基于SARAX/DAKOTA的快堆设计多目标优化框架开发与验证  

Research on Multi-objective Optimization of Fast Reactor andDevelopment of SARAX/DAKOTA Optimal Frame

作　　者：李晓琪 郑友琦[1] 杜夏楠 王永平[1] LI Xiaoqi;ZHENG Youqi;DU Xianan;WANG Yongping(School of Nuclear Science and Technology,Xi'an Jiaotong University,Xi'an 710049,China)

机构地区：[1]西安交通大学核科学与技术学院,陕西西安710049

出　　处：《原子能科学技术》2022年第1期96-105,共10页Atomic Energy Science and Technology

基　　金：国家重点研发计划(2019YFB1901102);国家自然科学基金(12005160)。

摘　　要：与具有大量运行经验、设计目的相对单一的压水堆相比,快堆设计经验少、设计目的更复杂,难以直接使用经验方法实现考虑多目标、多约束的堆芯设计以及布料方案的优化。本文将遗传算法用于快堆设计的多目标优化问题,对传统的遗传算法在编码方案和多目标的处理上进行了改进,使用中子学分析软件SARAX与优化工具箱DAKOTA搭建了针对快堆设计与布料方案优化的框架,并进行了功能上的验证和实际问题的应用。The optimization design of fast reactors faces two major problems:insufficient design or operation experience and more complicated optimization objectives,compared with pressurized water reactor(PWR).An optimization method which is independent of application banckground and experience is required.Intelligent optimization algorithm is a strategy for the search of optimal or satisfactory solution for given design objectives and satisfies above requirements.In the process of search,multiple design objectives and constraints are considered simultaneously and the times of search and calculation are limited.Genetic algorithm is a classical and representative method in numerous of optimization algorithms,which performs well in solving multi-objective problems with constraints.Genetic algorithm has inherent parallel characteristics as well for which the searching efficiency is able to be increased.Based on a genetic algorithm module in an open source toolkit DAKOTA and a self-developed neutron transport program SARAX,an optimal design frame for fast reactors with multivariable,multi-objective and multi-constraint problems was constructed in this work.Both genetic algorithm module and neutron transport program performed as a black-box.A coding system that could transform a decimal integer and a loading pattern to each other was suggested and applied to this optimal frame.This coding strategy realized the coverage of the multi-dimensional search space,so that the optimal frame can deal with the problems of continuous and discrete variables.For the constrained multi-objective problem,the direct ranking method and the weighted sum method were both applied and compared.Two cases were analyzed to test this frame:1)Rearrange the subassemblies of ABTR to find a scheme whose k_(eff) and power peak factor meet the given expectations,the searching result was compared with the enumeration results;2)Optimizing the design of the loading position of the ^(237)Np-containing subassemblies for China Experimental Fast Reactor(CEFR).The obj

关 键 词：快堆设计 多目标优化 遗传算法 SARAX DAKOTA 

分 类 号：TL329.2[核科学技术—核技术及应用]