Machine learning-guided optimization of nickel-based catalysts for enhanced biohydrogen production through catalytic pyrolysis of biomass

This study investigated the usage of machine learning (ML) to optimize nickel-based catalysts for biohydrogen production via catalytic pyrolysis (CP). While these catalysts offer cost, activity, and thermal stability advantages, they face deactivation through sintering and coke formation. Developing catalysts to combat these challenges is time-consuming and resource-intensive. ML models were used to predict hydrogen yield by leveraging a dataset of 159 points with 13 input features. The random forest (RF) regressor model demonstrated superior predictive capability, achieving an R 2 score of 0.78 and root mean square error (RMSE) of 0.47. Optimal catalyst formulation of 34.34 wt% nickel loading on carbon nanotube support was predicted to yield 75.74 vol% hydrogen. Analysis of feature importance revealed reaction temperature as the most significant parameter (0.14), followed by carbon content (0.11), calcination temperature (0.09), and nickel loading (0.085). This study provides valuable insights into optimizing nickel-based catalysts for enhanced hydrogen production from biomass. •ML optimizes Nickel catalyst for biohydrogen, tackling deactivation issues. •ML predicts 34.34 wt% Ni on CNT support, yielding 75.74 vol% hydrogen. •A 159-point dataset covers 14 key variables from literature on biomass pyrolysis. •Random forest led with R 2 = 0.78, RMSE = 0.47, excelling in hydrogen yield prediction.