Simulation to estimate the correlation of porous structure properties of secondary batteries determined through machine learning

Abstract

The negative and positive electrodes of lithium-ion batteries exhibit different structural characteristics. In this study, considering the characteristics of each electrode layer of a lithium-ion battery, the correlation equation of the effective ion conductivity was formulated using a machine learning model. In general, the tortuosity depends on the porous structure, and therefore, the morphology of the packed particles. The graphite particles that constitute the negative electrode have a flat shape, in terms of the aspect ratio. Therefore, the tortuosity of a structure likely depends on the aspect ratio. In contrast, because the positive electrode represents a secondary aggregate, the tortuosity depends on the particle morphology. In this scenario, the parameters representing the particle shape are unclear. Considering these aspects, the tortuosity for the negative electrode in terms of the particle aspect ratio was predicted through nonlinear regression based on a support vector machine. The tortuosity for the positive electrode was predicted using the cross-sectional image of the electrode, with the particle shape considered as a feature. This clarified the correlation between the tortuosity and other structural properties or images. The obtained findings can be applied in various fields pertaining to porous materials and facilitate the optimization of structural designs.