Study on Novel Grey Prediction Modified Model in Flood Control

• Main Authors: CHIU, CHIH-CHIANG, 邱志強
• Other Authors: Yong-Huang Lin
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/30407001998060426514

博士 === 國立臺灣科技大學 === 營建工程系 === 100 === The current global warming has led to dramatic climate changes, such as increased intensity and prolonged time of rainfall, resulting in problems of increased average daily rainfall and overloading of the drainage system. This has increased difficulty in resulted in challenges to rainfall prediction technology. In overview, the rainfall intensity has been increasingly and becoming more concentrated in Taiwan in recent years, posing great impact of major disasters as the maximum daily rainfall, rainfall in typhoon and water reservoir inflow are all above the warning values. In addition, the increase in rainfall and flow beyond the original design loads of the public drainage facilities has caused floods. Hence, building a high precision, real- time rainfall prediction model to accurately and timely predict rainfall has become an important topic for flood control at the present stage. This study attempts to use the grey system characteristics, along with the innovative grey prediction model proposed by Lin (2007), as the basis to propose the modification of novel grey prediction model. The novel grey prediction model is apparently more effective than traditional grey models in terms of prediction of the extreme values and delays of time series. However, it needs improvement in prediction precision and time in case of the prediction of the extreme values above two standard deviations in the time series. This study plans to propose two modification methods: the first is to use the dynamic index method and the innovative grey prediction model to build the modified dynamic grey prediction model; the second method is to combine the innovative grey prediction model with the fuzzy membership function to construct the fuzzy grey prediction modification model to improve the time sequence smoothness and prediction of extreme values with overly large fluctuations. In addition, the methods will be compared with the ANN and ARIMA for matching and analysis. The rationality of the research findings will be verified by the prediction of the annual maximum daily rainfall, the hourly rainfall in typhoon and the reservoir inflow amount.