Kriging代理模型的序列优化及其应用

发布时间：2018-05-15 12:45

  本文选题：多目标优化 + NSGA-Ⅱ算法　； 参考：《华东理工大学》2015年硕士论文

【摘要】：基于流程仿真模型的复杂化工过程优化,往往需要较长的优化时间,效率低下,本文致力于构造流程仿真模型的Kriging代理模型、提高Kriging代理模型的精度、研究基于Kriging代理模型的序列优化策略。代理模型的计算量比精确模型小得多,在模型精度得到保证的前提下,采用代理模型可以大大减少优化过程中的计算量,提高工程优化设计的效率。本文的主要工作与创新点有： 首先,本文简单地介绍了Kriging代理模型的基本组成项、优点以及近期学者关于Kriging模型的研究概况；对Kriging代理模型的机理、回归模型、相关模型等作了一一阐述；针对Kriging代理模型建模采样的方法、Kriging代理模型精度检验的标准作了相关说明。 其次,本文介绍了一种提高Kriging代理模型精度的方法——序列迭代优化,阐述了序列迭代优化的基本流程。然后考虑到原始序列优化方法EGO(efficient global optimization)的局限性和遗传算法的优越性,本文采用遗传算法搜索基于某类加点准则状况下模型更新所需要的迭代插值点,构造了基于遗传算法的序列优化的操作流程；基于EI(Expected Improvement)加点准则的加权思想,本文提出了一种全新的加点准则——DH值最大点插值准则,并利用遗传算法的全局搜索能力搜索模型迭代的插值点,提高了Kriging模型的建模精度。 另外,本文介绍了NSGA-Ⅱ算法的主要原理和基本流程,并将其成功地运用于对Kriging代理模型的操作优化。DH值最大点插值法综合了Kriging模型的预测误差和均方差,本文选择各子目标的DH值作为评价函数值,利用NSGA-Ⅱ算法,构造了一种新的多目标序列优化方法——MODH。 MODH算法利用NSGA-Ⅱ算法的遗传操作和非支配排序等操作算子寻找基于Kriging模型的非支配解集,将这些非支配解集作为迭代点实现原始Kriging模型的更新,提高了Kriging代理模型的精度,最后选用了POL测试函数证明了上述多目标序列优化方法的可行性。 最后,本文提出了一种Kriging代理模型与NSGA-Ⅱ算法两者相结合的算法——K-N算法。K-N算法将基于Kriging代理模型优化所得的Pareto解集作为新一轮NSGA-Ⅱ算法的初始种群,引导算法快速地在最优解集附近寻优。K-N算法综合了Kriging代理模型效率高的特点和流程仿真模型准确度高的优点,在PX氧化反应过程的操作优化中获得了很好的应用。
[Abstract]:Complex chemical process optimization based on process simulation model often needs long optimization time and low efficiency. In this paper, the Kriging agent model of process simulation model is constructed to improve the accuracy of Kriging agent model. The sequence optimization strategy based on Kriging proxy model is studied. The computational complexity of the agent model is much smaller than that of the accurate model. On the premise of ensuring the accuracy of the model, the calculation amount in the optimization process can be greatly reduced and the efficiency of the engineering optimization design can be improved by using the agent model. The main work and innovations of this paper are as follows: Firstly, this paper briefly introduces the basic components and advantages of Kriging proxy model, as well as the recent research situation of Kriging model, and expounds the mechanism of Kriging proxy model, regression model, relevant model and so on. In this paper, the Kriging proxy model sampling method and the accuracy test standard of Kriging agent model are introduced. Secondly, this paper introduces a method to improve the accuracy of Kriging proxy model-sequence iterative optimization, and describes the basic process of sequence iterative optimization. Then, considering the limitation of the original sequence optimization method EGO(efficient global optimization and the superiority of genetic algorithm, this paper uses genetic algorithm to search the iterative interpolation points needed for model updating based on certain additive criteria. The operation flow of sequence optimization based on genetic algorithm is constructed, and based on the weighted idea of EI(Expected improvement criterion, a new addition point criterion, DH maximum point interpolation criterion, is proposed in this paper. The global search ability of genetic algorithm is used to search the interpolation points of model iteration, which improves the modeling accuracy of Kriging model. In addition, this paper introduces the main principle and basic flow of NSGA- 鈪，

本文编号：1892515