Prognostication of lignocellulosic biomass pyrolysis behavior using ANFIS model tuned by PSO algorithm

• Published in: Fuel (Guildford) Vol. 253; pp. 189 - 198
• Main Authors: Aghbashlo, Mortaza, Tabatabaei, Meisam, Nadian, Mohammad Hossein, Davoodnia, Vandad, Soltanian, Salman
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01.10.2019; Elsevier BV
• Subjects: Activation energy; Adaptive network-based fuzzy inference system; Adaptive systems; Algorithms; Biofuels; Biomass; Cellulose; Constituents; Design optimization; Fuzzy systems; Heating rate; Hemicellulose; Kinetics; Lignin; Lignocellulose; Lignocellulosic biomass; Mathematical models; Nonlinear phenomena; Particle swarm optimization; Particle swarm optimization algorithm; Pyrolysis; Pyrolysis kinetics; Reaction kinetics; Statistical analysis; Statistical methods; Thermogravimetric analysis; Topology; Adaptive network-based fuzzy inference system; Pyrolysis kinetics; Thermogravimetric analysis; Lignocellulosic biomass; Particle swarm optimization algorithm
• ISSN: 0016-2361; 1873-7153
• DOI: 10.1016/j.fuel.2019.04.169

•Biomass pyrolysis behavior was estimated from its main constituents and heating rate.•The kinetic constants of lignocellulose pyrolysis were predicted by ANFIS–PSO model.•The developed framework could accurately estimate the biomass pyrolysis behavior.•A practical and handy software was designed for predicting biomass pyrolysis kinetics. In-depth knowledge on pyrolysis behavior of lignocellulosic biomass is pivotal for efficient design, optimization, and control of thermochemical biofuel production processes. Experimental thermogravimetric analysis (TGA) is usually employed to peruse the pyrolysis kinetics of biomass samples. In addition to that, the main constituents of biomass (i.e., cellulose, hemicellulose, lignin) as well as the process heating rate can excellently reflect its pyrolysis characteristics through modeling techniques. However, the application of statistical and phenomenological models for extremely complex and highly nonlinear phenomena like lignocellulose pyrolysis is challenging. To address this challenge, adaptive network-based fuzzy inference system (ANFIS) was consolidated with particle swarm optimization (PSO) algorithm to prognosticate the kinetic constants of lignocellulose pyrolysis. More specifically, the PSO algorithm was applied to tune membership function parameters of the ANFIS model. Three ANFIS−PSO topologies were designed and trained to estimate the kinetic constants of lignocellulose pyrolysis, i.e., energy of activation, pre-exponential coefficient, and order of reaction. The input variables of the developed models were biomass main constituents and the process heating rate. The developed models could predict the kinetic constants of lignocellulosic biomass pyrolysis with an R2 > 0.970, an MAPE < 3.270%, and an RMSE < 5.006. The pyrolysis behaviors of three different biomass feedstocks (unseen data to the developed models) were adequately prognosticated with an R2 > 0.91 using the developed models, further confirming their fidelity. Overall, the lignocellulose pyrolysis behavior could be reliably and accurately estimated using the trained ANFIS–PSO approaches as an alternative to the TGA measurements. In order to make practical use of the trained models, a handy freely-accessible software platform was designed using the selected ANFIS−PSO models for approximating biomass pyrolysis kinetics.