Research On Prediction For The Ship-to-anchor Freight Volume Based On EEMD And Phase Space Reconstruction

With the rapid growth of freight volume in the Yangtze River waterway transportation,the problem of limited design capacity in the Three Gorges ship-lock is arising,which can't satisfy the demands of the Yangtze River transportation.It also brings negative effect to the development of waterway transportation and economy along the river.Design capacity of the lock depends on the construction scale,besides,the construction scale lies on the future freight volume.Therefore,improving the prediction accuracy through scientific and effective methods has great significance on construction,management and program of the ship-lock.This paper discussed prediction model of the Three Gorges ship-lock daily ship-to-anchor freight volume,which based on the technique of Least Squares Support Vector Machine(LSSVM),Phase Space Reconstruction(PSR)and Ensemble Empirical Mode Decomposition(EEMD).The main contents discussed in this paper are as follows:Firstly,according to chaotic nature of the Three Gorges ship-lock daily ship-to-anchor freight volume,a prediction model was established based on PSR-LSSVM.In order to find its optimal parameters,this paper proposed a nonlinear decreasing inertia weight particle swarm algorithm.This model effectively improved the prediction accuracy compared with LSSVM model because the PSR theory avoided the blindness and randomness in determining the input set mode by experience.Secondly,on account of the non-stationary characteristics of the Three Gorges ship-lock daily ship-to-anchor freight volume,it was difficult to obtain a better prediction result by using the LSSVM model.And Ensemble Empirical Mode Decomposition,an adaptive non-stationary and nonlinear processing method,can decompose the original time series into a series of relatively stable components.In view of this,a hybrid prediction model was proposed based on ensemble empirical mode decomposition and the PSR-LSSVM(EEMD-PSR-LSSVM).Verifying validity of the hybrid prediction model and comparing with the prediction results of PSR-LSSVM prediction model by empirical analysis,the results showed that the prediction results based on EEMD-PSR-LSSVM hybrid forecasting model matches the actual data better,which verifies the validity of the EEMD-PSR-LSSVM hybrid forecasting model,and the prediction accuracy of EEMD-PSR-LSSVM hybrid forecasting model is much higher than the PSR-LSSVM model.