基于数据驱动与遗传计算的海域组合单元水质模型多参数分步耦合优化反演方法研究  

Research on Multi-Parameter Optimal Inversion for Marine Integrated Element Water Quality Model with Multi-Step Coupled Method by Data-Driven Model And Genetic Algorithms

作　　者：李明昌[1,2,3] 张光玉[1,3] 司琦[1] 尤学一[3] 梁书秀[2] 孙昭晨[2] 

机构地区：[1]交通运输部天津水运工程科学研究院水路交通环境保护技术实验室,天津300456 [2]大连理工大学海岸和近海工程国家重点实验室,辽宁大连116024 [3]天津大学环境科学与工程学院,天津300072

出　　处：《数学的实践与认识》2015年第12期167-175,共9页Mathematics in Practice and Theory

基　　金：国家自然科学基金(51209110);天津市科技兴海项目(KJXH2011-17);中央级公益性科研院所基本科研业务费专项资金项目(TKS130215;KJFZJJ2011-01)

摘　　要：考虑水质模型参数时域和地域差异性,建立了模型参数在各单元与时段内独立赋值的海域组合单元水质模型.结合多种反演方式的效率问题,通过数据驱动模型、遗传算法和海域组合单元水质模型的分步耦合,提出了水质模型多参数优化反演的新方法:将数据驱动模型同海域组合单元水质模型有机结合进行初步反演,获得多参数匹配关系;以误差函数为适应度,将海域组合单元水质模型嵌入遗传算法模型中,以多参数匹配关系初值为约束条件,进行多参数精确反演.最后以渤海湾海域组合单元水质模型多参数反演的"孪生"试验验证方法的有效性,数值结果表明分步耦合反演新方法具有较高的精度和效率;组合单元式的参数赋值与反演方式有利于提高模型验证精度.A marine integrated element water quality model is presented due to the difference of model parameters in space and time. A new multi-parameters optimal inversion method is established by the multi-step coupled method of data-driven model, genetic algorithms and marine integrated element water quality model for the efficiency .problem of some inversion means. In this method, firstly the primary inversion is applied for multi-parameter matching relationship by data-driven model and marine integrated element water quality model. Sec- ondly, marine integrated element water quality model is inverted into the genetic algorithms model for multi-parameters precise inversion by the fitness function of error function and con- straint of multi-parameter matching values. Case study in the Bohai Bay by twin experiment and an instance are presented for validating the presented method. Verification results show the multi-step coupled inversion method is efficient, and the integrated element means for assignment and inversion is advantageous to improving the simulation precision.

关 键 词：海域组合单元水质模型 数据驱动模型 遗传算法 多参数匹配 分步耦合 反演 

分 类 号：X55[环境科学与工程—环境工程] TP18[自动化与计算机技术—控制理论与控制工程]