Research On Ocean Spatio-temporal Sound Velocity Prediction Based On Argo Data

Based on the global ocean Argo grid data set(BOA_Argo),this paper starts with the empirical sound velocity formula and obtains the main characteristics that affect the ocean sound velocity,namely,the indicators of ocean temperature,salinity and depth.Underwater acoustic communication technology,which uses sound waves to transmit signals underwater,is a commonly used communication method to realize underwater long-distance transmission.More accurate sound velocity is of great significance to underwater target location and round-trip delay estimation in underwater network protocol design.Considering the layered structure of the underwater environment,the speed of sound propagation in the ocean varies with the environment.In order to predict the sound velocity profile accurately in space and time,the following research is carried out in this paper.Firstly,the spatial distribution density of data in BOA_Argo is considered,especially in the vertical direction,the distance between data samples increases with the depth increasing.So this paper proposes a high spatial resolution data processing scheme based on the artificial neural network model.In this scheme,multi time span data will be processed in a unified way.Compared with single time scale data,the former is at the same level with the latter in regression evaluation index,but the efficiency is greatly improved.In the comparison of other models including KNN regression,ridge regression and decision tree,the scheme based on artificial neural network is obviously better.Then,high spatial resolution temperature and salt data and smoother sound velocity profile are obtained.Based on the high spatial resolution data set,this paper proposes a prediction method of sound velocity profile based on convlstm cyclic neural network.Based on the prediction scheme of temperature and salt data,the experimental results show that the error between the predicted value and the real value is less than that of the previous study.And the average error of sound velocity in the two experiments is 0.47m/s and 0.62m/s.Considering the problem of efficiency and space area,this paper proposes a prediction scheme of sound velocity profile based on time and space information.The experimental results show that the scheme performs well in the generalization ability of time and space.After many experiments,the average sound velocity error is no higher than 0.68m/s,which can effectively predict the multi-point sound velocity profile.