Optimization of Redox Flow Battery Carbon Felt Electrode Mesostructures By Applying Lattice Boltzmann Method and Machine Learning

Carbon felt electrodes constitute state-of-the-art Redox Flow Batteries (RFBs) components because of their high electronic conductivity, high specific surface area, and high porosity. In order to further improve the electrochemical performance of these electrodes, many researchers have investigated different treatment methods, including plasma treatment, thermal, and chemical modifications [1]. However, the geometrical features of the fibrous electrode, which has a significant influence on mass transport (convection) and the anolyte/catholyte utilization rate, are often left from the discussion.<br/>This study presents an innovative computational approach that examines the fluid dynamic properties of the anolyte/catholyte flow separately from the electrochemistry behavior in various carbon felt electrode mesostructures. First, electrode mesostructures were generated stochastically based on realistic tuneable manufacturing parameters, including the fiber diameter, electrode density, amount of in-plane and through-plane fibers, and the compression ratio. Afterward, the Lattice Boltzmann Method (LBM) [2–4] was applied to simulate each electrode permeability and the reactive volume ratio, where the latter quantifies the dead volume due to the slow fluid velocity compared with the diffusion coefficient. Finally, based on the results obtained from the LBM, a Bayesian Optimization Algorithm was applied to analyze the datasets and propose promising parameter values corresponding to the optimized anolyte/catholyte utilization rate. This optimization workflow allows analyzing the impact of different manufacturing parameters on fluid dynamic properties and predicting a theoretical optimized carbon felt electrode mesostructure is also identified and tested by computational simulations for charging and discharging conditions.