Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Deep learning-based severe convection extrapolation method and system under multiple scales

A deep learning and strong convection technology, applied in the field of strong convection extrapolation methods and systems, can solve the problems of rapid deterioration in the quality of the extrapolation effect and poor MSE performance, and achieve the effect of improving performance and generalization ability

Active Publication Date: 2021-08-10
CHENGDU UNIV OF INFORMATION TECH
View PDF7 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0019] e) MSE does not perform well on extrapolation tasks
[0020] Most extrapolation-based studies use MSE loss, which will lead to a rapid decline in the quality of multiple extrapolation effects, so we need to propose an optimization method that is more in line with this business

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Deep learning-based severe convection extrapolation method and system under multiple scales
  • Deep learning-based severe convection extrapolation method and system under multiple scales
  • Deep learning-based severe convection extrapolation method and system under multiple scales

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0078] refer to figure 1 It is a structural diagram of a multi-scale strong convective extrapolation system based on deep learning in the present invention. Specifically, the system includes:

[0079] Feature extraction module 1 is used to receive radar image data and extract hidden state features of radar image data;

[0080] In this embodiment, the feature extraction module 1 includes a VGG16 network, and the VGG16 network is used to extract the Doppler radar base data of the radar image data, and then extract the hidden state features of the radar base data according to the Doppler radar base data; wherein,

[0081] Doppler radar base data includes radar basic reflectivity, radar combined reflectivity, radar basic radial velocity, and combined radial velocity.

[0082] Further, the VGG16 network adopts convolutional layers by stacking small-sized convolutional kernels.

[0083] The strong convective extrapolation module 2 is connected with the feature extraction module, a...

Embodiment 2

[0098] refer to figure 2 and image 3 , which is a flowchart of a multi-scale strong convective extrapolation method based on deep learning in this embodiment, image 3 In BN, normalization is to normalize each batch of data, and tanh is a hyperbolic tangent function. The method includes the following steps:

[0099] S1: Receive radar image data and extract hidden state features of radar image data;

[0100] In this step, the Doppler radar base data of the radar image data is extracted through the VGG16 network, such as the radar basic reflectivity at an elevation angle of 0.5°, the radar combined reflectivity, the radar basic radial velocity at an elevation angle of 0.5°, and the combined radial velocity Wait;

[0101] refer to Figure 4 It is the structure diagram of the VGG16 network model in this embodiment. Among them, the first one is the original diagram. The VGG16 network in this embodiment is different from the usual ones. The input Doppler radar base data is imp...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention provides a multi-scale severe convection extrapolation method and system based on deep learning, and the method comprises the following steps: receiving radar image data, and extracting the implicit state features of the radar image data; carrying out convolution on the implicit state features, inputting a convolution result into a TrajGRU network, and carrying out severe convection extrapolation to obtain a radar map; carrying out second convolution on the radar map and carrying out batch regularization at the same time to obtain extrapolation image data. According to the method, radar image data is trained to obtain an extrapolation image, and the extrapolation image is used for forecasting severe convection weather such as rainstorm, thunderstorm and hail.

Description

technical field [0001] The invention belongs to the technical field of computer artificial intelligence and meteorology, and specifically relates to a multi-scale strong convective extrapolation method and system based on deep learning. Background technique [0002] Severe convective weather generally refers to extreme weather phenomena with a disastrous nature, such as convective gale, hail, and short-term heavy precipitation accompanied by thunderstorms. The characteristics of this type of weather generally include: sudden occurrence, rapid movement, severe weather, and strong destructive force. It mainly occurs in small and medium-scale weather systems with a small spatial scale. The general horizontal range is about ten kilometers to two to three hundred kilometers. Generally less than 200 kilometers, and some horizontal ranges are only tens of meters to more than ten kilometers. Its life history is short and has obvious suddenness, about one hour to more than ten hours...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06K9/00G06N3/04G06N3/08G01W1/10
CPCG06N3/08G01W1/10G06V20/13G06N3/044G06N3/045Y02A90/10
Inventor 文立玉罗飞柴文涛卫霄飞
Owner CHENGDU UNIV OF INFORMATION TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com