基于Kriging代理模型的铅铋反应堆智能优化方法北大核心CSCD

铅铋反应堆广泛应用的需求要求研究人员在现有堆芯方案的基础上开展大量优化设计工作。针对铅铋反应堆多物理、多变量、多约束耦合影响的多维非线性约束优化设计问题,基于Kriging代理模型、正交拉丁超立方抽样和SEUMRE空间搜索技术构建铅铋反应堆智能优化方法,耦合物理蒙卡计算/热工分析程序,开发包含抽样、耦合程序前后处理、反应堆优化分析功能的优化平台,并以铅铋反应堆SPALLER-4,URANUS为原型分别开展最小燃料装载量的方案寻优与参数优化验证。验证结果表明,该智能优化方法用于铅铋反应堆设计方案寻优和堆芯参数优化可行、有效,相比传统蒙卡程序计算寻优,在保证预测精度前提下极大地降低了计算成本,与URANUS初始模型比较,燃料装载量、堆芯总质量、活性区体积、堆芯总体积分别优化10.8%,11.5%,18.1%,17.1%,为基于代理模型的智能优化方法应用于铅铋反应堆的优化设计提供参考。

Intelligent optimization method for lead-bismuth reactor based on Kriging surrogate model

The extensive application requirements of lead-bismuth reactors require researchers to carry out a lot of optimization design work on the basis of existing core schemes. Aiming at the multi-dimensional nonlinear constrained optimization design problem of lead-bismuth reactor with multi-physical, multi-variable and multi-constraint coupling effects, an intelligent optimization method for lead-bismuth reactor was constructed based on Kriging surrogate model, orthogonal Latin hypercube sampling and SEUMRE spatial search technology. Coupled with physical Monte Carlo calculation/thermal ranalysis code, an optimization platform including sampling, pre-and post-processing of coupling program and reactor optimization analysis function was developed. Taking SPALLER-4 and URANUS as prototypes, the scheme optimization and parameter optimization verification of minimum fuel load were carried out respectively. The verification results show that the core intelligent optimization method is feasible and effective for the optimization of lead-bismuth reactor design scheme and core parameters. Compared with the traditional Monte Carlo calculation optimization, the calculation cost is greatly reduced under the premise of ensuring the prediction accuracy. Compared with the URANUS initial model, the fuel loading, the total mass of the core, the volume of the active zone and the total volume of the core are optimized by 10.8%, 11.5%, 18.1% and 17.1% respectively, which provides a reference for the intelligent optimization method based on the surrogate model applied to the optimization design of lead-bismuth reactor.