2004-08-012024-05-16https://scholars.lib.ntu.edu.tw/handle/123456789/667476摘要：類神經網路為晚近崛起的控制理論，由於其具有多變性的數學結構，可用來模擬難以物理方程式描述的複雜非線性關係。因此近年來已經廣被應用於水文領域中，此種具有描述高度非線性及不確定性的模式亦逐漸被研究學者所接受並廣泛應用。關於類神經網路之研究，已經有多種形型式之網路架構及學習方法被提出，在眾多類神經網路中，又以倒傳遞類神經網路(back-propagation neural network)最廣為被熟知及應用。然而，倒傳遞類神經網路於訓練過程中會有許多問題。首先，它容易收斂到區域最小值。再者，它即使使用相同之訓練資料，亦會產生出不同之結果。此外，網路學習速度緩慢以及隱藏層結構不易訂定也是其主要缺點。由於輻狀基底函數網路(radial basis function network，簡稱RBFN)具有網路建構容易、學習速度快速及對外在環境具有快速之適應性等優點。有鑑於此，本研究嘗試針對RBFN 於水文系統中之適用性加以探討研究。本研究將分三年依序進行，敘述如下：第一年嘗試應用全面監督式訓練法則來建立RBFN之網路架構，作為洪水流量之預報模式；第二年則嘗試以RBFN 建立一個結合RBFN 與半變異元(semivariogram)理論之空間內插推估模式；第三年將結合RBFN 與自組織映射圖(self-organizing map)建立時間預測模式。上述之研究成果將能增加洪水預報結果之準確度、空間推估時之精確度以及改進時間序列預測上之效能。<br> Abstract: Within the last decade, artificial neural networks (ANNs) have experienced a huge resurgence due to the development of more sophisticated algorithms and the emergence of powerful computation tools. Extensive research has been devoted to investigating the potential of ANNs as computational tools that acquire, represent, and compute a mapping from one multivariate input space to another. Generally speaking, ANNs are information processing systems. Bypassing the model construction and parameter estimation phases adopted by most of the conventional techniques, ANNs can automatically develop a forecasting model through a simple process of the historic data. Such a training process enables the neural system to capture the complex and nonlinear relationships that are not easily analyzed by using conventional methods. Based on the structure of ANNs and the learning algorithm, various neural network models are frequently proposed to solve time series problems. The back-propagation neural network is the popular one. However, the back-propagation algorithm has several serious training problems. First, it tends to yield local optimal solutions. Second, it may produce different results after training process even when the same training data are used. Finally, its training rate is slow especially when the amount of training data is large. The architecture and training algorithms for radial basis function network (RBFN) are simple and clear, and the RBFN train more quickly than multiple layered perceptron networks. Therefore, RBFN is used instead of back-propagation algorithm in this project. The objective of the project is to apply RBFN in hydrosystem, and modify the learning algorithms. The project will be performed in three years. In the first year, the RBFN trained by the fully supervised learning algorithm is used to construct a rainfall-runoff model. In the second year, based on the combination of the RBFN and the semivariogram, a new spatial interpolation method is proposed. In the third year, based on the combination of the RBFN and the Self-Organizing Map (SOM), a time-series forecasting model is proposed. These models can improve the accuracy of flood forecasting, interpolation, and time series forecasting.輻狀基底函數網路全面監督式訓練法則水文系統半變異元自組織映射圖radial basis function networkfully supervised learning algorithmhydrosystemsemivariogramself-organizing map