Research On Prediction Algorithm Of Sea Surface Temperature Based On Space-time

The ocean has a significant impact on climate change and ecosystem development.As one of the most important indicators in the ocean hydrological system,ocean surface temperature has a significant impact on climate,precipitation and marine biological activities,and is a key factor leading to marine ecological disasters.The existing sea surface temperature forecasting methods suffer from the problems of inability to accurately mine the connection between data,low data processing speed and poor prediction accuracy.Therefore,this thesis accurately analyzes the trend of sea surface temperature based on the deep learning method and conducts a prediction study of both temporal and spatial aspects of the ocean surface temperature:Firstly,an SLA model of ocean surface temperature prediction model based on a decomposition-based fused attention mechanism is proposed to address the lack of study of potential features of data in ocean surface temperature prediction tasks.The model uses the STL decomposition algorithm to decompose the ocean surface temperature series and obtains the decomposed ocean surface temperature with trending features and periodic seasonal features;the temporal features are extracted from the ocean surface temperature by an improved LSTM model,and the improved LSTM model is fused with the attention mechanism to calculate the influence of the historical ocean surface temperature on the predicted ocean surface temperature Weights.Secondly,a CLED model based on spatio-temporal feature fusion for ocean surface temperature prediction task is proposed to address the problem of lack of consideration of spatial geographic location information.The model uses two encoders,a spatial encoder and a temporal encoder,to spatially encode and temporally encode the ocean surface temperature;the spatial and temporally encoded information from the two encoders are feature fused,and then the fused features are decoded by an improved LSTM,which solves the problem of spatial feature extraction in the sea surface temperature prediction task.Finally,the sea surface temperature data of longitude and latitude 112.125 E,10.125 N and longitude and latitude 119.875 E,28.875 N were taken in the South China Sea and Bohai Sea respectively,and compared with other models for analysis,which proved that the decomposition of sea surface temperature can improve the prediction accuracy of sea surface temperature,and verified the effectiveness of the sea surface temperature prediction model SLA proposed in this paper.Meanwhile,experiments were conducted on sea surface temperature data sets in the latitude and longitude ranges of 111.625 E to 112.625 E,9.625 N to 10.625 N and 119.625 to 120.625 E,28.625 N to 29.625 N,and compared and analyzed with other models to validate the effectiveness of the sea surface temperature prediction model CLED proposed in this paper with the fusion of spatial and temporal features.