Intelligent Nuclear Design Based On Hybrid Differentia Evolution

Keactor nuclear design is an important part of nuclear engineerng design and is essential for improving the economics and safety of nuclear power plants.Due to the large number of variables as well as many constraints,the optimization becomes very complicated.Optimization methods based on artificial intelligence have been widely used in nuclear design optimization,However,there is insufficient convergence and large sensitivity on control parameters,resulting in the optimization is not comparable with manual optimization.In this paper,the research on nuclear design optimization based on the Super Multi-functional Calculation Program for Nuclear Design and Safety Evaluation(SuperMC)will be performed.The main contents and innovations are summarized as follows:(1)The hybrid differential nuclear optimization method based on classical Differential Evolution(DE)was developed.There are variables with both continuous type and discrete type in nuclear design optimization.DE was originally designed to deal with continuous space optimization,thus it cannot be directly applied to solve the optimization of discrete variables.This work develops a hybrid differential evolution algorithm HTDE for nuclear design optimization.In HTDE,the coding method for discrete variables is designed independently.In the evolution,new mutation and crossover strategies are developed to deal with hybrid variables optimization.To further improve the performance of the algorithm,a new adaptive cross-probability is proposed,and the opposite-based learning method is adopted.The performance of HTDE was then verified in the continuous and discrete test problems,and the PWR core loading pattern case.The results showed the excellent convergence and robustness of HTDE.(2)The non-dominated hybrid differential multi-objective nuclear optimization method based on improved NSGA-II was developed.Multi objectives are often involved in nuclear design optimization.NSGA-II is a classical algorithm for solving multi-objective optimization,but it has the disadvantage of lacking diversity and convergence in optimization.This work combines HTDE with NSGA-II,and proposes a new sorting strategy with improved crowding definition and population dynamics adjustment strategy to achieve the non-dominated hybrid differential and multi-objective nuclear optimization method MHTDE.The performance of MHTDE was evaluate by the international multi-objective benchmark example sets.The results showed the efficiency of MHTDE in multi-objective optimization.The developed method was verified by core loading pattern optimization of the Westinghouse PWR benchmark and Commercial Reactor WBN1 located in Tennessee,and shielding design optimization of the Savannah nuclear power ship.For the Westinghouse PWR benchmark optimization problem,the power peaking factor(PPF)was reduced from 1.60 to 1.21 while ensuring that the effective multiplication factor(keff)of the core to meet the limit.The optimization goal of WBN1 is to maximize keff,and minimize PPF,considering the constraint of the reactor moderator temperature coefficient.The optimized pattern has an keff increased by 1.0%and PPF decreased by 2.6%.In the optimization of the shielding design of the US Savannah ship,the shielding weight was reduced by 25.5%,and the volume was reduced by 15%.The dose level was ensured to be within an acceptable range.The results fully prove the feasibility and effectiveness of the optimization method developed in this paper,which can be widely used in the design optimization of complex nuclear reactor systems.