基于核岭回归估计器的常压塔航煤干点推断控制

Inferential Control of Aviation Kerosene Dry Point in an Atmospheric Distillation Column Based on Kernel Ridge Regression Estimator

提出了一种基于核岭回归推断估计器的新型推断控制策略,来实现常压塔航煤干点的在线检测和控制.首先,对支持向量机与最小二乘支持向量机回归算法进行了分析,并提出一种直接优化核岭回归算法.其次,通过采集的二次变量数据和化验数据,用核岭回归方法建立了航煤干点的估计器模型.最后进行了仿真,结果表明,在相同样本集下,与支持向量机、RBF网络模型比较,所提建模方法调节参数少,预测精度高.

Based on kernel ridge regression （KRR） inferential estimator, a novel inferential control strategy is proposed to implement on-line examination and control of aviation kerosene dry point in an atmospheric distillation column. Firstly, regression methods of the support vector machine （SVM） and the least squares support vector machine （LS-SVM） are analyzed, and a kernel ridge regression method is presented by solving the optimization problem directly. Secondly, the kernel ridge regression method is used to set up a esitmator model for predicting the dry point of aviation kerosene through the collected samples of the primary and secondary variables. Finally, simulation is made using the same samples, and the results show that the proposed modeling method needs less adjusting parameters and obtains a higher estimation accuracy than the RBF neural networks and the SVM regression method.