Development of Predictive Models for Activated Carbon Synthesis from Different Biomass for CO2 Adsorption Using Artificial Neural Networks

Modeling the interactions between CO2 uptake and different parameters of activated carbon (AC) adsorbent synthesis in biomass types can be a way to create efficient adsorbents for CO2 capture. In the present work, several AC syntheses experiments, from 35 publications, have been used for the development of operability simulations, based on an artificial neural network (ANN). Four ANN structures were developed by multilayer perceptron (MLP) and radial-based function (RBF) algorithms to predict the specific surface area (BET) of the adsorbents and their CO2 adsorption capacity. The precursors, activators, pyrolysis temperatures, pour volumes, adsorption pressure, adsorption temperature, BET, and CO2 adsorption capacity have been considered as input and output variables. The Bayesian Regularization backpropagation algorithm has been chosen for the two hidden layers from the MLP and compared with the RBF algorithms. The number of neurons in the MLP and RBF algorithms was 35 and 45 for BET prediction, and 130 and 240 for CO2 adsorption capacity prediction, respectively, after an optimization process. MLP and RBF networks with high accuracy have the greatest MSE validation results (R-2 > 0.99). The ANN approach has been found to be a promising tool to accurately predict specific adsorbents of AC biomass types for CO2 capture.

Automated summary

The study focused on modeling the interactions between CO2 uptake and different parameters of activated carbon adsorbent synthesis in biomass types using artificial neural networks. The research developed four ANN structures to predict the specific surface area and CO2 adsorption capacity of the adsorbents, achieving high accuracy with MSE validation results exceeding 0.99. The ANN approach was found to be a promising tool for accurately predicting specific adsorbents of AC biomass types for CO2 capture.