Application of machine learning methods for lignocellulose biomass pyrolysis: Activation energy prediction from preliminary analysis and conversion degree
[Display omitted] •Machine learning was used to model the activation energy of biomass pyrolysis.•ANN, RF, and SVM models were applied using biomass properties and reaction degree.•With R2 of 0.911, the optimized RF model presented high accuracy for Eα prediction.•The contributions of variables were...
Saved in:
Published in | Fuel (Guildford) Vol. 343; p. 128005 |
---|---|
Main Authors | , , , , , |
Format | Journal Article |
Language | English |
Published |
Elsevier Ltd
01.07.2023
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | [Display omitted]
•Machine learning was used to model the activation energy of biomass pyrolysis.•ANN, RF, and SVM models were applied using biomass properties and reaction degree.•With R2 of 0.911, the optimized RF model presented high accuracy for Eα prediction.•The contributions of variables were determined as α > C > ash > N > H > S > method.•The constructed model could help reduce repetitive experiments and save resources.
The investigations of biomass pyrolysis kinetics are traditionally accomplished through experiments at the expense of extensive resources worldwide. The estimation of activation energy (Eα) of biomass pyrolysis is a crucial and essential aspect of process optimization and control. In this study, machine learning methods were applied for the first time to predict the Eα values of the pyrolysis process based on the preliminary analysis of the biomass feedstocks and conversion degree (α). A total of 1523 sets of Eα values calculated via three frequently-used model-free kinetic methods were collected from literature and modeled by using artificial neural network (ANN), random forest (RF), and support vector machine (SVM) algorithms. The training and optimization of the models showed that the RF algorithm exhibited satisfactory accuracy, making it a promising tool for a quick prediction of Eα values. Moreover, the feature importance analysis indicated that Eα mainly depended on the α, C content, and ash content of the biomass. This work suggested the effective application of machine learning in determining Eα accurately, which could help in understanding the pyrolysis mechanisms, reducing the experimental workload, and improving the optimization of the pyrolysis process. |
---|---|
ISSN: | 0016-2361 1873-7153 |
DOI: | 10.1016/j.fuel.2023.128005 |