Synthesis, characterization and machine learning based performance prediction of straw activated carbon

The research on the preparation of activated carbon co-activated by hydrothermal carbonization and pyrolysis was uncommon and less experiments used ultrasonic assisted impregnation. The synergistic effect of two catalysts had certain research value. There are still just a few studies on the performance prediction of activated carbon. In this paper, wheat straw, corn straw and sorghum straw were used as the raw materials. ZnCl2 and H3PO4 were used as the catalysts for the synergetic catalysis. Hydrothermal carbonization combined with pyrolysis was used to co-activate with the ultrasonic auxiliary impregnation method in order to prepare straw activated carbon. The straw activated carbon was characterized by different means and the principle of the method was analyzed. Then the performance prediction models of straw pyrolytic activated carbon using methylene blue number and iodine number as the main evaluation index based on Linear Regression, Support Vector Regression, Random Forest Regression were proposed and compared. The results indicated the three kinds of straw showed similar characteristics in the preparation of straw pyrolytic activated carbon whose specific surface area reached 1258.3927 m2/g, 1101.8430 m2/g and 1060.9723 m2/g respectively. The Random Forest Regression model was the most suitable. The n_estimators was set to 10. The evaluation indexes of the model were all good. It demonstrated the three kinds of straw were highly efficient precursor for the preparation of activated carbon used to remove dyes from wastewater. The preparation method in this paper combines the advantages of physical and chemical activation. The prediction model will accelerate the utilization of straw resources, realize the controllable and clean preparation of straw activated carbon and reduce environmental pollution.