Supercapacitors are a popular choice for energy storage in hybrid electric vehicles and portable electronic devices due to their quick charging/discharging rates, long lifespans, and low maintenance. This study examined the effect of various physicochemical characteristics of carbon-based materials on the capacitive performance of electric double-layer capacitors using an artificial neural network (ANN) model. The data was retrieved from published experimental datasets (over 300 articles) to build and evaluate ANN models. A total of 66 architectures were developed to estimate the effect of input variables on specific capacitance (F/g). Among all, the 13–9–9-1 architecture with 0.5/0.8 momentum term/learning rate and 15,000 iterations showed minimum error with high accuracy. The developed ANN model predicts the influence of input variables on specific capacitance through sensitivity analysis and qualitative assessment. The results are implemented in a modular, open-source application, enabling estimation of capacitance using only the properties of carbon-based electrodes as inputs. This study also presents a new comprehensive dataset of carbon electrodes for supercapacitors extracted from published research articles and highlights how electrochemical technology can benefit from ANN. • ANN models were developed to predict specific capacitance of carbon-based supercapacitors. • The best architecture was 13–9–9-1 with 0.5 MT, 0.8 LR, and 15,000 iterations. • A GUI was created based on the synaptic weights of the ANN model. • Quantitatively estimated the input variable's impact on specific capacitance.
Sung et al. (Tue,) studied this question.