One or Multiple BMS strings ? or why 1+1 ≠ 2 !

Grab peanuts and soda 'cause this one is going to be a long drag...

Per this van's requirements & priorities (see the "Planning the Electrical & Energy System" post) the battery system is organized like this:

This modular pack approach will allow the user to size the total pack capacity as needed.

Note: although I call these packs Summer and Winter, thanks to the modular trays in the garage it is actually possible to choose any combination to suit one's needs or season: S, W, or S + W.

If adventuring in a sunny place, not much battery capacity will be needed. So using only the smaller Summer pack will still provide enough reserve to, for instance, power the A/C unit at night. And it will save 100s of pounds in weight and free up a storage tray in the garage.

However, when traveling to, say, Alaska, much more capacity will be required for heating, showering, cooking. So, adding a complementary pack will boost capacity. But to the detriment of the van's weight and volume for toys in the garage.

So, the Summer pack will comprise 12 cells, in a 4S3P configuration, for a 12 x 3.2V x 280Ah = 10.7KWh capacity. Anticipated weight: ~200lbs (90kg). The Winter pack will hold the remaining 20 cells in 4S5P, adding 18KWh and ~300lbs (135kg).

Each pack must be able to deliver the system's 3000W max load, i.e. about ~250A at 12V. A BMS solution (battery management system) must be used to protect each pack against various cell-destroying events, such as over/under voltage, short circuit, over/under temp, etc.

In a pack with parallel strings, 2 different BMS strategies can be considered: 1 BMS per string (or set of strings), or 1 BMS total:

Parallel BMSs provide redundancy. Should 1 string fail, the others will still be active and you won't lose all the power.

That works great when the application's max current can be supported by each BMS. For instance, many solar installations use that approach as several strings are necessary to provide a very large capacity, without requiring the max output power to scale accordingly.

However, since the vast majority of BMSs do not support current sharing between them, a parallel BMS architecture does not guarantee power scaling. I.e. two 100A BMSs in parallel do not result in a 200A system.

Indeed, BMSs are just simple devices, even those marketed as "Smart". They care for and monitor only the cells they are connected to. Should the mean world out there do anything funky to their cells, the BMS will swiftly react by just disconnecting the whole string. Anything else a BMS can do (balancing, sing a song via bluetooth...) is just secondary.

That's why 1+1 ≠ 2 🤯

So, any resistance or capacity imbalance that develops between strings could result in 1 string outputting notably more current than the other(s).

Yes, it will happen, eventually. Cells will age, non-uniformally. Their internal resistance will increase, non-uniformally. Cells in the middle of the pack will experience higher temps, thus aging differently. Cells are also rated only for a finite number of charge / discharge cycles before they substantially degrade. Etc.

After years of use, all these factors will possibly lead to 1 cell with notably less capacity than the others. Once this weakest cell is depleted its BMS will open the circuit, leaving the other string(s) alone to do all the work. If each BMS is not sized for that, all the strings will now go offline.

Or, as an anonymous guy once erroneously concluded:

Did you spot the error ? Yes, the actual answer in our case is 6-3 = 0 ! All the strings go down.

Mitigations can be considered, though. Such as using BMSs with a higher current capability. Or overprovisioning the pack's capacity from the start to provide a few more troublefree years. Or setting up the load-side devices (like the inverter) to stop drawing current at a Low Voltage Cutoff value higher than the BMS's Discharge Low Voltage Cutoff. Or reducing the used capacity over the years. Or very carefully matching cell characteristics during assembly, and then measure / shuffle / replace cells regularly, as their characteristics drift over time between them. Etc...

Or simply ignore the situation, and lose power as soon as 1 string fails, if too much current is now demanded of the other strings / BMSs.

But since this van requires 300A (250A + design & safety margins), and most affordable BMSs top out at 100A, that made my choice simple. First, 100A per string is too little. Small developing imbalances between the 3 strings of the Summer pack will sooner or later trip that 100A limit on one of the strings, bringing all the strings down. Second, I did not relish the idea of buying, mounting, cabling and managing 8 BMSs...

Also, wiring cells in parallel provides the distinct advantage of boosting the apparent capacity of the weakest cells.

For example, let's consider 4 cells where one presents 276Ah of capacity and the other three 284Ah. If wired in series, only 276Ah can be drawn (at the 1C discharge current) before the weakest cell reaches 2.5V (LiFePo4) and the string gets shut off by its BMS.

However, if wired in parallel, the capacity is now 276+3*284 = 1128Ah. That's an apparent capacity per cell of 282Ah, before the 4 cells reach 2.5V together. I.e. 2.17% more !

So, when building a pack with parallel BMS strings, if getting all the capacity from the cells is critical, you'll want to put all the weakest cells on 1 string, then the next weakest ones on the next string, etc. Once the weakest cells are depleted their string will be shut down by their BMS but the others will keep delivering current, if the drawn current is not too high. Yes ?

Conversely, when building a pack with 1 string / 1 BMS / several cells in parallel, if getting all the capacity from the cells is critical, mix the weakest and strongest cells across the pools of parallel cells. Such that the per-pool capacities are all the same. Done.

So, then, 1 big BMS to rule all them cells ! Right ?

Well, maybe not. Might go with 2 😅 Explanation in the next post...