Modular Tidal Level Forecasting Based On Support Vector Machine

It is generally known that tidal forecast has been widely applied in ship navigation,port construction and costal production.With the trend of larger-sized vessel,the demand of forecast precision is higher and higher.The improvement of forecast accuracy has increasingly become a hot issue.In the past 20 years,Support Vector Machine has become one of the most popular machine learning algorithms.It has unique advantages in solving non-linear problem,so that Support Vector Machine provide a new research direction in tidal prediction.In this study,Support Vector Machine is used as a tool and its applications in tidal forecast are discussed.As traditional tidal forecast method,harmonic analysis method has been widely used at present.But as a result of harmonic analysis method only consider the effect of astronomical tide in forecasting.Prediction accuracy are influenced by environment greatly,which make the method could not meet the precision requirements in some occasions.As harmonic analysis method can not predict non-astronomical tide accurately.A modular tidal forecast model is proposed based on Support Vector Machine.Support Vector Machine is used in non-astronomical part,combined with harmonic analysis method which is stable,finally improves the prediction accuracy.Firstly,according to the analysis of tides,tides are considered as composing of astronomical tide caused by celestial bodies' movement and non-astronomical tide caused by environmental disturbances and other factors.Astronomical tide is predicted by harmonic analysis method.As non-astronomical tide shows strong non-linear characteristic,while Support Vector Machine has ability of non-linear regression.So Support Vector Machine is used to predict the tidal level of non-astronomical part.The treatment of modular method takes advantages of two methods,improves the forecast precision.Then,a graphical user interface is designed for program.The interface simplifies tidal forecast and simulation process greatly.Finally,the measured tidal data of port Honolulu is used to test the forecast model and the interface.Simulation results demonstrate that the modular method has higher accuracy than conventional harmonic analysis method.The modular method has high accuracy,convenient and easy to operate.