碩士 / 國立成功大學 / 水利及海洋工程學系碩博士班 / 100 / Current is one of the ocean motions and has an effect on human activities such as fishery, navigating, constructing and marine environment variations. Ocean current is even relating with the global climate changes. Current measuring is significant for offshore engineering planning, construction and operating. This research afford a current predicting and missing data supplying technique by using a datum station’s measured data to predict and supply missing data in surroundings.
The GMDH (Group Method of Data Handling) algorithm of self- organization network is being used as the basic configure in this paper to build up the GMDH current predicting and missing data supplying model with field observation current data of the datum station and surroundings. A recursive mode can assess the estimated error if it exceeds the identified threshold and then self adjusted the original model by updating the data input to make the model possible to achieve the long term prediction and accurate estimation. Once the recursive mode of this model could not enhance the estimation accuracy, the estimation error is then added in model organizing variables to reset up the error revised model, and then cooperates with the original GMDH model with recursive mode together to get reasonable current simulations.
Current data measured by ADCPs (Acoustic Doppler Current Profilers) about 7 meters beneath the sea surface at 3 locations in between Hsinchu to Miaoli maritime space and tide records obtained by Hsinchu data buoy and Yuankang fishing port were used as the input to set up the GMDH current predicting and missing data supplying model. The modeling procedure showed that GMDH algorithm is better than that of SGMDH (Stepwise regression GMDH) and revealed that events of Chungkang~Tunghsiao in Dec. 2010 of velocity simulation and direction simulation in Apr. 2011 are the optimum modeling. Hence trying to apply the data of a datum station (Chungkang station) to predict current and supply missing data in surroundings is the primary consideration, this paper proposed alternatively the sub-optimum modeling results which were set up by events of Hsinchu~Chungkang in May 2011 of velocity simulation and Chungkang~Tunghsiao in Dec. 2010 of direction simulation as the current predicting and missing data supplying model in Hsinchu~Miaoli maritime surroundings. The result showed that the overall average RMSE prediction errors of current velocity and direction are 0.06 m/s and 31.68 deg. respectively. The recursive mode was then proposed to test the facility of time variant property with the threshold of current velocity (± 0.05 m/s) and direction (± 15 deg.) by updating the measuring data and tests showed the simulation advanced the velocity estimation results by 13.57% improving and get a limit direction estimation improvement of 6.92% in overall RMSE comparisons. In order to get better direction forecasting, the estimated errors were treated as one of the model input variables to build up the error revised model and cooperate with the GMDH model to simulate once again, the analysis resulted that directions estimating achieve more superior improvement of 63.89~68.59% in overall RMSE range comparing with the original test and resulted a acceptable average RMSE value of 10.92 deg. So this GMDH current predicting and missing data supplying model proposed by this paper with the recursive mode as well as introducing the error revised model on time can be applied in practical usage of current predicting and missing data supplying in surroundings by using the current data of specific datum station.
| Identifer | oai:union.ndltd.org:TW/100NCKU5083129 |
| Date | January 2012 |
| Creators | Ming-YiLin, 林明毅 |
| Contributors | Pei-Hua Yan, 顏沛華 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 106 |
Page generated in 0.0078 seconds