基于差解结构的多程优化策略研究

Research on Multi-Pass Optimization Strategy based on Difference Structure

强制进化随机游走算法优化换热网络时,在个体选择阶段通过设置接受差解概率增强结构的进化能力,但优化过程中会出现个体在较短的迭代步数内未充分发挥其优化潜能而再次被差解取代的现象,使优化精度降低。针对两次接受差解间迭代步数较少的情况,提出差解结构的多程优化策略,将第二次接受差解前的结构传递给新种群。通过20流股(SP)和10流股算例验证,得到相较文献更低的年综合费用,分别为1396465和8705500$/a,表明引入该策略后,优化效率和质量得以提高。

Random walk algorithm with compulsive evolution is used to optimize the heat exchanger network,and the probability of accepting the imperfect networks solution is set to enhance the capability of structural evolution.But in the process of optimization,the heat exchanger network may not give full play to his optimization potential in a small number of iteration steps,and then be replaced by the difference network again,decreasing the optimization accuracy.In this paper,a multi-pass optimization strategy is proposed to improve the random walk algorithm with compulsive evolution,and the network before the second acceptation of imperfect networks is transferred to the new population.The cases of 20 streams(SP)and 10 streams(SP)show that the annual comprehensive cost is 1396465 and 8705500$/a,respectively.The results show that the optimization efficiency and quality can be improved by using this strategy.