带误差校正的常压塔汽油干点软测量模型  

Soft-sensing modei of gasoline endpoint of crude unit with error-corrected

作　　者：陈国初[1] 田学民[1] 

机构地区：[1]石油大学信息与控制工程学院,山东东营257061

出　　处：《计算机与应用化学》2002年第6期779-782,共4页Computers and Applied Chemistry

基　　金：中国石油天然气总公司石油科技中青年创新基金(科字1998年第3号)

摘　　要：根据误差反馈原理,设计出带误差校正的神经网络模型,并将其用于常压塔汽油于点软测量,建立带误差校正的软测量模型。它将开环软测量模型的预测误差反传,作为校正输入量送给带误差校正的软测量模型进行误差校正。最后结合常压塔工业实例,对汽油干点进行预测,±1℃的预测误差表明该模型比不带误差校正的软测量模型具有更高的精度。By applying the principle of error feedback, an ANN model with error-corrected has been designed by this paper. Based on the new ANN model, a new soft-sensing modei of gasoline endpoint of crude unit with error-corrected has been put forward. The error of soft-sensing model without error-corrected has been back propagated to the input of the soft-sensing modei with error-corrected to calculate gasoline endpoint. After comparing with industrial data, the absolute error of gasoline endpoint are within± 1℃, that shows that the soft-sensing modei with error-corrected established by this paper has higher accuracy than soft-sensing modei without error-corrected.

关 键 词：软测量模型 误差校正 神经网络 常压塔 汽油 干点 误差反馈 

分 类 号：TE626.21[石油与天然气工程—油气加工工程] TE962