Research And Implementation Of Forecasting Method Of Navigation Channel Current Based On Deep Learning

With the rapid development of economy,waterway transportation becomes more and more complex.Affected by weather and other factors,the voyage schedule of ocean transportation is not fixed,and the ships are easy to meet danger,in which navigation channel current is an important factor.Therefore,the real-time forecast of currents is of great significance for the risk avoidance of water transportation.This article starts from the prediction of the speed and direction of navigation channel current,and takes comprehensive consideration of main factors influencing the channel currents.Then,combining with periodic and random,etc.features of the current itself,we analyze the correlation between some factors and the current,such as temperature,humidity,wind speed,direction.Furthermore,we construct deep neural network forecast models of channel currents by taking these characteristics and channel current data as input so as to realize the channel currents accurate forecast.Considering the complexity of time series data prediction of navigation current,this paper bases on the Long Short-Term Memory(LSTM)to simulate the channel current prediction in order to achieve accurate and stable prediction results.This paper considers the influences of input time step,hidden layer number of LSTM,output time step and other factors on the accuracy and stability of the model,and input several groups of data to the model for testing so as to analyze the prediction results of the model.By comparing with BP and CNN,it is proved that the prediction accuracy and stability of LSTM are greatly improved under the premise of increasing the prediction step.Aiming at the disadvantage that the prediction accuracy of the LSTM model decreases with the increase of the prediction step,this paper further proposes the LSTM-CNN channel current forecast model based on the attention mechanism.In addition,this paper further proposes the Bi LSTM-CNN channel ocean current prediction model based on attention mechanism.The experimental results show that,compared with the LSTM neural network prediction model,the accuracy and stability of the LSTM-CNN model based on attention mechanism and the Bi LSTM-CNN model based on attention mechanism are effectively improved under the premise of increasing the prediction step.To solve the problem of channel current forecast,a current prediction system is developed,in which users can select appropriate prediction models according to the speed and direction of the corresponding currents.The experimental results show that the system can accurately forecast the ocean currents and provide guidance information for the stable operation of the water transportation.