Uncovering the battery direct current internal resistance puzzle: A machine learning-driven pore network approach

Direct current internal resistance (DCIR), as a fundamental characteristic of lithium-ion batteries, serves as a critical indicator for the accurate estimation and prediction of battery health. The DCIR of a battery is affected by the electrode structure. Despite its significance, the relationship between the electrode structure and the DCIR during charging and discharging remains unclear. Based on a pore network model of a lithium manganate cell, this work focuses on the cathode and quantifies the effects of cathode thickness ($L$), porosity ($\epsilon$), connectivity ($G$), average particle size ($d$) and specific surface area ($S/V$) on DCIR. Combined with machine learning, this work identify that cathode thickness, porosity and average particle size the primary determinants of the DCIR, and the formulas for calculating charging and discharging DCIR are derived, ${\text{DCIR}}_{\text{Charge}}=0.168{Ld}^4/{\epsilon }^{2.5}$ and ${\text{DCIR}}_{\text{Discharge}}=0.072{Ld}^3/{\epsilon }^2$. This work proposes a research framework for predicting DCIR from the electrode structure, which is applicable to most porous electrode batteries, providing a theoretical basis for calculating the DCIR and is of great significance for electrode design.