Research Of Temperature Prediction And Completion Method Based On Oceanic Timing Data

The ocean is closely related to human survival and sustainable development.On the one hand,the ocean is rich in various energy resources,including biological resources,energy resources,tourism resources;On the other hand,there are also many problems and potential risks in the ocean,including Marine environmental pollution,climate warming,loss of Marine biodiversity,sea level rise and other common problems that are closely related to our humanity and need to be solved.And the change of the index of ocean temperature,can be said to be a barometer of ocean health,through to the ocean temperature prediction as well as the history of ocean temperature data completion,to the climate characteristics of the ocean,we can change law of development trends have a further understanding,to the Marine meteorology,Marine fishery,Marine environment,Marine biological information subject has important significance.Based on the thermohaline depth data set in global ocean Argo(Array for real-time Geostrophic Oceanography)scatter data set(V3.0),this paper selects data from 2008 to 2018 to analyze and study the change trend of ocean surface temperature.This paper selected ARIMA(Autoregressive Integrated Moving Average model),LSTM(Long Short Term Memory)neural network model,and GRU(variant GRU)of LSTM model to predict the temperature in the ocean timing data.Through the comparison and analysis of the above three models,Combined with the above three models of temperature data prediction,comparative analysis,a LSTM-GRU combined multi-layer neural network model is proposed.The model has achieved good results in the prediction of ocean time series data temperature.The LSTM-GRU combined multi-layer neural network proposed in this paper takes autocorrelation into account and introduces periodic factors and seasonal to predict the temperature of ocean time series data.And according to the length of relevant effective data before and after missing data,an improved algorithm based on LSTM-GRU combined multilayer neural network was proposed,which realized the completion of missing data for a long time and also achieved good results.Compared with the traditional ARIMA,considering the time factor,not only considered the related to salinity,depth information influence on temperature prediction,compared with single LSTM or single GRU helped model,on the prediction precision and speed of training made a compromise,and validation of the model through the Argo data sets of different latitude range of ocean temperature data prediction is very effective.In addition,the model can not only be applied to the prediction of ocean water temperature,but also to multiple aspects of time series,including salinity,carbon content,dissolved oxygen,nitrate content and other data in Argo data,and can also be used to predict the corresponding data using LSTM-GRU combined multi-layer neural network.