Neural Network Models for Estimating the Impact of Physicochemical and Operational Parameters on the Specific Capacity of Activated-Carbon-Based Supercapacitors

Abstract

An artificial neural network (ANN) model is developed in this study to predict and analyze the specific capacitance of activated-carbon-based supercapacitors by utilizing a 12-dimensional data set related to physicochemical and operational parameters from the literature. A total of 61 ANN model architectures are constructed using a backpropagation algorithm to estimate the specific capacity. The 12−11−11−11−1 ANN architecture achieves good accuracy, with an average error of 5.8 and adjusted R 2 of 0.99. A standalone ANN software is developed for modeling specific capacity for infinite combinations of physicochemical and operational parameters. The findings illustrate the significance of structural characteristics and pyrolytic and oxidized N groups in determining the supercapacitor performance. Furthermore, the suggested approach can potentially guide future experiments to choose high-performance activated-carbon-based supercapacitor electrodes based on desired specific capacity.