0
\$\begingroup\$

I'm most interested in the application of replacing lead-acid packs with lithium-ion packs for UPS's and/or whole house/solar power banks. I've seen it stated Tesla's automotive packs are 6s with varying number parallel cells ~40-80p. I imagine the automotive application is on the extreme side of needing the most balancing.

I've not found good documentation on the ideal number of cells in series for balancing. Some concerns include:

  • What is a high count of series cells
  • Drawbacks and limitations of high count series.
  • Do drawbacks consist of slight reduction in performance(less capacity/lifetime), or are the consequences complete failure in short order.

EDIT: I've edited the first sentence for clarity(in bold). My original version was not clear. Also, I only mention lead-acid for setting the context of the use case.

Edit2: Tesla battery module with 444 cells, 6s74p 21v 7 wire bms enter image description here

Sources: https://www.youtube.com/watch?v=g_cC0NKvCDY https://www.youtube.com/watch?v=44uovhbMPxw

\$\endgroup\$
1
  • \$\begingroup\$ A rough estimate would be max capacity difference [Ah] divided by pack runtime [hours] = balancing current [A]. This is the tipping point for your system. The more peak balance current you have and the longer runtime you design for, the more the system can compensate. How many cells that works out to depends on how much miss-distribution you have and is not really N cell parallel dependent. In reality there are more things at play, for example dissipative balancing or (inductive) charge re-distribution. \$\endgroup\$
    – winny
    Commented Oct 10, 2021 at 18:05

2 Answers 2

5
\$\begingroup\$

Lithium-ion cells cannot tolerate overcharging at all. So the BMS balance channel is just one cell. Each cell or group of cells in parallel has to be managed independently of the others.

The cells may be arranged into groups or modules for convenience but each cell, or group of paralleled cells, will be managed separately. They are typically arranged in modules of 6-12 cells in series. The number of cells in parallel is not constrained and may be as done as a single cell up to many dozens as in the case of Tesla whose designs use small cylindrical cells. Since they are all in parallel they all have the same voltage so do not need to be treated individually.

In electric vehicles typically the total number of series cells are arranged to give about 400V (or 800v in a few eg Porsche Taycan and Kia). Usually each module of a few cells in series has its own circuitry that communicates with the central manager using opto-couplers or inductive coupling to isolate the hundreds of volts difference.

Balancing is usually only done while charging and diverts a small amount of the charging current from cells that reach full charge first. The balance may be fairly small, only tens of milliamps.

All implementations I know of use this passive balancing, although there are active balancing techniques where the excess charge is used to charge other cells by means of a DC-DC converter or capacitative charge pump they are in general too expensive,

Ni-Mh cells are much more tolerant about overcharge and can be managed as a series group of 6-12 cells. Toyota for example manages the NiMh cells in their hybrid vehicles in groups of 6 series cells with a nominal 7.2v.

Lead-acid may not need managing at the cell or cell group level at all as they have a convenient characteristic that their leakage current increases as they become fully charged. That leakage diverts current from the cell when it becomes fully charged, while other cells in series are still charging.

\$\endgroup\$
13
  • \$\begingroup\$ I've seen countless real world implementations where it's not 1 channel per series cell. And this is where my concern is. Ideally yes, 1 channel per series cell seems ideal, but I'm seeing otherwise being utilized. Are you familiar with Tesla's designs? Can you elaborate? \$\endgroup\$
    – nunya07
    Commented Oct 10, 2021 at 1:34
  • \$\begingroup\$ @nunya07 - That's surprising. The cells wouldn't last long. \$\endgroup\$ Commented Oct 10, 2021 at 3:39
  • \$\begingroup\$ @nunya07 - As I say in my answer all good implementations will sense every group of paralleled cells. I haven't seen the details of Tesla's implementation but I'm certain they do the same. The IC vendors make devices to do that easily. \$\endgroup\$ Commented Oct 10, 2021 at 3:46
  • \$\begingroup\$ @nunya07 - I just checked and for the Model S, Tesla uses the TI BQ75PL536A cell monitor as part of the BMS. \$\endgroup\$ Commented Oct 10, 2021 at 3:52
  • \$\begingroup\$ Agreed lithium cells don’t tolerate overcharging the way most others (lead acid, NiCd, NiMH etc) do, so any BMS that doesn’t have per-cell balancing is likely to result in reduced life at best. \$\endgroup\$
    – Frog
    Commented Oct 10, 2021 at 4:38
1
\$\begingroup\$

Batteries with N series cells like 6 or 12 for lead acid 12 or 24V are made with tight tolerances per cell, yet different cells may not match as well so balancing extends the life of series banks by equalizing the voltage. Thus VI=P bypass capacity to enable equal voltage on each under load charge or discharge.

Added

6 Li-Ion cells =6S is a standard "24V" module for BMS then x number of parallel is selected for the Ah required load. Cascading each xP6S module then can managed digitally by a Master BMS up to any voltage usually <1kV depending on the kW load to minimize conduction losses and cost of copper such that the ratio of source ESR / load R resistance is a relatively small ratio (like 1%) to minimize load regulation error on the battery voltage and Dissipation Factor in heat rise of the cells.

The resulting battery array for BMS is then (xP6S)x N modules. The Dissipation Factor DF may be expressed as a power loss and heat rise of the cells just as it is done for capacitors. However battery cooling is essential and the DF of load must be chosen by design with thermal resistance and max Temp. rise of cells.

( confirmed by consult today with son-in-law U of T Prof in EE Power Eng.)

Tesla S and BMW 3i both use a Siliconix IC for dual port bi-directional capacitive RF coupling of data from the HV BMS side to the LV Master BMS for SoC , V matching for each module.

====

The weakest cell (Ah) discharges and charges the fastest will need the most bypass to prevent over /under charge voltage. The parallel banks don't have this problem.

The optimum number of series cells in a string is a question of BMS bypass capacity per cell, deviation of capacity in the string, and temperature and reliability margin. Also the time available for balancing (if not 24/7).

It is unanswerable without specs for the actual BMS design of bypass capacity from a cost limit per cell and the deviation is based on the supplier specs of mismatch from new to the life of the module and the max temperature rise of the batteries due to BMS and operational use.

Thus, the cost of quality batteries and deviation of capacity are underdefined variables that must satisfy the limitations of the cell monitor and balancer.

These are not trivial problems when you consider some BMS systems support up to 256 cells in series for 1kV.

Here is a patent on one such system. and its website

The above appears to be a passive bypass method. Another I know uses a half bridge DC-DC to transfer excess energy with LC storage to the other cells in series, covered by a patent, which is not handy at this time.

\$\endgroup\$
8
  • 1
    \$\begingroup\$ Please see my OP again, I didn't make it clear what I was asking. \$\endgroup\$
    – nunya07
    Commented Oct 10, 2021 at 4:20
  • \$\begingroup\$ I hope you don't mind if I made a few corrections to your answer. Specifically, do not conflate "cell" and "battery". I know you know the difference, based on the high quality of your answer. Also, balancing only balances the SOC (and therefore voltage); not the current. \$\endgroup\$ Commented Oct 10, 2021 at 22:41
  • \$\begingroup\$ TY @DavideAndrea Yes, current must be bypassed to regulate V & SoC for equalization and temp. rise, \$\endgroup\$ Commented Oct 11, 2021 at 8:47
  • \$\begingroup\$ The 2014 BMW i3 I studied used LTC6801/2 devices. \$\endgroup\$ Commented Oct 11, 2021 at 21:59
  • 1
    \$\begingroup\$ Thank you for this informative answer Tony! \$\endgroup\$
    – nunya07
    Commented Oct 13, 2021 at 4:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.