and use fewer than 200. You are introducing a lot more points of failure, bus bars, wires, fuses, multiple BMS, etc. You can buy 1000Ah cells, so I would start with that. 16s or even 16s2p of 1000Ah is going to be a whole lot better than 5x as many 200Ah cells.

As a simple comparison between matched and unmatched cells. I bought 12 100Ah cells, though official channels not alixxx. Yes, they were expensive. They all tested at 115-117Ah. The IR was nearly exactly the same as well. No swelling. Those same cells, grade B from alixxx, could be between 80-100Ah, with very different IR, and varying degrees of swelling. I didn't test the self discharge rate, but that is likely the same across my cells as well. With Grade B cells, if they are within 20% of rated capacity, they will still sell them as that capacity. With Grade A, the factory specs include a margin, that the cells will be over the rated capacity.

Min-Max matching as PosplayR describes will maximize the capacity of the batteries built with Grade B cells. You won't get 100Ah per cell, because some cells are less than that. But, you will get more than 80Ah. Lets say you get 90Ah. I have 115Ah with Grade A cells.

Min-Max matching will do nothing to ensure that the cells stay in balance. If/When that happens, capacity of the battery will drop. There will be no damage, and you can top balance again to get that capacity back. You can add active balancers to continually balance the cells and prevent needing to top balance again later. Otherwise, unmatched grade B cells will likely go out of balance. However, matched and balanced grade A cells, properly installed with good connections, equal length wires etc., can go years without any balancing needed. Not even the passive balancing that BMS's include.

Also, the life expectancy of grade B cells is not known. Grade A cells are known to last (barring something happening to damage them) for 10-20 years.

Not sure if that answers your question. In summary, there isn't going to be cell damage if you connect unmatched cells, but they may require more maintenance and balancing in the future. They also may prove unreliable in the long term. That is an unknown, and you can read many posts about people that are unhappy with their grade B cells, but also as many that were very pleased.

I will respond separately RE: 2p16s vs 16s2p
 

wholybee

Solar Addict
Joined
Feb 11, 2021
Messages
364
But to not leave you hanging I plan on a 16S battery of 2P cells (280 A-hr cells). I believe it will be safer and cheaper than two parallel 16S batteries. These are matched cells (as reported above) because they are not going into the house but into a detached shop but they still need to be safe and so paying the premium for the top quality and highest expected longevity is worth it for hybrid grid-tie battery backup.
My battery is 3p4s. I believe it to be safe, simpler, cheaper, and most importantly in my case I could fit a bigger battery in a small space.

However safe as it is, it is most certainly *not* safer than multiple (in my case it would be 3) 4s batteries paralleled together. The reason is simple. With multiple batteries, each has its own BMS, and every single cells is monitored individually. Each battery has its own fuse, independent of the others, and any failure of a cell and the BMS will take one battery offline, and the system will still work uninterrupted but at a lower capacity. You could also fix that battery and bring it back online, without interrupting power. There is no failure mode of any cell that would cause unwanted current to flow or a fire.

In a 2p4s or (my)3p4s battery, the BMS is monitoring a paralleled group of cells. It does not have visibility to each cell, nor is there a way to stop current in that group if something goes wrong. If a cell fails to an open, the BMS will not know, only you now have a very out of balance battery. Your first indication would be when the BMS disconnects when you otherwise think you have 50% of the battery remaining. But the real issue is if a cell fails in a short. The other cell will now have a direct short, and with more than 1000A, will die in flames, possibly burning your house down. Both of those are very unlikely to happen, but both have happened. More likely, a cell will degrade and have much reduced capacity. There won't be any indication of anything wrong, except the battery will appear to be very out of balance. You would need to break it apart and individually test the cells to find the actual problem. If you had a really large battery (like 200 cells that lolailando is considering) that would be a lot of work.

I am comfortable with this because I have Grade A cells. With grade B cells, I think the additional cell monitoring and balancing possible with a series then parallel configuration is important because those cells are going to change in capacity and balance.
 

Posplayr

Solar Enthusiast
Joined
May 27, 2021
Messages
140
MIn Max pairing would mean you pair a 70Ah with a 100 Ahr resulting in the first 170 amp hr pair. The remaining cells would be paired say 80 Amp-hr and 90 A-hr resulting in 170 Ahr again. Obviously, this is a contrived case to get perfect matching between the two pairs. So if you have a complementary cell that is as high above the means as another cell that is lower than the mean (i.e. same absolute distance from the mean) then they cancel to the mean.

More realistically the worst case for the 70 A-hr cells is to be 5 higher to 75 A-hr and the 100 A-hr to also be 5 high to 105 A-hr so the sum is 180 A-hr only 5 A-hr from the cell mean. The mean of the original 4 is (70+70+100+100)/4 = 340/4=85 and the lowest cell is 70 or 15 A-hr low.

If you study the numbers in the spreadsheet you have exactly what you are talking about 8 cells ordered by capacity and MIN-MAX matched to yield a 4S battery of 2P cells.

How well this works is going to be affected by the actual distribution of the cells. A uniform distribution is best, and an asymmetric distribution is worst. From the plots, you can see that the matched cells seem to be pretty uniform although not perfect.

You responded after I posted the 32 cell matching data although the original was 8 cells were matched into 4 pairs. The principles are the same, you just have more opportunities to match the more cells there are.

This is a very recent video discussing Compression and post stresses and the Youtuber's frustration for finding a solution. There are several prominent names from this forum and others posting in the comments. Something to follow to see what jells.

 

Posplayr

Solar Enthusiast
Joined
May 27, 2021
Messages
140
My battery is 3p4s. I believe it to be safe, simpler, cheaper, and most importantly in my case I could fit a bigger battery in a small space.

However safe as it is, it is most certainly *not* safer than multiple (in my case it would be 3) 4s batteries paralleled together. The reason is simple. With multiple batteries, each has its own BMS, and every single cells is monitored individually. Each battery has its own fuse, independent of the others, and any failure of a cell and the BMS will take one battery offline, and the system will still work uninterrupted but at a lower capacity. You could also fix that battery and bring it back online, without interrupting power.. There is no failure mode of any cell that would cause unwanted current to flow or a fire
Without a diagram of your 3p4s system, it is hard to assess failure modes. While you might disagree with the author of the ORION BMS white paper referenced above (I will find it and post it here),


the primary concern with parallel strings is the energy from one entire string dumping into the other due to a single failed cell. In a parallel cell configuration, this simply can't happen. Yes, one cell in parallel configuration will dump into the mate if that mate fails (assuming no other protections), but that is much less energy than in say a 2P 16S configuration for example. So if there is any possibility this needs to be avoided. At the very least one would need separate breakers and BMSs per parallel string.

There are two separate issues in BMS monitoring in comparing serial vs parallel configurations; the first is whether separate parallel string BMSs per string can detect faults that a single BMS can not and the second is what happens worst-case. The first question is really beyond the scope of what I would want to get into here. But suffice to say, that even with separate voltage measurements per cell, I could envision cell faults that remain undetected within the BMS cell and array level voltage ranges (2.5 to 3.65V). With regard to the second, there is still potentially a whole lot of fault current and more importantly total energy that could be dissipated in a parallel string failure mode within the BMS fault limits. In contrast, the worst case for parallel cells is one dumping into the other. It can't get worse than that. This is 1/16 the energy for a 16S configuration comparing parallel cells vs parallel strings.


In a 2p4s or (my)3p4s battery, the BMS is monitoring a paralleled group of cells. It does not have visibility to each cell, nor is there a way to stop current in that group if something goes wrong. If a cell fails to an open, the BMS will not know, only you now have a very out of balance battery. Your first indication would be when the BMS disconnects when you otherwise think you have 50% of the battery remaining. But the real issue is if a cell fails in a short. The other cell will now have a direct short, and with more than 1000A, will die in flames, possibly burning your house down. Both of those are very unlikely to happen, but both have happened. More likely, a cell will degrade and have much reduced capacity. There won't be any indication of anything wrong, except the battery will appear to be very out of balance. You would need to break it apart and individually test the cells to find the actual problem. If you had a really large battery (like 200 cells that lolailando is considering) that would be a lot of work.
If we are really concerned about diagnostics (fault detection) and prognostics (detecting onset of failure of accelerated end-of-life) of a LiFePO battery array, I think it is far more effective to monitor crosscurrents in a parallel cell configuration. Voltage just does not tell you that much. But this means a different topology than a standard BLS which is probably beyond most people's capability to implement. That said, with an early warning, corrective action can be taken earlier and the resulting downtime may in fact be expected to be not only less but also planned.

Most grid-tied battery backup systems have separate critical loads panels. We plan on a transfer switch that has the critical loads being run as effectively off-grid with battery backup. By means of the transfer switch, the critical circuits can be switched to grid-tied and allows for plenty of time to perform any maintenance under scheduled planning allowed by the prognostics look-ahead times.
I am comfortable with this because I have Grade A cells. With grade B cells, I think the additional cell monitoring and balancing possible with a series then parallel configuration is important because those cells are going to change in capacity and balance.

We can think of cell balance both in series (vertically) and in parallel (horizontally). Parallel strings automatically balance the string average SOC but not at the cell level. With passive or active balancing, some level of string SOC balance will tend to bring all cells in parallel balance but that is indirect (by balancing all cells in all strings toward the same average voltage and SOC). In contrast, parallel cells will automatically balance in parallel without any BMS balancing. I don't have any experience or real data on the subject, but in my mind, it remains to be seen if even after only 5 years, matched cells remain so well balanced as to justify no form of active or passive balancing.

In summary, most of your points are quite valid, but I would counter with there are many ways to "skin a cat" (as I have proposed above), and so my main focus is on safety under an absolute worst-case scenario. At this point, I would consider worst-case to entail limiting the maximum possible energy loss under a single-point cell short scenario. To my mind parallel cells and good fire protection in a metal box is inherently the safest.
 
Last edited:

lolailando

Solar Enthusiast
Joined
Mar 9, 2021
Messages
130
Lolailando: I'm not an expert on LiFePO batteries but I am an EE with 40 years of experience much of it in systems engineering for complex systems. There are many issues that come to bear in what you are describing which amounts to ultimately a business plan for selling batteries.

If the only thing that, mattered was capacity matching, then as I stated originally as a goal, you could combine cells with MIN-MAX matching and essentially take B grade cell matching and turn them into A-grade matching. Some people will mention IR mismatching as a factor which I tend to think nothing of already it is already baked into the cake of capacity matching. The bigger issue in my mind is getting swollen Grade B cells and the likely expansion of those cells over their lifetime. Mine seem t have full capacity but what does the future hold is unknown.

The implication of this is the complexity of a compression fixture that you plan on incorporating into the battery pack design. You do not want to go through daily bulging cycles with cells pushing against their respective posts. I'm moving away from busbars entirely unless they are oriented orthogonally to the lateral bulging (i.e. end to end bus bars only). My small set of cells is bulged 2-3 mm and if these were clamp together with rigid busbars that would be putting a lot of pressure on the posts. The danger is damaging the posts, causing a short and then a thermal runaway as the shorted cell dumps all of its energy. Now we are getting into system-level safety which is beyond the scope of this post.

But to not leave you hanging I plan on a 16S battery of 2P cells (280 A-hr cells). I believe it will be safer and cheaper than two parallel 16S batteries. These are matched cells (as reported above) because they are not going into the house but into a detached shop but they still need to be safe and so paying the premium for the top quality and highest expected longevity is worth it for hybrid grid-tie battery backup.

On the other hand, my much smaller 4S2P 200 amp-Hr 12V battery is going to be abused in a moving vehicle exposed to the Tucson summer heat (usually above 110 degF outside). I'm not expecting them to last as long mainly because of the heat. In this case "more disposable" Grade B (swollen) cells matched MIN-MAX is the way I ended up but also feel comfortable with this situation.

As far as selling batteries, I think I would stick with commercial packs. where 2 100 amp-hr batteries are approaching $1000 especially if this is a sideline. For large battery packs (e.g. 32x 280 cells), this is prohibitive and much more endginning should go into the fire protection and reliability of the pack.

So to conclude I want to emphasize that I see the need to go far beyond what you see in almost all other commercials (integrated battery system) offerings to use some type of compression plate system and some type of thermal control(hot and cold). So depending on how well you are going to do this and what the life expectancy you want are going to drive the choice as to whether you want to buy Grade B vs paying the 30% premium for Grade A. MIN-MAX matching will improve either. Grade B will really need it and Grade A it is "gilding the lily" but in any event, it will always be better than what you start with.
Hi and thank you so much for your extensive answer. I do am planning to buy A cells with no swelling. But it was my understanding that suppliers basically will match capacity sets for people ordering 16 or 32 cells and those are the premium prices. I saw in this forum that some supplier answered that they will not/was not needed to capacity match a 200 pcs order. Me reading between the lines is that when they "match" cells they pick and choose from a 200 pcs pallets and organized them by capacity to ship bundles of 16 or whatever that are closer "matched".
I am also considering "built" batteries but 32x 310ah cells= 48v 620ah= 30kw, that is hard to match by built batteries: everyone wants to run the ACs all nite and there's cloudy days. One client wants 45kw. lifepo4 because he just bought 600ah at 48v deepcycle lead and already run out twice.
There's a lot of ways and as you said; lots of bussiness decisions; basically I do not plan to sell "batteries" per se, just a few banks to a few value added clients, no direct sales basically and nothing bellow 16x 280ah.
 
Top