换热网络优化中换热单元生成频次的影响分析及策略改进

换热单元的生成频次对换热网络优化有直接的影响,分析其机理可以指导求解算法的改进。本文采用节点非结构模型及强制进化随机游走算法,通过设置不同的换热单元生成概率与生成个数以改变换热单元的生成频次,观察并记录个体在优化过程中产生的年综合费用变化情况。发现不同生成频次下整型变量和连续变量的相对优化频率发生了变化,从而改变了整体优化过程,并且与个体当前优化状态不适应的生成频次会造成相对优化频率不平衡、优化结果较差。基于此,本文提出了一种具有生成参数动态调节策略的换热单元生成方式,在优化过程中根据个体状态实时调整换热单元的生成概率与生成个数,平衡整型变量与连续变量的相对优化频率。最后,采用15SP、10SP和20SP算例进行验证,换热网络的年综合费用较文献结果分别下降了1.06%、0.16%、0.68%,验证了该策略使得换热单元的生成频次更为合理,有效地提高了算法的优化效率。     

精细搜索策略应用于质量交换网络综合

质量交换网络是化工过程系统的重要组成部分,其优化设计对降低污染排放具有重要意义。采用启发式算法优化质量交换网络时,存在难以兼顾全局搜索和局部搜索的问题。通过分析不同精度优化参数下的优化结果,揭示了该问题的成因,并提出一种精细搜索策略用于基础算法所得结构的深度优化。该策略包含两种方法,方法1采用具有个体回代与分化的高精度强制进化随机游走算法,可保留个体结构变异能力;方法2采用确定性方法依次对多维目标函数中的每个变量进行一维搜索,具有精度高收敛快的优点。将该策略应用于焦炉气脱硫和空气除氨算例,得到的结果分别为407308 USD·a^-1和127807 USD·a^-1,经济性优于现有文献中的结果,验证了本策略的有效性。     

精细搜索策略应用于质量交换网络综合

质量交换网络是化工过程系统的重要组成部分,其优化设计对降低污染排放具有重要意义。采用启发式算法优化质量交换网络时,存在难以兼顾全局搜索和局部搜索的问题。通过分析不同精度优化参数下的优化结果,揭示了该问题的成因,并提出一种精细搜索策略用于基础算法所得结构的深度优化。该策略包含两种方法,方法1采用具有个体回代与分化的高精度强制进化随机游走算法,可保留个体结构变异能力;方法2采用确定性方法依次对多维目标函数中的每个变量进行一维搜索,具有精度高收敛快的优点。将该策略应用于焦炉气脱硫和空气除氨算例,得到的结果分别为407308 USD·a^-1和127807 USD·a^-1,经济性优于现有文献中的结果,验证了本策略的有效性。     

具有步长调整策略的强制进化随机游走算法优化换热网络

强制进化随机游走算法(random walking algorithm with compulsive evolution,RWCE)是一种优化换热网络的新方法,具有程序简单、算法适应性和全局搜索能力较强等优点。本文研究了最大步长对RWCE算法优化性能的影响,提出了抛物线函数的最大步长递减调整策略来平衡RWCE算法的全局搜索与局部搜索能力。将引入策略的RWCE算法与基础算法比较,发现加入最大步长递减调整策略的RWCE算法与基础RWCE算法相比,在进化后期能够跳出局部极小值,具有更强的局部搜索能力。采用10SP2、9SP和15SP换热网络实例检验加入此策略RWCE算法的有效性,其中10SP2和9SP算例的优化结果均好于文献最好结果,相比算例原始文献下降了20.98%和1.11%。对15SP算例优化找到了新的换热网络匹配结构,并好于多数无分流换热网络优化结果,且低于文献结果 4.60%,证明了此方法在换热网络优化中具有较强的优化能力。     

换热网络非结构模型中交叉结构的分析

采用强制进化随机游走算法优化换热网络非结构模型时易出现换热单元匹配相互交叉的现象,即同一对冷热流体上一个换热单元的冷、热流股的出口温度同时等于另一换热单元冷、热流股入口温度。这种交叉结构在热力学上是不合理匹配,其在优化中相互制约,从而影响算法效率和求解质量。针对这一问题,提出了交叉结构的禁忌策略,禁止在优化过程中出现交叉结构。通过两个不同规模的算例验证提出的策略,得到的结果分别为1 415 994和5 233 287$×year~(-1),相同时间内比改进前分别减少了40 645和182 899$×year~(-1)。结果表明,非结构模型的交叉结构禁忌策略能够提高算法效率和优化质量。     

棋盘模型在换热网络优化中的应用

针对复杂整数非线性的换热器网络综合和优化问题,文中应用换热网络流程模拟的棋盘模型及其相应的算法,建立了换热网络同步最优综合的数学模型,通过改变网络的结构次序的数学规划优化计算,以10SP1问题为例,获得了优于以往文献的最优换热网络结构。结果表明,文中应用棋盘算法建立的换热网络求解模型及其算法,具有简单、直观、灵活、易实现的特点,能够得到综合性能更优的网络结构。     

粒子群算法最优同步综合换热网络

采用粒子群算法,对无分流换热网络综合问题提出改进的优化策略.运用超结构建立换热网络模型,以最小年度总费用为优化目标,按顺序求解思路优化换热器换热量.该算法比遗传算法简单,所需调节参数少,且不易陷入局部最优解,能迅速得到换热网络的结构与参数,具有获得全局最优解的能力.仿真研究验证了方法的有效性.     

基于Kriging代理模型的气辅注射成型工艺优化

针对气辅成型工艺设定及控制的复杂性,使用Kriging代理模型近似拟合气辅成型工艺参数与气体穿透长度之间的非线性函数关系,建立起Kriging代理模型与APSO算法相结合的优化策略。应用所建立的优化方案,以气体的穿透深度为优化的目标,模拟了某汽车后视镜的气辅成型工艺优化过程。算例表明,基于Kriging代理模型与APSO算法耦合的优化策略可以在小样本情况下获取较高的求解精度和较快的收敛速度。     

基于自适应竞争群优化算法的无分流换热网络综合

换热网络模型具有非凸和非线性的特性,对于大规模超结构优化问题,采用经典的智能算法优化效率低,容易陷入局部最优值。以年综合费用为目标函数,基于自适应竞争群优化算法对无分流分级超结构换热网络模型进行优化。该方法采用对粒子平均位置的递减学习,通过自适应调节速度权重提高换热网络结构的全局优化能力和局部优化能力。通过两个典型算例分析表明,该方法相比量子粒子群算法大幅减少了模型调用次数,缩短了运行时间,并且找到了更好的优化结果。     

多孔错流喷射混合器内液体射流轨迹线

利用平面激光诱导荧光测试技术对多孔错流喷射混合器内液体混合过程进行了研究,考察了操作条件(射流速度比r、混合流股Reynolds数ReM)和混合器的结构参数(射流小孔直径d、孔径管径比d/D、射流小孔个数n)对射流轨迹线的影响。结果表明,混合流股Reynolds数对射流轨迹线影响较小,射流速度比和混合器的结构参数是影响射流轨迹线的主要因素。建立了射流轨迹线的数学模型,并利用实验结果回归了模型参数。模型预测的液体混合过程射流轨迹线与实验结果基本吻合。     

RWCE算法中采用单元重构策略激励换热网络结构优化

应用强制进化随机游走算法优化换热网络的过程中,通过监测最优结构中换热单元优化进程,探究总体结构进化特点,发现存在换热单元先入为主的现象,主体结构稳固难以被突破,导致换热网络优化进入“停滞”状态。基于此,提出了换热单元重构策略,该策略旨在增强结构变异能力,经过一定阶段概率性消除结构中既有换热单元,腾出结构进化空间,再依据约束条件重新构建换热单元,既丰富了结构多样性又扩大了搜索域。此外建立优化状态监测指标,维持结构较高的变异活性,增强算法全局寻优能力。最后将该策略加入算法流程优化多个换热网络实例,总结分析其作用机理,所得结果均优于现有文献结果,证实了该策略的有效性。     

Synthesis of cost-optimal heat exchanger networks using differential evolution

Heat exchanger network synthesis (HENS) has been one of the most-studied problems in process synthesis. Nevertheless, the complexity of the HENS problem provides enough scope for the development of novel algorithms involving the application of specialized optimization techniques. Evolutionary algorithms (EA) have emerged as viable alternatives to traditional methods for optimizing functions of both continuous and discrete variables. Differential evolution (DE) is one such evolutionary algorithm that promises simple, fast and robust optimization. The present study illustrates the application of this novel technique for the synthesis of heat exchanger networks. The HENS model proposed here considers stream splitting, does away with the simplifying assumption of isothermal mixing of the split streams and has the capability to handle compulsory and forbidden matching of streams. The DE-based model (DEM) does not rely on the decomposition of the problem into subproblems but employs a simultaneous method of approach to optimize the structure of the network of heat exchangers, the heat loads of these exchangers, the split stream heat flows and the minimum approach temperature. The proposed model has been applied to some case studies available in the literature and the results of these studies are very encouraging. The present work represents thus a step forward in the search for robust and efficient global optimization algorithms for the solution of the HENS problem.     

A monogenetic algorithm for optimal design of large-scale heat exchanger networks

The optimal design of large-scale heat exchanger networks is a quite difficult task not only due to its non-linear characteristics but also due to a great number of local optima in its solution space. An explicit analytical solution of stream temperatures for the superstructure heat exchanger networks was developed, which reduces number of decision variables significantly. Based on this solution, a mathematical model for synthesis of heat exchanger networks was formulated for searching the optimal configuration of a heat recovery system by a hybrid genetic algorithm. For large-scale heat exchanger networks, a monogenetic algorithm based on the optimization of sub-networks is proposed. In the first step of the optimization, the hybrid genetic algorithm is applied to the synthesis of the whole heat exchanger network for finding the functional groups (sub-networks) rather than the chromosomes (positions of the heat exchangers and splits of the streams) and genes (areas and heat capacity flow rates). Then the monogenetic algorithm for evolution of the functional groups is carried out to improve the HEN. This procedure was applied to examples taken from literature and better results were obtained.     

采用结构保护策略的强制进化随机游走算法优化换热网络

鉴于强制进化随机游走算法（random walk algorithm with compulsive evolution, RWCE）在优化换热网络时可能出现有潜力结构被差解代替,提出了一种结构保护策略,增加一个与原种群平行进化的新种群,新种群执行结构保护,个体与原种群一一对应并接收其当前最优解,同时采用一种降维邻域搜索的进化方式,设置个体各维变量更新概率确定搜索维数,充分挖掘个体结构进化潜力,提升了算法的局部搜索能力;原种群则沿用RWCE的主要操作,保留了较强的全局搜索能力。将采用结构保护策略的RWCE算法用于有分流换热网络优化,取得了优于现有文献的结果。     

基于子网络强制进化的大规模换热网络优化

换热网络优化是典型的混合整数非线性规划（MINLP）问题,此问题的非线性约束以及到处存在的局部极值,使得最优解的获得尤其困难。特别是对于大规模网络来说,当物流数目增加时,可行的结构数目呈指数增长,目前还没有一种有效的算法来解决此类问题。应用改进的混合遗传算法,首先对换热网络进行初始优化,对初步优化结果进行子网络的划分,然后进行基于官能团（子网络）的重组、分解和交叉操作,获得了很好的结果。     

Development of a new model for multiple effect evaporator system

A new simplified scalable mathematical model, based on concepts of stream analysis, temperature paths and internal heat exchange, has been developed for synthesis of a multiple effect evaporator systems. In this model, fresh feed is assumed to be composed of product and number of condensate streams, which come out from different effects and these are treated as separate streams. For the present work a septuple effect flat falling film evaporator system, used for concentrating black liquor in an Indian Kraft Pulp and Paper mill, has been considered. This system is being operated under backward sequence with condensate-, feed- and product-flashing as well as steam splitting in first two effects. The set of linear algebraic equations for this model are self-generated through programming and is solved simultaneously using Gaussian Elimination Method with partial pivoting. Results of the present approach are validated with published model and industrial data.     

OPTIMIZATION OF THE LOCATIONS OF SIDE STREAMS IN A FILTER CAKE WASHING PROCESS

A macroscopic model was developed to perform optimization studies for countercurrent, non-equilibrium, displacement filter cake washing on a belt filter with incoming side streams. Each washing stage was assumed to contain a number of perfect mixing cells connected in series. The optimum locations of the side streams, which yielded the minimum exit solute concentration for the washed cake, were evaluated. Three cases were considered: Incoming side stream added to the wash liquor flow, incoming side stream added to the cake flow and incoming side streams added to both flows simultaneously. A parameter study was performed systematically to determine the effect of washing parameters on the optimum side stream locations.