Research Of Visibility Prediction System Based On Deep Learning

With the frequent occurrence of severe haze weather,atmospheric visibility has become worse and worse.At present,more than 99% of the 500 largest cities in China that the air quality can not met the requirements of the World Health Organization.The continuous decline of visibility has brought many inconvenience and harm to people's production or life,even cause safety accidents.Therefore,the meteorological factors affecting visibility are studied and analyzed.Accurate prediction of visibility is helpful to reduce the occurrence of traffic accidents and prevent,and cure of the air pollution in cities.In order to solve the shortcomings of the original visibility prediction methods,such as weather map analysis,multiple regression model and fuzzy identification.this paper attempts to use the neural network model of deep learning to predict visibility.First of all,according to the meteorological factors that affect the visibility of the atmosphere,we choose wind speed,wind direction,air pressure,air temperature,humidity,PM2.5 of six elements as the network input and visibility as the network output.Secondly,the data are pretreated by min-max standardization method to get the normalized data.Then,the DBN atmospheric visibility prediction model of optimal structure which is the number of hidden layers and the number of nodes in each of the hidden layer are determined by trial and error.At the same time,DBN training algorithm is used to adjust the weight and bias of the model continuously,and the DBN atmospheric visibility prediction model with the least error is established.In addition,the traditional artificial neural network BP model is constructed by using the same sample set data.Compared the effect of atmospheric visibility prediction with DBN atmospheric visibility prediction model,it proves that the effect of depth learning DBN model under certain condition is better than the traditional BP neural network model in predicting visibility.Finally,a DBN model atmospheric visibility prediction system is designed and implemented by GUI,which intuitively shows the results of visibility prediction of DBN model.