改进精馏塔模型用于空分塔模拟计算

第18卷第2期计算机与应用化学Vol. 182001年3月28日d Applied ChemistryMarck改进精馏塔模型用于空分塔模拟计算李翠清任晓光殷国蓉陈茂伟北京石油化工学院化学工程系,北京102600)摘要∶在传统的机理模型(三对角矩阵法中的泡点法)基础上,将体系压力(P),平衡液相组成(x,x2,x3)为输入节点;体系温度(T),平衡汽相氮组成(y)及氩、氧的相平衡常数(k2,k3)为输岀节点的神经网络液体空气汽液平衡计算模型,取代三对角矩阵法中泡点计箅,从而建立改进的精馏塔模型。改进模型用于空分塔的模拟计算,不仅计算速度快,计算结果与机理模型结果非常接近,且符合设计要求,对实现产品质量的在线监控具有一定的指导意义。关键词∶改进模型;精馏塔;模拟计箅中图法分类号:TQ015.9,TQ018,TP183文献标识码:A文章编号:1001-4160(201)02-147-15The Simulation of Air Distillation Tower by UsingImproved Model of Distillation TowerLI Cui-qIngRENXⅰ ao-guauang YIN Guo-rong CHEN Mao-WeiDepartment of Chemical Engineering, Beijing Institute of Petroleum Chemical Technology, Beijing 102600)Abstract: Based on the traditional mechanical model( the buble point method of three diagonal matrix ) an neural network modelof gas-liquid equilibrium computation for liquid air, which makes system presure( P)and equilibrium component of liquidphrase( xi,x2,x3) as input nodes makes system temperature( T), equilibrium component of gas nitrogen( yi), equilibriumonstant of argon and oxygen k?, k3 )as output nodes, insteads of the buble point computation of the three diagonal matrix thusa improved distillation model is built. The improved model is applied to simulation of the liquid air distillation tower it not onlyhas rapid compulation speed but computation result is approximate to the mechanical model and meets the demand of the designto the realization of line control of product qu:Key words: improving model distillate1引言将空气进行低温精馏是获取高纯度氧气和氮气的主要途径,为此,人们进行了许多研究1-4。为了调节生产、控制产品质量,对空分塔进行模拟计算是非常必要的。传统旳杋理模型法必然涉及液体空气汽液平衡的计算问题5,但由于计算复杂,花费时间长,计算结果滞后于实际生产,不利于实现产品质量的在线监控。寻找一个更为有效的液体空气汽液平衡模型并用于空分塔的模拟计算显得十分重要。近年来迅速发展起来的人工神经网络,以其特有的自适应能力强、反映速度快的特点,在石油化工过程模拟、化工数据处理、化工过程控制等方面发挥了优势δ-]本文将神经网络汽液平衡计算模型与精馏塔机理模型相结合,对液体空气精馏塔进行模拟计算。2神经网络汽液平衡计算模型1利用BP神经网络建立液体空气汽液平衡计算模型。经过计算确定输入层节点为压力(P)各组分液相组成(x1、x2、x3);输出层节点为平衡温度(T)汽相氮组成(y1)及氩、氧平衡常数(k2、k3)采000-08-04日期:2000-02-12148计算机与应用化学18卷用带一个隐含层的三层B网络,隐节点数取10。神经网络汽液平衡计算模型不仅具有较强的记忆能力,对未参加训练的样本也能给岀满意的预测结果,可用于精馏塔旳模拟计算。3改进的液体空气精馏塔模型传统的杋理模型法对精馏塔进行模拟计算,通常采用三对角矩阵法中的泡点法5。它以平衡级为基本模型,建立MEHS方程。机理模型法存在双层迭代nitr。gn过程,内层迭代对每块板进行泡点计算以求解各板的塔板温度和塔板上各组分的汽相组成;外层迭代用于精馏塔全塔物料平衡和热量平衡计算。精馏塔expanded air模拟计算时间主要消耗在汽液平衡的计算上,当理论板数较多时,这一问题显得尤为突出,使计算结ifi.dquid nitro果严重滞后于实际生产过程,不利于产品质量在线监控。本文以某化工厂空分塔下塔为背景(流程简图见图1),尝试用神经网络汽液平衡计算模型代替机理模型中用于泡点计算的模块,将神经网络模型与机理模型相结合,建立改进的液体空气精馏塔模seE liquidriching oxycen4计算结果图1空分塔流程简图将改进的液体空气精馏塔模型用于空分塔模拟计算,其计算结果与机理模型的比较分别见图2-图7及表1、表2。101mechanic modelprayer meot1015dthe plate number图2温度随塔板变化曲线板变化曲线Fig 2 Temperature curve changing with plateg 3 Fluid curve of gas and liquid chang期李翠清等:改进精馏塔模型用于空分塔模拟计算1492.50E016.11图6液氧组成随塔板变化曲线图7塔板压力随塔板变化曲线Fig. 6 The compoent curve of liquid oxygen changing with plateig. 7 The pressure curve of plate changing with plat表1改进模型模拟计算结果及其与机理模型相对误差Table 1 The computation result of improved model and the relative error with the mechanic modelfluid rate offluid rate ofumberpressure,[ atm] temperature, [ K] enthalpy of liquidlative eor,% relative error, phrase, [J/Nm] phrase J/Nm] gas,[ Nm/h] liquid, Nm/hIrelative error, %o relative error, %o Relative error, %o relative error,%6.110054.4139112.684542,03975,89850,00000.01900.02110.18780.10036.113554.4086l12.67865.87140.00160.01330.06130.01240.10196.l17096.8378112.673275.84450.0000.03620.00420.07860.1033345696.8410-54.3953l12.668375.81760.00000.01360.07950.1046.123996.8449-54.3871112.664175.79060.00000.00110.0060.01090.08040.105296.8497-54.377998,43660.00000,00450,01760,08100.10556.1309l12.658498,409575.73590.00000.00720.03570,02340.08130.15596.86254.3570112.657798.381975.70720.00000.08126.1378112.659298.353275.67630.00000.00930.04660.08ll0.10486.141396.8831112.66498.32235.64120.08070.104112.674198.28725,59820,0065917354.3502l12.692998.24475,54150.00000.00350.02680.08180.11006.151754.3883112.724398.18750.01720.086.155254.474112.776375.34230.00000.132054.6474112.85775.16390.00000.00700.21930.01800.09850.13394.89190.00000.03080.1211l13.136497.537874.45070.00000.00580,0180.01710,09310.095373.68580.00000.02100,0725113.925196.331772.49580.00000.00790.07490.00880.09300.16816.176198.582860.4915114.780995.141650.54970.0.172099,5176-63,8563115,967493.507149.03500.00160,00560,01320.01770.0930.1053100,4270-66.7870l17,020491.992447,692150计算机与应用化学18卷表2改进模型模拟计算结果及其与机理模型绝对误差Table 2 The computation result of improved model and the absolute error with the mechanic modelthe equilibriumthe equilibriumthe equilibriumumbercomponent ofcomponent ofcomponent of gascomponent of gascomponent of gasliquid nitrogenliquid argonliquid oxygenargon0,000010.000010.9330.000000.000000.000000.000000.000000.0000012340.000010.000010.000010.000000.000000.000000.999960.000020,999980.000010.000000.000000,000000,000000.000030.00002000020.000000000.000040.000040.000020.000000.000000.000000.000000.000000.000060.000.000000.000000.000000.000000.999800.000090.000110.999900.000050.000050.0000.000010.000010.000010.999680.999850.000080.000010.000010.000010.00000.000000.99940.000330.999760.000100.000140.000010.00000.0000l00000000.000570.9996I0.000140.000250.000010,0000l0,000020.0000l0.000010.000390.0000430.0000l0.0000.000020.000010.001770.000300.000030.000010.000010.000020.000010.996060,0008200.000430.00135130.000020.993250.001190,996990.000620.000010.988370.001730.00090.004250.000040.000010,000050,000010,0000l0.000020.991100,001320.007570.000.000010.000030.000030.000040.965120.003630.984580.001920,013500.000070.000010.000060.000020.000000.000030.940230.005180.054600.973330.023900.00020.000010.000250.000050.000000.00000.899710.007230.093060.954370.041680.837570.152690.923700.00500.000060.000010.000050.000250.000000.00020.749850.012420.237730.000350.000020.000370.000010.0000l0.670600.014030.315380.83070.008760.16020.00010000000.0001L0.000230.000240.607960.014700.790420.200020.00006改进模型模拟计算结果与机理模型模拟计算结果及设计值的比较见表3。表3计算结果比较Table 3 The comparison of computation resultof modelgas nitrogen iniquid oxygen ingas nitrogerexpandedched liquid aiImproved model0.790420.377340.837570.99999Mechanic model0.790360,377190.83763李翠清等∶改进精馏塔模型用于空分塔模拟计算5结论将神经网络液体空气汽液平衡计算模型与空气精馏塔杋理模型相结合,建立了改进旳精馏塔模型。该模型用于空分塔模拟计算,不仅使模拟计算时间缩短,而且计算结果与杋理模型计算结果非常接近,符合工艺设计要求,对实现空分塔产品质量的在线监控具有一定的指导意义。此外,建立精馏塔混合模型的方法对其他精馏塔的模拟计算具有一定的借鉴作用。1张占柱,傅举孚,等,空分装置精馏系统的计算机设计.深冷技术,1983,41-52 Billingham J F, Olszewski W J, Ricotta JP. Cryogenic air separation system for feed air flow disturbances. U S US 5,950, 455. 1999, 14 SepTan Zimig Toppel K O. Process and apparatus for separating gas mixtures. USUS 5,906, 674, 1999, 25 May4 Harmens A. Cryogenics, 1977,9 519郭天民等.多元汽一液平衡和精馏.北京:化学工业出版社,1983*43-4596 Ge S$, Hang CC, Zhang T. Nonlinear adaptive control using neural networks and its application to CSTR systems. J Process Control, 1999, 4)317赵之山,等.B网络培训算法的改进及其AN在催化精馏水解塔模拟中的应用.化学工程,1998,266)-468肖剑,等.用人工神经网络模型模拟催化精馏塔.化学工程,198,26(2)29-409胡海峰,等.减压渣油评价中的人工神经网络分析方法.计算机与应用化学,2000,1X3)215-218周铁峰,等.人工神经网络在柴油馏分油加氬拖硫中的数学模拟.石油学报,200