最小二乘支持向量机用于常压塔汽油干点的软测量

Gasoline endpoint soft-sensing of crude column based on least squares support vector machine

最小二乘支持向量机(least squares support vector machine,LSSVM)为一种遵循结构风险最小化原则的核函数学习机器,其训练仅需求解一个线性方程组,且超参数较标准支持向量机更少。由于其实现简单且预测效果良好,近年来在化学、化工领域的应用日益广泛。本文研究了基于LSSVM的软测量建模过程中的数据预处理和优选超参数等问题。并将其应用于常压塔塔顶汽油干点的软测量建模。计算结果表明,其预测精度能够满足生产实际要求,是一种简单有效的非线性软测量建模工具。

Least squares support vector machine （LSSVM） is a kernel learning machine which obeys structural risk minimization （SRM） during training, which has less hyper parameters compared with standard support vector machine and its training is a linear equations set solving problem. Because of its simplicity and good prediction precision, LSSVM is applied widely in chemistry and chemical engineering field. The data pre-processing and hyper parameters selection problems for LSSVM are researched here, and soft sensing for gasoline end point of atmospheric column is used to test the performance of LSSVM soft senor. The result showed that the prediction performance of LSSVM can satisfy the manufacturing request and it＇ s an effective tool for nonlinear soft sensing.