基于模糊神经网络的甲醇合成塔转化率软测量模型  

Soft Sensor Model of Conversion Rate in Methanol Synthesis Tower Based on Fuzzy Neural Network

作　　者：缪啸华 宋淑群 王建华[3] 张凌波[1] 顾幸生[1] 

机构地区：[1]华东理工大学 化工过程控制与优化技术教育部重点实验室,上海200237 [2]兖矿集团国宏化工有限责任公司,山东邹城273500 [3]上海应用技术学院,上海201418

出　　处：《石油化工自动化》2012年第2期32-35,40,共5页Automation in Petro-chemical Industry

基　　金：国家"863"高科技研究发展项目(2009AA04Z141);中央高校基础研究项目;教育部博士点基金(200802510010);上海市自然科学基金(10ZR1408300;11ZR1409800);上海市重点学科项目(B504)

摘　　要：以某甲醇合成塔为研究对象,综合考虑甲醇生产过程各因素对甲醇转化率的影响,采用模糊神经网络技术,建立了甲醇转化率的软测量模型。为了提高模糊神经网络建模的精度,提出了一种改进的差分进化算法优化模糊神经网络的参数,从而获得了具有较高精度的软测量模型。以某甲醇合成塔的实际运行数据为样本进行甲醇转化率的软测量建模,结果表明该模型具有较高的精度,能够实现甲醇转化率的实时监测。The soft sensor model for methanol conversion is constructed by adopting the fuzzy neural network technology with consideration of all the influence factors on the methanol conversion rate during the manufacture process with one methanol synthesis column as the research object. To increase the accuracy for constructing fuzzy neural network model, one improved differential evolution algorithm （DE） is proposed to optimize fuzzy neural network parameters and obtain the soft sensor model with higher accuracy. The soft sensor model for methanol conversion rate is built by using actual running data of one methanol synthesis column as the sample. The simulation results show that the soft-sensor model is of great accuracy, and can be used to realize real-time monitoring of methanol conversion rate.

关 键 词：差分进化算法 单纯形 模糊神经网络 软测量 

分 类 号：TQ223.121[化学工程—有机化工]