A method of forecasting lightning

Abstract

The invention belongs to the technical field of lightning early warning and discloses a method for predicting lightning. It includes three stages of data preprocessing, lightning prediction and output prediction results. In the data preprocessing stage, images are used to represent the attributes of the data to be processed and fuzzy. In the lightning prediction stage, convolution operations are used to extract the features of the blurred image. The function reflecting the change of lightning properties is obtained after encoding by the cyclic neural network, and then the deconvolution operation is performed to decode the future changes of the lightning properties represented by the function into a dynamic grayscale image; this method eliminates non-convective echoes or non-lightning positions The redundancy and interference caused by the echo construct the local spatial characteristics and time series variation of thunderstorm clouds directly related to the occurrence of lightning, thereby improving the accuracy of lightning prediction.

Description

technical field [0001] The invention belongs to the technical field of lightning early warning, and specifically discloses a method for predicting lightning. Background technique [0002] Lightning is the product of strong convective weather process. The process of strong convective weather evolves rapidly, and the formation, dissipation, merging and splitting of cloud clusters are complicated, which makes it difficult to forecast and warn of thunder and lightning. At present, there are mainly the following methods for lightning forecasting: [0003] Using atmospheric electric field instrument or its network to forecast lightning. For example, Chinese patent CN110297284A "A Method for Tracking the Moving Path of Thunderstorm Cloud Based on a Three-dimensional Atmospheric Electric Field Instrument" discloses a method for detecting the azimuth of thunderstorm cloud and tracking its moving path by using an atmospheric electric field instrument. Chinese patent CN106353604B "A...

AI Technical Summary

Problems solved by technology

[0009] Generally speaking, the current lightning forecasting methods put forward relatively simple criteria, usually using the threshold or correlation formula of a certain physical parameter (such as atmospheric electric field strength or its rate of change, rate of change of cloud top brightness temperature, and radar echo dBZ) to determine the It is difficult to accurately describe the complex laws of thunderstorm cloud movement and generation, disappearance, accumulation, and change by predicting the occurrence or moving path of thunderstorms

Images