ANN-RSM based multi-parametric optimisation and modelling of H2 and syngas from co-gasification of residues from oil palm plants

Abstract

Despite their abundance, lignocellulose biomass co-gasification studies especially that of oil palm biomass are scarcely reported. In this study, the frond and trunk of the oil palm tree are co-gasified under different conditions. Experimental results were validated, modelled, and optimised with the aid of RSM and ANN for the syngas and H2 results. The optimum yield of syngas and H2 were found to be 49.01 %, and 23.26 % respectively when operated at 900°C with particle size of 2.6 mm and blending ratio of 1:1. ANOVA yielded satisfactory P-values in the case of RSM with 95 % confidence level. The Bayesian regularisation-based ANN with a 3–10–2 topology (3 inputs, 10 hidden neurons, and 2 outputs) has shown to be a very successful and resilient model, as indicated by its significant coefficient of determination R2 of more than 0.95. The selected ANN structure demonstrates an efficient framework for capturing complicated interactions among the data. The model's relevance is shown by its ability to provide statistically relevant predictions. Furthermore, its endurance under varying situations demonstrates its reliable effectiveness, implying a capacity to generalize effectively to new, existing data. Based on the findings of the suggested BP-based ANN, the proposed model may be used in co-gasification processing industries to make critical evaluations of process operating conditions.