基于简单模式的集合卡尔曼滤波与粒子滤波的比较研究

发布时间：2018-04-27 21:18

  本文选题：大气环境学 + 数据同化　； 参考：《热带气象学报》2017年05期

【摘要】：集合卡尔曼滤波和粒子滤波是大气海洋领域两种先进的数据同化方法。理论上讲,粒子滤波克服了集合卡尔曼滤波中先验分布的高斯假定。但现有的关于两种方法的比较研究不够全面和系统,基于简单的洛伦兹63模式,重点对基于确定性集合卡尔曼滤波和均权重粒子滤波的数据同化方法开展对比分析,通过对观测误差和模式误差的不同配置,设计了四组试验着重研究两种方法相同试验条件下的同化效果。试验结果表明:与采用最优膨胀系数的集合卡尔曼滤波的同化方法相比,均权重粒子滤波的均方根误差更加依赖于观测信息的质量,但最优膨胀因子的集合卡尔曼滤波的均方根误差低于粒子滤波同化方法。
[Abstract]:Aggregate Kalman filter and particle filter are two advanced data assimilation methods in atmospheric and oceanic domain. Theoretically, particle filtering overcomes Gao Si's assumption of prior distribution in ensemble Kalman filtering. However, the existing comparative studies on the two methods are not comprehensive and systematic. Based on the simple Lorenz 63 model, the data assimilation methods based on deterministic set Kalman filter and average weighted particle filter are compared and analyzed. Through different configuration of observation error and mode error, four groups of experiments are designed to study the assimilation effect under the same experimental conditions. The experimental results show that the root mean square error of average weight particle filter is more dependent on the quality of observation information than the assimilation method using set Kalman filter with optimal expansion coefficient. However, the root mean square error of the set Kalman filter with the optimal expansion factor is lower than that of the particle filter assimilation method.
【作者单位】： 哈尔滨工程大学自动化学院;国家海洋信息中心海洋环境信息保障技术重点实验室;
【基金】：国家重点研发计划(2016YFC1401800) 基于粒子滤波的海气耦合数据同化方法研究(HEUCF041705) 中国南海内波场数字化及其在数字化传播方面的应用(HEUCFP201708)共同资助
【分类号】：P732.6
，

本文编号：1812374