Prediction of activation energy of biomass wastes by using multilayer perceptron neural network with Weka

Abstract

The focus of this research was to find a link between the experimental activation energy, which was determined by a thermo gravimetric analyzer with varying temperature and constant heating rate using Coats and Redfern equation, and the theoretical activation energy, which was calculated using biomass input values and a machine learning technique. The activation energy of biomass waste is predicted using a Multilayer Perceptron Artificial Neural Network as a function of volatile matter, carbon, hydrogen, oxygen, ash, temperature, and heating rate in this study. The values of mean absolute error (MAE), root mean square error (RMSE), relative absolute error (RAE), root relative squared error (RRSE) were 0.1813, 0.2348, 0.37%, and 0.43% respectively for all biomass waste. Regression analysis was also carried out to determine the level of correlation between the experimental and predicted activation energy of biomass wastes. This method can be used to estimate the conditions required for the pyrolysis process by varying the temperature and heating rate and it can be used to improve the quality of the output pyrolysis products by utilizing biomass wastes. The authors concluded that these models can be a useful tool in the prediction of activation energy of biomass wastes to facilitate clean energy production.