Describing biomass pyrolysis kinetics using a generic hybrid intelligent model: a critical stage in sustainable waste-oriented biorefineries

Abstract The pyrolysis process is one of the most widely practised thermochemical pathways for converting biomass into biofuel. The most challenging aspect of the pyrolysis conversion is modelling the thermal decomposition kinetics of lignocellulosic biomass. Therefore, this study aimed to develop a generic hybrid intelligent model to describe biomass pyrolysis kinetics based on the ultimate analysis (carbon, hydrogen, oxygen, nitrogen, sulfur content) and process heating rate. First, an analytical model was fitted to the experimental data from thermogravimetric analysis reported in the published literature to determine the pyrolysis kinetic parameters of a wide range of biomass feedstocks. The derived kinetic parameters of biomass pyrolysis (i.e., reaction order, frequency factor, activation energy) were then modelled using three exclusive Adaptive Neuro-Fuzzy Inference System (ANFIS) models tuned by genetic algorithm (GA). The capability of the GA-ANFIS approach in modelling the kinetic parameters of biomass was also compared with that of the classical ANFIS model. The obtained results showed that the GA-ANFIS approach outperformed the classical ANFIS model in estimating the pyrolysis kinetic parameters of biomass. Generally, the highly nonlinear and extremely complex kinetic parameters of biomass thermal degradation were satisfactorily estimated using the GA-ANFIS models with a coefficient of determination exceeding 0.940 and a mean absolute error lower than 0.096. The pyrolysis reaction kinetics of five biomass materials, unexploited during the development of the GA-ANFIS models,‏ were estimated with a correlation coefficient higher than 0.811 and a mean absolute error lower than 0.7376 using the generic hybrid intelligent model. The promising agreement between the predicted and experimental kinetic data suggested that the generic hybrid intelligent model could be an alternative to the laborious experimental thermogravimetric measurements, thereby allowing pyrolysis process optimization, monitoring, and controlling to be more effectively conducted. Finally, an easy-to-use software package was developed based on the developed generic hybrid intelligent model to describe the devolatilization behaviour of biomass.‏