Air temperature forecast data correction method based on deep learning

Abstract

The invention belongs to the technical field of weather forecast, and particularly relates to an air temperature forecast data correction method based on deep learning. In a data preprocessing stage, a nearest neighbor interpolation method is used for converting air temperature forecast data into lattice point data, meanwhile, the spatial resolution is improved, and Gaussian filtering is adopted for carrying out smoothing processing on the air temperature data, so that Gaussian noise is removed; in the stage of constructing a deep learning network structure, the time resolution is improved by using up-sampling processing, meanwhile, time features are extracted by using LSTM, weighted fusion is performed on the time features and numerical forecasting features extracted by a UNet network, and the temperature forecasting precision is improved by using the nonlinear mapping capability of the deep learning network and the information extraction capability of lattice point data. In conclusion, according to the air temperature forecast data correction model, a more accurate correction value can be calculated, the temporal-spatial resolution of air temperature forecast can be improved, manpower consumption can be reduced, and a high-resolution and accurate-analysis correction service is provided for future refined grid point forecast.

Description

technical field [0001] The invention belongs to the technical field of weather forecasting, and in particular relates to a method for correcting temperature forecast data based on deep learning. Background technique [0002] In recent years, meteorological departments have vigorously developed smart grid forecasting business, and the leading position of high-resolution forecasting models has become more prominent. However, there are many limitations in high-resolution forecasting at present, mainly because initial value errors and model errors caused by initial conditions, boundary conditions, and physical processes cannot be eliminated, so the development of correction technology cannot be ignored. A reasonable, objective, and quantitative correction method is a bridge connecting numerical model forecasts and refined weather forecasts, and is also the key to high-resolution forecasts in the future. [0003] As one of the meteorological elements that have an important impac...

AI Technical Summary

Problems solved by technology

[0005] The present invention provides a temperature forecast data correction method based on deep learning, which uses the nonlinear mapping ability of the deep learning network and the information extraction ability of grid point data to solve the problem of relying on manual work and being unable to perform high-resolution forecasting in the statistical correction method problem and the problem of poor processing ability of a large amount of data in machine learning; Gaussian filter smoothing technology is used in the data preprocessing stage to remove Gaussian noise; at the same time, LSTM is used to extract time attributes and use it as the weight of the deep neural network. It can reduce the RMSE (root mean square error) of the forecasted temperature and effectively improve the accuracy of the temperature forecast

Images