基于粒子群优化的VB-LSSVM算法研究辛烷值预测建模

Forecasting model of research octane number based on PSO-VB-LSSVM

针对现有红外线分析仪表无法实现阶段在线检测车用汽油调合中,MMT抗爆剂对辛烷值的影响问题,考虑到样本数据较少的因素,提出一种基于粒子群优化算法的矢量基最小二乘支持向量机方法,首先以粒子群优化的方法来选取最小二乘支持向量机的模型参数,然后用矢量基判据选择支持向量,使最小二乘支持向量机的解具有稀疏性。该方法不但克服了常用的交叉验证法的耗时与盲目性问题,发挥了最小二乘支持向量机的小样本学习和计算简单的特点,而且提高了最小二乘支持向量机模型的泛化能力,将其应用于汽油调合系统中研究法辛烷值的预测,仿真结果表明,该方法是可行且有效的。

Octane number can be improved by adding MMT when gasoline is blended. But the octane number can not be obtained by infrared analyzer. Considering the samples are few, a model is proposed in this paper, which integrates PSO （ particle swarm optimization）, VB （vector base） and LS-SVM （ least square support vector machine）. Firstly, the model parameters are selected using PSO algorithm; then, the support vectors are selected using vector base learning, which makes the support vectors sparse. The method not only overcomes the time-consuming and blind problems that commonly used cross-validation method has, but also makes use of the small sample learning ability of LS-SVM and features simple calculation. The proposed method was applied to the prediction of research octane number （RON） in a system of gasoline blending. The obtained results demonstrate that the model based on PSO-VB-LSSVM can achieve good accuracy.