New Numerical Approach to Determine the Optimum Mixing Ratio of Electrode Materials for Maximum Li-ion Battery Performance by the Hierarchical Homogenization and Feedforward Neural Networks

Abstract

The effective electrical conductivity (EEC) and capacity of the electrodes are altered by the composition of electrode materials, leading to a significant impact on the performance of the Li-ion battery (LIB) cells. This study aims to develop a more efficient numerical optimization method that integrates hierarchical homogenization and feedforward neural networks (FNN) to identify the optimal composition of electrode materials. Currently, this determination heavily relies on conducting multiple experiments. The cathode's EEC, as per its formulation, is assessed through hierarchical homogenization of its components. The optimization is expedited using FNN in the homogenization. The LIB cell's performance is evaluated based on the cathode formulation via the hierarchical homogenization and the Doyle/Fuller/Newman model. The multi-objective optimization problem is formulated and solved using the modified NSGA-II. The resulting Pareto-optimal solutions identify the power optimized and energy optimized cells. The power density of the former is increased by 51% while maintaining the same energy density and the latter cell's energy density is increased by 68% while maintaining the same power density, as compared to the initial cell.