Leveraging machine learning hybrid framework and multi-objective optimization for efficient catalytic carbon dioxide hydrogenation to methanol

Catalytic CO2 hydrogenation offers a sustainable path for methanol production, mitigating greenhouse gases (GHG) emissions and enabling a circular carbon economy. However, optimizing this process for efficiency and yield remains challenge due to its inherent complexity and lack of effective modeling and optimization tools. This study addresses this challenge by proposing a novel two-step approach that integrates machine learning (ML) with a powerful optimization algorithm. Four ML models, namely SVM, GPR, GBR, and ANN, were trained on data generated by a physics-based process simulator. Among these, GPR demonstrated the highest performance with R2 values exceeding 0.99 and lower error metrics for both CO2 conversion and methanol yield. The best-performing GPR model was coupled with the Non-dominated Sorting Genetic Algorithm II (NSGA-II) for multi-objective optimization. This identified a set of 35 Pareto-optimal solutions, achieving a balance between CO2 conversion (30.76 %–31.23 %) and methanol yield (72.85 %–80.49 %). Finally, validation confirms the chosen solution's effectiveness, with deviations within 4.19 % for CO2 conversion and 2.21 % for methanol yield. This research not only presents an effective hybrid ML strategy as surrogate model for optimizing methanol production, but also paves the way for a more sustainable future by promoting efficient CO2 conversion.