碩士 / 國立交通大學 / 土木工程系 / 89 / Genetic algorithm was applied to finding the fittest amplitude and phase lag of each tidal constituent and then to forecasting long-term tidal levels. The present model needs only 540-720 hours’ tidal data instead of a continuous tidal record for 369 days used in harmonic analysis method. The variation of astronomical tides due to topography and temperature was found by moving Gaussian average method. An artificial neural network model was proposed to deseasonalize tidal levels related to temperature. That the forecasted tidal levels after moving average of 720 hours display a variation with large period or not is suggested to be an examination for prediction accuracy. A 11-constituent model is recommended to forecast tidal levels when only tide record of 540 hours is needed to be input and has an equivalent prediction capability as the harmonic method that needs more than three-month tidal data does. The other 23-constituent model with an input of only 720-hour data for forecasting capability is compared with the harmonic analysis method with an input of 6-month data.
Identifer | oai:union.ndltd.org:TW/089NCTU0015020 |
Date | January 2001 |
Creators | Hao-Cheng Wang, 王皓正 |
Contributors | Hsien-Kuo Chang, 張憲國 |
Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
Language | zh-TW |
Detected Language | English |
Type | 學位論文 ; thesis |
Format | 108 |
Page generated in 0.0087 seconds