A multi-step parameter identification of a physico-chemical lithium-ion battery model with electrochemical impedance data

Physico-chemical battery models are widely used in the design and simulation of lithium-ion batteries due to their physically descriptive modelling approach. The model accuracy highly depends on the accurate identification of model parameters, yet accurate and feasible model parameterisation with state-of-the-art experimental techniques is laborious, expensive and tied to inherent measurement and estimation errors. Multi-step voltage-based data-driven parameter identification techniques are widely adopted in the literature to tackle this challenge. However, impedance-based parameter identification schemes lack a similar level of detailed analysis. Therefore, we propose a novel multi-step electrochemical impedance spectroscopy (EIS) based data-driven parameter identification framework to identify kinetic parameters of a physico-chemical battery model utilising particle swarm optimisation featuring metaheuristic optimisation capabilities. The parameter optimisation framework is designed methodically to identify parameter clusters with distinct sensitivities to specific EIS impedance regions, significantly improving identification accuracy. The generic particle swarm optimisation is fused with nature-inspired Darwinian events cross-generating new particles using selected parents to improve the algorithm's predictions. The proposed data-driven parameter identification framework is benchmarked under multiple cell operating conditions and achieves a voltage prediction improvement of 28% for constant current discharge and 65% for transient duty cycles compared to an experimentally derived parameter set from the literature.

Automated summary

Physico-chemical battery models are widely used in the design and simulation of lithium-ion batteries. However, accurate identification of model parameters is challenging and expensive. This study proposes a novel electrochemical impedance spectroscopy (EIS) based data-driven parameter identification framework to improve the accuracy of a physico-chemical battery model using particle swarm optimization. The results show significant improvements in voltage prediction compared to previous experimental parameter sets.