基于机器学习对机场能见度预测模型研究

Research on Airport Visibility Prediction Model Based on Machine Learning

为进一步精确预测茅台机场能见度变化趋势,论文以2018年全年机场气象室资料,先对茅台机场2018年全年能见度变化趋势进行统计分析和对所观测到的气象要素进行Pearson相关性分析,发现相对湿度与机场能见度变化呈现明显的正相关性,温度露点差与机场能见度呈现显著负相关,然后通过多元线性回归(MLR)和BP神经网络、径向基(BBF)神经网络分别对茅台机场的能见度进行预测,对比拟合优度、误差平均值、均方误差和均方根误差发现,相较于MLR和BP神经网络,RBF神经网络模型不管是在误差控制还是预测精度上都有着较好的表现,所以可选取其作为预测茅台机场能见度的模型,对茅台机场的安全运行有着重要意义。

In order to further accurately predict the visibility change trend of Maotai Airport,this paper first conducts statisti⁃cal analysis on the visibility change trend of Maotai Airport in 2018 and Pearson correlation analysis on the observed meteorological elements based on the annual airport meteorological office data in 2018.It is found that the relative humidity is significantly positive⁃ly correlated with the airport visibility change,and the temperature dew point difference is significantly negatively correlated with the airport visibility.Then,the visibility of Maotai Airport is predicted by multiple linear regression(MLR),BP neural network and radi⁃al basis function(BBF)neural network respectively.By comparing the goodness of fit,average error,mean square error and root mean square error,it is found that RBF neural network model performs better in error control and prediction accuracy than MLR and BP neural networks.Therefore,it can be selected as the visibility prediction model of Maotai Airport,which is of great significance to the safe operation of Maotai Airport.