An integrated decision method for prediction of tropical cyclone movement by using genetic algorithm

摘要

In this study an integrated decision method for the multi-prediction results of tropical cyclone movements by using genetic algorithm (GA) has been adopted. A genetic learning artificial neural network (GLANN) as a learning mechanism for track prediction was developed through training with the huge samples of typhoon tracks from 1884 to 2002. The prediction of typhoon motion is to consider how to suit the dynamic change of environment, where typhoon passed through. The neural network structure is described by GA. A fitness function for choice of GA is given to approach the maximum of fitness (the minimum of the errors). Parallel comparing analyses for prediction abilities of unusual typhoon tracks in recent years by GLANN with multi-forecast methods were conducted. More types of cases were employed in the comparing analyses, which consist of the typhoon, Winnie, No.11, in 1997, characterized by a long life over the ocean and in the mainland with severe effect on China, the 10 unusual typhoon tracks in the 1990s, the tracks with landfall and affected action in 1998 and the typhoons born in the Northwestern Pacific and the South China Sea in 2000, including the 24—48 h forecast, etc. The results indicated that for prediction ability of typhoon tracks, specially for unusual typhoons, GA decision-making developed in this study is superior to the others.