小波分解与SOFM-BP网络结合的西太平洋副高数值预报产品优化技术  

NUMERICAL FORECAST PRODUCTS OPTIMIZATION OF THE WEST-PACIFIC SUBTROPICAL HIGH BASED ON THE WAVELET DECOMPOSITION AND SOFM-BP ARTIFICAL NEURAL NETWORKS

作　　者：徐海斌[1] 张韧[1] 刘科峰[1] 王彦磊[1] 洪梅[1] 

机构地区：[1]解放军理工大学气象学院海洋与空间环境系

出　　处：《热带气象学报》2007年第3期265-270,共6页Journal of Tropical Meteorology

基　　金：国家自然科学基金项目(40375019);热带海洋气象科学研究基金开放课题(200609)共同资助

摘　　要：基于1995～1997年夏季(5～8月)T106数值预报场资料,研究讨论了夏季西太平洋副热带高压面积指数的预报误差修正与预报优化问题。首先通过小波分解对预报目标进行频域分解和高频滤波,随后引入了人工神经网络BP模型与自组织特征映射网络(SOFM)相结合的方法,对副高指数的数值预报结果进行了预报优化与误差修正的训练建模。试验结果表明,所建模型能够较为客观、有效地修正副高指数的数值预报误差,优化和改进副高预报效果。Based on the T106 numerical forecast model output product （May-August, 995- 1997）, an idea of forecast optimization technique for subtropical high＇s characteristic index and its errors revisal route were presented and discussed in the paper. Firstly, the time series of prediction target was decomposed into different frequency section by using wavelet method and some high frequency signals and noises were filtered then a synthetical technique of combining BP （Back-Propagation Network） lgorithm with SOFM （Self-Organizing Feature Map） model of ANN （Artifical Neural Network） was introduced, and some numerical forecast samples were well optimized and its errors were effectively revised by using the technique, experimental results showed that the methods here was promising in practice.

关 键 词：副热带高压 预报优化 神经网络 小波分解 

分 类 号：P426.62[天文地球—大气科学及气象学]