Disaster weather forecasting method based on energy generation antagonistic predictor

Abstract

The invention relates to a disaster weather forecasting method based on an energy generation adversarial predictor. The disaster weather forecasting method comprises the following steps: S1, acquiringa radar echo image sequence for weather short-term and short-term forecasting; and S2, inputting the radar echo image sequence into an energy-based EBGAN predictor to generate a weather short-term and short-term forecasting result, and inputting a radar echo image sequence used for carrying out disastrous weather proximity prediction into a model of the EBGAN predictor. According to the method, the image feature information in the radar echo image sequence is acquired through the model, so that the image feature information can be analyzed, the weather short-term and short-term forecast can be acquired more conveniently, and the forecast is more accurate.

Description

technical field [0001] The invention relates to the field of weather forecasting, more specifically, to a disaster weather forecasting method based on an energy generation confrontational predictor. Background technique [0002] The existing disastrous weather nowcasting methods mainly include the optical flow method, single centroid method, and cross-correlation method, and these methods have large errors in the forecast results in a short period of time, and the error of the forecast results after a certain period of time Even bigger. [0003] Due to the highly dynamic and chaotic nature of meteorological movement, severe weather nowcasting, especially precipitation nowcasting is an important task in weather forecasting. In the mission, the fundamental problem is radar echo image extrapolation. [0004] Once the extrapolation is determined, precipitation prediction can be achieved through traditional methods such as the Z-R relationship. As classical backbones, optical ...

AI Technical Summary

Problems solved by technology

[0005] However, these methods suffer from three important disadvantages of optical flow:

However, GA-ConvGRU suffers from the inherent disadvantages of many GAN methods, i.e. the generator and discriminator are not easy to coordinate training, suffer from training instability, which often leads to poor performance

Also, the problem with the blur effect still occurs

Images