應用ANFIS模式於台灣東部海域之季節風波浪推算

Abstract

本文利用中央氣象局所提供之2002-2003年位於台灣東部海域之測站浮標波浪資料，分別為龍洞、蘇澳及花蓮等三個測站。配合歐洲中尺度預報中心之模式所預報風速，分析季節風與示性波高之特性，以暸解兩者之間的關係，進而應用適應性類神經模糊推論系統，來架構季節風推算波浪之模式。 本文以風速、延時與風向等參數作為架構季節風推算波浪之模式之輸入值，以建立單站模式季節風波浪預報。其單站模式所推測波高與實測值之相關係數約在0.8以上，其推測波高之平均誤差值則為0.3 米以下。並以此單站模式作為基準，應用此模式加以延伸於非學習測站之波浪推算，得到其相關係數約在0.76以上，平均誤差值則為0.3米左右。此結果証實本模式在波浪推算上有不錯之精度，未來可供海岸工程施工、港灣船隻作業參考應用。

Both computed monsoon data and corresponding measured waves data during 2002-2003 at three positions, Long-Dong, Su-Ao and Hua-Lien harbor in the eastern waters of Taiwan, are collected. The paper applies an adaptive network-based fuzzy inference system (ANFIS) to set up a monsoon wave model. The proposed wave model adopting wind velocity calculated by European Centre for Medium Range Weather Forecasts as input parameters can provide good simulating wave heights. Three parameters, wind velocity, 12-hour time-lag effect and wind direction, are used to construct the present wave model. The proposed model has high accuracy of calculating waves by a correlation coefficient over 0.8 and by a mean absolute error less than 0.3m. The proposed wave model is also extended to a point neighboring to the interesting point where the wave data were used to train the model. An example is examined to have a correlation coefficient of 0.76 between the predicted wave heights and observed data. Therefore, this proposed model is applicable for a eference of marine construction and navigating due to fast and accurately simulating waves.