深度学习方法在玉溪市烤烟种植区冰雹预测中的初步应用

PRELIMINARY APPLICATION OF DEEP LEARNING METHOD IN HAIL PREDICTION IN YUXI CITY TOBACCO PLANTING AREA

冰雹预测最核心的是判断所提供的数据是具有降雹特征还是非降雹特征,同时冰雹云的发展,跟时间序列有密切关系,因此可以用深度学习方法建立冰雹预测模型。将玉溪市2006—2018年的降雹过程雷达参数通过数据标准化,利用深度学习方法,学习降雹雷达回波特征,建立冰雹预测模型。将2019—2021年的降雹过程和非降雹过程数据用于预测检验。经验证:降雹预测模型的准确率达82.6%。证明了在气象领域使用深度学习方法预测冰雹天气的可行性。

The core of hail prediction is to determine whether the provided data has hail or non hail characteristics,and the development of hail clouds is closely related to time series.Therefore,deep learning methods can be used to establish hail prediction models.Standardize the radar parameters of the hail process in Yuxi City from 2006 to 2018 through data standardization,use deep learning methods to learn the radar echo characteristics of hail,and establish a hail prediction model.Use hail process and non hail process data from 2019 to 2021 for prediction testing.After verification,the accuracy of the hail prediction model reached 82.6%.Proved the feasibility of using deep learning methods to predict hail weather in the meteorological field.