Research On Vertical Continuation Technology Of Marine Hydrological Element Data Based On Neural Network

As an important parameter of ocean elements,seawater temperature is usually obtained by measuring ships or buoys.The measured sea temperature is usually only a certain depth,and the temperature data of the whole sea depth at a certain time cannot be obtained.Therefore,the method of constructing a seawater temperature model can be adopted to obtain all sea depth temperature data at the same time by using a short piece of seawater temperature data.In recent years,neural network,as a research hotspot in the field of artificial intelligence,has made remarkable achievements in many aspects.At the same time,it also provides a variety of possibilities for the study of sea water temperature.Aiming at the problem of vertical extension of seawater temperature,this paper uses the method of neural network to establish the seawater temperature model,and realizes the high precision prediction of the whole sea deep water temperature through part of seawater temperature.Firstly,this paper introduces the following aspects of neural network: the general topology of the network,the setting of training parameters and the learning method of the network.On the basis of introducing the basic principle of artificial neural network,the characteristics of feedforward neural network and feedback neural network are summarized and analyzed.Combined with the project,the specific neural network algorithm is selected,and the mathematical expression of the corresponding algorithm is given.In addition,the experimental neural network is analyzed and compared according to the characteristics of different algorithms.Secondly,based on BP neural network,RBF neural network and the traditional piecewise fitting method of seawater temperature,this project constructed seawater temperature models respectively,and carried out vertical extension experiment for seawater temperature of a section depth.On the basis of the experimental results,the overfitting phenomenon of the neural network method is solved.In addition,three kinds of extension methods are analyzed and summarized.Finally,based on the improvement of the hidden layer structure of RBF neural network,this project proposes to introduce the characteristics of neuronal activation activity and adjust the structure of the neural network to obtain a neural network algorithm with more compact structure and stronger generalization ability.At the same time,the same background field is used to verify the algorithm.Experimental results show that the improved RBF neural network algorithm has higher adaptability and effectiveness than the other three experimental methods.The vertical extension method of sea water temperature based on neural network proposed in this paper solves the problem of low precision of deep sea water temperature caused by the discontinuity of sea water temperature collection in the vertical direction,and innovates and expands the application of neural network in the field of atmosphere and ocean.