Research On Marine Meteorological Prediction Model Based On Spatio-temporal Data Model

Weather forecast has become an indispensable information service in people's production and life.Providing precise and accurate marine meteorological forecasts is of great practical significance for ensuring the safety of various types of marine operations and avoiding the occurrence of marine accidents.At present,modern information technology represented by computer technology has made meteorological data explosively accumulated,and puts forward higher requirements on the level of marine meteorological prediction technology.Based on the above research background,this paper studies the regression and prediction of marine meteorological elements,including wind speed,air temperature,air pressure,relative humidity,rainfall and visibility data,and proposes a marine meteorological prediction model based on spatio-temporal data model,which mainly includes the following three Work on:(1)Autocorrelation judgment of time series and model variable selection.The autocorrelation of time series data is the premise of regression prediction.Based on the autoregression function and partial autoregression function,this paper judges that the marine meteorological time series data has autocorrelation.In order to increase the prediction accuracy of the model,this paper considers the correlation between multiple meteorological elements and the spatial characteristics of the meteorological system.Based on the autoregressive model,external variables and spatial variables are introduced,and correlation analysis between different variables is carried out To select external variables and spatial variables of the model.(2)Data outlier processing and missing value filling.In order to improve the stability of the model,this paper deals with the outliers of marine meteorological data based on the box chart method.In order to improve the sample data information,this paper uses the Kalman filter,K-nearest neighbors and random forest algorithm to fill in the missing values of the sample data,and compares the interpolation effect to select the most effective missing value filling algorithm.(3)Establishment and prediction of spatiotemporal data model.In this paper,unit root test is used to judge the stationarity of the data and non-stationarity data is treated with differentiation.Based on the idea of multiple linear regression,it is analyzed from two perspectives: time and space,with time variables,external variables,and space variables as model inputs.In the future,meteorological data will be used as the model output to establish a spatio-temporal data model,and the least square method will be used to estimate the unknown parameters of the model,and the Akichi information criterion will be used to select the optimal delay order of the model to obtain the optimal model structure.Compared with the time series analysis method commonly used in Marine meteorological prediction technology,which only evaluates the properties of the predicted variables,this paper also considers the implied correlation and spatial correlation among the meteorological variables,and proposes a Marine meteorological prediction model based on the spatio-temporal data model.The results of numerical experiments show that the spatio-temporal data model proposed in this paper has higher prediction accuracy than the autoregressive model and other traditional regression models,which verifies the validity of the spatio-temporal data model.