Best way to manage this LFP house battery

[max2k]

At the end of the day, certainly, everyone will have a different safety level that allows them to sleep at night. Myself, I'm primarily worried about something so exothermic and rapid that it prevents me from physically escaping a runaway before it becomes deadly.

OK, that's your choice

I'm convinced LFP can't achieve this, while conversely I think the cobalt chemistries are all too risky. There are plenty of other combustion mechanisms on board my RV that can also result in the whole thing ending up on fire; I can't fully eliminate all of the risk in any case.

there's a difference- other 'dangerous' things on board of your RV probably passed certain certification before you bought them. That would mean they have history of things going wrong and those certifications reflecting this. You creating LFP bank yourself and I simply wish you to avoid learning one or two small details the hard way. I have no doubt you know what you're doing electrically but safety is different kind of expertise often learned off failed designs.

You're talking about AC bleeding back into the DC side of the system? I would expect that to be well-filtered, but also I think if it were causing measurement issues we would hear about them (?). My DC SOC estimator seems very, very accurate so far at least at 1/6-scale (AC and DC) loads. I can put my scope on my shunt to see what we see, but practically speaking I don't think there is much of an effect, if I'm understanding what you're saying. I'll put it on the list to test, though... it's easy enough.

In abstract sense- yes, AC bleeding. What I meant is inverter feeding AC loads would consume energy according to sine wave on the load side. It would cause equivalent unipolar sine wave on DC side in the consumed current as the voltage would remain relatively constant.

Depending how big that AC component on top of DC component of the current going through shunt is it could be ignored for the purposes of coulomb counting or not. If you could hook up scope parallel to the shunt you'd see this right away.

[nebster]

worth a thousand words. These burnt battery boxes were not DIY:

It's certainly effective marketing...

[nebster]

Mike's examples are different as I believe they show relatively fresh banks and it is hard to tell the exact reasons for those failures. My concern with Mike's cases is the empty battery cabinets- looks like there were no design measures taken to prevent chain reaction.

At the end of the day, certainly, everyone will have a different safety level that allows them to sleep at night. Myself, I'm primarily worried about something so exothermic and rapid that it prevents me from physically escaping a runaway before it becomes deadly. I'm convinced LFP can't achieve this, while conversely I think the cobalt chemistries are all too risky. There are plenty of other combustion mechanisms on board my RV that can also result in the whole thing ending up on fire; I can't fully eliminate all of the risk in any case.

I'm underwhelmed by the LFP combustion evidence that's available so far. I want to see documented examples of thermal events in LFP systems that we can show are reasonably-managed, and then those events need to exhibit the kind of high-speed reactions that differentiate them from traditional electrical fires or propane fires or similar, common occurrences to really justify an aggressive additional layer of physical protection.

That and also in another thread I was asking karrak details of his data logger and came to realization AC inverter must introduce AC component on the current shunt at double AC frequency, should look similar to the AC component on top of DC coming from common diode rectifier connected to a capacitor. I never worked with LFP batteries myself but I think hooking up scope there and checking how big is that AC component relatively to DC is worth the trouble. In karrak's case his data logger is taking samples at about 1 Hz intervals and if 120Hz AC component is present it would be also present in the measured data but at much lower frequency due to aliasing. Depending on AC 'noise' amplitude it might present a problem or not.

You're talking about AC bleeding back into the DC side of the system? I would expect that to be well-filtered, but also I think if it were causing measurement issues we would hear about them (?). My DC SOC estimator seems very, very accurate so far at least at 1/6-scale (AC and DC) loads. I can put my scope on my shunt to see what we see, but practically speaking I don't think there is much of an effect, if I'm understanding what you're saying. I'll put it on the list to test, though... it's easy enough.

[Mike90250]

worth a thousand words. These burnt battery boxes were not DIY:

Li-IonFires.png

[sensij]

That and also in another thread I was asking karrak details of his data logger and came to realization AC inverter must introduce AC component on the current shunt at double AC frequency, should look similar to the AC component on top of DC coming from common diode rectifier connected to a capacitor. I never worked with LFP batteries myself but I think hooking up scope there and checking how big is that AC component relatively to DC is worth the trouble. In karrak's case his data logger is taking samples at about 1 Hz intervals and if 120Hz AC component is present it would be also present in the measured data but at much lower frequency due to aliasing. Depending on AC 'noise' amplitude it might present a problem or not.

Take a look a @BackwoodsEE's mppt thread for evidence of this. He is plotting the input side of the mppt at 4096 Hz sampling, but if you can see it there, you can be sure it is on the battery side as well.

[max2k]

So I am going to switch back to physical switching.

I'd too + fuses

Yeah, the energy contained in these cells coupled with their oxidative stoichiometry makes them a big fire/burn risk. However, I don't believe there are any documented runaway explosions, and actually physically isolating these things would be prohibitively heavy and expensive I am afraid. I mean, think about it: so what if you put a steel plate between two adjacent cells? The steel plate will get red-hot in a true runaway, and then the next cell will overheat just as well. You'd probably just be better off put the whole pack in a big steel box and evacuate it.

Me, I'm content to know that I will have time to respond if there is an emergency thermal event. I'll put the whole pack in an enclosure, but if a cell goes nuclear, the pack is possibly going to be taken out by ongoing burning. I can't find anyone who's actually had this happen with LFP, and there are a ton of multi-cell packs in lots of boats and RVs now, so I'm not too worried.

You can put steel separators between cell compartments and then some fire resistant 'pads' between cells and the steel separators to create air gaps and prevent adjacent cells heating up as a 'whole'.

Check out messages #21 & 23 - Mike posted pictures and gave some details on how those LFP batteries failed:

https://www.solarpaneltalk.com/forum/off-grid-solar/off-grid-solar-panel-systems/361015-proposed-off-grid-design-attached-would-love-your-input/page2

While I agree there's not much noise about LFP banks catching fire given their widespread it also might be due to not enough years ppl put on them. Mike's examples are different as I believe they show relatively fresh banks and it is hard to tell the exact reasons for those failures. My concern with Mike's cases is the empty battery cabinets- looks like there were no design measures taken to prevent chain reaction. I myself used much more dangerous Li-po cells in my lawn mower for several years but that thing was always charged from balancing charger, had controller inside preventing current overload or over discharge and most importantly saw probably 50 cycles over its lifetime. While it worked I wouldn't put this as a proof someone can scale that up to house or RV case or even simply repeat the design. Your 112 cell bank will be cycled daily and it weights 600+ pounds so in case of fire I doubt you'll be able to carry it out in time. If it was me I'd buy few extra smaller similar cells and killed them myself just to see how it goes and sleep better at the end .

I've been pretty casual with numerical precision in this thread, I admit. I don't have precision data for the ADCs on the battery monitor or inverter hardware. I know my Fluke and Klein meters are 2k and 6k counts respectively, and their measurements do cross-check.

I think you might be saying that voltages under different loads are not directly comparable, and that's certainly true. There will be huge swings when the full-size pack is under full-size loads, so the emergency cutoff voltages will have to account for that for sure.

That and also in another thread I was asking karrak details of his data logger and came to realization AC inverter must introduce AC component on the current shunt at double AC frequency, should look similar to the AC component on top of DC coming from common diode rectifier connected to a capacitor. I never worked with LFP batteries myself but I think hooking up scope there and checking how big is that AC component relatively to DC is worth the trouble. In karrak's case his data logger is taking samples at about 1 Hz intervals and if 120Hz AC component is present it would be also present in the measured data but at much lower frequency due to aliasing. Depending on AC 'noise' amplitude it might present a problem or not.

[nebster]

MOSFETS have 3-4 mOhm resistance and the lower it goes the lower voltage they can sustain in the closed state. At 100A it becomes 0.003 x 10, 000 = 30W heat to dissipate somehow. Due to positive temp coefficient of this resistance they can be connected in parallel without using any extra components. 10 such things would bring that heat down to 3W and spread the current between themselves 'automagically': the one with higher current will heat up more leading to reduction of current.

The (fully engineered) product I was using had a huge heat sink, so it probably had numerous FETs chained up to handle the current. The problem wasn't heat, it was design: they act as an open circuit in the other direction, so I'd have to at a minimum put two back-to-back opposed to each other. The vendor wasn't willing to agree that they would function correctly in that configuration.

Another concern I have with FETs is their failure mode: they fail as a closed circuit, with a non-linear, cascading breakdown in the transistors. Not only is that the wrong way to fail for an emergency disconnect, but it could fail quietly, so I might not even know I have lost my protection mechanism!

So I am going to switch back to physical switching.

- fire proofing your setup. From what I read when something goes wrong you want to contain the problem within failed cell and with those 100Ah ones you'll have plenty of smoke & heat to deal with. You mentioned PVC frames and I hope these will be replaced with something more fire/heat resistant in your final product so when 1 cell melts the others will remain in their places and won't be affected by heat from the failed cell preventing chain reaction.

Yeah, the energy contained in these cells coupled with their oxidative stoichiometry makes them a big fire/burn risk. However, I don't believe there are any documented runaway explosions, and actually physically isolating these things would be prohibitively heavy and expensive I am afraid. I mean, think about it: so what if you put a steel plate between two adjacent cells? The steel plate will get red-hot in a true runaway, and then the next cell will overheat just as well. You'd probably just be better off put the whole pack in a big steel box and evacuate it.

Me, I'm content to know that I will have time to respond if there is an emergency thermal event. I'll put the whole pack in an enclosure, but if a cell goes nuclear, the pack is possibly going to be taken out by ongoing burning. I can't find anyone who's actually had this happen with LFP, and there are a ton of multi-cell packs in lots of boats and RVs now, so I'm not too worried.

your apparent trust in voltages you observe on your 16 cell string. When you're saying '54.2V' it is 3 digit number requiring better than 0.1% precision from components which supposed to maintain it, including cells themselves. That is not possible IMO especially at these power levels: even charging/discharging will introduce 'noise' exceeding 0.1V resolution: 10A over 20mOhm is already 0.2V.

I've been pretty casual with numerical precision in this thread, I admit. I don't have precision data for the ADCs on the battery monitor or inverter hardware. I know my Fluke and Klein meters are 2k and 6k counts respectively, and their measurements do cross-check.

I think you might be saying that voltages under different loads are not directly comparable, and that's certainly true. There will be huge swings when the full-size pack is under full-size loads, so the emergency cutoff voltages will have to account for that for sure.

If I was doing this I'd rely more on coulomb counter of sorts making sure I reset it on every cycle due to its integrating nature and more direct relevance to what matters while using voltages for alarms/safety cut offs.

Yes, I agree, and that's exactly what I'll do. Probably trigger charge at something like 20% SOC and stop charge based on a voltage (since reverse load is constant there, voltage becomes a stable indicator).

There's danger in half knowledge: you know enough to connect things together but you lack experience in those small details which matter one day but impossible to see at the beginning. I sincerely wish you success with your project and please share your progress / problems as you go, I think it will be very valuable for all of us.

Completely agree -- I bet there are ten more "small" things I'm not aware that I'm not aware of yet! That's something I'd love to glean from the collective expertise here.

[max2k]

Each s16 string is fused MRBF at the positive pole, AWG 1 to a pair of 1000A busbars. Main feeders are 2x AWG 2/0, also MRBF-fused at the busbar at 300A each, so DC design peak draw is 600A. Disconnects are Gigavac MX14s with PTC-fused override. I haven't settled on a string-level disconnect yet, because the ones I thought I was going to use turned out to be an asymmetric MOSFET instead of what I assumed was a symmetric design, so they can't interrupt charge current. Probably will use a smaller set of Gigavacs, though.

MOSFETS have 3-4 mOhm resistance and the lower it goes the lower voltage they can sustain in the closed state. At 100A it becomes 0.003 x 10, 000 = 30W heat to dissipate somehow. Due to positive temp coefficient of this resistance they can be connected in parallel without using any extra components. 10 such things would bring that heat down to 3W and spread the current between themselves 'automagically': the one with higher current will heat up more leading to reduction of current.

It may seem off-putting because of how I've arrived on the scene and seemingly skipped a few years of muddling about with smaller arrays -- but, hey, at least I haven't adopted a scornful, dismissive attitude. You still haven't shared any of the anecdote or hard data that several of us requested to help justify some of your bolder assertions, nor have you really attempted to answer the original question that started this thread (except to suggest that everything is "wrong," in your opinion and with ample boldface and underlining).

I don't know at what point it will become clear to you that I am a bit more resourced and prepared than you have insinuated, but I think everyone would appreciate if you'd lay off the melodrama going forward.

not to me although I'm lacking particular LFP experience. My only 2 concerns are:

- fire proofing your setup. From what I read when something goes wrong you want to contain the problem within failed cell and with those 100Ah ones you'll have plenty of smoke & heat to deal with. You mentioned PVC frames and I hope these will be replaced with something more fire/heat resistant in your final product so when 1 cell melts the others will remain in their places and won't be affected by heat from the failed cell preventing chain reaction.

- your apparent trust in voltages you observe on your 16 cell string. When you're saying '54.2V' it is 3 digit number requiring better than 0.1% precision from components which supposed to maintain it, including cells themselves. That is not possible IMO especially at these power levels: even charging/discharging will introduce 'noise' exceeding 0.1V resolution: 10A over 20mOhm is already 0.2V. If I was doing this I'd rely more on coulomb counter of sorts making sure I reset it on every cycle due to its integrating nature and more direct relevance to what matters while using voltages for alarms/safety cut offs.

There's danger in half knowledge: you know enough to connect things together but you lack experience in those small details which matter one day but impossible to see at the beginning. I sincerely wish you success with your project and please share your progress / problems as you go, I think it will be very valuable for all of us.

[nebster]

Nebster - it's time to get off the paper and into the real world.

I love it, except the tone of your response. I'll respond, however, for the benefit of future readers.

Since your installation is mobile, and not fixed, have you properly connected all 224 terminals with Nord-locks and not just regular hardware? We're also assuming you've properly cleaned all the terminals and nickel interconnects and a light coating of No-Alox or similar.

All Nordlocks. No noalox until the final install.

Are the casings of those cells "hot" with any polarity? That is, if you scrape or breach the thin coverings, will you be in trouble, especially if those cells are prone to movement in an RV environment?

The casings are not designed to be hot, but of course they may well have some leakage. They are wrapped in a thin plastic sheathing and each cell is separated with a set of PVC frames that create the aforementioned spacing. It's the best of both worlds: aluminum-shell energy density and excellent convective cooling.

You haven't done your homework. You are "winging it", and will end up with wing-it results, no matter how many graphs and charts and speculation you cast upon it.

This is where you've allowed yourself to descend into the classic, puerile abyss of so many discussion forums (and so many historical threads on this one). Step back for a minute, and then stop it.

But, for total clarity: of course I'm winging it. That's part of the challenge, and it's part of what makes it so rewarding. Of course I could just go buy something pre-engineered, but what fun would that be? By the way, I can't believe you're faulting me for providing graphs and data to support engineering reasoning.

Let's talk infrastructure. What are you using for DC disconnects, fusing and the like? No matter the chemistry, there is a LOT of power at your disposal, and make a mistake here you end up with a plasma ball and not an RV. Even the early EV pioneers running high-end agm's in the bed of their trucks know this. Look up how "Plasma Boy Racing" got his name. Here we are just talking agm, long before lithium came along for the consumer.

Each s16 string is fused MRBF at the positive pole, AWG 1 to a pair of 1000A busbars. Main feeders are 2x AWG 2/0, also MRBF-fused at the busbar at 300A each, so DC design peak draw is 600A. Disconnects are Gigavac MX14s with PTC-fused override. I haven't settled on a string-level disconnect yet, because the ones I thought I was going to use turned out to be an asymmetric MOSFET instead of what I assumed was a symmetric design, so they can't interrupt charge current. Probably will use a smaller set of Gigavacs, though.

There's also a precharge mechanism and numerous other components, and a full schematic that's been on my drawing board for nearly a year now.

Unfortunately, I'd have to say that this is more of a wallet-emptying exercise, ignoring those who have gone before, and ultimately an UNSAFE installation for both you and your fellow travelers.

It may seem off-putting because of how I've arrived on the scene and seemingly skipped a few years of muddling about with smaller arrays -- but, hey, at least I haven't adopted a scornful, dismissive attitude. You still haven't shared any of the anecdote or hard data that several of us requested to help justify some of your bolder assertions, nor have you really attempted to answer the original question that started this thread (except to suggest that everything is "wrong," in your opinion and with ample boldface and underlining).

I don't know at what point it will become clear to you that I am a bit more resourced and prepared than you have insinuated, but I think everyone would appreciate if you'd lay off the melodrama going forward.

[PNjunction]

Nebster - it's time to get off the paper and into the real world.

Since your installation is mobile, and not fixed, have you properly connected all 224 terminals with Nord-locks and not just regular hardware? We're also assuming you've properly cleaned all the terminals and nickel interconnects and a light coating of No-Alox or similar.

Are the casings of those cells "hot" with any polarity? That is, if you scrape or breach the thin coverings, will you be in trouble, especially if those cells are prone to movement in an RV environment?

You haven't done your homework. You are "winging it", and will end up with wing-it results, no matter how many graphs and charts and speculation you cast upon it.

Let's talk infrastructure. What are you using for DC disconnects, fusing and the like? No matter the chemistry, there is a LOT of power at your disposal, and make a mistake here you end up with a plasma ball and not an RV. Even the early EV pioneers running high-end agm's in the bed of their trucks know this. Look up how "Plasma Boy Racing" got his name. Here we are just talking agm, long before lithium came along for the consumer.

Unfortunately, I'd have to say that this is more of a wallet-emptying exercise, ignoring those who have gone before, and ultimately an UNSAFE installation for both you and your fellow travelers.

[nebster]

Today I tested a charge from about 50% SOC with a CV threshold of 54.2v, or 3.39vpc. I left CC at 50A, although in real life it'll need to be closer to 25A. This time, we hit the CV after 8 minutes, and the minimum of 1 hour absorption brought the pack up to 87% SOC. Getting closer!

I did another quick test with absorption and float both set very low, 52.8v. There is essentially no current flow into the cells at this setting, so I believe this gives me one way to eliminate any absorption without resorting to extra relays or logic.

After the pack discharges again tomorrow, I'll test that and see how far we get under just CC.

I think my preference for day-to-day pack use is going to be something like:

1) On shore power, charge pack to ~60-80% SOC and stop.
2) When not on shore power, start generator at 20% SOC and charge to 90% SOC. Turn generator off.
3) If SOC falls below 20%, alarm.
4) Disconnect pack at 5% SOC.

When we add PV, I'd like to set the charge voltages on the MPPT to be below the shore charge settings, but just enough to make extra current available for loads. I imagine this will be a bit of a balancing act to get it "just right."


Screen Shot 2017-09-20 at 10.32.24 PM.jpg

[karrak]

For the graph you included, is this a bottom-balanced set of cells? Because that's pretty much what mine look like, too. Notice how all your cells are super close in voltage right before you reach CV? How about I just set up the charging logic to never go into CV? I can still set a HVD or charge-terminate relay at the pack voltage where I see the worst cell get up to 3.6V or so, but I don't even think I need to go that far. My minimum acceptable capacity is 25kWh, and this pack is capable of at least 40% more than that at the outset.

That graph is for a top balanced set of cells that is slightly out of balance, probably around 1% out, certainly no more that 2% out.

If you are not charging from solar I would definitely say try to terminate the charge at the start of the CV phase. If you really want to extend the lifespan of the battery try to keep the SOC between 70% and 30%.

I would look at using your coulomb counting SOC meter to start and stop the charge cycle. If your SOC meter can compensate for the coulomb losses of the battery you only have to reset it around once a week by charging to nearly 100%.

Simon

Off grid 24V system, 6x190W Solar Panels, 32x90ah Winston LiFeYPO4 batteries installed April 2013
BMS - Homemade Battery logger github.com/simat/BatteryMonitor/wiki
Latronics 4kW Inverter, homemade MPPT controller

[karrak]

I'm currently evaluating Victron Quattro 48/15000/200/100-100. That decodes to 48vdc nominal, 15kva inverter, 200a charger, dual 100a AC feed-through. I believe it has the exact same programmability as all of the Quattros and perhaps the Multipluses as well. Which is to say, less every day, because they seem to be removing features from the software year over year.

So am I correct that you will only be charging your battery with a CC of 200A?

I think I read this discussion too, somewhere. I'll have to go look it up and see if there's more to learn from his experience. I don't find it hard to imagine that a cell might do something like that at some point, but I'm also expecting that it will happen gradually and that I'll be able to observe the side effects of it with careful monitoring. It's possible that I'm wrong on one or both of those expectations.

What concerns me is that these sort of problems only show up at the end of the charge cycle or when the battery is nearly empty. I think it is very easy to get complacent after things have been running well for a long time and miss the signs of things going amiss.

Like you, I'm less concerned about catastrophic shorts. Were one to occur, though, the parallel-then-serial topology makes that pretty exothermic without cell level fusing.

This is one reason I quite like the Tesla approach of lots of small cells in parallel, all each individually fused.

And have you seen a complete failure? Could you also comment on the general type of work you've needed to do and at what frequency?

I only had any problems with my battery right at the start. I had asked the supplier to balance my battery as I didn't have the equipment to do it at that time. This had not been done properly and I didn't pick it up. I think it was the first night that The batteries were being used that the low voltage alarm on the cellog 8 that I was using at the time went off at about 4am. I turned the inverter off and turned it back on when I was getting enough solar power to charge the battery. It took me a while to get the battery balanced as it was in service. I estimate that it was out of balance by around 20%.

The only other major service/upgrade was when I installed my own BMS. The battery has been in service on a continuous basis since it was installed.

The battery I installed for a friend has been in operation just over four years. He is not as vigilant about power usage as me and has had a few occasions where he has had to turn off his inverter during the night. There have also been a few occasions where his battery has got out of balance enough to disconnect the solar panels from his solar controller. Due to balance issues I thought at one time that one of his cells might have lost capacity so purchased a new one and did a substitution. I tested the one I removed and found to my surprise that it still had its rated capacity! I have installed one of my BMS's on his system so I can now check his battery remotely which gives me peace of mind.

Both batteries are manually balanced. I have found that as they have aged they stay in balance much better than they used to. My battery balance hasn't changed much in the last eighteen months. Still having to do minor balancing of my friends battery which is complicated by the fact that he lives about 100 miles from me.

I'm surprised this is so challenging. Maybe I'm being too blase about this SOC thing, but if I'm not planning on running from 0 to 100 (or anywhere close to either of those boundaries), ever, then I feel like I have a ton of room for slippage, even though on top of that we observe very little in real packs where I have seen data. Heck, I'll barely even get these cells out of their linear regime, it seems like.

It doesn't have to be challenging if the system is set up and commissioned properly in the first place. Unfortunately there is allot of "devil in the detail" that has to be addressed. My system and my friend's are pretty well hands free these days. With top balancing both our system's solar controllers charge to 3.45V/cell(27.6V for my system and 55.2V for my friends system), my MPPT controller terminates the charge when the current drops to ~C/50 which give about 10-15 minutes of absorb, my friend's commercial controller had a minimum absorb time of one hour which is not ideal but works OK. Both our systems float at 3.35V/cell. This charge regime charges our batteries to around 98%-99% and holds them at around this SOC if there is enough sunshine until nightfall. At the low end my BMS generates an alarm when any cell drops below 2.8V/cell.

This regime is only good if you are using the battery on a continuous basis. If the battery is standing idle for any length of time you should keep the SOC below 70% and below 50% for long term storage.

Just remember that the "linear region" can go all the way up to around 98%-99% when charging with solar or low charge currents.

Simon

Off grid 24V system, 6x190W Solar Panels, 32x90ah Winston LiFeYPO4 batteries installed April 2013
BMS - Homemade Battery logger github.com/simat/BatteryMonitor/wiki
Latronics 4kW Inverter, homemade MPPT controller

[nebster]

thanks for the link; while I didn't find much details in there about why parallel LFP strings would wear out quicker I found interesting external link there: http://www.smartgauge.co.uk/batt_con.html It considers effects of the way batteries wired in parallel on the amount of current they pass. As author rightfully notes in low internal resistance applications resistance of the wires between batteries becomes comparable to the internal battery resistance leading to significant imbalance of currents flowing through parallel batteries. Ideally all batteries would need to be connected to the load points with individual wires of the same length. With LFP cells lower Ri this is x10 times more applicable so OP might want to take a look there.

Yep, I'm on it. I'll be looking to break down the interconnect resistance we've calculated in this thread to see where it's coming from, but the plan all along has been to ensure identical wire runs to the busses.

I'm already of a weakly-founded belief, though, that the interconnect R swamps the battery Ri such that string load imbalance won't amount to much in practical terms.

[max2k]

While I don't think this referenced thread provides any hard data, and much of the headline is buried since the diagrams are apparently broken in it, I do think there is merit in being concerned about parallel strings having load imbalance issues. However, my back-of-the-envelope calculations suggest that there is enough lossy resistance in the interconnect (fusing, contactors, lugs, wire) to make the current division pretty minimal across a set of six or seven large strings. It's possble, though, that some degree of wear leveling may be warranted at some point.

There are lots and lots of paralleled banks in commercial, engineered LFP packs (and in tons of EVs). These seem to do okay? They benefit from state estimation in each module ("string") but, nevertheless, I have no doubt that they experience some uneven wear over time.

what makes currents more balanced is higher internal resistance inside strings compare to lower- between them. For 2 strings connected one to another and then to load it's something like Ri / (Ri + Rwires in between) If you can make inter- string wires resistance fraction of internal string resistance or change your wiring to more star looking shape it will balance the currents. That link is an interesting read.