Intelligent optimization algorithms such as particle swarm optimization (PSO) have been introduced into four-dimensional variational assimilation of atmospheric data to solve complex optimization problems. The time-varying double compression model can solve the problem of accuracy well. But when confronted with the problem of high accuracy, the long training time will become the weakness. Parallelization acceleration is one of the effective ways to solve the conundrum. And applying Graphic Processing Unit (GPU) to accelerate PSO algorithm in parallel has the advantage of low hardware cost. In this paper, a parallel time-varying double compression factor PSO algorithm based on GPU acceleration is proposed. The parallel operation of particle swarm optimization algorithm is carried out by GPU, in which the time can be improved with the same precision kept. Compared with the dynamic inertia weight algorithm and time-varying double compression factor algorithm, the experimental results display that the accuracy is better than the former and the consuming time is shorter than the latter, which proves that the method can process the prediction in a faster and more accurate way.
CITATION STYLE
Chen, K., Liu, Y., Liu, L., Yu, Y., Dong, Y., & Tong, Y. (2020). Research on Atmospheric Data Assimilation Algorithm Based on Parallel Time-Varying Dual Compression Factor Particle Swarm Optimization Algorithm with GPU Acceleration. In Communications in Computer and Information Science (Vol. 1205 CCIS, pp. 87–96). Springer. https://doi.org/10.1007/978-981-15-5577-0_7
Mendeley helps you to discover research relevant for your work.