Please use this identifier to cite or link to this item:
標題: 類神經網路在兩測站間波浪資料互補之應用
Application of Neural Network for Wave Data Complement between Two Recording Stations
作者: 楊志斌
Yang, Chin-Pin
關鍵字: 類神經網路
Artificial Neural Network
summation function
activity function
transfer function
Back-propagation Neural Network
root mean squared error
correlation coefficient
出版社: 土木工程學系
摘要: 摘要 本文旨在以倒傳遞類神經網路,建立同一風場相鄰海域兩測站間一簡便可行的波浪推算模式,並加以應用於波浪資料之互補上。因往昔模式在單一測站用之前發生過的波浪資料,推測之後的波浪資料,其精度較不理想;故本文乃嘗試利用兩測站在同一風場所受影響相似之因素,依倒傳遞類神經網路模式將兩測站的波浪資料分別設為網路中的輸出入層,經由倒傳遞類神經網路依波浪特性學習而擬合出一簡便可行的波浪推算模式,並可應用於測站間資料之互補。本研究以二個波浪觀測站,一以基隆港內測站,另一以基隆港外相鄰海域鼻頭角測站之波浪實測資料分成多個範例進行實例操作,以驗證模式之精度。研究顯示,不論是西南季風時期或東北季風時期,只要用兩測站之同季風時期45天示性波高資料,藉由倒傳遞類神經網路訓練出一組權重值及閥值,利用該值就可準確地進行另一測站的波浪。本研究模式亦可應用在颱風時期兩測站示性波高的互補上。
ABSTRACT Accurate prediction for the wave climate is an essential part in the ocean engineering. The precision of forecasting the time series of wave data using the past wave records in the same station is not good in general. Thus this paper attempts to apply the back-propagation neural network (BPN) for the complement of wave data between two recording stations under the same condition of wind field. The model does not only be able to forecast wave, but also be used in supplementing the wave data. The field data used in the testing model were measured in two stations, one is in Keelung Harbor and the other is in Bitoujiao. The performance of the neural network model was first discussed by using two indices, root-mean square error and the correlation coefficient. It is found that the neural network model performs well for the wave complementary when a 45- days wave record was is used in the training process of back-propagation neural network for the situation of season wind. For typhoon waves, it is also found that the neural network model could also be applied well in the complementary of significant wave heights.
Appears in Collections:土木工程學系所



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.