Data efficient open circuit voltage hysteresis modelling – Transfer fitting the trajectory correction hysteresis (TCH) model from SOH-to-SOH and different li-ion cell chemistries

Abstract

The novel trajectory correction hysteresis model (TCH) is based on measuring the first-order reversal branches (FORBs). As the enormous measurement effort required for parameterisation hinders a real-world application, this paper presents the data-efficient transfer fit (TF) method. The TF methodology is validated through two application cases: ageing update and cell chemistry adaptation. Remarkably, using only 12 measurement points on the open-circuit voltage (OCV) envelopes instead of hundreds of measurement data points, the ageing update TF model attains a mean absolute error (mae) of 4.1 mV, closely approaching the accuracy of a newly parameterised target model (3.6 mV mae). Similarly, adapting an NCA cell model to an NMC target cell using selected OCV envelope points yields a 5.3 mV mae, which further reduces to 3.2 mV with an additional discharge FORB starting at 10% SOC. In addition to the selective model adjustment using continuous OCV measurement trajectories, the much more realistic adaptation by measurement points randomly distributed within the hysteresis window was successfully demonstrated. The presented TF methodology overcomes the hurdle of data efficiency while maintaining model accuracy and paves the way for the future application of the TCH model for voltage-based SOC correction.