| Visibility is very important for the normal operation of airports.A continuous low visibility process not only increases the operating cost of airports but also significantly reduces the air transportation capacity of airports.Therefore,it is of great significance to improve visibility forecasting to ensure safe airport operations.With the development of artificial intelligence technology,deep learning has been widely used in natural language processing,image recognition,and other fields,which can fully explore the mapping relationship between data and thus improve the forecasting performance of the model.In this paper,we take Shenzhen airport as the research object and build an airport visibility forecast model based on the meteorological information provided by METAR(Meteorological Terminal Aviation Routine Weather Report)reports and ERA-Interim(ECMWF Re-Analysis-Interim)reanalysis data.The objective is to improve the visibility forecasting level of Shenzhen airport.First,the LSTM(Long Short-Term Memory)visibility forecasting algorithm based on the historical observation data of the airport is proposed.In this paper,a visibility forecasting model is built based on the LSTM algorithm and the meteorological information provided by the METAR message of Shenzhen airport,and the low visibility process of Shenzhen airport in March 2016 is tested using the model.The experiments show that the model can effectively predict the visibility trend,among which,the model with a forecast duration of 3 hours has the best forecast effect,and the combination of wind field,barometric pressure,relative humidity and temperature has the best effect on the optimization of the visibility forecast of the model.Secondly,an LSTM visibility forecasting algorithm based on the optimization of sensitive areas is proposed.In this paper,the correlation between the meteorological fields provided by ERA-Interim reanalysis data and the visibility of Shenzhen Airport is analyzed based on the Person correlation coefficient.The meteorological elements obtained based on the sensitive area analysis method are combined with the LSTM algorithm,and then the visibility forecast model is constructed and tested for the low visibility process at Shenzhen Airport in March2016.The experiments show that the 3-hour forecast is the best among all models of forecast time,and among all combinations of meteorological elements,the 10-meter wind field,sea level pressure,2-meter relative humidity,and 2-meter temperature combinations have the best forecast effect,while the forecast effect is significantly improved after comparing this model with the visibility forecast model based on the historical observation data of the airport only.Finally,the visibility forecasting algorithm based on CNN and LSTM is proposed.In this thesis,the visibility forecasting model is constructed using CNN and LSTM algorithms combined with meteorological fields provided by ERA-Interim reanalysis data,and the model is used to test the low visibility process at Shenzhen Airport in March 2016.The experiments show that the 3-hour model has the best forecast effect among all forecast time models,and the combination of 10-meter wind field,sea level pressure barometric field,2-meter temperature field,and 2-meter relative humidity field has the best forecast effect among all combinations of meteorological elements.Finally,after comparing this model with the visibility forecasting model based on the combination of single-station meteorological elements,it is found that the forecasting effect is significantly improved. |
