【作者】 陈理；

【导师】 姚平经；

【作者基本信息】 大连理工大学 ， 化学工程， 2008， 博士

【摘要】 质量交换网络(MEN)综合作为过程集成的一个分支,具有提高资源利用效率和降低环境污染的双重效能。本文从热力学角度分析了多组分MEN综合的特点及研究难点,提出了解决多组分MEN综合的策略,对多组分质量交换网络综合的方法进行了研究。着重开展了以组合法和超结构法基于传质浓度差同步优化多组分质量交换网络的研究。论文的主要研究内容及结果如下:(1)分析并考察了传质浓度差对质量交换网络各种费用的影响,发现传质浓度差对MEN的操作费用和投资费用的影响趋势相反,且程度不同,作者认为只有采用同步优化的方法才可设计出年度总费用最小的MEN,所以本文重点研究了以传质浓度差为变量权衡MEN操作费用和设备投资费用的同步优化方法。(2)为解决相容多组分MEN综合问题,提出了浓度问隔表法与数学规划法相结合的同步优化方法—组合法。该方法以传质浓度差作为变量,根据浓度间隔表(CID)建立数学模型,以传质浓度差权衡操作费用和设备投资费用,可同步获得年度总费用最小的MEN和对应的最优传质浓度差。研究表明,夹点分析可发现过程瓶颈,剔除一些不合理的结构,降低数学模型的维数和求解难度,所以采用优化软件GAMS就可以求解。将该方法应用到文献实例中,年度总费用比文献均有降低,最大降幅达20.1%,表明该方法对相容多组分MEN综合问题是有效的和可行的。(3)为解决复杂的多流股和不相容多组分MEN综合问题,提出了超结构同步综合法,可以同时考虑MEN中流股的各种匹配情况及组分间相互作用等约束。与文献不同,将传质浓度差作为变量权衡MEN的操作费用和投资费用,以流股进出单元设备的流率来判断传质单元的取舍,建立了NLP数学模型,优化变量数目明显减少,并避免MINLP模型中0-1变量带来的离散性,采用自适应模拟退火遗传算法(ASA/GA)进行求解,可同步获得最优的传质浓度差和最优的MEN结构。将该方法应用于两个文献实例的计算,年度总费用比文献结果分别降低了9%和4%,表明所提出的超结构方法能以更大的概率获得最优解。(4)采用超结构同步综合法对国内某合成氨厂工艺凝液回收系统进行了优化设计。该系统因具有单元传质负荷大,操作温度、压力高,且具有强非理想性的特点,故属于比较复杂的多组分MEN综合问题。因此,与传统MEN综合问题不同,必须将流股进出单元的流率作为变量,同时需要考虑组分间的相互作用以及温度/压力对传质相平衡方程的影响。本文分别采用串级超结构和多级超结构法对该系统进行了优化计算,结果表明,将流率、温度/压力作为变量优化是必要的。温度/压力的优化可以改变传质相平衡及传质推动力,可获得年度总费用更小的MEN,验证了该方法对指导工程实际问题的优化设计是有一定理论指导作用的。(5)本文采用自适应模拟退火遗传算法对超结构同步综合法所形成的NLP模型进行求解,并对算法的评价函数进行了改进。采用自适应调整步长和温降的策略可提高算法的运算速度和搜索解的质量。对各实例的求解表明该算法在收敛速度及以更大概率获得全局最优解方面具有良好的性能,可以有效地求解由多组分MEN综合所形成的NLP问题。更多还原

【Abstract】 As one of the branches of process integration, synthesis of mass exchange network (MEN) has been provided with dual benefits of improving the efficiency in resource utilization and reducing the environmental pollution. In this thesis, the strategies to solve synthesis problems for multi-component MEN are proposed based on the analysis of its characteristics and difficulties of the study from the views of thermodynamics. And the synthesis methods for multi-component MEN are studied in detail and the work focuses on the studies of simultaneous optimizing methods for the multi-component MEN based on the mass transfer composition differences (e) between the rich and lean streams with combined and superstructure methods respectively. The main contents and results of this paper are as follows:(1) Having studied the influences of e on the various costs of a MEN, it is found that the e affects the operating cost and capital cost reversely but in different degrees, so the optimal MEN targeting minimum total annual cost (TAC) can be obtained only by means of simultaneous methods. As a result the simultaneous methods have been studied thoroughly by taking the e as the variables to trade-off the operating cost and capital cost of the MEN.(2) A combined method of composition interval diagram (CID) with mathematical programming is presented in this thesis in order to synthesize compatible multi-component MEN. Having taken the e as optimal variable, the CID is established according to the corresponding composition scales. Then the nonlinear programming (NLP) model is formed based on the CID. So the optimal values of e and the structure of MEN with minimum TAC can be obtained simultaneously by solving NLP model. The proposed method based on the pinch technology is helpful to find the bottleneck of the process, eliminate unreasonable solutions and cut down the dimensions of the mathematical model. So the NLP model can be solved by the software of GAMS. The proposed method is demonstrated with examples and the TAC of the calculating results have dropped with maximum 20.1% comparing with that of literatures, which validated the feasibility and effectivity of the combined method for the synthesis of compatible multi-component MEN.(3) A superstructure simultaneous method to consider various feasible streams matching and constraints is presented in order to solve the synthesizing problems of complicated multi-stream and incompatible multi-component MEN. Different from literatures, a NLP model is formed rather than MINLP model, which can reduce dimensions of mathematical model and avoid the discrete of binary variables 0-1 in the MINLP, so that the solution difficulties are decreased. The adaptive simulated annealing genetic algorithm (ASA/GA) is used to solve this model. At last the minimum TAC and the e values of a MEN can be optimized simultaneously by ASA/GA. The proposed method is illustrated with two examples and the TAC of the calculating results have decreased 9% and 4% respectively comparing with the results of literatures, which demonstrated that the proposed method can be applied to multi-component MEN effectively and with the higher probability to find the global optimum solution.(4) The process condensate recovery system for an ammonia plant in our country is optimized by the superstructure method proposed in this thesis. The system has the characters of relatively larger mass transfer loads between the rich and lean streams, higher operating temperature and pressure, and stronger non-ideality, so it is belong to the problems of complicated multi-component MEN synthesis. Different from most ordinary MEN, the inlet and outlet flowrates of streams must be taken as variables. Besides the interactions of the components in the system and effects of operating temperature/pressure in each unit on the phase equilibrium equations are also considered. The superstructures in series-stage and in multi-stage are applied respectively. The results show that it is necessary to take flowrates, operating temperatures/pressures as optimizing variables, and the operating temperature/ pressure can affect the phase equilibriums obviously. At last, the minimum TAC close to global optimum can be obtained. The results illustrate that the proposed method can be successfully used in the guidance of optimizing the process of industry.(5) The ASA/GA is developed to solve the NLP model formed and evaluating function is modified according to the synthesis problems. The strategies of adjusting step size and temperature decreasing adaptively improve the operation speed and the ability to find solution with high quality of the hybrid algorithm. The results of case studies show that the ASA/GA has good capability in convergence speed and finding the global optimum with a high probability and can successfully solve the NLP model established from synthesis of multi-contaminant MEN.更多还原