An artificial neural network model for flood forecasting in Kemaman, Terengganu / Tuan Asmaa Tuan Resdi

Flood is the most common natural hazard in Malaysia. Flood hazard brings damage to life and property in Malaysia. This hazard happens almost every year in the eastcoast and the southwest of Peninsular Malaysia. Kemaman district, Terengganu is one of the flood prone area, and was considered in the pr...

詳細記述

保存先:
書誌詳細
第一著者: Tuan Resdi, Tuan Asmaa
フォーマット: 学位論文
言語:English
出版事項: 2016
オンライン・アクセス:http://ir.uitm.edu.my/id/eprint/17847/2/TM_TUAN%20ASMAA%20TUAN%20RESDI%20EC%2016_5.pdf
http://ir.uitm.edu.my/id/eprint/17847/
タグ: タグ追加
タグなし, このレコードへの初めてのタグを付けませんか!
その他の書誌記述
要約:Flood is the most common natural hazard in Malaysia. Flood hazard brings damage to life and property in Malaysia. This hazard happens almost every year in the eastcoast and the southwest of Peninsular Malaysia. Kemaman district, Terengganu is one of the flood prone area, and was considered in the present study. Using historical hourly data of rainfalls, evaporation, temperature, mean relative humidity, tidal and river stage for the year 2009, the performance of Feed Forward Back-Propagation (FFBP), General Regression Neural Network (GRNN), and Radial Basis Function Neural Network (RBFNN) model were evaluated. Results of network training show that RBFNN model performs best. Hydrological variables including temperature, humidity and evaporation are shown to be important in the determination of river stage in the sensitivity study. However, this network model is incapable of reproducing the river stage accurately in the validation stage. In subsequent investigation, it is shown that the Nonlinear Autoregressive Network with Exogenous (NARX) model performs satisfactory in both the training and validation stages. Using representative set of hourly data, with optimal time delay for both the input and output, it is shown that the model with 13 hydrological inputs variables performs slightly better compared to a model which takes into consideration the tidal data. For one-step ahead prediction, the model performs satisfactorily for simultaneous hydrological simulations at multiple gauging stations.