一维河流水质模型多参数识别的反演优化通用算法  

General inversion optimization algorithm for multi-parameter identification of 1D river water quality model

作　　者：刘晓东[1,2] 华祖林[1,2] 谢增芳[2] 顾莉[1,2] 

机构地区：[1]“浅水湖泊综合治理与资源开发”教育部重点实验室,河海大学,南京210098 [2]河海大学环境科学与工程学院,南京210098

出　　处：《水力发电学报》2012年第2期122-127,共6页Journal of Hydroelectric Engineering

基　　金：中央高校基本科研业务费专项资金项目(2009B17814,2009B16714);国家自然科学基金项目(50979026,50679019);国家重点基础研究发展计划(973)项目(2008CB418202);国家水体污染控制与治理科技重大专项(2009ZX07103-006);水利部公益性行业科研专项经费(201001028);江苏省六大人才高峰计划(08-C)

摘　　要：为了克服传统参数识别方法在进行河流水质模型多参数识别时的不足,以一维纵向离散方程为基础,构建了同时确定河流水质模型水力和水质参数的反演优化通用算法。利用反演优化法重点探讨了初值选取、观测噪声、测量精度等因素对参数计算结果的影响,比较了变尺度法和单纯形法两种优化算法的反演效果。算例计算结果表明,单纯形法在可靠性与精度方面均要优于变尺度法。结合瞬时源和连续源两个算例验证了该方法的可靠性。两个算例的计算结果表明,采用反演优化算法对瞬时源河流水质模型和连续源水质模型都能给出较好的反演结果。A general inversion-optimization algorithm was constructed for determination of hydraulic and water quality parameters of a water quality model such as longitudinal dispersion coefficient and cross-sectional average velocity to overcome the difficulties of traditional algorithms in parameters identification.It was used to examine in details the influences of observation noise,instrument precision and initial value on estimation of parameters.A comparison was made between BFGS quasi-Newton algorithm and Nelder-Mead simplex algorithm via case studies and the results indicate superiority of the former in stability and accuracy.The new algorithm gives good accuracy of inversion results in the water quality models of instantaneous source and successive source.

关 键 词：水资源 参数识别 水质模型 反演 优化 

分 类 号：TV122[水利工程—水文学及水资源] X502[环境科学与工程—环境工程]