Predictive modelling of the higher heating value in biomass torrefaction for the energy treatment process using machine-learning techniques

Abstract

[EN] Torrefaction of biomass can be described as a mild form of pyrolysis at temperatures typically ranging between 200 and300°C in the absence of oxygen. Common biomass reactions during torrefaction include devolatilization, depolymer-ization, and carbonization of hemicellulose, lignin, and cellulose. Torrefaction of biomass improves properties likemoisture content as well as calorific value. The aim of this study was to obtain a predictive model able to perform an earlydetection of the higher heating value (HHV) in a biomass torrefaction process. This study presents a novel hybridalgorithm, based on support vector machines (SVMs) in combination with the particle swarm optimization (PSO) tech-nique, for predicting the HHV of biomass from operation input parameters determined experimentally during the tor-refaction process. Additionally, a multilayer perceptron network (MLP) and random forest (RF) were fitted to theexperimental data for comparison purposes. To this end, the most important physical–chemical parameters of this industrialprocess are monitored and analysed. The results of the present study are two-fold. In the first place, the significance of eachphysical–chemical variables on the HHV is presented through the model. Secondly, several models for forecasting thecalorific value of torrefied biomass are obtained. Indeed, when this hybrid PSO–SVM-based model with cubic kernelfunction was applied to the experimental dataset and regression with optimal hyperparameters was carried out, a coefficientof determination equal to 0.94 was obtained for the higher heating value estimation of torrefied biomass. Furthermore, theresults obtained with the MLP approach and RF-based model are worse than the best obtained with the PSO–SVM-basedmodel. The agreement between experimental data and the model confirmed the good performance of the latter. Finally, weexpose the conclusions of this study.