J4 ›› 2005, Vol. 23 ›› Issue (4): 1-6.

• 论文 •    下一篇

气象要素时间序列的演化建模分析与短期气候预测

俞康庆1周月华1杨荆安1康卓2   

  1. 1.中国气象局武汉暴雨研究所,湖北武汉430074; 2.武汉大学计算中心,湖北武汉430072
  • 收稿日期:2005-09-30 修回日期:2005-11-07 出版日期:2005-12-31 发布日期:2005-12-31
  • 作者简介:俞康庆(1944),女,浙江镇海县人,正研级高级工程师,主要从事数值预报模式研究.E-mail: yukq@whihr.com.cn
  • 基金资助:

    国家自然科学基金项目(4207504)资助

The Evolutionary Modeling and Short一range Climatic Prediction  for Meteorological Element Time Series

SHU Kang-Qiang1, ZHOU Ru-Hua1, YANG Jing-An1, KANG Zhuo2   

  1. 1.  Institute of Heavy Rain  China Meteorological Administration  Wuhan 430074,China;
              2.  Computation Center  Wuhan University  Wuhan 430072  China
  • Received:2005-09-30 Revised:2005-11-07 Online:2005-12-31 Published:2005-12-31

摘要:

以武汉站(5 -9月)汛期降水量观测数据序列为例,将这类具有明显的不规则性(混沌特征)时间序列分解为宏观气候尺度周期的波动部分和迭加其上的微观气候尺度周期的波动部分,分别采用演化建模方法和自然基小波方法模拟逼近。特别强调由演化建模方法得到的非线性常微分方程较之传统的线性建模具有更好的分析预测能力。

关键词: 时间序列, 演化建模, 短期气候预测

Abstract:

The evolutionary modeling(EM),which is developed from genetic; programming(CP),is a relatively new technique that is an adaptive method for solving(omputational problems in(omplex systems which are of(haoti<(harac;ter and nonlinear variation with tlme  In  many  flelds. The  use  Of   EM  f01' the  Obsel'Ved  tlme  sel'les  Of  pl'eelpltatlOn  In flOOd  season(May一September) at Wuhan station is studled  In  thls  papel'.  The  tlme   sel'les  Of pl'eelpltatlOn is  split lnt0  tW0 pal'ts:  One lneludes  mael'O(limatic; timesc;ale period waves that are affected by some relatrve steady、limatic; factors such as astr'onomic;al factors(like sunspot number et<)as well as other factors that are known and unknown } the other includes micro(hmatle  tlmeseale pel'lOd  W3VPS  supel'1mpOSed  On  the mael'O One.  The  eVOlutlOnal'y  mOdeling(EM)is  suppOSed  t0  be  adept  at  slmulatlng the f01'mel'  pal't  because  It、reates the nonlinear ordinary diferential equation(NODE) based neon the observed trine series.  The natural frac;tals(NE}  al'e used t0  slmulate the lattel' pal't.  The flnal pl'edletlOn is  the  sum Of values from both methods } so the model(an reflects multi一time、(ale satisfactory for(limatic; prediction operation.  The NODE(an suggest the data vary with time which is benefit to think over(limatic; analysis and short一range(limatic; prediction.  Comparison in principle between EM arid a lineal' modeling ARC p) mdn;ates that the EM is a much better method to simulate the、OmpleX  time  sel'les being of nonlinear(harac;teristi<、

Key words: evolutionary modeling , NODE, time series, short一range(limatic, prediction

中图分类号: