Ocean data assimilation method and system based on high-performance parallel optimization

Abstract

The invention provides an ocean data assimilation method and system based on high-performance parallel optimization, and the method comprises the steps: obtaining to-be-assimilated ocean data and a mathematical model, calculating the data assimilation complexity of each ocean lattice point according to background field data and an ocean lattice point distribution diagram in the ocean data, and calculating the data assimilation complexity of each ocean lattice point; generating a calculation topological graph based on data assimilation complexity; according to a preset longitude range, grid points in the ocean grid point distribution diagram are grouped, the overall assimilation complexity corresponding to each group is counted according to the calculation topological diagram, a plurality of calculation nodes are distributed for each group in a balanced mode, and the calculation amount responsible for each calculation node is divided into calculation processes in the calculation nodes in a balanced mode; and after the calculation processes in the calculation nodes complete respective data assimilation tasks, assimilation result data is obtained, and the assimilation result data is written back to the mathematical model to serve as an ocean data assimilation result. According to the method, the load balancing of each process in the distributed computing system is realized by taking the computing topological graph as a basis for load balancing division.

Description

technical field [0001] The invention relates to the technical fields of high-performance computing and marine data assimilation, and in particular to a method and system for marine data assimilation based on high-performance parallel optimization. Background technique [0002] Ocean data assimilation is a prediction and correction method developed from the field of climate numerical prediction, which combines physical models with observational data, and is widely used in many fields such as atmosphere, ocean and surface. At present, the most common data assimilation methods are the ensemble Kalman filter algorithm (FnKF), the ensemble optimal interpolation algorithm, and the three-dimensional and four-dimensional variational methods. The ensemble-optimal interpolation is developed from the ensemble Kalman filter. [0003] Since ocean data assimilation uses high-resolution ocean observation data, there are problems such as large amount of computation, high memory requirement...

AI Technical Summary

Problems solved by technology

When the program in the prior art is tested on a scale of more than 1,000 cores, when the number of cores is greater than 2,000, the overall running time does not decrease significantly or even increases as the number of cores increases.

Images