Study On The Forecasting Algorithm Of The Qiantang River Tidal Bore Based On Gaussian Process

The Qiantang River is influenced by the special bell-mouth topography of Hangzhou Bay and the gravity of celestial bodies and the rotation of the earth,forming the Qiantang River tidal,which is a natural wonder of the world and has a very high ornamental value.However,the tidal phenomenon is also a process of energy release.When the tidal bore comes,the front almost stands steeply,which is harmful and destructive.Accurate tidal bore prediction can provide an important basis for ensuring the safety of residents along the river and the safety of ship navigation.In this paper,a method of tidal bore prediction based on gaussian process is put forward,which is based on discrete wavelet transform of tidal level sequence,statistics of tidal lag to tidal moveout data and statistical analysis of the regularity of tidal bore height.The main tasks of forecasting the tidal level,tidal time and tidal bore height of Qiantang River are as follows:(1)Forecast of tidal level based on gaussian process regression.Firstly,the missing values in the tidal level monitoring data are interpolated by the moving average window method,and then the tidal level sequence is decomposed by db4 wavelet function according to the Mallat algorithm.The correlation between the trend series and the approximate sequence under the condition of 1 ~ 8 days lag with the original tidal level sequence was studied.And according to the results of correlation analysis,the wavelet decomposition sequence,which has a good correlation with the original tidal level sequence,is superposed as the basis for constructing the input sample of the model.Then the influence of covariance function on gaussian process regression prediction distribution is analyzed,and on this basis,the covariance function suitable for the trend of tidal level data itself is selected,and then the trend regression prediction model of tidal level gaussian process is established.The tidal level of Yanguan,Cangqian and Ganpu stations was predicted in one step.The experimental results show that the preprocessing of discrete wavelet transform and the reasonable selection of covariance function are effective in the prediction of tidal level series.(2)Particle swarm optimization optimization gaussian process regression model.In view of the shortcomings of traditional conjugate gradient method in searching superparameters such as too strong dependence on initial value,difficult to determine the number of iterations,and easy to fall into local optimum,the particle swarm optimization(PSO)algorithm is proposed to intelligently search the superparameters of Gaussian process.The regression coupling model of particle swarm gaussian process is established.(3)Prediction of tidal level,tidal time and tidal bore height of Qiantang River based on particle swarm optimization gaussian process regression model.Firstly,the tidal level series is predicted by using the particle swarm gaussian process regression model,and the prediction results are compared with those of the traditional Gaussian process regression model.The feasibility of the regression model of Gaussian process in prediction of tidal level is obtained.Then,on the basis of the statistics of the interval lag to tidal time difference,the Gaussian process model of tidal time delay particle swarm optimization is established,and the prediction results are compared with the traditional time-lag model.Finally,on the basis of statistical analysis of the internal implication law of tidal bore height series at stations along Qiantang River,a regression model of gaussian process for tidal bore height is proposed.The tidal height of the upstream station is predicted according to the height of the tidal bore at the downstream station.In addition,the accuracy of the model is evaluated by a certain error index.The experimental results show that the prediction error of gaussian process regression model is small,and it is feasible in Qiantang River tidal prediction.