甲醇四塔精馏建模与变负荷能耗优化研究
|
摘要
甲醇被广泛地应用于有机合成、涂料、医疗和燃料等工业,其作为煤化工产业最主要的产品之一和非常重要的替代能源,在当今全球化工市场起着非常重要的作用。我国以煤气化为龙头、以甲醇为产品的煤化工市场格局正在形成。甲醇精馏工艺对整个甲醇生产流程的生产能力、产品质量、能源消耗、原料消耗、环境保护都有重大影响。精馏过程装置属于化工生产过程中的高能耗装置,甲醇精馏能耗占甲醇生产总能耗的20%左右,且能源利用率不高。随着近年以来节能减排的问题越来越受到人们的关注,能源节约问题成为全球关注的热点问题之一。甲醇精馏过程中,由于上游生产条件的波动与改变,进料组分会发生变化;而近年来,随着市场竞争的激烈,生产厂商往往需要根据市场调节生产负荷,系统进料负荷会改变。因此,对于实际投入运行的甲醇四塔精馏系统,当进料条件变化较大时,尤其是在较低负荷下,如何确定各情况下的最优操作参数使得在保证产品质量的前提下能耗较低,成为生产中不可避免的问题。针对这一问题,本文在充分研究甲醇四塔精馏建模的基础上,提出一个系统的解决方案,解决进料负荷及组分变化情况下寻找系统的最优操作参数,从而在保证产品质量前提下能耗最低的问题。
本文的研究针对实际投入运行的甲醇四塔精馏系统,提出了进料负荷及组分变化情况下寻找系统的能耗最优操作参数的方案。对四塔精馏的节能优化方案进行了讨论:通过基于动态规划最优性原理进行系统分级以降低建模与寻优的计算维度;选择出合理的与实际工艺流程相符合的控制方案;结合神经网络与Aspen模拟软件,提出了适合于变负荷寻优的建模方法,并通过分析结果探讨了其可行性;使用自适应遗传算法寻找可能存在的能耗最优操作条件,最终实现仿真平台并验证,说明其结果的准确性。
本文的主要内容和创新点主要包括:
1、提出一个系统的解决方案,对实际投入运行的甲醇四塔精馏系统,解决在进料负荷及组分变化情况下,寻找系统的稳态最优操作参数,从而保证产品质量前提下能耗最低的问题。该方案可以实时、定量地寻找变负荷能耗最优参数,用于甲醇四塔精馏系统的开环优化控制。
2、以某化工厂实际运行的甲醇精馏系统为基础,充分分析了甲醇四塔精馏工艺,提出了基于动态规划最优性原理的系统分级方法,将系统分解为相互联系的3级子系统,然后依次对每一级系统寻优,最终获得整个系统的最优操作参数。通过该方法,可对子系统分别进行建模与寻优,大大降低了系统的维度与计算的复杂度。同时,对划分的子系统做了的分析,根据实际情况确定相关的控制变量与被控变量,将实际的精馏变负荷最优化操作问题转化成为最优化的数学问题。
3、重点介绍了化工模拟软件Aspen的建模方法,对利用Aspen模拟结果进行了论证;分析了使用神经网络进行数据建模的方法与可行性。在此基础上,分析两种建模方法在变负荷优化中的优缺点,提出使用Aspen Plus采出合适的实验数据,再利用神经网络拟合建立相关快速反应模型;并将该方法运用于该化工厂的甲醇精馏系统,通过结果分析其可行性。
4、得到相关模型后,介绍了使用改进的自适应遗传算法与神经网络模型求解最终的变负荷能耗最优参数的方法,并进行验证。
5、对以上方法进行总结,提出进一步完善的相关思路。
Methanol is widely used in organic synthesis, coatings, medical, fuel industry, etc. As one of the most important products of coal chemical industry and alternative energy sources,it is very important in today's global chemical market. Methanol is one of the main productions in the coal chemical industry methanol in China. Methanol distillation process has a significant impact on whole the methanol production process such as the production capacity, product quality, energy consumption, material consumption and environmental protection. Distillation process is highly energy intensive. Methanol distillation energy consumption accounts for about 20% of total methanol production energy consumption. With the issue of energy conservation in recent years taking more and more attention, it has become a hot topic of global concern. In methanol distillation process, due to fluctuations in upstream production conditions, the components in the feed changes a lot. And in recent years, with the fierce market competition, manufacturers often need to adjust the production load according to market, so the feed load changes. Therefore, for the operational four- towers methanol distillation system, when the feed conditions changes, especially at lower loads, it becomes an inevitable problem that how to determine the optimal operating parameters to ensure product quality and to reduce energy consumption at the same time. This paper researched on a solution which aims to find these optimizing manipulation parameters.
This paper is limited in the research of the operational four- towers methanol distillation system with a solution of determining the optimal operating parameters to ensure product quality and to reduce energy consumption at the same time when the feed conditions changes. Energy optimization solution was discussed, and the system dimension was reduced by the dynamic programming principle. Then Aspen Plus and artificial neural network were used to simulate and analyze the system respectively. A modeling method applied to optimization was proposed which combines Aspen Plus and artificial neural network. Then adaptive genetic algorithm was used to find out the possible optimal operating parameters. The main contents of this paper are as follows:
1. It proposes a solution to calculate the optimizing manipulation parameters of four-tower rectification system under varying load condition, which ensures product quality and reduces energy consumption at the same time. This quantitative and real-time calculation method is used in open-loop control of four-tower rectification system.
2. It fully analyses the methanol distillation process, based on an actual a chemical plant. It proposes a method that reduces the modeling and optimization system dimensions based on dynamic programming principle, which greatly reduced the system's analysis and computational complexity. Meanwhile, the analysis of the subsystems was done and the relevant control variables and controlled variable were determined. Then the practical problem was transformed into a mathematical optimization problem.
3. It analyzed neural network modeling methods and feasibility. It studied on the chemical simulation software Aspen Plus modeling method, and used Aspen simulation results to analyze; on this basis of the analysis of the advantages and disadvantages in variable load optimization of the two modeling approaches, it proposes an approach which uses Aspen Plus to adopt a suitable experimental data and then train the BP neural network to get a rapid response model. To analyse its feasibility, the method was used in the chemical methanol distillation system.
4. It introduces the adaptive genetic algorithm. Then the algorithm was used to find the optimal value of the operation after getting the rapid response model, and the result was analyzed.
5. It gives the brief conclusion of the form work. And it brings up the ideas of further optimization.
本文的研究针对实际投入运行的甲醇四塔精馏系统,提出了进料负荷及组分变化情况下寻找系统的能耗最优操作参数的方案。对四塔精馏的节能优化方案进行了讨论:通过基于动态规划最优性原理进行系统分级以降低建模与寻优的计算维度;选择出合理的与实际工艺流程相符合的控制方案;结合神经网络与Aspen模拟软件,提出了适合于变负荷寻优的建模方法,并通过分析结果探讨了其可行性;使用自适应遗传算法寻找可能存在的能耗最优操作条件,最终实现仿真平台并验证,说明其结果的准确性。
本文的主要内容和创新点主要包括:
1、提出一个系统的解决方案,对实际投入运行的甲醇四塔精馏系统,解决在进料负荷及组分变化情况下,寻找系统的稳态最优操作参数,从而保证产品质量前提下能耗最低的问题。该方案可以实时、定量地寻找变负荷能耗最优参数,用于甲醇四塔精馏系统的开环优化控制。
2、以某化工厂实际运行的甲醇精馏系统为基础,充分分析了甲醇四塔精馏工艺,提出了基于动态规划最优性原理的系统分级方法,将系统分解为相互联系的3级子系统,然后依次对每一级系统寻优,最终获得整个系统的最优操作参数。通过该方法,可对子系统分别进行建模与寻优,大大降低了系统的维度与计算的复杂度。同时,对划分的子系统做了的分析,根据实际情况确定相关的控制变量与被控变量,将实际的精馏变负荷最优化操作问题转化成为最优化的数学问题。
3、重点介绍了化工模拟软件Aspen的建模方法,对利用Aspen模拟结果进行了论证;分析了使用神经网络进行数据建模的方法与可行性。在此基础上,分析两种建模方法在变负荷优化中的优缺点,提出使用Aspen Plus采出合适的实验数据,再利用神经网络拟合建立相关快速反应模型;并将该方法运用于该化工厂的甲醇精馏系统,通过结果分析其可行性。
4、得到相关模型后,介绍了使用改进的自适应遗传算法与神经网络模型求解最终的变负荷能耗最优参数的方法,并进行验证。
5、对以上方法进行总结,提出进一步完善的相关思路。
Methanol is widely used in organic synthesis, coatings, medical, fuel industry, etc. As one of the most important products of coal chemical industry and alternative energy sources,it is very important in today's global chemical market. Methanol is one of the main productions in the coal chemical industry methanol in China. Methanol distillation process has a significant impact on whole the methanol production process such as the production capacity, product quality, energy consumption, material consumption and environmental protection. Distillation process is highly energy intensive. Methanol distillation energy consumption accounts for about 20% of total methanol production energy consumption. With the issue of energy conservation in recent years taking more and more attention, it has become a hot topic of global concern. In methanol distillation process, due to fluctuations in upstream production conditions, the components in the feed changes a lot. And in recent years, with the fierce market competition, manufacturers often need to adjust the production load according to market, so the feed load changes. Therefore, for the operational four- towers methanol distillation system, when the feed conditions changes, especially at lower loads, it becomes an inevitable problem that how to determine the optimal operating parameters to ensure product quality and to reduce energy consumption at the same time. This paper researched on a solution which aims to find these optimizing manipulation parameters.
This paper is limited in the research of the operational four- towers methanol distillation system with a solution of determining the optimal operating parameters to ensure product quality and to reduce energy consumption at the same time when the feed conditions changes. Energy optimization solution was discussed, and the system dimension was reduced by the dynamic programming principle. Then Aspen Plus and artificial neural network were used to simulate and analyze the system respectively. A modeling method applied to optimization was proposed which combines Aspen Plus and artificial neural network. Then adaptive genetic algorithm was used to find out the possible optimal operating parameters. The main contents of this paper are as follows:
1. It proposes a solution to calculate the optimizing manipulation parameters of four-tower rectification system under varying load condition, which ensures product quality and reduces energy consumption at the same time. This quantitative and real-time calculation method is used in open-loop control of four-tower rectification system.
2. It fully analyses the methanol distillation process, based on an actual a chemical plant. It proposes a method that reduces the modeling and optimization system dimensions based on dynamic programming principle, which greatly reduced the system's analysis and computational complexity. Meanwhile, the analysis of the subsystems was done and the relevant control variables and controlled variable were determined. Then the practical problem was transformed into a mathematical optimization problem.
3. It analyzed neural network modeling methods and feasibility. It studied on the chemical simulation software Aspen Plus modeling method, and used Aspen simulation results to analyze; on this basis of the analysis of the advantages and disadvantages in variable load optimization of the two modeling approaches, it proposes an approach which uses Aspen Plus to adopt a suitable experimental data and then train the BP neural network to get a rapid response model. To analyse its feasibility, the method was used in the chemical methanol distillation system.
4. It introduces the adaptive genetic algorithm. Then the algorithm was used to find the optimal value of the operation after getting the rapid response model, and the result was analyzed.
5. It gives the brief conclusion of the form work. And it brings up the ideas of further optimization.
引文
[1]刘兴高.精馏过程的建模、优化与控制[M].北京:科学出版社,2007.
[2]李健.精馏塔机理—神经网络混合建模的研究[学位论文].南宁:广西大学, 2006.
[3]张洪流.化工原理[M].上海:华东理工大学出版社,2006.
[4]宋维端,肖任坚,房鼎业.甲醇工学[M].北京:化学工业出报社, 1991.
[5]许波.粗甲醇的双塔精馏与三塔精馏[J].化工催化剂及甲醇技术, 2002(1): 22-24.
[6]马俊睿.国际国内甲醇市场分析与预测[J].油化工技术经济, 2006(2): 41-47.
[7]常虹.甲醇精馏系统模拟与优化[学位论文].大连:大连理工大学, 2010.
[8]从山.甲醇精馏系统工艺模拟与优化[学位论文].天津:天津大学, 2005.
[9]许桂忠.合成甲醇精制过程的优化[学位论文].浙江:浙江大学, 2005.
[10]王锡臣.有机化学[M].北京:科学技术出版社, 1991.
[11] A. P .Douglas. A process integration approach to the design of the two-and three-column methanol distillation schemes[J].Applied Thermal Engineering, 2006, 26(4): 338-349.
[12] C. D. Holland. Multi-component Distillaiton. McGraw-Hill.Inc,1963.
[13] D. S. Billingley. On the Numerical Solution of Problem in Malticomponent Distillationath the Steady State[J]. AIChEJ, 1966, 12(6): 1134-1147.
[14]郑洪裕.多规格产品甲醇精馏塔的建模与仿真[学位论文].浙江:浙江大学, 2002.
[15]晓任坚.粗甲醇精馏浅谈[J].中氨肥, 1997(11): 1-11.
[16] Ohsakik, Shojjk, Okuda O. A. Large scale methanol plant based on ICI Process[J]. CEER, 1985, 17(5): 31-38.
[17]石东坡.甲醇双塔精馏流程的模拟与分析[J].广州化工, 2009, 37(5): 202-204.
[18]黄洁,郭志强,高如燕,张震欧.甲醇双塔精馏过程的模拟与分析[J].化肥设计, 2004, 42(1), 21-23.
[19]宋佰顺,李鸽,刘喜保.甲醇精馏采用高效塔板和工艺路线的改进[J].小氮肥设计技术, 2004, 25(1): 28-31.
[20]郭扬. 6万吨/年甲醇精馏系统节能环保扩产技术研究[学位论文].北京:北京化工大学,2009.
[21]周志超.一种新型甲醇双效精馏系统建模、动态行为与控制的若干问题研究[学位论文].浙江:浙江大学, 2008.
[22]陈心路.甲醇精馏能耗建模和优化研究[学位论文].上海:上海交通大学, 2011.
[23]韩晓芹.预精馏塔优化操作经验总结[J].中氮肥, 2007, 5: 23-25.
[24]冯国楠.最优控制理论与应用[M].北京:北京工业大学出版社. 1991.
[25]杨文鹏,贺兴时,杨选良.新编运筹学教程[M].西安:山西科学出版社, 2005.
[26] Eric V. DeNardo. Dynamic Programming: Models and Applications. US: Dover Publications, 2003.
[27]孙红先,赵听友,蔡冠梁.化工模拟软件的应用于开发[J].计算机与应用化学, 2007, 24(9): 1285-1289.
[28]赵琛琛.工业系统流程模拟利器—Aspen Plus[J].电站系统工程, 2003, 19(2): 57-59.
[29] Aspen Technology Inc. Aspen Plus 11.1 Unit Operation Models. US: Aspen Technology Inc, 2002.
[30]袁曾任.人工神经元网络及其应用[M].北京:清华大学出版社, 1999.
[31]靳蕃.神经计算智能基础原理方法[M].成都:西南交通大学出版社, 2000.
[32]张立明.人工神经网络的模型及其应用[M].上海:复旦大学出版社, 1993.
[33]钟珞,饶文碧,邹承明.人工神经网络及其融合应用技术[M].北京:科学出版社, 2007.
[34] Funahshi K. I. On the Approximate Realization of Continuous Mappings by Neutral Networks[J]. Neural Network, 1989, 2: 183-192.
[35] Aspen Technology Inc. Aspen Plus 2006 User Guide. US: Aspen Tecology Inc, 2006.
[36]赵弘,周瑞祥.基于Levenberg-Marquardt算法的神经网络监督控制[J].西安交通大学学报,2002(5): 523-527.
[37]张文修,梁怡,遗传算法的数学基础[M].西安:西安交通大学出版社, 2000.
[38]陈国良.遗传算法及其应用[M].北京:人民邮电出版社, 1996.
[39] Koza J. R. Genetic Programming II,Atuomatic Discovery of Resuable Programs[M]. USA: MIT Press, 1994.
[40]阎平凡,张长水.人工神经网络与模拟进化计算[M].北京:清华大学出版社, 2000
[41]李敏强,寇纪淞,林丹.遗传算法的基本理论与应用[M].北京:科学出版社, 2002.
[42]费烨,李楠楠,郑夕健,谢正义.基于结构和参数自适应的改进遗传算法[J].沈阳建筑大学学报, 2006, 22(2): 338-340.
[43]郭毓,林喜波,胡维礼.基于代沟信息的自适应遗传算法[J].东南大学学报:自然科学版,2004,34(增刊): 53-57.
[44]张义长,杨加明,鲁宇明.集合保优策略和移民政策的自适应遗传算法[J].计算机工程与应用, 2010, 46(31): 36-38.
[45] Eiben A. E. , Aarts E. H. , Van Hee K. M. Gloabal convergence of genetic algorithms: an infinite Markov chain Analysis[C]. Parallel Problem Solving from Nature. Berlin: Spronger-Verlag, 1991: 4-12.
[46]恽为民,席裕庚.遗传算法的全局收敛性和计算效率分析[J].控制理论与应用, 1996, 13(4): 455-460.
[47]何琳,王科俊,李国斌,金鸿章.最优保留遗传算法及其收敛性分析[J].控制与决策, 2000, 15(1): 63-66.
[48]张思才,张方晓.一种遗传算法适应度函数的改进方法[J].计算机应用于软件. 2006, 23(2): 108-110.
[49] Kvan T. Collaborative design: what is it?[J]. Automation in Constraction, 2000, Vol9, Na 4: 409-415.
[50]刘瑞国,邵诚.一种可自适应调节参数的改进遗传算法[J].信息与控制, 2003, 32(6): 556-560.
[51]李欣.自适应遗传算法的改进与研究[学位论文].南京:信息工程大学, 2008.
[2]李健.精馏塔机理—神经网络混合建模的研究[学位论文].南宁:广西大学, 2006.
[3]张洪流.化工原理[M].上海:华东理工大学出版社,2006.
[4]宋维端,肖任坚,房鼎业.甲醇工学[M].北京:化学工业出报社, 1991.
[5]许波.粗甲醇的双塔精馏与三塔精馏[J].化工催化剂及甲醇技术, 2002(1): 22-24.
[6]马俊睿.国际国内甲醇市场分析与预测[J].油化工技术经济, 2006(2): 41-47.
[7]常虹.甲醇精馏系统模拟与优化[学位论文].大连:大连理工大学, 2010.
[8]从山.甲醇精馏系统工艺模拟与优化[学位论文].天津:天津大学, 2005.
[9]许桂忠.合成甲醇精制过程的优化[学位论文].浙江:浙江大学, 2005.
[10]王锡臣.有机化学[M].北京:科学技术出版社, 1991.
[11] A. P .Douglas. A process integration approach to the design of the two-and three-column methanol distillation schemes[J].Applied Thermal Engineering, 2006, 26(4): 338-349.
[12] C. D. Holland. Multi-component Distillaiton. McGraw-Hill.Inc,1963.
[13] D. S. Billingley. On the Numerical Solution of Problem in Malticomponent Distillationath the Steady State[J]. AIChEJ, 1966, 12(6): 1134-1147.
[14]郑洪裕.多规格产品甲醇精馏塔的建模与仿真[学位论文].浙江:浙江大学, 2002.
[15]晓任坚.粗甲醇精馏浅谈[J].中氨肥, 1997(11): 1-11.
[16] Ohsakik, Shojjk, Okuda O. A. Large scale methanol plant based on ICI Process[J]. CEER, 1985, 17(5): 31-38.
[17]石东坡.甲醇双塔精馏流程的模拟与分析[J].广州化工, 2009, 37(5): 202-204.
[18]黄洁,郭志强,高如燕,张震欧.甲醇双塔精馏过程的模拟与分析[J].化肥设计, 2004, 42(1), 21-23.
[19]宋佰顺,李鸽,刘喜保.甲醇精馏采用高效塔板和工艺路线的改进[J].小氮肥设计技术, 2004, 25(1): 28-31.
[20]郭扬. 6万吨/年甲醇精馏系统节能环保扩产技术研究[学位论文].北京:北京化工大学,2009.
[21]周志超.一种新型甲醇双效精馏系统建模、动态行为与控制的若干问题研究[学位论文].浙江:浙江大学, 2008.
[22]陈心路.甲醇精馏能耗建模和优化研究[学位论文].上海:上海交通大学, 2011.
[23]韩晓芹.预精馏塔优化操作经验总结[J].中氮肥, 2007, 5: 23-25.
[24]冯国楠.最优控制理论与应用[M].北京:北京工业大学出版社. 1991.
[25]杨文鹏,贺兴时,杨选良.新编运筹学教程[M].西安:山西科学出版社, 2005.
[26] Eric V. DeNardo. Dynamic Programming: Models and Applications. US: Dover Publications, 2003.
[27]孙红先,赵听友,蔡冠梁.化工模拟软件的应用于开发[J].计算机与应用化学, 2007, 24(9): 1285-1289.
[28]赵琛琛.工业系统流程模拟利器—Aspen Plus[J].电站系统工程, 2003, 19(2): 57-59.
[29] Aspen Technology Inc. Aspen Plus 11.1 Unit Operation Models. US: Aspen Technology Inc, 2002.
[30]袁曾任.人工神经元网络及其应用[M].北京:清华大学出版社, 1999.
[31]靳蕃.神经计算智能基础原理方法[M].成都:西南交通大学出版社, 2000.
[32]张立明.人工神经网络的模型及其应用[M].上海:复旦大学出版社, 1993.
[33]钟珞,饶文碧,邹承明.人工神经网络及其融合应用技术[M].北京:科学出版社, 2007.
[34] Funahshi K. I. On the Approximate Realization of Continuous Mappings by Neutral Networks[J]. Neural Network, 1989, 2: 183-192.
[35] Aspen Technology Inc. Aspen Plus 2006 User Guide. US: Aspen Tecology Inc, 2006.
[36]赵弘,周瑞祥.基于Levenberg-Marquardt算法的神经网络监督控制[J].西安交通大学学报,2002(5): 523-527.
[37]张文修,梁怡,遗传算法的数学基础[M].西安:西安交通大学出版社, 2000.
[38]陈国良.遗传算法及其应用[M].北京:人民邮电出版社, 1996.
[39] Koza J. R. Genetic Programming II,Atuomatic Discovery of Resuable Programs[M]. USA: MIT Press, 1994.
[40]阎平凡,张长水.人工神经网络与模拟进化计算[M].北京:清华大学出版社, 2000
[41]李敏强,寇纪淞,林丹.遗传算法的基本理论与应用[M].北京:科学出版社, 2002.
[42]费烨,李楠楠,郑夕健,谢正义.基于结构和参数自适应的改进遗传算法[J].沈阳建筑大学学报, 2006, 22(2): 338-340.
[43]郭毓,林喜波,胡维礼.基于代沟信息的自适应遗传算法[J].东南大学学报:自然科学版,2004,34(增刊): 53-57.
[44]张义长,杨加明,鲁宇明.集合保优策略和移民政策的自适应遗传算法[J].计算机工程与应用, 2010, 46(31): 36-38.
[45] Eiben A. E. , Aarts E. H. , Van Hee K. M. Gloabal convergence of genetic algorithms: an infinite Markov chain Analysis[C]. Parallel Problem Solving from Nature. Berlin: Spronger-Verlag, 1991: 4-12.
[46]恽为民,席裕庚.遗传算法的全局收敛性和计算效率分析[J].控制理论与应用, 1996, 13(4): 455-460.
[47]何琳,王科俊,李国斌,金鸿章.最优保留遗传算法及其收敛性分析[J].控制与决策, 2000, 15(1): 63-66.
[48]张思才,张方晓.一种遗传算法适应度函数的改进方法[J].计算机应用于软件. 2006, 23(2): 108-110.
[49] Kvan T. Collaborative design: what is it?[J]. Automation in Constraction, 2000, Vol9, Na 4: 409-415.
[50]刘瑞国,邵诚.一种可自适应调节参数的改进遗传算法[J].信息与控制, 2003, 32(6): 556-560.
[51]李欣.自适应遗传算法的改进与研究[学位论文].南京:信息工程大学, 2008.