DIY LiFePO4 bank: individual prismatic cell fusing

[aybabtme]

Hey there,
I'm building a DIY LiFePO4 bank, using multiple 100Ah cells in parallel, to provide a way to individual cell failures. In order to achieve this, I need to individually fuse them, so that a failed, short-circuiting cell in a parallel block doesn't end up receiving the full load of the other cells in the parallel block, causing these cells to then fail in a cascading manner. This diagram (from Nordkyn Design's series on LFP banks) shows what I'm looking for:




I've been looking at what I could use to individually fuse each cell. I'm using CALB 100Ah FI cells, and I think replacing the bus bars with ANL fuses would work perfectly, dimension wise. See this CAD drawing:

It's a rough model but it seems to prove the ANL fuse would work perfectly: https://a360.co/2OUkVai

Since the cells are 100Ah with max discharge of 2C (200A), I thought using a 300A ANL fuse (closest available/affordable) would work well. However I'm just wondering about the electrical characteristic of these fuses. Will they heat up under normal operations (0.3C to 1C)? Will they cause undue voltage drop (their resistance seems to be around 300µΩ according to this Victron document)?

Has anyone done this before?

[tanglewood]

I think it's a really tough call. The fuses and connections will heat some, and will introduce imbalances that might not otherwise be present. You will experience those issues/risks every day of use.


This has to be balanced against the probability of a short circuit failure of one cell, and the subsequent damage to the other cells in that parallel block. And also the risk that the heating from the failed and short circuited cells will subsequently damage adjacent cells.


Keep in mind that pretty much all cells are tested and pass UN 38.3 which includes a short circuit test. This is what the other cells in the parallel group would experience if one cell failed short circuit. To pass, the cell cannot catch fire or explode.


Also, is a short circuit failure an expected or likely failure mode with LFP? I don't know, but that seems to be part of the calculus too.

[Steve_C]

I think if I had to use that many cells I would put them in series first and then parallel. It would require much fewer fuses.

I may not be thinking about this correctly, but I think the fewer fuses would be preferable?

[Steve_C]

Also it would seem the way you have it configured any cell problem would involve current flow thru 2 fuses? Not sure what the implications are, but certainly worth checking out. If for no other reason to consider fuse sizing.

The size you suggest seems really large.

[aybabtme]

Originally Posted by Steve_C

I think if I had to use that many cells I would put them in series first and then parallel. It would require much fewer fuses.

I may not be thinking about this correctly, but I think the fewer fuses would be preferable?

Series first is dangerous with Li-ion, unless the series strings have their own BMS. The problem is that with series first (without a BMS per series), there's no way to keep the cells in balance, and some of them will eventually end up either overcharged or overdischarged.

Also the cost of having dedicated BMS per string (~500$) is much higher than using individual ANL fuses per cell (10~20$ per cell). Series first also is less resilient to individual cell failure: the whole series string goes down if a single cell in the string fails, whereas in a parallel first, fuse-per-cell topology, the bank is losing no more than the cell's individual capacity (and its current intake/outake capacity).

[hzcruiser]

Originally Posted by aybabtme

Series first is dangerous with Li-ion, unless the series strings have their own BMS. The problem is that with series first (without a BMS per series), there's no way to keep the cells in balance, and some of them will eventually end up either overcharged or overdischarged.

Also the cost of having dedicated BMS per string (~500$) is much higher than using individual ANL fuses per cell (10~20$ per cell). Series first also is less resilient to individual cell failure: the whole series string goes down if a single cell in the string fails, whereas in a parallel first, fuse-per-cell topology, the bank is losing no more than the cell's individual capacity (and its current intake/outake capacity).

Hello Antoine,

it seems you've made up your mind already, so here goes nothing:

How likely do you think cell failure really is? Also, how often do you think you need to balance the cells?
There are a few people (incl me) on this forum that are running LFPs for years on their boats and the consensus seems to be that both events are very rare.
Hence you don't need to overdo the fusing between cells. Also, keep in mind the fuses are there to protect the cabling from overheating. Do you really expect to draw the 200A ? If so, you need the cabling to those consumers to support the current.
I would rather put 4 well-balanced cells in series and connect them via a circuit breaker to the bus. Do the same for the next batch of 4.

Install each batt where convenient and required. I have the biggest two LFPs next to the inverter in the salon, one next to the windlass and one in the wheelhouse for all the radios, lights, nav equipment.

Now you can easily disconnect each battery (of 4 cells) and test it individually, say once every 6 or 12 months, plus you get the protection from overcurrent. The bus would still be powered from the other batts.



The BMS for each batt should not set you back $500 either. At a minimum it only needs to implement high and low voltage cut-offs. Since you seem to know which end of a multimeter goes where, you can do the cell checking yourself at your leisure. If they really need balancing, you can put a resistor across the highest cell and bring that down to the other ones whithin an hour or so.

[aybabtme]

Well I agree that there might not be a need to fuse the cells individually ("all cells are tested and pass UN 38.3"), so I'm not 100% sold on the ANL fuses. However I already have the BMS here (an Orion Jr BMS runs ~500$ all accounted for), and I'm definitely in the camp of using a BMS I'm sure the pros and cons of using a BMS or not have been discussed at length and I won't embark on this here since it's not really the topic of my question.

[aybabtme]

I asked the same question on the article that I quoted in my opening post, and got a thoughtful answer from the author. TL;DR: the fuses aren't really effective in a 3p setup where each cell possibly provide large amounts of current. His actual answer is here.

So I'll do without the per-cell fuses and carry on with life!

[Cpt Pat]

Originally Posted by Steve_C

I think if I had to use that many cells I would put them in series first and then parallel. It would require much fewer fuses.

I may not be thinking about this correctly, but I think the fewer fuses would be preferable?

I agree. I don't see the point to connecting the 4S packs in parallel at every cell. It looks like unnecessary complexity. Then, you only need 3 fuses at the end of each 4S pack.


Every fuse will produce some voltage drop. Fuses aren't "free," they each cause some loss of power, and I don't see any gain in safety over connecting the packs conventionally.

[hzcruiser]

Originally Posted by aybabtme

[...] However I already have the BMS here (an Orion Jr BMS runs ~500$ all accounted for), and I'm definitely in the camp of using a BMS [...].

Sorry to break it to you, Antoine, but deep inside you might already know that this BMS of yours is not ideally suited to the job. Good luck shoe-horning the rest of your LFP setup around this device (rather than re-selling it while it's new).

And by that I'm not saying this BMS is useless, but it's spec'd for 1-16 cells in series.

[CatNewBee]

Originally Posted by Cpt Pat

I agree. I don't see the point to connecting the 4S packs in parallel at every cell. It looks like unnecessary complexity. Then, you only need 3 fuses at the end of each 4S pack.


Every fuse will produce some voltage drop. Fuses aren't "free," they each cause some loss of power, and I don't see any gain in safety over connecting the packs conventionally.

If you connect the cells into a pack and then add a fuse, its meaningless, because then you end up having 4 fuses in series, each fuse will cut the cuurent by herself, no additional benefit. You cam ommit all 3 and rely on the main battery fuse.

The electric problem of having parallel cells and one shorts is not solved by these fuses, still 2 healthy cells discharge over the one faulty without passing a fuse.

Second issue is in the case of one cell failure without a short, the 2 femaining cell get faster charged and may go up over the permitted cell voltage already at 60% SOC, not having a BMS will cause a run off. Even balancing will not help on the long run. Without a BMS, relying only on pack voltage and Ah counter will cause serious overcharge problems.

You can run a 3P4S system without fuses, but definitely need a BMS and some knowledge what it means when a OVP or LVP trips when SOC is far away from full or empty. The battery pack should be easy accessible and you should monitor cell temperatures to get an early indicator of uprising issues.

Safest way indeed is to chose the right size of cells and not parallel them. But if you have no other choice, than watch them properly an get ready to disassemble a pack if one cell overheats.

[StoneCrab]

General comment to all.

I learn quite a bit from the forum posts and make a point of reading on topics that I am not familiar with. To that end abbreviations and acronym make some posts confusing if they aren't spelled out at some point in the posts.

I know it takes a little more effort but I for one would appreciate the benefit. I do spend time researching to fill in the gaps, but sometimes the poster's point is missed.

Thanks,

[CatNewBee]

Sorry

LVP Low Voltage Protection
OVP OVer Voltage Protection
SOC State Of Charge
BMS Battery Management System
3P4S three cells paralleled (positives connected to each other and negatives connected to each other to build a cell of a larger capacity with the same voltage) and 4 of those packs connected in series, (meaning positive pole goes to the fuse and then to the main breaker, solenoid and then to the power distribution point, the negative side connects to the positive of the next cell, then this negative goes to the positive of the third, and the negative of the third to the positive of the forth one, finally the negative of the forth cell goes to the shunt and from there to the negative bus bar.

[aybabtme]

Just want to reiterate that paralleled strings are generally accepted to be a big no-no in Li-ion banks, no matter the chemistry. Don't want to enter into arguments about this, but I think it's worth reiterating for anyone else reading by.

Originally Posted by hzcruiser

Sorry to break it to you, Antoine, but deep inside you might already know that this BMS of yours is not ideally suited to the job. Good luck shoe-horning the rest of your LFP setup around this device (rather than re-selling it while it's new).

And by that I'm not saying this BMS is useless, but it's spec'd for 1-16 cells in series.

Apologies, not displayed in the CAD but my pack is 3p8s. However the Orion Jr would work for a 4p4s, each cell in the parallel block is at the same voltage and counts as 1 series. Maybe I misunderstood your comment, you say "it's spec'd for 1-16 cells in series": a 4p4s pack, or 3p8s pack, definitely is in the 1-16 series range.

In any case, this BMS is used in existing installations using the same topologies.

Originally Posted by CatNewBee

Safest way indeed is to chose the right size of cells and not parallel them,

CALB doesn't recommend cells above 200Ah for non-stationary installations, that's why I went with 3p and 100Ah cells. On larger cells, the internal content of the cells is subject to a wider range of unrestricted motion. I know you installed 1000Ah cells in your Lagoon, and I think your setup speaks for itself and works. However I wanted to be able to do the individual cell-fusing down the line and be able to add capacity as needed later. Also this particular series of CALB cell (the FI series) is designed for paralleled installations.

[CatNewBee]

Originally Posted by aybabtme

Just want to reiterate that paralleled strings are generally accepted to be a big no-no in Li-ion banks, no matter the chemistry. Don't want to enter into arguments about this, but I think it's worth reiterating for anyone else reading by.
.

If you use a BMS per string and disconnectors, you will be fine with paralleled strings, this is how drop in LFP batteries with internal BMS used to be connected in parallel to increase capacity....

Nothing wrong with your idea under this premises, a battery project is always a compromise, you have to chose your priorities and make an educated decision. So go for it, it will work.

I had similar considerations, originally I wanted to build 2P4S with 700Ah cells for 1400Ah, or a pack of 3P4S with 400Ah cells for 1200AH but then decided, the 1000Ah should be sufficient and the cheapest and safest, and simplest approach.