Artificial Neural Network-Based Prediction and Optimization of Polymer Membrane for Alkaline Water Electrolysis

Aiming at a polymeric porous membrane applied in the field of electrochemistry, especially alkaline water electrolysis, this paper combines polymer network microstructure prediction, characterization, high-throughput computation, and artificial neural networks to predict the performance of the membrane by material intrinsic characteristics and manufacturing parameters. Through the joint use of principal component analysis, fully connected neural networks, and convolutional neural networks, the microstructure tortuosity and maximum pore size can be predicted at the accuracy of R2 = 0.746 and R2 = 0.886, respectively. The influence of input parameters on performances is further analyzed, and several algorithms are utilized for parameter optimization of membrane manufacturing. The optimal parameters are implemented to a hand-cast membrane, which surpasses a commercialized membrane in certain aspects.