Demand for higher energy density and faster charging capabilities in lithium-ion batteries has driven researchers to engineer thicker electrodes. However, as cathode thickness increases, severe kinetic limitations arise: sluggish ionic and electronic transport leads to non-uniform reaction distributions, increased polarization, and incomplete active-material utilization.
To overcome these structural bottlenecks, a research team from Huazhong University of Science and Technology, alongside collaborators from South China Normal University, has developed an innovative deep-learning-guided inverse design framework. Published in
ENGINEERING Energy, the study demonstrates that optimizing heterogeneous porosity distributions and electrode thickness can significantly boost the discharge capacity of LiNi
0.6Co
0.2Mn
0.2O
2 (NCM622) cathodes across a wide range of discharge rates (0.5 C to 4 C).
The researchers established a physics-based pseudo-two-dimensional (P2D) half-cell model to construct a database containing nearly 100,000 data samples linking electrode thickness, discharge rate, and spatial porosity distribution to voltage-capacity responses. Using this dataset, they trained a conditional generative adversarial network (cGAN) to serve as a high-fidelity surrogate model, which was subsequently integrated with particle swarm optimization (PSO) for rapid inverse structural design.
DOI:10.1007/s11708-026-1091-z