Cell Balancing

Cell balancing is all about the dissipation or movement of energy between cells. The aim being to align them all with respect to state of charge. Aligning the state of charge of all of the cells in a pack will allow the pack to deliver the most energy and power. This becomes more crucial as the pack ages and differences between cells become more significant.

A set of cells connected in series will all see the same current, whether they are generating it when discharged or accepting it when charged. If all of the cells have the same initial SoC and the same capacity then they should charge and discharge equally.

However, if one cell has a slightly lower capacity it will be discharged further in SoC and it will go slightly overvoltage each time it is charged up. Over time this will result in a further reduction in it’s capacity.

When we then come to recharge this group of cells the centre cell will reach the maximum cell voltage first and the BMS will prevent further charging.

We need to remove charge from this cell so that we can charge the other cells in the string.

If we balance the cells then we can recharge them all to their maximum SoC. This means we need to ideally move charge from the higher SoC cell to the lower SoC cells.

After balancing the centre cell will have a lower SoC, but the 3 cells will all be at the same SoC and so we can now charge them up to the maximum voltage.

Fundamentally there are four methods of cell balancing:

1. Passive balancing
2. Active balancing
3. Runtime balancing
4. Lossless balancing

Passive Balancing

This simple form of balancing switches a resistor across the cells. In the example shown with the 3 cells the balancing resistor would be switched on for the centre cell. Discharging this cell and losing the energy to heat in the balance resistor (typically 30Ω to 40Ω).

This is ok when the balancing requirements are small. However, as the cells age the amount of balancing required to optimise the available energy is likely to increase. Resulting in increasing amounts of energy being lost to heat. This can also increase charge times when trying to reach maximum SoC for the pack.

Active Balancing

The idea here is to redistribute the energy across the cells. Give energy from the cells with the highest SoC to the cells with the lowest SoC. This is the ideal cell balancing approach.

However, this means the system has to be able to move energy between cells in the pack. Ideally between any two cells in the pack. Lots of wires and lots of switches means more weight, complexity and cost.

In practice, active balancing hardware splits into three families depending on how the energy is stored in transit between cells: switched-capacitor, switched-inductor, and transformer or converter-based (flyback / DC-DC) circuits. → Active Balancing Topologies

Switched-Capacitor Balancing

A capacitor is switched across a pair of adjacent cells, charging from the higher-voltage cell and discharging into its neighbour. Multi-capacitor designs place one capacitor per cell pair so every adjacent pair balances at once rather than sequentially. The control logic is minimal — essentially a switch matrix — which keeps cost and reliability advantages over inductor- or transformer-based designs, but energy can only hop between neighbouring cells, so correcting an imbalance between the first and last cell of a long string takes many sequential hops, each losing a little energy and taking time. This suits smaller packs rather than long series strings.

Switched-Inductor Balancing

Switched-inductor balancing follows the same adjacent-cell transfer pattern, but stores the transferred energy in an inductor’s magnetic field rather than a capacitor’s electric field: two switches alternate to pull current from the higher-voltage cell into the inductor, then push it into the lower-voltage neighbour. Inductors handle higher currents and larger voltage differentials than capacitors do, so this topology balances faster while keeping similar control simplicity. A peer-reviewed 2025 prototype built to balance 22 series-connected cells per submodule measured 84% energy transfer efficiency between adjacent cells at a 908 mA average balancing current, cutting an initial 1.18V imbalance to 0.47V in around two and a half hours (Pinto et al., Batteries 2025).

Transformer-Based (Flyback) and DC-DC Converter Balancing

Both approaches remove the adjacent-cell-only limitation. A multi-winding transformer uses one primary winding and a secondary winding per cell, so energy pulled from the string redistributes to every cell in parallel rather than hopping pair by pair — effective for high-power applications, though winding count (and cost) scales with cell count, and the core must be switched at the right frequency to avoid saturation. A dedicated DC-DC converter (commonly buck-boost or flyback) instead moves energy directly between any two points in the string. This any-cell-to-any-cell reach is why converter-based topologies dominate long series strings: a weak cell near the far end of a 200-cell string doesn’t need a hundred sequential hops to reach it. The trade-off is firmware complexity, since the BMS has to plan transfers across the whole string rather than simply reacting to the largest neighbouring voltage gap. Flyback circuits with an LC absorbing network have been reported achieving around 89% efficiency at balancing currents near 1.8A between non-adjacent cells.

Choosing between them comes down to string length, required balancing speed, and budget per pack: short strings can tolerate a slow, cheap switched-capacitor circuit; strings that must correct imbalance within a single charge cycle need a converter- or transformer-based design; switched-inductor sits between the two, faster than switched-capacitor but still adjacent-cell only.

Runtime Balancing

Each cell is connected to an individual low DC-DC power converter, then each converter is connected in series. This then allows the power delivered and received by each cell to be completely controlled based on their capability.

This system can provide a much higher level of robustness, but comes with a significant cost increase.

Lossless Balancing

This approach switches cells in and out of the circuit during charging. This means we have a lot of switches and that these switches have to be designed to carry the full current.

In the image we see that the middle cell has been bypassed.

Thoughts

This immediately raises some questions that we need to think about:

• How do I size the balancing resistor?
  • Balancing currents are small. In a 100kWh pack they are typically 100 to 300mA for each of the set of parallel cells (~280Ah). This equates to 1 to 3mA/Ah.
  • This does depend on the quality of the cells.
• How out of balance can the cells be before needing to be balanced?
  • If the cells are within the error of what you BMS can measure then don’t balance. If the error is measurable and say reducing the capacity or power capability by 10% then you should balance the cells.
  • There is a fine line between balancing to improve the pack performance and balancing continuously. Therefore it is important to set limits on when to start and stop balancing. Any algorithm needs testing on new and old packs to ensure that it is stable.
• Do we balance when charging or discharging?
  • Balance when charging and at a low charge rate. You want the cells to be as close together in terms of temperature and current flow with no extreme or fast changes that may take time to settle.
• What is the impact on cell lifetime?
  • Balancing will improve the overall pack lifetime as you will not be pushing some cells over voltage in order to charge the pack to 100%.
• How different are a batch of new cells?
  • Cell production quality is improving all the time. The quantity often measured is cell capacity and this is getting tighter. One reason for this improvement is that cells are sold based on capacity and hence it is important for the manufacturer to control this spread.

References

1. Cell Balancing Techniques and How to Use Them, Circuit Digest
2. Battery Cell Balancing: What to Balance and How, Yevgen Barsukov, Texas Instruments
3. Bortecene Yildirim, Mohammed Elgendy, Andrew Smith, Volker Pickert, Evaluation and Comparison of Battery Cell Balancing Methods, Newcastle University
4. Pinto, J.G.O.; Miranda, J.P.D.; Barros, L.A.M.; Afonso, J.A. Development of Modular BMS Topology with Active Cell Balancing, Batteries 2025, 11, 421, MDPI (peer-reviewed)
5. Active Balancing Hardware Topologies Compared, SunLith Energy

See also: Cell Balance and End of Life, 2015 Nissan Leaf, Weakest Cell

SoC Estimation Techniques

A look at the estimation of State of Charge (SoC) using voltage profiling and coulomb counting. These two methods give a good overview of the difficulty and errors associated in estimating this critical battery parameter.

Cell Matching

What level of cell matching do you 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.

1. none, force the cell supplier to deliver cells matched to within +/-0.02V
2. none, gross balance the pack during first charge once built
3. preselect and group cells prior to build
4. pre-charge/discharge all in-coming cells to a set voltage/SOC
5. average-balance cells in parallel group prior to building in series
6. average top-balance cells in parallel group prior to building in series