基于神经网络方法的短邻大雾预测

Prediction of Short-Term Heavy Fog Based on Neural Network Method

摘要: 机器学习已经在气象领域强对流天气的研究中取得了进步，但是在大雾短临预测的研究中较少。针对该问题，本文利用神经网络方法建立模型实现对未来时刻大雾的预测。首先根据气象学对有雾，无雾的定义建立不同短时邻近时刻样本集。然后，利用神经网络方法对样本集进行训练建立不同短时临近时刻雾预测模型并通过测试数据集检验神经网络的预测能力。最后使用实时天气预报系统中常用的气象学方法TS评分来评估预测模型的性能。

Abstract: Machine learning has significantly made progress in the researches of strong convection weather, but there is little research on the prediction of short-term fog. For this problem, we adapt the model of neural network to solve the problem of short-term fog prediction. Firstly, dataset of different short-term time is established based on the meteorological definition of fog and non-fog weather. Then these data are inputted into neural network to train and develop the corresponding short-term predictive model, respectively. Meanwhile, we use some test data to test the predictive performance of the model. Eventually, we use the meteorological approach—TS Score to validate the predictive capability on neural network.