一个多参数优化系统在简单模型中的应用  

Application of a multi-parameter optimization system in an idealized model

作　　者：祖子清 杨庆[2] 夏江江[2] 张蕴斐[1] 朱学明[1] ZU Zi-qing;YANG Qing;XIA Jiang-jiang;ZHANG Yun-fei;ZHU Xue-ming(Key Laboratory of Marine Hazards Forecasting,National Marine Environmental Forecasting Center,Ministry of Natural Resources,Beijing 100081 China;Key Laboratory of Regional Climate-Environment for Temperate East Asia(RCE-TEA),Institute of Atmospheric Physics,Chinese Academy of Sciences,Beijing 100029 China)

机构地区：[1]国家海洋环境预报中心自然资源部海洋灾害预报技术重点实验室,北京100081 [2]中国科学院东亚区域气候-环境重点实验室,中国科学院大气物理研究所,北京100029

出　　处：《海洋预报》2021年第3期11-18,共8页Marine Forecasts

基　　金：国家重点研发计划(2017YFA0604203);国家自然科学基金(41506040)。

摘　　要：针对数值模式的物理参数,发展了一个多参数优化系统,可以根据观测数据对模式的多个参数进行同时调整。该系统具有设计简单、无需伴随模式和便于移植等特点。在孪生试验中,针对盒子模型的3个参数,经过约10次迭代,多参数优化系统可以收敛到预先给定的参数真值。通过对比同时优化和分别单独优化3个参数增量,发现分别单独优化参数增量存在一定的局限性。将多参数优化系统应用到简单模型的尝试和检验,结果表明:优化算法的收敛速度和模型的积分次数处于可接受的范围内,因此具有同时调整多个复杂模式物理参数的潜力。A multi-parameter optimization system is developed in this study,which can be used to simultaneously adjust numerical model parameters based on observation data.The system is characterized by simple framework and high portability without the need of adjoint model.In the twin experiments,the three parameters of a box model is convergent to the prescribed true values after about ten times of iteration.By comparing the parameter increasement in single-parameter optimization and multi-parameter optimization,the limitation of single parameter optimization is revealed,which possibly yields the wrong directions of parameter adjustment.This is case study to apply the multi-parameter optimization system to a simple numerical model,which shows its encouraging performance in the convergent speed and model iteration times.As a result,the multi-parameter optimization system can be potentially applied to adjust the parameters of complex numerical models.

关 键 词：数据同化 参数估计 模式偏差 多参数优化 

分 类 号：P73[天文地球—海洋科学]