Octane's number prediction analysis using multilayer perceptive neural network based on k-means clustering

The index of octane's number is important to evaluate the gasoline’s quality. During the desulfurization and olefin reduction process of catalytically cracked gasoline, the octane's number is generally reduced. Establishing prediction model is significance to predict octane's’s number and help enterprises to optimize technological process and then improve octane's number content of refined oil. In this paper, multi-layer perceptual neural network based on k-means clustering for octane's number prediction analysis is proposed. Specific as follows: The experimental data are the historical data accumulated from the operation of fluid catalytic cracking gasoline refining desulfurization unit in a petrochemical enterprise for four years. In this paper, the method combining clustering and correlation analysis is used to process the data to achieve the purpose of data dimension reduction. This paper classifies all variables according to the basic idea of k-means clustering, selects 16 independent and representative variables from 366 variables by combining the results of clustering analysis and comparing the Spearman correlation coefficient between variables, and finally uses the selected variables to establish the prediction model of octane's number based on multi-layer perceptual neural network. The experimental results show that the model has high prediction accuracy, which not only reveals the nonlinear mapping relationship between variables and octane's number, but also provides a new way of thinking and method for octane's number prediction.