Short-term prediction of residual water level by using LSTM model with regard to time difference
At present,the residual water level prediction adopts a single-station method,which is only based on the autocorrelation of the residual water level.Aiming at the prediction of residual water level at large-scale coastal tidal stations,this paper combines the spatial correlation and the time difference information of residual water levels between stations.In the form of a"reference station-auxiliary station"tidal gauge station group,we use historical synchronous residual water level data to train a multivariable LSTM(Long Short-Term Memory)neural network to achieve residual water level prediction at reference stations.The results of four long-term tidal stations along the Bohai coast shows that the prediction accuracy of the method proposed in this paper is better than that of the single station independent prediction method,which can significantly increase the prediction time length.The method proposed in this paper can forecast the residual water level in large-scale mautical dynamic water level assurance.
ocean tidedynamic water levelresidual water level predictionlong short-term memory network modeldifference of tidal time
万方数据
维普
