基于NARMAX模型对连续搅拌反应釜建模与优化设计  

NARMAX Modelling and Optimal Design for Continuous Stirred Tank Reactor(CSTR)

作　　者：郑思童 史洪岩 Zheng Sitong;Shi Hongyan(Department of Information Engineering,Shenyang University of Chemical and Technology,Shengyang 110142,China)

机构地区：[1]沈阳化工大学信息工程学院,辽宁沈阳110142

出　　处：《山东化工》2022年第5期193-198,共6页Shandong Chemical Industry

基　　金：国家重点研发计划(2018YFB1700200);辽宁省自然科学基金(2019-ZD-0069)。

摘　　要：针对连续搅拌反应釜(CSTR)化学反应过程的机理模型,提出了一种基于正向回归最小二乘(FROLS)的系统建模方法;在机理模型产生的输入输出数据的基础上,分别用普通最小二乘算法(OLS)与FROLS算法对系统进行辨识,采用模型预测输出(MPO)方法作为验证指标。在非线性自回归滑动平均(NARMAX)模型的基础上,利用周期操作技术对系统的操作参数进行响应面优化设计,得到使系统性能指标最优化的参数。仿真结果表明,FROLS算法克服了普通算法在辨识过程中未考虑模型项顺序而导致辨识结果不充分、模型精度低等缺点,得到更准确的辨识模型,在此基础上,通过优化操作系统的参数提高了反应物的转化率,证明了该设计具有重要的理论和现实意义。This paper mainly focuses on the mechanism model of the continuous stirred reactor(CSTR)chemical reaction process,and proposes a system modeling method based on forward regression least squares(FROLS).Based on the input and output data generated by the mechanism model,respectively the ordinary least square algorithm(OLS)and FROLS algorithm are used to identify the system,and the model predictive output(MPO)method is used as the verification index.On the basis of the nonlinear autoregressive moving average(NARMAX)model,the periodic operation technology is used to optimize the response surface design of the operating parameters of the system to obtain the parameters that optimize the system performance indicators.The simulation results show that the FROLS algorithm overcomes the shortcomings of insufficient identification results and low model accuracy due to the lack of consideration of the sequence of model items in the identification process of ordinary algorithms,and obtains a more accurate identification model.On this basis,the parameters of the operating system are optimized.The conversion rate of the reactants is improved,which proves that the design has important theoretical and practical significance.

关 键 词：连续搅拌反应釜 参数辨识 正向回归最小二乘算法 参数优化 

分 类 号：TQ018[化学工程] TP273[自动化与计算机技术—检测技术与自动化装置]