基于初等效应的全局敏感性量化方法优化及应用  

Optimization and Application of Global Sensitivity Quantization Method Based on Elementary Effects

作　　者：谢林峰 江鹏浩 李冬 XIE Linfeng;JIANG Penghao;LI Dong(School of Energy and Mechanical Engineering,Shanghai University of Electric Power,Shanghai 201306,China)

机构地区：[1]上海电力大学能源与机械工程学院,上海201306

出　　处：《核科学与工程》2024年第4期894-903,共10页Nuclear Science and Engineering

摘　　要：最佳估计加不确定性(BEPU)分析方法已广泛应用于核电系统安全评价,敏感性分析对重要现象和参数进行识别和排序,是BEPU方法中的重要环节。由于通常采用的局部敏感性分析方法对多维非线性系统适用性有限,而全局敏感性分析方法运算成本较高,因此本文在对比了方差分解、矩独立和初等效应的全局敏感性分析方法后,选择基于初等效应的方法通过变化百分比的欧式距离和轨迹法对其进行优化,并通过数值算例验证了优化方法的准确性。最后将方法应用于大破口事故完整性试验台架(LOFT),结果表明优化方法可以在抽样次数大大降低的情况下准确识别重要参数,并得到与原始方法相似的重要度排序结果。The best estimate plus uncertainty(BEPU)analysis method has been widely used in safety evaluation for nuclear power system.The sensitivity analysis is an important part of the BEPU method to identify and rank important phenomena and parameters.Due to the limited applicability of local sensitivity analysis methods to multi-dimensional non-linear systems,and the high computational cost of the global sensitivity analysis method,this paper compares the global sensitivity analysis methods of variance decomposition,moment-independence and elementary effects.Select the method based on elementary effects to optimize it through the Euclidean distance and the trajectory method of percentage change,several numerical examples are given to verify the accuracy of the optimized method.Finally,the proposed method is applied to LOFT,a large break loss of coolant accident.The result shows that the optimized method can accurately identify important parameters under the condition of greatly reduced sampling times,and the importance ranking results are similar to the original method.

关 键 词：BEPU 全局敏感性分析 初等效应 轨迹法 LOFT 

分 类 号：TL48[核科学技术—核技术及应用]