本研究旨在探討反傳遞模糊類神經網路(CFNN)來建立抽水站抽水機組操作模式之可行性;CFNN原理係模擬人類依經驗法則的學習及判斷方式建立-網路演算架構,網路架構包括一層輸入層、一層隱藏層及一個輸出層,在學習階段首先將輸入的資料依資料點相似程度自動予以分類,架構其隱藏層,在輸出層方面則依輸入訊息的對應輸出值與網路訓練輸出值所得比較,逐步調整網路輸出層的連結權重,以獲得最佳的描述結果。 本研究以台北市玉成抽水站為研究區域,歷年颱風暴雨事件中,抽水站在不同降雨量、內水位及閘門操作下之抽水機組操作資訊,以CFNN類神經網路進行訓練學習,再將其應用於抽水機組操作推估上,經實例比較驗證,確實獲得合理的結果,顯示CFNN類神經網路應用於抽水站抽水機組操作有相當優越的能力。
This study uses counterpropagation fuzzy-neural network (CFNN) to build pumping control model. Based on a rule-base control, a modified self-organizing counterpropagation network, and a fuzzy control predictor, the CFNN can automatically generate rules by increasing the training data to improve its accuracy. The network is developed as a three-layer network, consisting of an input layer, a Kohonen (hidden) layer, and a Grossberg (output) layer. The network is trained to build a pumping operation model of the Yu-Cheng pumping station in Taipei. The historical records of several flood events are used to train and verify the proposed model. The results demonstrate that the network has a great ability to control the pumping station and provide a human-like operation mechanism.