An intelligent approach for the modeling and experimental optimization of molecular hydrodesulfurization over AlMoCoBi catalyst.

Abstract The accurate prediction of sulfur content is an essential role to ensure proper operation and product quality control in the hydrodesulfurization process (HDS) for environmental protection. Machine learning (ML) techniques have proven their predictive capability in solving challenging problems in the petroleum industry. In this work, an optimized support vector machine (SVM) model was constructed to predict the degradation of the sulfur content in the HDS process. To improve the predictive ability of the models, a genetic algorithm (GA) was used to select the SVM optimal parameters when building the (GA-SVM) model. Thus, a catalyst consisting of alumina modified with 15% molybdenum, 5% cobalt and 2% of bismuth (AlMoCoBi) was prepared by a thermal process and its catalytic performance for HDS was evaluated using a batch reactor. The experimental results demonstrated that our proposed model has promising potential with correlations of R2 = 0.99 being achieved during the testing stages of the prediction of the sulfur concentration. The experimental validation demonstrates that the predicted values obtained from the (GA-SVM) model correspond closely with the experimental results with average experimental errors of <5%. This proves that the AlMoCoBi catalyst holds great promise for obtaining the sulfur-clean production of fuels. Highlights • Designing of an accurate and reliable predictive model for sulfur content estimation in desulfurization process. • Integrating support vector machine with a genetic algorithm to select the optimal variables. • Validating the constructed model prediction accuracy using real experimental data. [ABSTRACT FROM AUTHOR]