頁籤選單縮合
題 名 | 倒傳遞類神經網路在波浪時序列預報之應用=Forecasting of Wave Time Series Using Back-Propagation Neural Network |
---|---|
作 者 | 蔡清標; 李宗霖; 朱良瀚; | 書刊名 | 中國土木水利工程學刊 |
卷 期 | 11:3 1999.09[民88.09] |
頁 次 | 頁589-596 |
分類號 | 443.1 |
關鍵詞 | 倒傳遞類神經網路; 波浪預報; 波浪時序列; Back-propagation neural network; Wave prediction; Wave time series; |
語 文 | 中文(Chinese) |
中文摘要 | 本文係藉由倒傳遞類神經網路建立波浪預報模式,預測單一測站波浪時序變化 特性。文中先以 Bretschneider 波譜之試驗波浪時序列,探討倒傳遞類神經網路參數對網 路結構收斂性之影響,再以臺中港及高雄 LNG 港之波浪實測資料進行實例操作,驗證模式 之可行性;結果顯示,以單一月份短時期的歷史資料作為網路訓練學習,即可相當準確地預 測未來數個月的波高變化。颱風波浪之預報,則以使用有颱風經過的月份做訓練學習時,可 獲得較佳之預測結果。 |
英文摘要 | An artificial neural network was established for forecasting a time series of waves in this paper. The back-propagation procedure for minimizing the error of predictive output was used in the learning process of the neural network. A wave time series with Bretschneider wave spectrum type performed in the laboratory was first adopted in this study to optimize the structure of the network model. The field wave data measured at Taichung Harbor and Kaoshiung LNG Port were then used to verify the accur cy of the model. Results showed that the prediction model performed well when very short-term observed data were used as a training set. Good performance was also found for the storm-wave prediction when measured storm-wave data were used in the learning process. |
本系統中英文摘要資訊取自各篇刊載內容。