基于支持向量机算法的芳烃抽提装置优化模型研究  

STUDY ON OPTIMIZATION MODELING FOR AROMATIC HYDROCARBON EXTRACTION UNIT USING SUPPORT VECTOR MACHINE METHOD

作　　者：杨善升[1] 陆文聪[2] 陆治荣[1] 刘太昂[2] 

机构地区：[1]北京石油化工设计院上海分院,上海200030 [2]上海大学理学院

出　　处：《石油炼制与化工》2011年第2期65-68,共4页Petroleum Processing and Petrochemicals

基　　金：国家自然科学基金资助项目(20503105)

摘　　要：将适合小样本数据建模的支持向量机算法(包括支持向量分类算法SVC和支持向量回归算法SVR)用于某石化公司芳烃抽提装置优化建模,建立了装置优化目标与有关工艺参数间的定性、定量模型。结果表明,抽余油中芳烃含量SVC模型的分类和预测正确率皆为100%;SVR模型对抽余油中芳烃含量的拟合与预报的均方根误差(RMSE)分别为0.072和0.060;抽余油中芳烃含量的SVR模型对128个训练集及32个测试集拟合和预测的R^2和q^2分别为0.820和0.867。应用所建优化模型,制定了装置生产优化方案,优化后抽余油中芳烃质量分数从0.82%降至0.74%,下降了9.8%。In this work, the support vector machine （SVM） method, including support vector classification （SVC） and support vector regression （SVR）, which especially appropriate for the modeling of small data set was applied to the process optimization of an aromatic hydrocarbon extraction unit. The qualitative and quantitative models correlated between objective function and some technical parameters are summarized. The optimal results are showed as follows using SVC model the correct rates of classification of the aromatic content of raffinate based on training and predicted data sets are both 100%;the root mean square errors （RMSE） of the aromatic content of raffinate calculated by SVR trained/predicted model are 0. 072 and 0. 060, respectively. The SVR model from a training set consisting of 128 samples have good determination coefficient （R2 =0. 820）. The SVR model is tested using an external test set consisting of 32 samples, which shows satisfactory external predictive ability （q^2 =0. 867）. In practice, the aromatic content of raffinate is reduced from 0.82% to 0. 74%;a 9.8% decrease in aromatic content is obtained as compared with that of prior to optimization.

关 键 词：芳烃抽提 支持向量机 支持向量分类 支持向量回归 优化建模 操作指导 

分 类 号：TQ203[化学工程—有机化工]