柴油调合人工神经网络模型研究  

STUDY ON DIESEL BLENDING MODELS BASED ON ARTIFICIAL NEURAL NETWORK METHOD

作　　者：李为民[1] 华贲[2] 

机构地区：[1]江苏石油化工学院江苏省精细化工重点实验室,江苏省常州市213016 [2]华南理工大学强化传热与过程节能教育部重点实验室,广东省广州市510640

出　　处：《炼油设计》2001年第4期44-46,共3页Petroleum Refinery Engineering

基　　金：教育部强化传热与过程节能重点实验室研究基金资助!项目编号 :5 2 10 83 0

摘　　要：根据某炼油厂所提供的柴油调合凝点和冷滤点基础数据 ,用人工神经网络的反向传播方法进行建模预测。提出了人工神经网络适宜的拓扑结构 ,通过生产数据的检验 ,证明了用该方法建立的柴油调合模型能有效地给出预测信息。研究表明 ,凝点预测绝对值平均误差为 1.0 5℃ ,冷滤点预测绝对值平均误差为 0 .91℃ ,较常用的调合系数模型、指数关联模型和凝点换算因子模型 ,能更准确地预报调合柴油的凝点和冷滤点。Based on the data of solidifying point(SP) and cold filter plugging point(CFPP) of blending diesel oil from a refinery,diesel blending models were established with back-propagation method of an artificial neural network(ANN).An appropriate topological structure of ANN was put forward.It was proved that the predicted information about blending diesel oil can be successfully obtained.The average errors of predicted SP and CFPP are 1.05℃ and 0.91℃ respectively.It was shown that the ANN model can predict both SP and CFPP more accurately than the blending coefficient model,index correlation model and the SP conversion factor model.

关 键 词：人工智能 神经网络控制 柴油 调合 凝点 凝滤点 降凝剂 

分 类 号：TE626[石油与天然气工程—油气加工工程]