观测误差对非线性环境模型参数识别的影响.pdfVIP

第 23 卷第 6 期 环 　境 　科 　学 　学 　报 Vol . 23 ,No . 6 2003 年 11 月 ACTA SCIENTIAE CIRCUMSTANTIAE Nov. ,2003 ( ) 文章编号 2003 　　　中图分类号 :X32 　　　文献标识码 :A 观测误差对非线性环境模型参数识别的影响 1 (清华大学环境科学与工程系 ,北京 　100084) 杜鹏飞 ,邓义祥 ,陈吉宁　 摘要 :采用 SCEUA 方法和 RSA 方法 , 比较了在不同的观测误差条件下 ,优化方法和不确定性分析方法对于非线性模型参数 识别的影响. 分析表明,RSA 方法是在具有观测误差的数据条件下进行参数识别的一种有效手段. 同时 ,通过比较分析 ,发现 RSA 方法预测的浓度平均值和概率最大值与“真实值”并不完全一致. 因此 ,在使用 RSA 方法时 ,应该充分考虑预测区间 , 以 降低决策风险. 关键词 :观测误差 ;不确定性分析 ;SCEUA ;RSA Effects of observation errors on parameter identif ication of a nonlinear environmental model ( DU Pengfei ,DENG Yixiang ,CHEN Jining 　Department of Environmental Sciences and Engineering Tsinghua University ,Beijing 　100084) Abstract :Parameter identification is a crucial procedure for model application . Observation errors are widely present in environmental data . This paper focused on the effects of observation errors on parameter identification of nonlinear environmental models. It specifically compared the identification results by the optimization and uncertainty analysis approaches , taking the SCEUA and RSA methods as representatives respectively. It was found that the RSA was an effective approach to reduce the impacts of observation errors. It was also showed that the idntified average values or the statistically most likely values of the RSA were often not in consistence with the “true values ”. It is there fore suggested that the intervals be considered as identified parameter uncertainty so as to reduce the decisionmaking risks. Keywords :observation errors ; uncertainty analysis ; environmental model ; SCEUA ; RSA 数学模型是环境规划与管理的重要手段 ,参数是数学模型必不可少的组成部分. 一般情况 下 ,数学模型的参数必须通过识别加以确定. 因此 ,参数识别是数学模型