全球电离层模型的OpenMP多线程并行解算

Multithreaded Parallel Estimation for Global Ionospheric Model by Using OpenMP

探讨了OpenMP多线程技术在全球电离层建模中的应用。在日固地磁参考系下采用15阶次的球谐展开建立全球电离层模型,并对1天解、3天解两种方案的结果与IGS电离层产品进行了对比,电离层图偏差的均方根约3～5TECU,且3天解的方案首尾两组电离层图与IGS产品符合得更好;卫星差分码偏差和接收机差分码偏差与IGS的差异分别约为0.2ns和2ns,仅有少数几个接收机差分码偏差在少数几天与IGS差异较大,超过3～4ns。实验中使用Dell服务器R730（配置：128GB内存、2个CPU、8个核心和32个线程数）,采用OpenMP多线程并行计算能够明显提高全球电离层模型的建模效率,单天解算仅需约7 min,3天解算需约22min,效率提升近8倍。使用3d观测数据并采用OpenMP多线程并行计算来建立全球电离层模型可有效节省建模时间,同时还能提高首尾两组模型系数的精度以进一步提升全球电离层模型的精度,对建模算法的测试、电离层产品的快速发布以及模型后续检验和预测等带来了便利,也为后续实现利用多卫星导航系统观测数据快速建立全球电离层模型提供了参考。

The application of multithreaded parallel technology using OpenMP for global ionospheric modeling is discussed.The vertical total electron content（TEC）is modeled in a solar-geomagnetic reference frame using a spherical harmonics expansion up to degree and order 15.Comparison results between 1-day,3-day solutions in this paper and IGS products are investigated.Root mean square of difference of TEC maps is approximately 3-5 TECU.There is a better agreement between the first and the last TEC maps of a 3-day solution and IGS products compared with those of 1-day solution.The difference in differential code bias（DCB）for satellites and receivers,comparing our solutions and IGS products were around 0.2 ns and 2 ns,respectively.Only a few DCB values for receivers on a few days are beyond 3-4 ns.In this study,a Dell R730 with 2 CPUs,8 cores,32 threads,and 128 GB memory was used for testing ionosphere modeling.Results show that multithreaded parallel computing using OpenMP promotes the efficiency of global ionospheric modeling.The time consumption of 1-day solution and 3-day solution are approximately 7 minutes and 22 minutes,respectively.The efficiency of modeling is promoted by nearly eight times.Thus,multithreaded parallel technology with 3-day solution could not only save time costs efficiently,but also improve the precision of the ionosphere model.It also brings convenience for modeling algorithm testing,fast release of ionosphere products,and verification and prediction of products.This technology provides a reference value for fast modeling of the global ionosphere using multi-GNSS measurements.