Application of Nonlinear Parameters Predictive Model Based on LS-SVM in Pipeline of Mixed Transportation
Abstract In order to meet the refinery’s requirements on the properties of raw materials, improve the hydraulic stability of pipeline transporting process, and reduce the energy consumption, it’s necessary to mix up different crude oils transported in pipeline. The traditional method to obtain the physical properties of mixed crude oil, especially the nonlinear parameters such as viscosity and pour point, is experiment test or empirical formula, but the disadvantage is the heavy workload or limited use conditions. In this paper, a nonlinear parameters prediction model based on LS-SVM is put forward for transporting mixed crude oil in pipeline. The modeling method of the nonlinear parameters prediction based on LS-SVM : 1 normalize the different batch sample data; 2 the kernel function of SVM is Gaussian RBF; 3 regress coefficients by genetic algorithm; 4 train parameters and then obtain the nonlinear parameters prediction model. The nonlinear parameters prediction model based on LS-SVM has superiority in aspects of reliability, modeling efficiency, universality and adaptability. It can adapt to the dynamic requirements of refinery on crude oil properties, make full use of the accumulated mixed crude oils’ physical properties and their mixing ratio. The prediction accuracy of non-linear parameters model can be improved continuously. And then, the rationality and efficiency of online mixed transportation in pipeline can be improved effectively.