基于字典学习的稀疏约束型数据同化海洋污染预报研究  

Research on Sparse Constrained Data Assimilation for Marine Pollution Prediction Based on Dictionary Learning

作　　者：任燕 徐洪蕾 苏轼鹏[3] 杜振彩[4] Ren Yan;Xu Honglei;Su Shipeng;Du Zhencai(92493,51 st Detachment,Huludao 125000,China;91208,Qingdao 266000,China;Dalian Naval Academy,Dalian 116018,China;Institute of Atmospheric Physics,Chinese Academy of Sciences,Beijing 100029,China)

机构地区：[1]92493部队51分队,辽宁葫芦岛125000 [2]91208部队,山东青岛266000 [3]海军大连舰艇学院,辽宁大连116018 [4]中国科学院大气物理研究所,北京100029

出　　处：《环境科学与管理》2025年第2期62-67,共6页Environmental Science and Management

摘　　要：研究采用数据驱动紧框架和结构字典稀疏约束型变分同化模型进行海洋流场四维重建,提高参数重现精度和算法收敛速率。引入正则化方法处理不确定性,考虑正则化项、数据精度、算法和观测特性等因素。通过基于涡量结构字典的稀疏约束数据同化模型,实现二维流场准确重构。利用稀疏约束正则化操作,平衡数据拟合和稀疏惩罚,通过优化算法提高解的稀疏性,保留流场结构。数值模拟验证了方法在时空域缺失和混合结构观测数据仿真的有效性,提高流场重建速度和改善伪影现象。This study adopts a data-driven tight framework and a structural dictionary sparse constrained variational assimilation model for four-dimensional reconstruction of ocean flow fields,improving parameter reproduction accuracy and algorithm convergence rate.Introducing regularization methods to handle uncertainty,taking into account factors such as regularization terms,data accuracy,algorithms,and observation characteristics.By using a sparse constrained data assimilation model based on a vortex structure dictionary,accurate reconstruction of two-dimensional flow fields can be achieved.Utilizing sparse constraint regularization operations to balance data fitting and sparse penalty,improving the sparsity of the solution through optimization algorithms,and preserving the flow field structure.The numerical simulation has verified the effectiveness of the method in simulating missing and mixed structure observation data in the spatiotemporal domain,improving the reconstruction speed of the flow field and improving the phenomenon of artifacts.

关 键 词：字典学习 稀疏约束 海洋污染 二维海表流场重建 

分 类 号：X55[环境科学与工程—环境工程]