Research On Sst Intelligent Prediction Method Of South China Sea Based On Improved EMD Algorithm

The prediction of sea surface temperature is not only helpful to the exploitation of marine resources,but also plays an important role in China’s national defense and security construction.There are two methods to forecast sea surface temperature: numerical prediction and statistical prediction.Among them,statistical prediction is a very important method of sea surface temperature prediction.As long as the amount of temperature data is large enough,it can not be affected by physical laws.With the continuous development of sea surface temperature data and reanalysis data,the coverage of sea surface temperature data is becoming wider and wider,besides,there are great increasing data.Therefore,this paper studies the prediction of sea surface temperature in the South China Sea by using statistical forecasting method and selecting suitable forecast factors.Firstly,this paper introduces the background and significance of sea surface temperature prediction in the South China Sea,and summarizes the research status of sea surface temperature prediction at home and abroad,as well as the research status of deep learning.Secondly,due to the nonlinear and non-stationary characteristics of sea surface temperature series in the South China Sea,the empirical mode decomposition(EMD)method and the improved empirical mode decomposition(EMD)method are introduced.In the first place,the process and theory of empirical mode decomposition(EMD)are introduced,and the problems of this method are pointed out.Then two Improved EMD methods are introduced,one is ensemble EMD,the other is complementary ensemble EMD.These two methods solve the mode aliasing problem of the original method.Thirdly,the traditional statistical prediction method and intelligent prediction method are introduced.Firstly,the algorithm principle and mathematical model of multiple linear regression are introduced.Then the algorithm principle and model structure of BP neural network are introduced.Then the network structure and BPTT algorithm of LSTM deep neural network,a variant of RNN,are introduced.Then,the temporal and spatial distribution characteristics of sea surface temperature in the South China Sea are studied,and the variation law of sea surface temperature in the South China Sea in the past six years is analyzed.Finally,the sea surface temperature prediction models of the South China Sea are established by multiple linear regression method and BP neural network method respectively,and the prediction effects of the two models are compared.It is proved that the prediction effect of the intelligent prediction model is better than that of the traditional statistical prediction.Then,a combined forecasting model based on the Improved EMD algorithm and BP neural network is proposed.Firstly,the original sea surface temperature series in the South China Sea is decomposed into various components.Then,each component series is predicted separately by BP neural network method.Then,the predicted results of each component are reconstructed and added to obtain the final sea surface temperature prediction value of the South China Sea.The experimental results show that the prediction effect of the hybrid model of empirical mode decomposition and BP neural network is better than that of the single BP neural network.Finally,for the shortage of BP neural network in the South China Sea SST medium and long-term prediction,a hybrid prediction model of Improved EMD algorithm and LSTM deep neural network is proposed.The experimental results show that the proposed model has smaller prediction error and improves the performance of medium and long-term prediction of sea surface temperature.