Title:
Utilizing Artificial Neural Networks and Regression Models to Enhance H/C-biocrude Thermochemical Liquefaction Contents
Author(s):
Okoro, O.V., Marquet, B., Lei, N., Shavandi, A.
Document(s):
Paper
Abstract:
This study compares the predictive and optimization capabilities of artificial neural network (ANN) models coupled with a multi-objective genetic algorithm (MOGA) and regression models (RMs) combined with desirability function (DF) optimization, aiming to maximize hydrogen and carbon (HC) contents in biocrude produced via thermochemical liquefaction of major food waste streams in Belgium. The ANN models demonstrated superior predictive accuracy, with higher Rē values (0.926 for carbon, 0.940 for hydrogen) and lower error metrics than the regression model, which showed Rē values of 0.650 and 0.776, respectively. These findings show the improved accuracy of the ANN models and their potential for optimizing HC content in biocrude. Optimization results revealed differing preferred feedstock blends between the models. The RM-DF approach identified an optimal blend of 0.430 g spent coffee grounds, 0.010 g spent tea leaves, and 0.560 g bread waste, yielding biocrude with maximum C and H contents of 72.29 wt.% and 9.22 wt.%, respectively. In contrast, the ANN-MOGA approach proposed 0.530 g spent coffee grounds, 0.110 g spent tea leaves, and 0.360 g bread waste, producing biocrude with higher C and H contents of 74.17 wt.% and 9.82 wt.%, respectively. Nonetheless, both approaches resulted in relatively low yields, indicating a potential trade-off between maximizing HC content and overall biocrude production, and suggesting the need for further research to optimize both.
Keywords:
agricultural residues, biorefinery, environment
Topic:
Biomass Conversion to Intermediate Bioenergy Carriers and Sustainable Biofuels
Subtopic:
Hydrothermal processing
Event:
33rd European Biomass Conference and Exhibition
Session:
5CV.1.14
Pages:
1036 - 1041
ISBN:
978-88-89407-25-7
Paper DOI:
10.5071/33rdEUBCE2025-5CV.1.14
Price:
FREE
