Innovative hydrogen production from waste bio-oil via steam methane reforming: An advanced ANN-AHP-k-means modelling approach using extreme machine learning weighted clustering

Steam Methane Reforming (SMR) is an established, cost-effective technique where methane or hydrocarbons react with steam, producing mostly hydrogen and carbon monoxide. This research explores hydrogen production through SMR applied to bio-oil, particularly from pyrolysis of various biomass sources. The methodology employs similarity analysis to select suitable bio-oils, which are then tested for hydrogen production using SMR. The results are analyzed through Pearson's R correlation plot to establish relationships, while the Analytic Hierarchy Process (AHP) prioritizes different outcomes. This prioritization is applied in k-means clustering to categorize bio-oils, enabling comparative performance assessments. Correlation analysis shows a strong positive correlation between CH4 conversion and energy efficiency (r = 0.97219), indicating that optimizing methane conversion improves the overall process efficiency. AHP analysis ranks CO yield (0.5) as the most significant performance factor, followed by hydrogen yield (0.35), CH4 conversion (0.25), and energy efficiency (0.15). k-Means clustering identified Jatropha Press Cake, Hemp Residue, and Eucalyptus Leaves as efficient bio-oils for hydrogen production. In the Artificial Neural Network (ANN) prediction model, Jatropha Press Cake is recognized as the most effective biomass for hydrogen production through SMR, achieving an RMSE of 0.48, an R2 value of 0.93, and a MAPE of 2.40%. Following closely is Hemp Residue, which has an RMSE of 0.52, an R2 of 0.91, and a MAPE of 2.80%. The study identifies Jatropha Press as the leading choice for hydrogen production from bio-oil, yielding 3.6 mol Hi/mole of biomass with a methane (CH4) conversion rate of 82% and an energy efficiency of 66%. In comparison, Rice Bran demonstrates the least effective performance, achieving only 2.8 mol Hi/mole of biomass.