【作者】 魏关锋；

【导师】 姚平经； 罗行；

【作者基本信息】 大连理工大学 ， 化学工程， 2003， 博士

【摘要】 多流股换热器网络是一种新型的换热器网络。与双流股换热器相比，多流股换热器以其高效率、结构紧凑及投资低在气体加工和石油化工等一些过程工业中有着广泛的应用。但是过去的20多年里人们把大部分精力用于传统的双流股换热器网络综合的研究，有关多流股换热器网络综合问题的数学模型及求解技术研究目前仍然很有限。本文在德国DFG项目和中国石油天然气股份有限公司辽阳石化分公司的支持下进行了多流股换热器网络综合的理论和实践研究，主要包括如下内容： 1) 建立了带有多流股换热器的换热网络综合数学模型，该模型改进了文献中等温混合的不合理假设，建立了非等温混合的网络综合超结构数学模型，使得模型更加严格、符合工程实际。等温混合的假定使部分约束条件线性化，但是也缩小了问题优化的解空间，建立在这样前提之下的数学模型更加导致局部极值的出现。本文提出的多流股换热器网络综合数学模型不依赖于夹点技术的网络温差、夹点位置和子网络的划分，对于过程流股的传热膜系数没有限制，做到了换热器网络的同步最优综合。 2) 多流股换热器网络综合问题本质上是一个混合整数非线性规划问题(MINLP)，这类问题的非凸非线性的特性使得目标函数存在多个局部最优解。传统的基于梯度的搜索方法在处理这类问题时由于计算规模庞大且极易陷于局部最优解而不再适用，而遗传算法却为解决这类问题提供了很有希望的一个方向。本文对遗传算法求解多流股换热网络综合问题进行了研究，主要有： (1) 遗传算法提供了一种求解复杂系统优化问题的通用框架，但是构造遗传算法时需要考虑的一个重要问题是遗传算子的设计。对不同的优化问题需要使用不同的编码方法和不同操作的遗传算子，它们与所求解的具体问题密切相关，因而对所求问题的理解程度是遗传算法应用成功与否的关键。本文将遗传／模拟退火(GA/SA)算法应用到多流股换热器网络综合数学模型的求解问题上去。提出了多流股换热器网络初始可行解的发生方法、超结构中匹配换热量的确定、个体相似性的判断方法，设计了遗传算法综合多流股换热器网络问题的交叉算子、变异算子、保留父代良好信息的OCX算子、多样性保持EC算子。 (2) 用遗传算法求解约束优化问题的难点在于：在遗传算法的计算过程中，遗传因子作用于可行解后，一个可行解在经过各种算子的作用之后常常会变成不可行解，即违反约束。用遗传算法求解此类问题需要考虑一些对策。与传统的双流股换热网络综合问题相比，多流股换热网络综合问题中有更多的约束，大量的流股分支流量是未知的优化变量，问题更加复杂，规模也更大，因此这给遗传算法的应用带来许多的困难。大连理工大学博士学位论文针对多流股换热器网络综合问题约束条件苛刻带来的问题，根据换热网络问题具体的特点，提出了适合于HENS问题进化过程中面向特定问题的约束处理策略及进化过程计算中较小热量匹配去除以及最小传热温差控制等与网络的可控性和经济性相关的一系列策略。本文成功地将G刀SA与HENS问题结合起来求解，计算结果证明本文提出的这一整套方法对解决多流股换热器网络综合问题是有效的。 3)论文中对多流股换热网络和普通的双流股换热网络综合方法进行实例计算分析，说明应用多流股换热器的优势。作为工程实例，本文将GA/SA综合多流股换热器网络程序应用于辽阳石化分公司的原油预热换热器网络的改造，所提出的换热网络满足了厂方的改造要求，降低了操作费用，可以带来较高的经济效益。作为对比，同时还用商品化的换热网络综合软件HExTRAN进行了综合。两者综合结果的对比表明，GA/SA进行多流股换热器网络综合的方法是有效的，可以应用于中等规模的生产实践过程中去。更多还原

【Abstract】 Multi-stream Heat Exchanger Network is a new type of heat exchanger networks. Compared with traditional two-stream heat exchangers, multi-stream heat exchangers possess many advantages such as high efficiency, compact structure and low cost. Although multi-stream heat exchangers are widely used in process industries such as gas processing and petrochemical industries, much work during the last two decades is focused on conventional two-stream heat exchanger networks. Few studies have been done on mathematical models and their solution technique for heat exchanger network (HEN) synthesis with multi-stream heat exchangers.Supported by the Deutsche Forschungsgemeinschaft (DFG, No. RO294/9) and Liaoyang Petrochemical Co. Ltd., we done theoretical and engineering application study on multi-stream heat exchanger network synthesis problem. The work in this paper involves following subjects:1) Based on concept of stage-wise heat exchanger network superstructure, a synthesis model of heat exchanger network with multi-stream heat exchangers supporting non-isothermal mixing of splits is established. The assumption of isothermal mixing in literature reduces non-linearity, eases solution difficulty, but also narrows the problem space, leading to local optima. The presented model does not rely on any heuristics such as pinch technology, and no constraint for process stream heat transfer film coefficients is needed. Through solving the model, we can optimize utility cost, exchangers’ area, number of units and stream split fractions simultaneously.2) The synthesis of HEN with multi-stream heat exchangers is modeled as an MINLP problem. Traditional gradient-based optimization technique fails to find the global optima due to the non-convexity and non-linearity and many local optima of this kind of problem. Alternatively, genetic algorithm (GA) and simulated annealing (SA) provide a promising approach for this problem. In the paper, to solve the synthesis problem with an algorithm (GA/SA), some combination of algorithm with the specific synthesis problem is considered:(1) Genetic algorithm provides only a general framework for solving complex optimization problem. The genetic operators are often problem-dependent and are of critical importance for successful use in practical problem. Specifically, to the synthesis problem of HEN with multi-stream heat exchangers, an approach for initial HEN generation, heat load determination of a match within superstructure are given. Some operators such as crossover operator, mutation operator, OCX (Orthogonal Crossover) and EC (Effective Crowding) operators are appropriately designed to adapt to the HENS problem.(2) Another difficulty for genetic algorithm is the constraints handling. During genetic evolution, the individual of the population often turn into some infeasible solution after manipulated by genetic operators. This often leads to failing to find a feasible solution during evolution, especially to the strict constraints optimization problem. Hence, some strategy should be contrived for constraints guarantee inIIIgenetic computation. The HENS problem of multi-stream heat exchangers is a more lage scale combinatorial optimization problem with more strict constraints and complexity. It brings forth more difficulty for constraints handling. According to characters of the heat exchanger network of multi-stream heat exchangers, we design an operator to transform the infeasible solution to feasible ones. Some strategies related with the network operability and cost-reducing are also devised in the paper. The feasibility of proposed GA/SA approach for HENS with multi-stream heat exchangers is demonstrated successfully through varying scale synthesis problems.3) The economical advantage of multi-stream heat exchangers network over traditional two-stream heat exchangers network is illustrated through calculations. At last, as engineering practice, we applied our approach to an industrial case, a heat exchangers network synthesis problem for crude oil pre-heating. Using our program更多还原