Artificial neural networks modelling: Gasification behaviour of palm fibre biochar

Abstract

Growing interest in sustainable energy has created focus in gasification technology. Biochar can be used in gasification process, where, when the biochar is exposed to gasifying agent, it creates syngas as the product. This study involves the usage of palm fibre wastes to be converted into useful product like syngas. In this study, palm fibre biochar was subjected to temperatures ranging from 600 to 1000 ℃ using carbon dioxide (CO2) as gasifying agent. Carbon conversion and char reactivity were determined from the thermogravimetric analysis for each temperature profile. Understanding and predicting the biochar yield and biochar reactivity produced from different feedstock is critical for biomass screening and process design. ANN is one of the modeling technique that has been successful in predicting the process response. Hence in this study, 11 training algorithms of artificial neural networks (ANN) were used to predict the outputs of the model which are weight loss, char reactivity and carbon conversion of the palm fibre biochar during the gasification process. Biochar pyrolysis temperature, gasification temperature, and time were used as the inputs for the models. The objective of this study was to identify the training algorithm that provides the best results for prediction. Mean Squared Error (MSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE) and regression (R) are the three performance criterions used to determine the ability of the ANN model to predict the experimental results. It can be concluded that among the 11 training algorithms, Levenberg Marquardt (LM) provided the lowest MSE, MAE and MAPE values for validation results and highest R value of 0.99, proving that LM is most suitable trianing alogorithm for the gasification data.

Keywords

• Artificial neural networks
• Gasification
• Modelling
• Palm fibre