改进递推最小二乘支持向量机及在过程建模中的应用

Improved Recursive Least Squares Support Vector Machine and Its Applications in Process Modeling

针对流程工业存在多变量、非线性和数据动态性等问题,提出一种改进递推最小二乘支持向量机。该算法首先利用K均值算法(Kmeans)将训练样本分类,然后针对各聚类用人工鱼群算法(Artificial Fish Swarm Algorithm,AFSA)对最小二乘支持向量机参数进行优化,以避免人为选择最小二乘支持向量机参数的盲目性,最后在各聚类基础上建立相应在线递推最小二乘支持向量机模型。在加氢裂化反应过程蒸馏塔航煤干点的软测量建模研究中,表明所提出算法的有效性和优越性。

Considering the problem of multivariable, nonlinear and dynamic date in industry process, an improved recursive least squares support vector machine was proposed. First, the algorithm used Kmeans to divide the training sample into several clusters. Then, for each cluster, this paper separately used artificial fish algorithm to calculate the optimal parameters of least squares support vector machine, avoiding the blindness of selecting the parameters of least ,squares support vector machine. Finally, online recursive least squares support vector machine model in each cluster was set,up. In distillation tower of hydro cracking reaction, the soft measurement modeling of Jet fuel obtained highly precise and effective prediction.