CIESC Journal ›› 2022, Vol. 73 ›› Issue (2): 801-813.DOI: 10.11949/0438-1157.20210909

• Process system engineering • Previous Articles     Next Articles

A hybrid algorithm based on parallel computing for heat exchanger network optimization with stream splits

Zhiqiang ZHOU1,2(),Guomin CUI1,2(),Ling YANG1,2,Xiubao MA1,2,Yuan XIAO1,2,Qiguo YANG1,2   

  1. 1.School of Energy and Power Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China
    2.Shanghai Key Laboratory of Multiphase Flow and Heat Transfer in Power Engineering, Shanghai 200093, China
  • Received:2021-06-30 Revised:2021-09-14 Online:2022-02-18 Published:2022-02-05
  • Contact: Guomin CUI

一种基于并行计算的混合算法优化有分流换热网络

周志强1,2(),崔国民1,2(),杨岭1,2,马秀宝1,2,肖媛1,2,杨其国1,2   

  1. 1.上海理工大学能源与动力工程学院,上海 200093
    2.上海市动力工程多相流动与传热重点实验室,上海 200093
  • 通讯作者: 崔国民
  • 作者简介:周志强(1997—),男,硕士研究生,470242099@qq.com
  • 基金资助:
    国家自然科学基金项目(21978171);中国博士后科学基金项目(2020M671171)

Abstract:

Heat exchange network optimization is a research difficulty in the field of chemical process system. Its mathematical model is highly nonconvex and nonlinear, so it often has limitations when using a single heuristic algorithm. The objective of the research is to minimize the annual cost of heat exchange network. In order to solve the problem of individual independent evolution and lack of communication between individuals when random walk algorithm compulsive evolution (RWCE) is used to optimize heat exchanger network, a hybrid algorithm with genetic algorithm (GA) and RWCE is proposed. The mixed algorithm maintains the individual evolution of the individuals in the first half of the dominant population, and replaces the inferior population by generating offspring through periodic crossover and mutation operations, thereby enhancing the original algorithm's ability to optimize integer variables. It makes up for the lack of renewal of vulnerable individuals. In order to improve the computational efficiency when optimizing heat exchanger network with splits under large population and save cost of time, the parallel design of the hybrid algorithm is realized by OpenMP system. The parallel hybrid algorithm is verified by three heat exchange network problems of different scales. The results show that the algorithm can greatly shorten the calculation time compared with the serial algorithm on the premise of effectively improving the optimization quality, and two of the examples have broken through the current optimal solution in the literature.

Key words: heat exchanger network, parallel computing, optimization, hybrid algorithm, multithreading

摘要:

换热网络优化是化工过程系统工程领域的研究难点,其数学模型具有高度的非凸、非线性,在使用单一启发式算法优化时,往往具有局限性。研究以换热网络的年综合费用最小为目标,针对强制进化随机游走(RWCE)算法在优化时由于个体间独立进化,导致优化过程中信息缺乏交流的问题,提出将遗传算法(GA)与其混合。混合后的算法在保持前一半优势种群中的个体单独进化的基础上,通过周期性的交叉、变异等操作产生子代来替换掉劣势种群,从而增强了原有算法的整型变量优化能力,并弥补了弱势个体无法更新的不足。为了兼顾算法在大种群下优化有分流换热网络的计算效率,节约时间成本,使用OpenMP系统将混合算法实现了并行化设计。通过三个不同规模的换热网络问题对并行后的混合算法进行验证,结果表明该算法能在有效提升优化质量的前提下相比串行算法大幅缩短计算时间,其中两个算例突破了目前文献最优解。

关键词: 换热网络, 并行计算, 优化, 混合算法, 多线程

CLC Number: