优化、RSA和GLUE方法在非线性环境模型参数识别中的比较

参数识别是数学模型应用的前提,本文对非线性环境模型常用的3种参数识别方法进行了比较分析.优化方法是出现最早、应用最广泛的参数识别方法之一.但在观测误差存在的情况下,采用优化方法识别的最优参数进行模型预测,存在很大的决策风险.考虑到这种不足,RSA方法和GLUE方法摒弃了识别单一最优参数的传统思维,而把识别参数扩大到多点组成的参数集.RSA方法与GLUE方法不同的是,RSA方法把可行的参数点看成是同等接受的,而GLUE方法则根据模拟值与实测值的差别,确定其似然度,代表参数的可信度水平.除参数识别以外,RSA方法和GLUE方法也是全局灵敏度分析的重要工具.

Comparison of Optimum,RSA and GLUE Methods in Parameter Identification of a Nonlinear Environmental Model

Parameter identification plays a key role in environmental model application.The optimization method is one of the earliest and most widely used methods.However,as the parameters by optimization may not fully fit the observations,there is a risk that the errors may be enhanced in the decision|make stage.With this deficiency in consideration,the RSA and GLUE algorithms search for the feasible parameters not only to the optimum but also around the neighbors.The difference between RSA and GLUE is that the RSA accepts the estimated parameters equally as the candidates for application;while the GLUE keeps the difference among the parameters as measured by likelihood.In addition for parameter identification,both RSA and GLUE are efficient tools for global sensitivity analysis.