文献类型：期刊文献

中文题名：基于抽样平均近似的双层改进粒子群算法的无分流换热网络综合

英文题名：Sample Average Approximation Based Double-Layer Improved Particle Swarm Optimization for Heat Exchanger Network Synthesis without Split Streams

作者：陈帅[1];罗娜[1]

机构：[1]华东理工大学化工过程先进控制与优化技术教育部重点实验室,上海200237

年份：2018

卷号：32

期号：3

起止页码：620

中文期刊名：高校化学工程学报

外文期刊名：Journal of Chemical Engineering of Chinese Universities

收录：CSTPCD;;EI(收录号：20183205654267);Scopus;北大核心:【北大核心2017】；CSCD:【CSCD2017_2018】；

基金：国家自然科学基金(61403140);中央高校基本科研业务费重点科研基地创新基金(222201717006)

语种：中文

中文关键词：换热网络综合;量子粒子群算法;竞争群优化算法;抽样平均近似

外文关键词：heat exchanger network; quantum particle swarm optimization; competitive swarm optimization; sample average approximation

摘要：针对单层优化算法求解换热网络不易找到最优解以及传统的双层优化算法计算成本较高的问题,研究以年综合费用最小值为优化目标,提出了基于抽样平均近似的双层改进粒子群算法求解无分流换热网络综合。该算法在优化初始阶段对内层变量优化时以抽样平均近似方法为基础,根据随机挑选的部分粒子信息快速评价外层的换热网络结构,而外层优化通过结合变异算子的量子粒子群算法对模型结构进行优化。待缩小最优网络结构的范围后再通过竞争群算法深度优化热负荷等连续变量,从而得到最优的年综合费用。研究所提出的算法通过抽样平均近似方法对外层结构模型进行初步筛选,避免了大量非最优解网络结构下热负荷等连续变量的优化,从而减少了双层优化过程的计算量,提高了优化效率,同时在缩小网络结构范围后采用竞争群优化算法提高了优化精度,保证了优化效果。通过对两个典型算例的测试可知,研究的方法相比传统的双层算法在求解时间上缩短了近90%,在求解效果上也能达到较好结果。
It is difficult to find the optimal solution using single-layer optimization algorithms for heat exchanger networks（HENs）. Meanwhile, normal double-layer methods are computationally expensive in the synthesis of heat exchanger network. This study proposed a new two layer algorithm based on sample average approximation and quantum particle swarm optimization for heat exchanger networks without split streams.During the beginning of optimization, sample average approximation（SAA） was used in the inner layer to generate random particles for the evaluation of structure variables in the out layer of HEN. Quantum particle swarm optimization combined with the mutation operator was used to optimize the structure variables.Competitive particle swarm optimization algorithm was used after shrinking the range of the optimal structure variables to optimize heat load and other continuous variables with higher accuracy. Better network structure can be found with much less simulation after initial estimation from SAA to reduce computation cost. The competitive particle swarm optimization algorithm can improve the optimal of the inner layer in the late stage, which increases result accuracy. The proposed algorithm was tested on two typical heat exchanger networks. The results show that the calculation time is shortened by nearly 90% comparing with conventional two-layer algorithms.