|国家科技期刊平台
首页|期刊导航|原子能科学技术|基于Cholesky分解法的LHS放射性废物处置场安全不确定性分析

基于Cholesky分解法的LHS放射性废物处置场安全不确定性分析OA北大核心CSTPCD

Cholesky-factorization-based Latin Hypercube Sampling for Safety Uncertainty Analysis of Radioactive Waste Disposal Sites

中文摘要英文摘要

放射性废物处置是一项与国土环境、公众安全、核工业健康以及可持续发展有关的重大问题.安全全过程系统分析是保障放射性废物处置设施从选址、建设、运行到关闭后安全性的重要手段,不确定性分析是其中重要一环.环境变化、人员行为等事前无法控制的外部因素都将对放射性废物处置设施的安全产生重大影响,需要对其进行不确定性评估.在放射性废物处置库的不确定性分析中,参数不确定性分析的计算过程相较于常见的蒙特卡罗模拟的运用场景,其涉及输入的随机参数多、运用模型庞杂,势必需求更小的抽样样本以减少运算时间、提高抽样效率.拉丁超立方抽样(LHS)是不确定性分析中常用的方法,但该方法应用于多维抽样时由于排序质量较低,使得小样本条件下的相关性要求不能得到满足.本文采用Cholesky分解法对拉丁超立方抽样过程中的排列构造过程进行了改进,通过对排列矩阵各行向量进行解耦,以最小化其各维度间的相关性.此改进方案显著降低了拉丁超立方抽样对样本相关性的影响,加速了计算结果的收敛速度.在本文的使用场景下,改进后的抽样方法只需要使用改进前所需样本规模的1/10,提高了计算效率.

Safety-focused process system analysis is an important approach to ensure the safety of radioactive waste disposal facilities throughout their lifecycle,including siting,construction,operation,and closure.In recent years,the uncertainty in the analysis process receives increasing attention,and uncertainty analysis and management become integral parts of safety-focused process system analysis and safety evaluation.Interna-tional documents such as the Safety-Focused Process System Analysis and Safety Evaluation for Radioactive Waste Disposal by IAEA and domestic regulations like the Safety Regulations for Near-Surface Disposal of Low and Intermediate-Level Radio-active Solid Waste emphasize the importance of uncertainty analysis in these processes.Properly addressing uncertainty in safety-focused process system analysis and safety evaluation is crucial for ensuring the credibility of the evaluation results.Therefore,conducting uncertainty analysis on case studies for long-term safety evaluations is partic-ularly important.External factors,such as environmental changes and human behavior,which cannot be controlled in advance,significantly impact the safety of various nuclear facilities,including radioactive waste disposal facilities,and thus require uncertainty assessment.Many countries and international organizations in the nuclear industry con-ducted related research,focusing on probability,risk guidance,and uncertainty analy-sis.They developed some general evaluation methods,such as Monte Carlo simulation and Latin hypercube sampling,for uncertainty analysis according to specific circum-stances.In the uncertainty analysis of radioactive waste disposal repositories,the com-putational process of parameter uncertainty analysis is more complex compared to the common application scenarios of Monte Carlo simulation.It involves a large number of random input parameters and complex models,which necessitates smaller sample sizes to reduce computation time and improve sampling efficiency.In this case,the sampling process with small sample sizes and multiple random parameters will inevitably face a challenge:Whether the uniformity and correlation between the dimensions of the sam-pled samples meet the computational requirements.To address the issue of uniformity in the sampling process,adopting Latin hypercube sampling(LHS)coupled with Monte Carlo simulation is a common choice.In this study,LHS was improved by applying the Cholesky decomposition method in the parameter uncertainty analysis of safety-focused process system analysis for radioactive waste disposal repositories.The improved LHS method through this approach can use smaller samples to meet the correlation require-ments of the computation,thereby enhancing sampling efficiency.In the scenario of this study,the improved sampling method only requires one-tenth of the sample size needed in the original method.

赵润才;玉宇;陈涛

华北电力大学核科学与工程学院,北京 102206

核科学

不确定性分析Cholesky分解法拉丁超立方抽样放射性废物处置

uncertainty analysisCholesky factorization methodLatin hypercube sam-plingradioactive waste disposal

《原子能科学技术》 2024 (004)

731-741 / 11

10.7538/yzk.2023.youxian.0466

评论