Cell parameters can be used as features for ML method for Automatic battery problem discovery.

(Relatively) fundamental parameters:

• Estimated Loss of lithium inventory (via methods from [2])
• Estimated Loss of active material: anode and cathode (via methods from [2])
• Coulombic efficiency
• Cell self-discharge rate
• Cell internal resistance
  • Charge internal resistance
  • Discharge internal resistance
• Diffusion resistance and polarisation (modelled as resistance and capacitance in Thevenin equivalent circuit model; see Cell diffusion voltage )

These parameters are only "relatively" fundamental because, e. g., diffusion resistance and capacitance describe the actual cell kinetics very crudely.

Some more conventional parameters that can be derived from the parameters above:

• (Remaining) Cell capacity
  • Charge capacity
  • Discharge capacity
• Open-circuit voltage curve is one of the cell(group) parameters. For example, it was used as a feature for a model predicting cell cycle life in [1].
  • The open-circuit voltage at 0%, 100% SoC (unless SoC is defined through voltage levels itself).
  • Anode and cathode potential functions of SoC, separately.

In theory, models that use only the "fundamental" parameters and not the "derived" ones should be regularised better (see ‣). However, there are two practical problems with using only the fundamental parameters:

• Estimating those fundamental parameters is very non-trivial itself and the resulting estimates can include significant errors. Estimation of more conventional parameters is comparatively straightforward. For example, it's harder to estimate the loss of lithium inventory than the remaining cell capacity.
• The changes in the derived parameters are not only caused by the changes in the fundamental parameters: these fundamental parameters represent some cell model, not the ultimate reality, and hence are necessarily limited in the ability to describe the changes in the empiric cell parameters such as capacity and OCV(SoC) function. (See ‣)

See also Simple vs. complicated cell models for a related discussion.

Estimate most cell parameters at once

References

[1] "Data-driven prediction of battery cycle life before capacity degradation" (2019)

[2] Degradation diagnostics for lithium ion cells