摘要
针对线性预测方法难以有效描述云团的非线性、非平稳变化的困难,基于经验正交函数分解(EOF)和遗传算法参数优化结合的思想,提出了一条云团非线性预测模型反演的方法途径。首先将卫星云图序列作EOF的时、空分解;在此基础上,引入遗传算法对EOF的时间系数序列进行了动力模型重构和模型参数反演,建立了EOF时间系数的非线性微分方程组;再通过时、空函数合成,构造了云团演变的动力预报模型。试验结果表明,反演的云团预报模型能较为合理地描述特定季节区域内云团演变的基本趋势,预测结果与实际云图的主要特征基本相符,尤其是实现了云图3h以上的中、长时效的客观预测。
Due to the fact that linear prediction method is difficult to describe the nonlinear, non-stationary changes of cloud clusters, a technique of retrieval nonlinear clouds clusters forecast model, based on the idea of combining the decomposition of empirical orthogonal function (EOF) and the genetic algorithm optimization parameters, was presented. Firstly, satellite image sequences were temporal-spatially decomposed by EOF. On this basis, genetic algorithms were introduced to make the dynamic model reconstruction and model parameters optimization retrieval of EOF time coefficients sequence, and a nonlinear differential equations of EOF time coefficients were established. Then, by the EOF temporal-spatial functions synthesis, a dynamic forecast model of cloud clusters evolution was structured. The experimental results showed that the retrieved clouds dynamic forecast model was more reasonable in describing the cloud evolution of the underlying trend in particular seasons and region, and the forecast results were better accorded with the basic characteristics of actual satellite cloud pictures. Especially, a middle-long period over three hours objective cloud clusters predictions was achieved.
出处
《国防科技大学学报》
EI
CAS
CSCD
北大核心
2007年第5期41-47,共7页
Journal of National University of Defense Technology
基金
国家部委资助项目
关键词
卫星云图
云团预测
正交分解
遗传算法
参数优化
satellite cloud pictures
cloud clusters forecast
empirical orthogonal function
genetic algorithms
parameter optimization
