Prediction of Monthly Discharge and Typhoon Rainfall based on Hilbert-Huang Transform and ELM

• Main Authors: Chiao-Hung Wang, 王僑宏
• Other Authors: Fu-Kuo Huang
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/81905251233117022026

碩士 === 淡江大學 === 水資源及環境工程學系碩士班 === 101 === Water is essential to life, but the water resource of Taiwan is limited and hard to retain for most of the rivers run from high mountains in short and steep courses. In addition, the temporal and spatial distribution of rainfall is very uneven. How to allocate the water resources rationally becomes an important issue and a better prediction method with a higher accuracy is necessary in response to global climate change. In this study, several hybrid prediction models are employed based on autoregressive-integrated-moving average (ARIMA、SARIMA), empirical mode decomposition(EMD) of Hilbert-Huang transform (HHT), and extreme learning machine(ELM). Two case studies are presented according to the data of the upstream monthly riverflow of Shihmen Reservoir in Tahan River Basin and the rainfall of Laonong River Basin during typhoon Morakot and Herb. It is shown that for long-time scale discharge, ELM model has the better performance of prediction. However, HHT_ELM is superior to ELM at the section of the larger data values. On the other hand, for short-time scale rainfall of typhoon, ELM and HHT_extension models behave well in view of sequence characteristics of rainfall.