似然度函数对GLUE方法的影响  

Effects of Likelihood Functions on the Results of GLUE Method

作　　者：邓义祥[1] 郑丙辉[1] 苏一兵[1] 李子成[1] 卢少勇[1] 

机构地区：[1]中国环境科学研究院,北京100012

出　　处：《环境科学研究》2008年第2期44-48,共5页Research of Environmental Sciences

基　　金：国家"十五"科技攻关计划项目(2003BA614A)

摘　　要：GLUE方法是不确定性条件下参数识别的重要方法,似然度函数的选择是GLUE方法的关键.为克服观测误差和模型结构误差对研究结果的影响,利用合成的数据序列,以污染物衰减模型为基础,分析了似然度函数对参数识别和模型预测结果的影响.从识别结果来看,由于似然度函数的变化改变了参数似然度之间的对比,因此对参数识别及灵敏度分析的结果都产生了影响,甚至能够改变参数全局灵敏度的相对排序.从预测结果来看,似然度函数的变化影响了模型预测的分布,但即使预测结果的模糊性在减少,精度并没有相应地提高.因此,似然度函数的恰当选择对分析结果具有重要的影响,应结合研究问题的具体特点和对模拟结果的要求,采用概率论等相关方法选择出尽可能体现参数真实重要性的函数.The GLUE method is an important means of parameter identification under uncertain information. The likelihood function plays an important role in GLUE. The effects of likelihood functions on the parameter identification and model forecast results were discussed with a substance decay model by artificial composite data to avoid the effects of observation errors and model structure errors. From the identification results, the variation of likelihood functions changes the comparison of the likelihood values of the parameters, so the results of parameter identification and sensitivity results are both influenced, even the order of the global sensitivity of the parameters may be different. From the forecast results, the variation of the likelihood functions influences the distribution of the modeling results. On the other hand, although the illegibility of forecast results is decreasing, the accuracy may not increase at the same time. Because of the important effects of likelihood functions on the analysis results, this functions should be determined on the basis of the detailed characteristics of studied results and the requests of the modeling by the probability theory and other metbods to reflect the real importance of the parameters.

关 键 词：GLUE方法 参数识别 灵敏度分析 预测 

分 类 号：X522[环境科学与工程—环境工程] X143