Development of Catalyst Grading System for Diesel Hydrodesulfurization Using Machine Learning Techniques

Abstract

Improved catalyst performance for diesel hydrodesulfurization and meeting various requirements from each refinery was investigated using the machine learning technique to effectively and efficiently estimate the optimized grading ratio of selected hydrodesulfurization catalysts. A catalyst grading optimization program was developed based on “design of experiment” and “multi-objective optimization,” in which virtual experiments and multi-objective optimization were carried out to estimate the grading ratio of the optimum catalyst to minimize the refined oil sulfur concentration under specific conditions. The calculated evaluation value of the optimized grading system obtained from our program showed quite good agreement with the pilot experimental results using the same system. In addition, the pilot evaluation showed superior HDS activity of the grading system obtained from our program compared to the single catalysts used for the grading system as well as other catalyst systems with other grading ratios.