Development of Parallel Genetic Algorithm and Application to Small Modular Fast Reactor Design Optimization

Genetic Algorithm;Multiobjective Optimization;Nondominated Sorting;Valuable Phenotype;Small Modular Fast Reactor;622.33

공과대학 에너지시스템공학부 ;

University:서울대학교 대학원

关键词: Genetic Algorithm;    Multiobjective Optimization;    Nondominated Sorting;    Valuable Phenotype;    Small Modular Fast Reactor;    622.33;   

Others  :  http://s-space.snu.ac.kr/bitstream/10371/137375/1/000000145244.pdf

美国|英语

来源: Seoul National University Open Repository

【 摘 要 】

In multi-objective nuclear reactor design problem, instead of implementing a single-objective optimization ;;scalarized” from the multi-objective problem, for example, by assigning each objective an importance, it is beneficial to provide a trade-off-surface to the decision maker for further consideration. However, the relatively expensive calculation in the nuclear reactor design prevents the true Pareto front to be established. Instead, the ;;pseudo-trade-off surface” is usually provided. Thus, when a preferred solution has been decided, the decision maker comes to face the question that whether this solution is the non-dominated solution. The Genetic Algorithm with the ;;valuable phenotype” archival rule developed in this work abnegates the logic that higher quality individuals should have the priority to be selected. The new rule addresses more about of the balanced accomplishment of the objectives rather than pitch into the elitism. This Optimized Logic Genetic Algorithm has demonstrated its efficiency and robustness in assisting the designer to obtain the better flexibility by providing the diverse potential solutions that can dominate or are similar to the interested solution on the ;;pseudo-trade-off surface”.