Synergetic effect of N/O functional groups and microstructures of activated carbon on supercapacitor performance by machine learning

Heteroatoms-rich activated carbon (AC) can effectively promote the pseudo-capacitance of AC-based electrodes used in supercapacitors. The well-known microstructural properties of AC immensely contribute to electric double layer (EDL) capacitance. The synergistic role of these multi physio-chemical features of ACs material can enhance the capacitance performance. In this work, Artificial Neural Network (ANN) achieves a superior model to interpret how microstructural features, N/O configurations, and operational parameters influence the EDLCs performance. Multilayer perceptron neural network (MLP-NN) model predicts the in-operando performance of N/O co-doped AC, using electrode materials. The training algorithms in the MLP-NN model show the high feasibility of the model in predicting AC performance with the least error. Among them, the trainbr and trainlm provide the best prediction accuracy. The sensitivity analysis exhibits the maximum effects of microstructural characteristics (<2 nm); and oxidized, pyrolytic of N groups on EDLCs performance by excluding features based on the MLP model. The utilized model presents insights into possible procedures to fabricate the N/O co-doped porous carbon electrodes with maximized specific capacitance. Finally, the genetic algorithm predicts the optimal AC in 6 M KOH at a three-electrode system with a specific capacitance of 550 Fg(-1) at 1 Ag-1.

Automated summary

This study investigates the influence of microstructural features, N/O configurations, and operational parameters on the capacitance performance of activated carbon electrodes through an artificial neural network. It provides theoretical support for the fabrication of N/O co-doped porous carbon electrodes with high specific capacitance.