本研究研發一個網路雨水下水道水位預警系統(ANFIS for sewer stage forecasting system,簡稱ASFS系統)。以適應性模糊類神經網路系統(adaptive-network-based fuzzy inference system, ANFIS)建構都市雨水下水道水位站水位預警模式(ANFIS for sewer stage forecasting model,簡稱ASF模式),並採用Web GIS架構作為預測結果展示之平台,建立ASFS系統。ASFS系統以臺北市第四分區中港抽水站系統作為研究區域,使用2004~2006年颱風暴雨事件每分鐘監測水位紀錄作為建構ASF模式之資料。研究中探討雨量與雨水下水道人孔水位的相關性,以及在不同超前預測時距(如提前5、10、60分鐘)下ASF模式預測結果的精確性。經由模擬結果顯示,雨量對於雨水下水道人孔水位的相關性較低,不適合作為雨水下水道人孔水位模擬預測之輸入資料,而ASF模式在超前預測5分鐘與10分鐘有較佳之預測效果。
The study develops a web-based sewer stage forecasting and warning system (ASFS) in which an adaptive-network-based fuzzy inference system (ANFIS) is used to construct a sewer stage forecasting model (ASF); web-GIS (geographic information system) technology is then integrated with ASF to display the water stage forecasted by ASF in real time. The Jhong-Gang drainage system of the fourth drainage area of Taipei is employed as a case area and minute records of water stages of storm and typhoon events from 2004 to 2006 are used to build an ASF model. This research investigates the correlation between rainfall and water stage and the accuracy of predicted water stage by ASF model within different forecast periods (5 minutes, 10 minutes and 1 hour, respectively). The simulated results show that rainfall data is not appropriate to use as inputs to the ASF model due to weak correlations between rainfall data and forecasted water stages; ASF provides excellent forecasts for leading 5-minute and 10-minute water stages as well.