Development of A Typhoon Inundation Warning System by Integrating Clustering and Neural Networks

• Main Authors: Ching-Yun Ou, 歐靚芸
• Other Authors: 林國峰
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/57531159034791286107

碩士 === 國立臺灣大學 === 土木工程學研究所 === 100 === During typhoons, flood inundation caused by rainfall often leads to loss of life and property damage. For inundation mitigation, development of a regional inundation warning system has been recognized as an important task in hydrology. In this study, an accurate and effective regional inundation warning system by integrating k-means clustering and support vector machine (SVM) is proposed. The proposed regional inundation warning system consists of three parts: classification, forecasting and extension. Firstly, the inundation depth hydrographs are clustered by k-means clustering, which is a useful technique for solving classification problems. The inundation depth hydrographs with specific different characteristics are classified and the center of each cluster is seen as a control point in this study. Secondly, the rainfall and inundation depth are used as inputs to develop the SVM-based inundation forecasting model for each control point. Thirdly, the point forecasts resulting from the SVM-based inundation forecasting model are extended to the spatial forecasts by using the SVM-based extension model. The input variables of the SVM-based extension model are the coordinates, the elevation and the rainfall of forecasted point and the forecasting result of the control point. An actual application of the proposed regional inundation warning system in the Xiluo Township is conducted to demonstrate the advantages of the proposed system. The results show that the proposed regional inundation warning system can effectively forecasting the inundation depth, and the proposed regional inundation warning system is expected to be useful to mitigate the inundation damage.