The Study on the Optimization Model for Producing Olefin by Coupling Reaction of Ethanol Based on Clustering Analysis and Multiple Regression

Abstract

As a basic chemical raw material, olefins play a very important role in optimizing product structure and promoting economic development. In recent years, the olefin industry has faced unfavorable external environmental challenges such as low oil prices and the spread of the epidemic, and the uncertainty of industry development has greatly increased. Due to the huge demand, technological innovation in olefin production has become a top priority. Ethanol, which is the feedstock for C4 olefins, has an impact on the selectivity and yield of C4 olefins under different catalyst combinations and temperatures. Therefore, the study of ethanol-coupled preparation of C4 olefins under different conditios is conducive to improving the synthesis efficiency of C4 olefins and providing decision-making suggestions for the optimization and nconfiguration of chemicals. In this paper, 120 sets of experimental data are analyzed and processed, which provides a universal calculation step and solution for the parameter selection of ethanol-coupled to prepare C4 olefins under specific conditions. First, the by-product relationship was determined by cluster analysis in order to analyze the factors affecting ethanol utilization and C4 yield. Then, using multiple linear regression and partial least squares regression, the quantitative relationship between the variables was found, and the yield of C4 olefins at a temperature of 400 degrees, catalyst quality of 200 mg Co/SiO2-200 mg HAP, Co support of 1.1 Wt%, Co/SiO2 concentration of 1.06 wt%, and C4 olefin yield of 1.15 ml/min of ethanol concentration reached a maximum value of 77.21%. Finally, after error analysis and validity test, it is concluded that it is suitable for chemical reactions in most industrial production and can be used for efficient extraction of the target product in actual production. In the future, it will be optimized in a larger amount of experimental data, further optimized in combination with deep learning, to improve accuracy and accuracy.