现有的节点非结构模型（Nodewise nonstructural superstructure，NWNSS）在优化换热网络时需要预设固定的节点分流数量，难以满足结构进化过程对于求解空间和计算效率的需求，易造成换热单元生成空间受限，影响算法全局优化性能。本文提出一种流股分流动态调节策略，该策略基于实时结构的整型变量分布信息，动态增加结构进化所需的必要分流并减少无效结构对优化的阻碍，辅助算法以更高的效率跳出局部极值，提升优化质量。将策略应用于16SP、20SP算例，分别得到年综合费用为6 653 940和1 711 886 $/a的最优换热网络结构，较文献最优结果降低了3 140和3 202 $/a。

The existing nodewise nonstructural superstructure (NWNSS) needs to set fixed number of split flow for all nodes in advance when optimizing heat exchanger networks,which is difficult to meet the needs of solving space and computational efficiency in the process of structural evolution,tending to limit the generation space of heat exchanger unit and reduce the global optimization quality of algorithm.Therefore,this paper proposes an optimal strategy to dynamically adjust the number of split flow for each node.Based on the distribution information of integral variables of realtime structure,this strategy dynamically increases the necessary split flow required by structural evolution and reduces the obstacle of ineffective structural evolution to optimization,improves the efficiency of the algorithm to jump out of the local search scope,and enhances the quality of optimization results.The strategy is applied to 16SP and 20SP cases,and the optimal heat exchanger network structures with the overall annual costs of 6 653 940 $/a and 1 711 886 $/a are obtained respectively.Compared with the optimal results in the literature,the results of this paper are reduced by 3 140 $/a and 3 202 $/a.

TK124

国家自然科学基金（21978171）