软计算技术在环境复杂模型参数识别中的应用研究

以WASP模型在密云水库水质模拟中的应用为例,研究考察分析了软计算技术用于环境复杂模型参数识别的性能和效率.参数分析表明,WASP模型存在一些低灵敏参数且受相关参数的影响明显.为对比参数灵敏度和相关性的影响,共设计了4组数值试验.数值试验表明,全局搜索法能较好解决参数全局寻优问题,是获取参数辅助信息的重要手段.MCMC法可有效对参数后验分布进行采样,采样序列稳定收敛到参数后验分布上,同时MCMC法较好地处理了相关参数采样的问题.最后给出案例表明基于软计算技术的复杂模型参数识别技术路线是高效的、实用的.

Parameter Identification of Complicated Environmental Model Using the Soft-computing Approach

In order to study the performance and efficiency of soft-computing technology in application to parameter identification of complicated environmental model,a case study was presented,which was an application of water quality simulation in the Miyun Reservoir using WASP model.Parameter analysis indicated that there were many low sensitivity parameters in WASP model and several groups of parameters were correlative apparently.For analyzing effects of sensitivity and correlativity,four numerical experiments were designed.Resulted indicated that global searching method could solve global optimization well and also was an efficient step to acquire auxiliary information of parameters,And MCMC method could sample posterior distributions of parameters effectively;moreover,sampling series could converge to the posterior distributions of parameters eventually.Case study indicated that it was efficient and practicable for the soft-computing approach presented by the paper to identify parameters of complicated model.