ARTIFICIAL NEURAL NETWORKS BASED MODELING AND OPTIMIZATION OF THERMOCHEMICAL CONVERSION OF BIOMASS

Abstract

Biomass is considered one of the most promising and feasible renewable energy sources. Over time, the exploitation of biomass feedstock for various industries has been grown significantly. Most experimental techniques, however, require equipment that is extremely complex and costly. In this research, a novel approach aiming to predict the most desirable outputs of different biomass thermochemical conversion processes has been adopted. The main goal of this study is to utilize the machine learning techniques specifically deep learning in the field of biomass energy recovery through the development of artificial neural network models that can predict the higher heating value of biomass feedstock, lower heating value of gasification product, bio-oil and bio-char weight percentages for fast and slow pyrolysis respectively. The main input parameters used are obtained using both proximate and ultimate analysis as well as operating conditions for gasification and pyrolysis. This study also introduces deep learning aside with optimization as a magic tool for identifying different biomass feedstock that have a high potentiality for further processing technology or investigation.