Cloud layer change trend prediction method based on deep learning

Abstract

The invention discloses a cloud layer change trend prediction method based on deep learning. The method comprises the following steps: carrying out data preprocessing on a cloud picture sequence collected from a ground station; classifying the cloud picture sequence data according to cloud layer conditions above the ground station, and determining a training set, a verification set and a test set;building a deep prediction network model based on a convolution-long short-term memory network, wherein the deep prediction network model comprises a prediction network layer, a time slice full-connection layer and a full-connection layer; training the deep prediction network model; feeding a training set into the deep prediction network model; and carrying out offline training on the model. Thecalculation overhead is reduced, the time correlation between the model and the cloud layer image sequence to be predicted is increased, and the accuracy of predicting the cloud layer change trend over the ground station in a future period of time is higher, so that an important guarantee is provided for realizing uninterrupted transmission of satellite-ground laser communication.

Description

technical field [0001] The invention is applied to the uninterrupted transmission system of satellite-to-ground laser communication, and realizes a method for predicting cloud layer change trends based on a deep learning model. Background technique [0002] Compared with the inter-satellite laser link, the satellite-ground laser link needs to pass through the atmosphere and will be absorbed and scattered by the clouds, which will seriously attenuate the laser transmission signal. Therefore, the influence of clouds on the laser link must be considered when performing satellite-ground laser communication. . Knowing the status and movement information of the clouds above the ground station in advance can provide important prior information on whether the satellite-ground is suitable for laser link building and the expected link quality after link building, thus providing important information for realizing uninterrupted transmission of satellite-ground laser communication. Ass...

AI Technical Summary

Problems solved by technology

[0004] Purpose of the invention: In order to solve the problems of low temporal and spatial resolution and high implementation cost of satellite cloud image prediction results, and overcome the shortcomings of traditional prediction methods that require modeling of clouds and cloud movement trends and high computational complexity, the present invention provides a method based on deep learning The cloud layer change trend prediction method makes the prediction results reliable and credible, and provides an important guarantee for the uninterrupted transmission of satellite-ground laser communication

Images