碩士 / 國立臺灣海洋大學 / 海洋環境資訊學系 / 101 / This study constructed back propagation neural network model for predicting typhoon waves at the Shinchu coast. Information on typhoon passed near Shinchu as well as wave data measured by the Shinchu buoy between 1997 and 2010 were collected. After examining all the data, it was found that there were eighteen typhoons having sufficient typhoon parameters and wave height for model building. Among the 18 typhoons, 15 of them were used for training and 3 for validation. We tested three models. The input parameters for the best model included: radius of Beaufort scale 7 wind of the typhoon, the speed of the typhoon relative to the wave station, distance between the wave station and typhoon center, bearing of the typhoon as well as effect of land topography after typhoon landing. It is to note that the latter three parameters were represented by fuzzy membership functions. For the training run, the correlation coefficient between the model wave height and measured wave height was 0.989 with a RMSE (Root Mean Square Error) of 11.8 cm. The validation run for the three typhoons showed that the peak wave height errors as a percentage of respective peak wave height were 4.2%, 26.6% and 5.7%. The errors in the time of occurrence of the peak wave height were -4, 4 and -4 hours respectively. The minus sign means that the model peak wave height occurred before the measured peak wave height. This study also showed that our model can predict satisfactorily the wave height for typhoons with a route coming from the Pacific Ocean passing through the middle of Taiwan and exiting to the Taiwan Strait at a east to west track, also known as the track III typhoons.
| Identifer | oai:union.ndltd.org:TW/101NTOU5276031 |
| Date | January 2013 |
| Creators | 高慈媛 |
| Contributors | Tsai, Chen-Han, Tsai, Jen-Chih, 蔡政翰, 蔡仁智 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 54 |
Page generated in 0.0103 seconds