類神經網路模式推測侵台颱風路徑之研究

Forecasting of Typhoon Path Raided in Taiwan by Using Neural Network Models

吳信宏(Shin-Hung Wu)

指導教授 ： 柯亭帆

國立屏東科技大學/工學院/土木工程系所/碩士(2014年)

https://doi.org/10.6346/NPUST.2014.00044

摘要

台灣位於亞熱帶，每年約有3至4個颱風侵襲而來，往往會造成台灣本島不少損失，若能預測颱風的路徑，必能減少災害和損失的發生。本研究旨在利用類神經網路建立颱風路徑預測模式，進一步而言，本研究蒐集中央氣象局所提供的西元1972~2012年間十個路徑類型合計100筆侵台颱風資料，利用靜態和動態兩種類神經網路模式，每一類路徑建立一個模式共20個模式。兩種網路模式共同採用經、緯度、暴風半徑、中心風速和移動速度等5個輸入參數和3個時間延遲數為輸入參數，經訓練、驗證及測試階段建立了可供不同類型颱風路徑之預測模式。從研究結果顯示類神經網路模式對第一類到第九類颱風路徑之預測均有良好的表現，其相關係數平方值都在0.8以上，而第十類颱風屬特殊類路徑，則可能因紀錄資料過少，無法建立可信賴之模型。整體而言，本研究採用方法對颱風路徑之預測則有些許創新性，研究結果可供後續研究之基礎並可提供相關單位之參考。

關鍵字

颱風路徑 ； 靜態類神經網路 ； 動態類神經網路 ； 預測模式

並列摘要

Due to its location in typhoon prone area, Taiwan is facing annually 3-4 typhoon events that cause many damages and losses for the country. Forecasting typhoon occurrence is highly desired for reducing negative effects of this extreme event. Therefore, this study aimed to predict typhoon path using artificial neural network that is widely recognized as suitable and effective tool for weather prediction. One hundred data (from 1972 to 2012) obtained from the central weather bureau, and both static and dynamic neural networks were employed. Each of both networks were used to develop ten types of model ranged from type 1 to type 10, and five parameters (longitude, latitude, storm radius, typhoon path speed, wind speed), and three delays were involved. From the statistical results, except type 10 model, the coefficient of determination (R2) was higher than 0.8 for the other used models showing that these model were able to predict successfully typhoon events. This study helped for identifying different types of model that may be used to predict typhoon occurrence, but it is suggested the use of large number of data for reliable forecasting.

並列關鍵字

Typhoon Path ； Static neural network ； Dynamic neural network ； Forecast Model

參考文獻

4.周昆炫，黃柏智，2010，「渦旋植入對不同降水物理參數法颱風路徑系集預報之影響研究」，大氣科學，第38 卷 4 期，第301 – 329。

7.黃清勇，羅璋盛，2006，「颱風路徑預報誤差EOF分析」，大氣科學，第 34卷，第4期，第277 – 290頁。

12.C.M. Tseng, C.D. Janb, J.S. Wang, C.M. Wang, 2007, “Application of artificial neural networks in typhoon surge forecasting”, Ocean Engineering, Volume 34, Issues 11–12, pp. 1757-1768.

13.Chun-Chieh Wu, Kun-Hsuan Chou, and Po-HsiungLin , 2007 ,“The Impact of Dropwindsonde Data on Typhoon Track Forecasts in DOTSTAR”, Weather And Forecasting, Volume 22, pp. 1157–1176.

14.Fi-John Chang, Yen-Ming Chiang, Wei-Guo Cheng, 2013, “Self-organizing radial basis neural network for predicting typhoon-induced losses to rice”, Paddy and Water Environment, Volume 11, Issue 1-4, pp 369-380.