Research And Application Of Tidal Current Prediction Based On Sparse Linear Prediction

The processing and prediction about the tidal current signal has very important meaning and value in many aspects. In the practical work, the accuracy of the tidal current is very valued. The harmonic analysis and prediction method of tidal current is more important in analysis of characteristics of the main constituent of tidal current, but this is limited by the selected main constituent, so the accuracy of harmonic analysis model of tidal current is also restricted by the selected constituent. This text introduced signal sparse theory, then it is builded a sparse AR model. The first is builded regular AR model with the measured tidal current,then a set of complete sparse matrix are obtained, Secondly the incomplete sparse matrix extracted randomly through this complete sparse matrix, a underdetermined equations are established, it used sparse optimization calculation to obtain the most sparse AR coefficients, so to achieve reconstruction or predict tidal current. Compared with the traditional harmonic analysis and sparse AR model, this method has simple algorithm and high forecast precision.This paper first described the whole theory. These are the theory of AR model, sparse optimization principle,algorithm of AR model, algorithm of sparse optimization. Sparse linear model is set up,with AR model and sparse optimization.This model is for the tidal current prediction.Measured data of tidal current is used to this model.It is introduced that how to collect data.The tidal current in some station can stand for some area. It is explained the instrument which collect data in some area,then this model is tested by these tidal current data.Different stage,different sparse level,different accumulative total times are choosed,these are used to the AR model of sparse optimization.The model is to work out.Comparing mean square error of sparse model coefficient in various circumstances,it is to estimate the precision of the tidal current prediction model.This model contrast to harmonic analysis model of the tidal current.Analyzing the these experiment, the mean square error is less than 10 cm in the east or north component of tidal current, the AR model based on sparse linear of tidal current prediction is better the harmonic analysis method.In some area,the tidal current is more regular,the tidal amplitude is small,the wave number is little,the advantage of this model is not obvious.The mean square error only decreases 2 cm. But in some area,the tidal current is not regular,the tidal amplitude is large,the wave number is more,the advantage of this model is obvious. The mean square error decreases more than 5 cm.That is to say, the method using for the situation of other flow disturbance into the tidal current change, the deviation of this model is more small,the precision is better.