Following on from his article in the Autumn 2022 issue of BEST, Chris Hale looks at the variety of methods for selecting and balancing cells in a battery pack. The defects are there (within the cells), finding them is a different matter, and can be a key difference between packs available at different price points.
When assembling a battery pack with cells of varying quality, there are a number of things you can do to mitigate problems caused by underperforming cells or combining cells at a different state of charge. These remedies can take an awful amount of time and money. Unfortunately, not all battery companies allocate enough resources to this task and their batteries can fail prematurely as a result. That can have knock-on effects in terms of warranties. It doesn’t have to be this way.
The issue of how long a battery pack will remain in useful operation, to guarantee a warranty, is a challenge faced by manufacturers and may result in either:
- very short warranty periods
- higher warranty repair costs or
- restrictions on usage to invalidate a warranty.
Knowing when a battery’s health can suddenly decrease, known as the ‘knee-point’ effect, is a hotly studied problem among lithium-ion battery researchers. It is fairly well understood that avoiding high temperatures at full charge, deep discharges and reducing the amount of fast-charging, enables the battery to perform and last the expected lifetime of the battery pack.
Of course, there is a lot more to it. The length of time in storage, storage temperature, load stressing, low temperature charging, balancing issues from uneven quiescent drain from attached electronics, not to mention of course the quality of cells used.
It might also be assumed that all cells in the battery pack undergo the same degree of degradation from any of the above factors (except battery management system (BMS)) – given that they are all subject to the same environmental and load/charge conditions. In reality, that’s not entirely true as it is likely there will be thermal gradients across the pack and uneven heating of cells in the centre compared to those on the periphery (or vice-versa if external heat sources are applied). The state-of-balance between cells in a series string and the capacity of the weakest cell will also play a significant part.
The weakest cell will typically reach the top-of-charge (ToC) first and be subject to the highest degree of balancing, it may also have the highest internal resistance and generate more self-heating. Alternatively, a weaker cell with a higher self-discharge may exacerbate balancing issues. It is also likely that a parallel collection of cells will mask a weakening cell, making it harder to detect, until a critical failure or the knee point is reached, dragging down the other cells.
Good pack design and BMS
Good pack design and BMS can improve a number of the challenges related to; thermal spread, balancing, operation, depth-of-discharge (DoD) and state-of-charge (SoC). It may not help with issues of cell quality or ageing spread due to manufacturing defects/variations on the other hand. As reported in the Autumn issue of BEST, cell quality during the manufacturing process is certainly a factor relating to the spread of cell ageing. Fig 1 demonstrates a normal distribution of cell failures and lends itself to the question of what spread can be applied for a good or bad cell supplier.
If we know we can expect a spread of cell failures through the life of a pack and that in some cases the weakest cell can promote an acceleration of its own demise through BMS charge/balancing management, can we mitigate the outliers through initial production quality testing? What are the most reliable or cost effective tests and will they find cells with manufacturing defects as well as general outliers?
Cell matching before assembly
A 2013 paper on battery safety from NASA said testing of a cell’s self-discharge properties to identify subtle manufacturing defects on its own failed to be reliable for defect detection. It advised screening methods such as internal resistance, capacity, 3-sigma open cell voltage (OCV), mass and dimensions to determine outliers.
It adds that a “very critical aspect” in the understanding of subtle defects is to carry out destructive analysis of cells from every lot. This is to confirm the quality of production and screen all cells and batteries in a stringent manner to have a high quality set of flight cells. Albeit that destructive testing is not a viable approach for general manufacturers.
So, what level of cell matching do you actually do prior to assembling a battery pack? Assuming the battery pack will be balanced the first time it is charged and in use. Also, assuming the cells are assembled in series, there are eight options.
- None, force the cell supplier to deliver cells matched to within +/-0.02V
- None, gross balance the pack during first charge, once built
- Preselect and group cells prior to build based on capacity, voltage, and resistance measurements
- Match cells based on OCV alone
- Pre-charge/discharge all incoming cells to a set voltage/SoC
- Average-balance cells in parallel group prior to building in series
- Average top-balance cells in parallel group prior to building in series
- Matching of beginning-of-life (BoL) cell self-discharge.
If the cells are very different in SoC when assembled the BMS will have to gross balance the cells on the first charge. This can take a long time as the maintenance balancing currents are generally very small compared to the Ah ratings of the cells (typically 1 to 3mA/Ah).
1. Supplier delivers matched cells
If the cell manufacturer can deliver cells with a proven quality history of OCV within +/-0.2V then you will be able to assemble and charge these cells without gross balancing.
However, you will need to consider a few things:
- Cell manufacture, formation, ageing, end-of-line testing, all have reporting and metrics
- Logistics around transport minimises time and temperature swings
- Local storage of the cells carefully controls storage time and temperature
- What is measured at cell manufacturing end-of-line should be remeasured at goods receipt.
Cells will be placed on sample test, built into modules & packs, or placed in storage and, at a bare minimum, an OCV confirmation check will be necessary prior to any test, build, or selection from storage.
The manufacturer will, in general, specify OCV, SoC and (delta OCV/time), insulation resistance, etc at defined environment in case of anomalies in transport and/or storage or buffered lot builds.
- This is the preferred volume build approach, provided data on batch anomalies can be made available.
- A number of smaller pack manufacturers will not get this level of support from the cell manufacturer
- The testing is reliant on only basic metrics and will not necessarily identify an outlier cell that may shorten the life of a battery pack
- There is less ability to properly grade the cells
- Warranty failures will be harder to keep to a minimum or warranty periods will likely be shortened
- SoC tolerance ranges may still provide a spread of capacity requiring higher cell balancing requirements.
2. Gross balance pack
This is what you are probably trying to avoid as it can take hours or even days for the pack balancing to remove large SoC differences. Even with good cell manufacturers, a +/- 3-4% capacity spread can be generally stated for the cells that, in theory, could represent a worst case capacity spread of 6-8% within the pack.
For lower quality cell manufacturers, 10% or more could be possible. For a 100Ah cell (or parallel collection of cells), it could take 100 hours to balance with a 100mA balancing current. This means you will need to put the pack on charge for more than four days at the end of the manufacturing process. This is a high cost in capital equipment, energy required to charge the pack and inventory costs of holding packs post production for four days.
In many ways this is the worst approach of all as it hides problems, in a belief that as long as the pack gets balanced the pack is good.
- It is cheap, easy, and only requires end-of line-balancing.
- Outliers will not be screened (if significant balancing is required, then it is likely that the cells will have fairly mixed quality)
- Charge balancing during operating life may become more of an issue, especially if the pack is left in storage for any lengthy duration, as capacity spread is likely to be more pronounced in packs with poorly matched cells at the start
- Warranty. Cell spreads, outliers, hidden defects, and balancing will all go to reduce pack longevity and reliability
- Safety. Weak cells can be held-up by stronger cells, in parallel with tell-tale signs being masked, until critical failure is reached. Monitoring the spread of cells at top-of-charge/end-of-discharge and balancing requirements during charge are straightforward means of identifying cells showing signs of weakening, unless the pack already houses a large potential for spread of capacity
- Some cells in a parallel configuration may be subjected to high stress in compensating for weaker cells.
3. Select best cells
This is a more widely recommended approach used within industry targeting reliability. The cells undergo a number of checks from visual inspection, capacity, and internal resistance measurement before finally selecting the best cells.
- Reliability, pack longevity, and maximised capacity.
- Fewer warranty issues
- Increased safety. Weaker and outlier cells can be rejected, tighter balancing and cell spread tolerances can be applied during operation to identify weakening cells.
- This is an expensive approach in both time, equipment and number of cells that are processed and rejected.
4. Preselect and group cells
Similar to option 3, but using just OCV to group cells such that the initial SoC of the cells in a pack will not require gross balancing.
- Reduces risks associated with gross balancing
- Enables some outlier detection
- Improves reliability and safety associated with anomaly detection due to tighter tolerances on cell capacity spread.
- Doesn’t identify all outliers or cells with some potential faults
- Each cell needs to be measured at goods in and binned according to a voltage range
- OCV only has a limited capability for matching capacity, as the voltage represents a ‘percentage of full’ capacity, eg. 3.60V measured on two separate cells may well show they are matched and equate to 50% SoC on each; but is that 50% of 4Ah or 50% of 4.2Ah? How long were the cells in storage and did the higher capacity cell have a higher rate of self-discharge?
- A cell’s internal resistance isn’t measured. Cells with a higher resistance will lead during charge or discharge and may still represent balancing issues.
5. Pre-charge/discharge cells
Prior to assembling the battery packs, you can charge/discharge all of the cells to a defined voltage. This ensures all of the cells are matched in SoC prior to assembly.
- Capacity matching will have a big impact on pack balancing and enable much easier fault detection during operation
- Some outliers can be rejected and cells are less stressed being in matched parallel configuration.
- Capacity won’t indicate other underlying issues, such as higher internal resistance or high self discharge rates
- Cost, time and energy usage associated with cell cycling.
6. Average-balance cells
Before assembling cells in series you connect them in parallel. This will discharge the higher SoC cells and charge the lower SoC cells. All cells will settle to the same voltage depending on the time left connected.
- Cells require a little less balancing at top-of-charge than gross balancing
- Initial cell capacities are set to 30% SoC, with overall individual capacity spreads of cells around +/- 3 to 4% at full charge. By the time the cells are charged there will still be a reasonable requirement for balancing
- No outliers are identified.
7. Average top-balance cells
Before assembling the cells in series you connect them in parallel with a charger that brings them all to an upper SoC/voltage. This alignment at a higher voltage will produce a better balanced battery pack with less balancing required during the first charge cycle.
- Cells require less balancing at top of charge.
- No outliers are identified.
8. Matching of BoL cell self-discharge
Charge/discharge all cells to a set voltage, ideally towards the lower SoC end and leave for a period of two weeks, measuring the change in OCV at the end of the period. Alternatively, using a potentiostatic method for directly measuring a cell’s internal self-discharge current may be achieved in hours rather than weeks.
- Prevents rapid cell divergence during cycling.
- Time and cost to test and match each cell.
Outliers and quality
How does an outlier represent a problem for quality and longevity? Can we quantify the impact outliers will have on the battery pack life, safety, and reliability? Only three of the test methods above identified any outliers, with ‘select best cells’ offering the most reliable results. Destructive analysis carried out by NASA in 2013 for the cells used in its soft-short detection tests, identified a varied degree and a fair number of defects from tears in the separator to dry anodes, foreign particles, and porosity.
Defects are there, finding them is a different matter. On a final wrap up, Figs 2 and Fig 3 From the 2013 NASA document offer a good example that defect detection is possible.