文献类型：期刊文献

中文题名：基于多神经网络模型的石脑油干点软测量

英文题名：Soft Sensor of PDU Naphtha Dry Point Based on Multiple Neural Network

作者：张笑天[1];颜学峰[1];钱锋[1]

机构：[1]华东理工大学自动化研究所,上海200237

年份：2004

卷号：11

期号：S2

起止页码：52

中文期刊名：控制工程

外文期刊名：Control Engineering of China

收录：CSTPCD;;CSCD:【CSCD2011_2012】；

语种：中文

中文关键词：模糊C均值聚类;软测量;初顶石脑油干点;多神经网络

外文关键词：FCM; soft sensor; dry point of naphtha; multiple neural network

摘要：应用多神经网络建立初顶石脑油干点软测量模型,首先采用模糊C均值聚类法将样本集分成具有不同聚类中心的子集,每个子集运用BP神经网络训练得出子模型,然后根据聚类后产生的隶属度将各子模型的输出加权求和获得初顶石脑油干点软测量值。同时为了克服因炼制原油性质无法及时获得而造成对初顶石脑油干点预测偏差的影响,在于模型建立时将前一时刘初顶石脑油干点分析值作为网络模型的自变量。实际应用表明,所建模型具有良好的预测精度,泛化能力强,效果令人满意。
Based on the idea of combining models to improve prediction accuracy and robustness, the soft sensor model of the dry point of the first top naphtha is built by using FCM to divide a whole training dataset into several clusters with different centers. Each subset is trained by BP neural network. Then the membership degrees are used for combining the outputs of these models to obtain the final measuring result. Higher approaching precision and better generalization capability are gained. Meanwhile the former analysis values of the first top naphtha dry is point to be the independent variables of the network model. The result is satisfying when this method is applied to soft sensor of PDU naphtha dry point. Practice application shows that this method is worthy of further application.