區域型淹水警示技術研發應用-以北部區域為例

機構典藏 > College of Engineering > Graduate Institute & Department of Water Resources and Environmental Engineering > Research Paper > Item 987654321/76134

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/76134

Title:	區域型淹水警示技術研發應用-以北部區域為例
Other Titles:	The development of regional flood inundation warning system - a study case of North Taiwan
Authors:	張麗秋
Contributors:	淡江大學水資源及環境工程學系
Keywords:	類神經網路;聚類分析;區域淹水預測模式;artificial neural networks;clustering analysis;regional flood inundation estimation model
Date:	2011
Issue Date:	2012-05-02 10:10:20 (UTC+8)
Abstract:	近年來水文極端事件發生情形明顯偏多，所造成的災害範圍與程度也都遠較以往來的劇烈，96年柯羅莎颱風與97年辛樂克颱風，其所挾帶的豪大雨量皆分別對台灣北部地區造成重創，故強化防災體系之應變措施與建置淹水預警機制，有助於水患災情防救之緊急應變，期以達到減少洪災損失之效。本計畫以北部地區為研究區域，包括基隆市、台北縣市、桃園縣等地區，以二維水理模式模擬之淹水歷程資料做為區域淹水資料，進行區域淹水特性分析，建置區域型淹水深預測模式，可提供颱風暴雨時期進行即時線上預測區域淹水情況，以延伸淹水潛勢圖應用範圍，並提升淹水潛勢資料在災害防救計畫之參考價值，有助於防災單位掌握區域之淹水程度及範圍。本計畫以聚類分析為基礎之複合型淹水模式，分別在基隆市、台北縣市、桃園縣建置區域型淹水預警模式。模式建構過程分為資料前處理階段與模式建置階段；前者以K-means聚類分析將不同淹水特性之區域範圍進行分類，並求得各分群之控制點作為該分群之淹水特性指標；後者則分別建置控制點之淹水預測模式、線性網格點之線性迴歸淹水預測模式、非線性網格點之倒傳遞類神經網路多點淹水預測模式。模式建置完成後，經由資料庫設計、自動化作業模組、網頁展示介面，搭配Google Map API模組，可建立自動化即時淹水預報系統，再搭配Google Map豐富的地理資訊圖層，展示各鄉鎮、村里未來1~3小時之最大淹水深與發生位置，具體地呈現本計畫所建置淹水預測模式可即時淹水預測之時效性。最後，將研究成果彙整投稿於國際期刊。 In recent years, extreme hydrological events obviously increase in frequency and accompany severe disasters. The 2007 typhoon Krosa 2008 typhoon Sinlaku struck North Taiwan with stunning rainfall that caused serious flood disasters. Therefore, strengthening the disaster defense systems and establishing the flood inundation forecasting systems will be helpful to deal with some contingencies and emergencies, then alleviate the risk and losses of the disasters. The study areas of this project include Keelung City, New Taipei City, Taipei City and Taoyuan County. In each study area, the characteristics of regional flood inundation will be analyzed, and then the regional flood inundation depth forecasting model will be built based on the flood inundation potential database. Hence, the forecasting model can provide the forecasting information, including the flood depth and extents, to the disaster defense departments during storm periods. In this project, we propose a clustering-based hybrid inundation model to construct four individual regional flood inundation forecasting models for these four cities/counties. The procedure of this approach is divided into two stages: data preprocessing and model building. In the data preprocessing stage, K-means clustering is used to categorize the flooding area of the different flooding characteristics in the study area and to determine a control point from individual flooding cluster(s). In the model building stage, three types of models were built in each group: a single-grid back-propagation neural network (BPNN) for each control point, the linear regression models for the linear grids and a multi-grid BPNN for the nonlinear grids. Moreover, this project will integrate the regional flood inundation forecasting models and Google Maps in an on-line forecasting system to display the inundation coordinates and depths in next 1~3 hours of villages and towns in these four areas. Finally, we would submit this methodology to the international journals.
Appears in Collections:	[Graduate Institute & Department of Water Resources and Environmental Engineering] Research Paper

Files in This Item:

There are no files associated with this item.

Loading...