Hybrid SoC Estimation Methods

None of the individual SoC estimation methods on this site’s SoC Estimation Techniques overview is free of a weakness: voltage-based methods need the cell at rest and struggle on a flat OCV curve, coulomb counting drifts without a periodic reset, and the Kalman filter needs a battery model and well-tuned noise covariances to be worth the extra computation. Hybrid methods pair two approaches so that one covers the other’s weak point, rather than trying to make a single method work everywhere on its own.

The combinations named on the SoC Estimation Techniques overview are:

Coulomb counting and EMF (OCV) combination — the ampere-hour integral runs continuously, and every time the cell reaches a genuine rest point its running total is corrected against the OCV-SoC curve, bounding how far the drift can go before it is caught.
Coulomb counting and Kalman filter combination — coulomb counting supplies the filter’s process model (how SoC should change given the measured current), and the Kalman filter’s voltage-based correction step keeps that integral from drifting, without needing the cell to actually stop and rest the way a pure OCV reset does.
Per-unit system and EKF combination — cell or pack quantities (voltage, current, capacity) are first normalised into a per-unit system so the same filter tuning behaves consistently across cells or packs of different capacity and voltage, before an extended Kalman filter is applied on top.

The trade-off is complexity: a hybrid estimator has more parameters to tune and more code to validate than any one of its component methods, so it is usually only worth the extra engineering once a single method’s error rate is the limiting factor in a real design.