Co-gasification of coal, biomass with waste plastic in a fluidized bed: Machine learning model optimized and gasification scheme

Shuai Yan, Jingyue Mao, Xueqin Wei, Weiwei Li,
Co-gasification of coal, biomass with waste plastic in a fluidized bed: Machine learning model optimized and gasification scheme,
Journal of Environmental Chemical Engineering,
Volume 13, Issue 6,
2025,
119916,
ISSN 2213-3437,
https://doi.org/10.1016/j.jece.2025.119916.
(https://www.sciencedirect.com/science/article/pii/S2213343725046135)
Abstract: Co-gasification of coal, biomass, and waste plastic not only mitigates the environmental issues caused by waste plastic but also curtails the CO2 emissions from coal gasification. Machine learning (ML) models: support vector machine, random forest, multi-layer perceptron (MLP), and artificial neural network (ANN) were firstly conducted to support these processes and the parameters for MLP and ANN were further optimized. The MLP and ANN models demonstrated superior performance. Feed-forward Back Propagation neural network model with genetic algorithms showed the best performance of coefficient (R2) of 0.998, mean square error (MSE) of 0.0002, mean absolute error (MAE) of 0.024, and mean absolute percentage error (MAPE) of 0.001. The refined ANN model was subsequently employed to compared with experimental data from the literature and later predict the content of H2, CO, H2/CO ratio, lower heating value (LHV), syngas yield (Qs) and tar emission (Ctar) under different gasification agent scenarios, such as air gasification, oxygen-enriched air gasification, air with steam gasification, and oxygen with carbon dioxide gasification. At small equivalence ratio (ER), the H2, CO, H2/CO, LHV and Ctar decreased, and Qs increased. With the increased of steam flow rate, H2, H2/CO, LHV and Qs increased, but CO decreased. Highest hydrogen concentration of 25.6 % and H2/CO ratio of 1.64 were achieved through air with steam gasification. The highest LHV of 9.3 MJ/m3and gas yield Qs of 2.8 m3/kg was obtained during oxygen with air gasification. The simulation trend showed strong agreement with the experimental data from the literature, which could be used as a tool to guide the industry operation.
Keywords: Co-gasification; Machine learning; Fluidized bed; Gasification scheme; Parameter Optimization