I wounder if I could design a safe series/parallel Li-ion battery pack by using different cells with different characteristics: different capacity, SOC, age.
I know that is VERY VERY VERY DANGEROUS at first sight but, my question is:
- Is it possible to design sufficient safety devices to obtain the same level of risk as with an homogeneous battery pack (same capacity, same SOC, same model, etc.) with dedicated BMS and PCB?
- If yes, what would be the possible side-effects: eg. the cells with smallest SOC or capacity will see their lifetime dramatically reduced?
- If no, is it possible with less in-homogeneities between the cells: eg. same capacity but different SOC and different cell model or same capacity, same model but different SOC?
- An underlying question is therefore: what are the undesirable phenomena (in terms of safety aspects and also cell lifetime) that can occur when charging or discharging a series/parallel pack composed of in-homogeneous cells? Among these phenomena, which could be avoided by the use of passive protection (diode, capacitor) or active (a kind of special BMS)? Which cannot be avoided?
Naive first pseudo-suggestion
Considering a 3s/3p Li-ion battery pack coupled with an appropriate BMS and PCB (let say a "classical" BMS/PCB model commercially available at less than 1$ on the internet):
- I wonder if I could use a device to protect each cell individually, as for example, one "Diodes Incorporated Battery Protection" (find here, see the following figure and main device features at the end) for each cells (so, 6 devices in this example) in addition to the global BMS/PCB (this individual cell protection device will be called "Cell Protection" in the following text and figures):
- Would it also be relevant to connect a diode in parallel with each cell in order to avoid polarization inversion?
- Would it also be relevant to connect a capacitor across the terminals of each parallel module (maybe to prevent high current variation coming from one cell to another if there is significant SOC differences between cells) ?
Here an example of the use of a "Cell Protection" for each cells (called "Cell + Cell Protection" in the figure below) coupled with a global BMS/PCB including a kind of "Supervisor" (not shown in the picture) which is able to tell the user that one "Cell Protection" is "activated" (and so tell the user that one cell is not working and so that the remaining cells, in the related parallel module, have to provide more current):
So, in my opinion, in this example, each cells should be carefully selected in order to be able to provide the full load current (during discharge phase) if the two other cells in the parallel module are not working because of a "Cell Protection" "activation". Or, the "Cell Protection" should be able to measure the cell current and "disconnect" the cell from the circuit if the current is above a defined threshold. In this last case, the "activation" of one "Cell Protection" in a parallel module could induce a chain reaction effect and "disconnect" the whole parallel module (it depends on the current threshold level) and so no current will be provide to the load anymore.
During charge phase of the whole pack, is this design able to prevent overcharging of the cells with lowest capacity or cells with highest SOC ?
Features of the "Diodes Incorporated Battery Protection" example cited above (find here) :
High voltage CMOS process, up to 24V (VDD to VM)
Low quiescent current (+25°C)
In normal mode, 3.0μA (typical), 4.5μA (maximum) VDD = 3.5V
In power-down mode, 0.1μA (maximum)
High-accuracy voltage detection circuit (+25°C)
Overcharge detection voltage: 3.5V to 4.5V (5mV steps) accuracy -15mV, +25mV
Overcharge hysteresis voltage range: 0.1V to 0.4V (50mV steps) accuracy ±50mV
Overdischarge detection voltage: 2.0V to 3.4V (10mV steps) accuracy ±35mV
Overdischarge hysteresis voltage range: 0V to 0.7V (40mV steps) accuracy ±65mV
Discharge overcurrent detection voltage: 0.025V to 0.2V (10mV steps) accuracy ±12mV
Short current detection voltage: 0.12V to 0.45V (50mV steps) accuracy ±50mV
Charge overcurrent detection voltage: -0.2V to -0.025V (10mV steps) accuracy ±12mV
Overcharger detection voltage: 8.0V (fixed) accuracy ±2V
Overcharger release voltage: 7.3V (fixed) accuracy ±2V
Built-in fixed detection delay time (+25°C), accuracy ±20%
Power-down mode selectable (Yes or No)
0V battery charge selectable (Permission or Inhibition)
Lead-free and fully RoHS compliant
Halogen and antimony free; "Green" device
Best Answer
Contrary to popular belief, a battery with mismatched cells is safe, but only if it includes a properly configured and installed BMS.
It is ineffective, but it is safe.
Yes. Set the BMS limits to the minimum range that is common to all cells in the battery. For example, if you're mixing LFP (3.2 V) and LCO (3.6 V) cells, configure the BMS for the narrower range of 3.0 to 3.5 V, which is safe for both Li-ion chemistries.
Reduced capacity because some cells have charge left in them that you can't access because doing so would overcharge or over-discharge other cells.
Faster degradation of the best cells because they do most of the work.
Series string:
What you are describing is simply an unbalanced series string. Any decent BMS, properly configured, will eventually bring the series string into balance.
Parallel cells:
Connecting cells at different SoC level is bad as there is a damaging inrush current that at best degrades the least-charged cell. Having said that, cells in parallel will naturally swap charge until they are balanced.
Safety: none, as long as the BMS is properly installed and configured.
Lifetime: as I said, the best cells degrade faster than they would in a well-matched battery.