張斐章陳彥璋梁晉銘Chang, Fi-JohnFi-JohnChangChen, Yen-ChangYen-ChangChenLiang, Jin-MingJin-MingLiang2009-02-232018-06-292009-02-232018-06-292001-12http://ntur.lib.ntu.edu.tw//handle/246246/139282感潮河段內河川水位受潮汐及諸多水文因子影響，難以利用一般數學或物理模式對其水位進行預測。本研究藉由類神經網路之強大學習功能，以淡水河系感潮河段之水位資料建構臺北橋站之水位預測模式，並用於預測未來一小時之水位。首先網路以團塊分類演算法預先將訓練範側依性質相近程度進行聚類，使歷史資料之規律性得到區隔並降低複雜度；其次，對非線性基底函數作簡易之線性組合運算，使重現機率較低之極端事件亦能得到適當的權重推論。 為求對模式建構成效作檢測，本研究取近年來淡水河感潮河段內多個水位站時序列為背景資料，經參數檢定、模式驗證與適用性測驗三階段之模式模擬預測，結果顯示模式能對水位漲落之趨勢掌握的相當良好，足見以類神經網路模式對感潮河段水位進行預報模式之建構具有實用之潛力。To forecast the water stage in open-channel under tidal effects is always a tough task. Even those sophisticatedly conceptual and mathematical models cannot do a good job. In this study we propose an artificial neural network model to forecast the one-hour-ahead water stage. One of the advantages of the artificial neural network is its powerful learning ability. During the training scheme, the training data with same similarities are clustered together at the beginning. Then the least squares method is used to estimate the weights of the model. Thus it can reduces the complexity of the system. The water stage data of the Tanshui River under tidal effects are used to construct a water stage forecasting model. The data is split into three independent subsets, namely, the training, validation, and testing subsets. The training subset is used for parameter estimation and model development. The validation subset is applied to choose the best model from the candidate ones. The testing subset is devoted to show the performance of the selected model. The results show that the artificial neural network model is a reliable and accurate tool for forecasting the water stage in an open-channel under tidal effects.en-US感潮類神經網路水位團塊Tidal effectArtificial neural networkWater stageClusterjournal articlehttp://ntur.lib.ntu.edu.tw/bitstream/246246/139282/1/以類神經網路預測淡水河感潮河段水位.pdf