Lithium Batteries (for the rest of us)

[alanapar]

ebaugh - If your objective is to protect, using fuses, against major upset when a cell fails shorted, then you are going to need a fuse per cell, directly in series with each cell. This assumes you are paralleling the cells as shown at the 3V, 6V and 9V intermediate points.

The three different thumbnail diagrams all suffer from the same problem, that if any of the cells fail shorted (or for the third case, any of the cells in the middle group), then many other cells can dump into the failed cell and the cell directly in series with it. Forcing a large charge rapidly into the intact cell (especially if it is already fully charged) may possibly be a much worse thing to have than forcing that current through the shorted cell.

Anything short of doing this properly is probably just false security. (I will add that there are some subtleties here, such as the fault current limiting effect of the fuses used to parallel the strings, etc.)

Another consideration when fusing is how do you know a fuse has opened, possibly with the effect of significantly unbalancing the bank?

I'm not a battery expert (although I have some experience in power engineering), but I guess the issue here is, what are the risks of serious failure and the consequences thereof, if you have a large number of unfused, paralleled cells? With two cells in parallel, the chance of a random failure is only doubled, and the consequences are probably not much greater than for a single cell suffering an internal short.

But with say 48 cells in a bank, the probability of a single cell failure is increased 48 times, and the consequences of 11 cells all dumping into one shorted cell (or into an intact cell in series with a shorted cell) are quite possibly disastrous.

So while one possible solution is to remove the fuses altogether, you would want to be very sure of the failure risk/failure consequence situation before doing so. I would also add that I'm not familiar with likely LIFP cell failure modes, but there may be other failures (other than dead shorts) that are relevant here.

[alanapar]

I will add that there could be a way of effectively fusing the cells individually, which might be more compatible with the physical layout of your bank. Starting with the layout of your second thumbnail diagram, I would remove all the paralleling busbars at the 6V level. Then I would replace one of those busbar sets with a set of fuses. This will require almost as many fuses as individually fusing all the cells. Several fuses might have to open to isolate a shorted cell, but at least isolation would be possible.

One more thought about using fuses for paralleling - you need to be sure that the balancing currents flowing through the fuses do not produce enough voltage drop to impair balance across parallel groups. I suspect this is OK, but it would be good to run the numbers.

[alanapar]

BTW, what is wrong with having to blow several fuses to isolate a fault in this case (vs. the fuse directly in series with each cell scenario) is that you get into a situation where the fault current may divide over several paths, only part of the current flowing through each fuse involved. A dead short will no doubt blow all the fuses required, but a less than total short runs a risk that the current flowing through the faulted cell, while large and damaging, after being divided among several paths may not blow the fuses, or only blow them after an excessive delay.

[T1 Terry]

I think a proper risk analysis is required here, adding failure points like fuses within a parallel series pack in my eyes is asking for more trouble than you are trying to avoid in the first place.
Let's look at how an individual cell is constructed.
There are multiple smaller capacity pouches connected in parallel, each with positive and negative foil tab which is part of the anode and cathode of that pouch. If it was possible for one of these pouches to develop a big enough short between the anode and cathode foil the tabs where that section of the foil pouch would be the next weakest link. Just the capacity within the other non shorted pouches, and within the non effected section of pouch containing the shorted anode and cathode, would vaporise one or the other foil tabs on the shorted pair. A catastrophic dead short failure with a cell is highly unlikely and even if it did, it would sort itself out in a very short period.

Lithium ferrous cells don't fail like that, they fail with a minor internal short that only drags a few amps. If this short continues it will generate heat within that cell, all the other cells in the parallel pack will feed this minor short and that parallel packs voltage will drop compared to the other non shorted parallel packs, this will show up if the parallel pack voltages are monitored. If the current flowing through the minor shorted section is enough to build up sufficient heat but not enough to melt one of the foil tabs, the plastic separator between the 2 plates will melt and block the path between the two plates, the short will stop and that cell will have lost some of it's capacity. If the heat build up is enough to melt all the separators within that pouch, quite a noticeable loss of capacity will result, this will show up as on pack always being the first to the high charge voltage cut off and the same one that hits low voltage first. Still not a catastrophe if the voltages are monitored, more frequent recharges will be required until the problem cell is located and replaced.

To fit fuses between the cells with high enough current carrying capacity not to fail during normal service would not fail with a minor or I doubt a major internal short. About the only thing it would protect is a spanner dropped across the terminals, a possible scenario but the protection fuse network would offer more possible fail event causing the spanner to be there in the first place... a self generating dead short scenario?

Linking all the cells in parallel, one link for the series connection, the link between the 6v pack and the 9v pack could be an ANL fuse say 300 amps, this limits any short circuit greater than 300 amps to a max of 6v, 300 amps would create quite a spark at 12v, but any bigger amps would only be at 6v

T1 Terry

[ebaugh]

To fuse or not to fuse? It's a thorny path. I would have not done it by default, but the battery supplier recommended fusing the parallel paths as shown in the attached thumbnail. So I did it. Still searching for the optimal answer.

I can tell you understand the problem, this solution leads to overcharging the adjacent cell and doing nothing dumps into the shorted cell. Fusing every cell serially does solve the problem, or at least isolates it to the failed cell.

I have no way to detect a blown fuse other than inspection with a multimeter to check the resistance down the path of the fuses to tell if one has failed. This could be automated in a BMS, but not implemented yet.

I agree with Terry, a dead short is not a likely event. Much more likely is a low current short that the fuses do nothing to help with and would be detected by the BMS eventually as a cell out of balance. He is also correct, every prismatic (rectangular) cell is made up of multiple cells connected internally in parallel. They make single 1000Ah cells, that's almost the same energy as my 12 cells without any internal fusing. Can't hardly since the spec for that cell would allow for a 3000A discharge.

Are 48 cells more prone to failure than 4? Probably, but the failure can potentially be better contained. Is it needed? I don't know. It is attractive to limit a fault to 3.2V at 100Ah, versus the 10x that in the 1000Ah cell.

[ebaugh]

Originally Posted by alanapar

I will add that there could be a way of effectively fusing the cells individually, which might be more compatible with the physical layout of your bank. Starting with the layout of your second thumbnail diagram, I would remove all the paralleling busbars at the 6V level. Then I would replace one of those busbar sets with a set of fuses. This will require almost as many fuses as individually fusing all the cells. Several fuses might have to open to isolate a shorted cell, but at least isolation would be possible.

One more thought about using fuses for paralleling - you need to be sure that the balancing currents flowing through the fuses do not produce enough voltage drop to impair balance across parallel groups. I suspect this is OK, but it would be good to run the numbers.

I don't understand your suggestion. I attached the diagram I think you mean, but it seems to meet your description?

[ebaugh]

Originally Posted by alanapar

BTW, what is wrong with having to blow several fuses to isolate a fault in this case (vs. the fuse directly in series with each cell scenario) is that you get into a situation where the fault current may divide over several paths, only part of the current flowing through each fuse involved. A dead short will no doubt blow all the fuses required, but a less than total short runs a risk that the current flowing through the faulted cell, while large and damaging, after being divided among several paths may not blow the fuses, or only blow them after an excessive delay.

If the failure is insufficient to blow fuses, it would also be insufficient to cause much harm. Soon enough it would show up as a cell imbalance and be caught by monitoring the cell voltages.

[ebaugh]

I think there are flaws in all the diagrams I've posted so far. The only solution I've found that looks like it will work in the desired manner is this one. But it's a real "busy" solution, and not one I'm jumping over yet.

[alanapar]

ebaugh - Re your post #141, I was actually referring to the lower diagram in your post #122 as the starting point. Remove all the busbars at the 6V level, and replace one set with fuses. I mentioned this as I thought it might be an easier change to your existing setup than going to the fuse-per-cell model you have in post #143. But frankly, it's not the greatest solution as I pointed out.

If you are going to fuse within the bank, I think the fuse-per-cell model in post #143 is the way to go. I agree this is "busy", a consequence of having many small cells paralleled. The question is, is it necessary?

I'm interested in the answer as I hope one day to do a LiIon conversion myself. My "straw man" design would be 4 parallel Mastervolt 24V/180Ah batteries (24V/720Ah bank). This would be dandy, and is no doubt well-engineered (the batteries would be paralleled with one fuse and one electrically-operated safety switch per battery). But it will cost me an arm and a leg, and DIY is definitely in my mind if I can convince myself the result will be safe. The Mastervolt solution effectively has 4 parallel strings of 8 cells in series, with each string individually balanced/managed. So it side-steps the "massively parallel" issue. But the required balancing/management setup is not conducive to DIY.

I did some thinking and reading based on Terry's post. If one regards internal short circuits as limited by the anode/cathode foil tabs in stacked-plate cells, then the question is limited to what? At Li-Ion BMS - Prismatic cells there is a dissection of a 160Ah prismatic cell. The tabs are shown as 65 x 0.05mm foils which gives 3.25mm^2, roughly equivalent to #12 AWG wire. In copper that has a fusing current of 1kA for 1 sec and 5kA for 32ms. (The aluminum tabs would fail at somewhat lower currents, but I didn't have data handy). BTW, these numbers are roughly the same as for a 275A ANL fuse. The foil tabs would probably fail at higher currents due to having greater heat loss to their surroundings than a long copper wire in air (particularly so for the lower-current/longer-time case). So if you believe the fault will only involve one pair of plates, and will be cleared gracefully by their tabs fusing, you could argue that an added fuse in series with the cell is only adding protection to the extent it opens below the fusing current of the tabs.

If you were adding external fuses to individual cells, I do think you could set the fuse rating to well below this for a marine house bank (no propulsion load), say around 1C or 150A for a 160Ah cell. (You do want these fuses to comfortably carry any normal load, and to have a main bank fuse in series with the positive bus that opens well before any cell fuse will fail in normal operation.) This would allow a 1200Ahr bank to operate safely up to a very generous 1000A or more continuous current (allowing for some imbalance amongst cells causing higher currents in some fuses). Depending on your application (e.g., if you don't have a bowthruster or other very heavy load), you could possibly use still smaller per-cell fuses. But as Terry said, there may be cases where the fault is still not low enough resistance to blow the fuse. (N.B., if you make the cell fuses too small, their resistance will degrade the performance of the bank.)

I haven't managed to find much information about internal cell shorts. There are some research papers on the web on test methods, etc., but very little real-life data. The Dreamliner incidents are interesting. I know it is a different chemistry, but the NTSB report suggested multiple short circuits occurred in one cell, and that the bank voltage declined by about one cell's-worth, which sounds like the faults did not clear. In this case it was game over once that cell ran away thermally. Now hopefully a single LiFePO4 cell would not run away in the same situation, but what if other paralleled cells dump energy into it?

The problem is that this is a very-low-probability/very-high-consequence situation, and I would really want to be convinced that the failure modes are understood and the risk was under control.

[alanapar]

Originally Posted by ebaugh

If the failure is insufficient to blow fuses, it would also be insufficient to cause much harm. Soon enough it would show up as a cell imbalance and be caught by monitoring the cell voltages.

Well, if the cell fuse is say 70A, that fuse could carry maybe 100A indefinitely without opening. At 3.5V that is 350W, which is a lot of heat for one cell (assuming the energy is not absorbed electrochemically but goes into joule heating). That is worst-case of course.

If you see the imbalance caused by a fault that doesn't blow a fuse, but is still heating a cell, what do you do? I guess disconnecting that cell is all you can do. At least with the fuse-per-cell setup, you only have to unbolt one end of the relevant fuse.

BTW, I posted a longer response on other things, but it seems to have to go through the moderators for some reason. Could that be because I put a URL in it? (I'm new here!)

[alanapar]

Another thought about the fuse-per-cell. A small side benefit is that you can monitor the per-cell currents by measuring the voltage drops across the fuses. I don't have accurate data handy, but order-of-magnitude, a low voltage fuse of this size could be expected to drop maybe 50-100mV at rated current. So with a meter of 10uV resolution, you could measure cell currents with resolution of 10mA. The usefulness of this would depend to some extent on how uniform the fuse resistances were.

(I assume that the way your cells are interconnected, it is impossible to get a clip-on DC ammeter around the cell leads.)

This could be a useful diagnostic, as otherwise you might not have much visibility into what is going on with each cell in a parallel group.

[ebaugh]

The fuses only really exist to keep one bad cell from impacting the rest of the bank. The bad cell, if it hard shorts internally, it's going to die. No way to stop that.

The significant difference with LFP versus the Boeing chemistry is there is no chemical reaction possible to add additional heat. This is a big difference.

[ebaugh]

I don't see any safety reason to use the commercial marine solutions if you are able and willing to add a minimum of cell level monitoring and an automated disconnect in the event of overcharging or over discharging. The commercial solutions use the same cells with added electronics that integrate with their overall product offerings, but at 3x the cost.

Working on a project now to build a BMS, based on an industrial PLC with flexibility to adapt to most any installation requirement, keep track of SOC and automate the initial balancing of the bank. I'm hoping it can be truly plug and play.

[ebaugh]

Originally Posted by alanapar

Well, if the cell fuse is say 70A, that fuse could carry maybe 100A indefinitely without opening. At 3.5V that is 350W, which is a lot of heat for one cell (assuming the energy is not absorbed electrochemically but goes into joule heating). That is worst-case of course.

If you see the imbalance caused by a fault that doesn't blow a fuse, but is still heating a cell, what do you do? I guess disconnecting that cell is all you can do. At least with the fuse-per-cell setup, you only have to unbolt one end of the relevant fuse.

BTW, I posted a longer response on other things, but it seems to have to go through the moderators for some reason. Could that be because I put a URL in it? (I'm new here!)

Is this the fusing arrangement you suggested? On first blush, I like it. It essentially duplicates what Genasun does with a single serial fuse in their builds. And duplicates the parallel fusing at the cell level instead of the bank level like Mastervolt recommends.

The fusing values vary greatly based on the cell count, but with my 12P cells, the serial connections can be 25-30A each and the parallel connections probably much less, if the cells are equally used, theoretically almost zero current flows across the internal parallel connections. I thought of that when I built my bank, the outside bus bars are thicker towards the taps to carry the additional current and hopefully distribute the load equally. I attached a very busy diagram I used in my design phase that shows C1 to C4, representing the number of bus straps in each connection. But I think the "matrix" possible with LFP paralleling all the cells that can't be done with flooded strings is probably the most significant equalizer.

My clamp meter won't work on the bus bars, so I've never been able to measure this, but you're right, the single serial fuse would allow an easy measurement point.

[nimblemotors]

might be better to invest in general fire control system than fuses..