系统工程理论与实践
繫統工程理論與實踐
계통공정이론여실천
Systems Engineering—Theory & Practice
2012年
6期
1362~1368
,共null页
MCMC方法 参数优选 后验分布 改进SCEM-UA算法 不确定性分析 概念性降雨径流 模型
MCMC方法 參數優選 後驗分佈 改進SCEM-UA算法 不確定性分析 概唸性降雨徑流 模型
MCMC방법 삼수우선 후험분포 개진SCEM-UA산법 불학정성분석 개념성강우경류 모형
Markov chain Monte Carlo; parameter optimization; posterior distribution; modified shuffled complex evolution metropolis algorithm; mlcertainty analysis; conceptual rainfall-runoff model
在系统研究SCEM-UA算法基础上,借鉴自适应Metropolis算法的思想,对SCEM-UA算法中推荐分布的协方差策略和接受率策略进行了改进,改进后的算法能有效确保样本多样性的丰富和保持,增强算法全局搜索能力,避免＂早熟收敛＂.以岷江流域降雨-径流模型参数优选研究为例,分析了改进算法在推求参数后验分布的搜索性能和效率,计算结果表明,改进SCEM-UA算法的参数推荐分布和接受率随着计算过程自动更新,整体计算效率和解的精度要明显优于传统的SCEM-UA算法.
在繫統研究SCEM-UA算法基礎上,藉鑒自適應Metropolis算法的思想,對SCEM-UA算法中推薦分佈的協方差策略和接受率策略進行瞭改進,改進後的算法能有效確保樣本多樣性的豐富和保持,增彊算法全跼搜索能力,避免＂早熟收斂＂.以岷江流域降雨-徑流模型參數優選研究為例,分析瞭改進算法在推求參數後驗分佈的搜索性能和效率,計算結果錶明,改進SCEM-UA算法的參數推薦分佈和接受率隨著計算過程自動更新,整體計算效率和解的精度要明顯優于傳統的SCEM-UA算法.
재계통연구SCEM-UA산법기출상,차감자괄응Metropolis산법적사상,대SCEM-UA산법중추천분포적협방차책략화접수솔책략진행료개진,개진후적산법능유효학보양본다양성적봉부화보지,증강산법전국수색능력,피면＂조숙수렴＂.이민강류역강우-경류모형삼수우선연구위례,분석료개진산법재추구삼수후험분포적수색성능화효솔,계산결과표명,개진SCEM-UA산법적삼수추천분포화접수솔수착계산과정자동경신,정체계산효솔화해적정도요명현우우전통적SCEM-UA산법.
Covariance of proposal distribution and the acceptance rate of SCEM-UA algorithm are modified based on the principle of adaptive Metropolis. The modified SCEM-UA algorithm can effectively ensure the diversity of the population and capability of the global search capability as well as computational effciency, avoiding the premature convergence. The efficiency and effectiveness of modified SCEM-UA algorithm for sampling the posterior distribution of model parameters is discussed based on the case study of tile Min River Basin. The results show that modified SCEM-UA algorithm is consistent, effective and efficient in inferring the parameter posterior distribution in the aspects of computation efficiency and accuracy. and is much better than the original algorithm
